poly_int: vect_get_constant_vectors
[official-gcc.git] / gcc / ChangeLog
blobc25eb15242cb6f84e06e040df3fb02d48bac7b46
1 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
2             Alan Hayward  <alan.hayward@arm.com>
3             David Sherwood  <david.sherwood@arm.com>
5         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
6         constant and extern definitions for variable-length vectors.
7         (vect_get_constant_vectors): Note that the number of units
8         is known to be constant.
10 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11             Alan Hayward  <alan.hayward@arm.com>
12             David Sherwood  <david.sherwood@arm.com>
14         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15         of units as polynomial.  Choose between WIDE and NARROW based
16         on multiple_p.
18 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
19             Alan Hayward  <alan.hayward@arm.com>
20             David Sherwood  <david.sherwood@arm.com>
22         * tree-vect-stmts.c (simd_clone_subparts): New function.
23         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26             Alan Hayward  <alan.hayward@arm.com>
27             David Sherwood  <david.sherwood@arm.com>
29         * tree-vect-stmts.c (vectorizable_call): Treat the number of
30         vectors as polynomial.  Use build_index_vector for
31         IFN_GOMP_SIMD_LANE.
33 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34             Alan Hayward  <alan.hayward@arm.com>
35             David Sherwood  <david.sherwood@arm.com>
37         * tree-vect-stmts.c (get_load_store_type): Treat the number of
38         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
39         for variable-length vectors.
40         (vectorizable_mask_load_store): Treat the number of units as
41         polynomial, asserting that it is constant if the condition has
42         already been enforced.
43         (vectorizable_store, vectorizable_load): Likewise.
45 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
46             Alan Hayward  <alan.hayward@arm.com>
47             David Sherwood  <david.sherwood@arm.com>
49         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
50         of units as polynomial.  Punt if we can't tell at compile time
51         which vector contains the final result.
53 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
54             Alan Hayward  <alan.hayward@arm.com>
55             David Sherwood  <david.sherwood@arm.com>
57         * tree-vect-loop.c (vectorizable_induction): Treat the number
58         of units as polynomial.  Punt on SLP inductions.  Use an integer
59         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
60         cast of such a series for variable-length floating-point
61         reductions.
63 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
64             Alan Hayward  <alan.hayward@arm.com>
65             David Sherwood  <david.sherwood@arm.com>
67         * tree.h (build_index_vector): Declare.
68         * tree.c (build_index_vector): New function.
69         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
70         of units as polynomial, forcibly converting it to a constant if
71         vectorizable_reduction has already enforced the condition.
72         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
73         to create a {1,2,3,...} vector.
74         (vectorizable_reduction): Treat the number of units as polynomial.
75         Choose vectype_in based on the largest scalar element size rather
76         than the smallest number of units.  Enforce the restrictions
77         relied on above.
79 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
80             Alan Hayward  <alan.hayward@arm.com>
81             David Sherwood  <david.sherwood@arm.com>
83         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
84         number of units as polynomial.
86 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
87             Alan Hayward  <alan.hayward@arm.com>
88             David Sherwood  <david.sherwood@arm.com>
90         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
91         * target.def (autovectorize_vector_sizes): Return the vector sizes
92         by pointer, using vector_sizes rather than a bitmask.
93         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
94         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
95         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
96         Likewise.
97         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
98         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
99         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
100         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
101         * omp-general.c (omp_max_vf): Likewise.
102         * omp-low.c (omp_clause_aligned_alignment): Likewise.
103         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
104         * tree-vect-loop.c (vect_analyze_loop): Likewise.
105         * tree-vect-slp.c (vect_slp_bb): Likewise.
106         * doc/tm.texi: Regenerate.
107         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
108         to a poly_uint64.
109         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
110         the vector size as a poly_uint64 rather than an unsigned int.
111         (current_vector_size): Change from an unsigned int to a poly_uint64.
112         (get_vectype_for_scalar_type): Update accordingly.
113         * tree.h (build_truth_vector_type): Take the size and number of
114         units as a poly_uint64 rather than an unsigned int.
115         (build_vector_type): Add a temporary overload that takes
116         the number of units as a poly_uint64 rather than an unsigned int.
117         * tree.c (make_vector_type): Likewise.
118         (build_truth_vector_type): Take the number of units as a poly_uint64
119         rather than an unsigned int.
121 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
122             Alan Hayward  <alan.hayward@arm.com>
123             David Sherwood  <david.sherwood@arm.com>
125         * target.def (get_mask_mode): Take the number of units and length
126         as poly_uint64s rather than unsigned ints.
127         * targhooks.h (default_get_mask_mode): Update accordingly.
128         * targhooks.c (default_get_mask_mode): Likewise.
129         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
130         * doc/tm.texi: Regenerate.
132 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
133             Alan Hayward  <alan.hayward@arm.com>
134             David Sherwood  <david.sherwood@arm.com>
136         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
137         * omp-general.c (omp_max_vf): Likewise.
138         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
139         (expand_omp_simd): Handle polynomial safelen.
140         * omp-low.c (omplow_simd_context): Add a default constructor.
141         (omplow_simd_context::max_vf): Change from int to poly_uint64.
142         (lower_rec_simd_input_clauses): Update accordingly.
143         (lower_rec_input_clauses): Likewise.
145 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
146             Alan Hayward  <alan.hayward@arm.com>
147             David Sherwood  <david.sherwood@arm.com>
149         * tree-vectorizer.h (vect_nunits_for_cost): New function.
150         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
151         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
152         (vect_analyze_slp_cost): Likewise.
153         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
154         (vect_model_load_cost): Likewise.
156 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
157             Alan Hayward  <alan.hayward@arm.com>
158             David Sherwood  <david.sherwood@arm.com>
160         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
161         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
162         from an unsigned int * to a poly_uint64_pod *.
163         (calculate_unrolling_factor): New function.
164         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
166 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
167             Alan Hayward  <alan.hayward@arm.com>
168             David Sherwood  <david.sherwood@arm.com>
170         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
171         from an unsigned int to a poly_uint64.
172         (_loop_vec_info::slp_unrolling_factor): Likewise.
173         (_loop_vec_info::vectorization_factor): Change from an int
174         to a poly_uint64.
175         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
176         (vect_get_num_vectors): New function.
177         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
178         (vect_get_num_copies): Use vect_get_num_vectors.
179         (vect_analyze_data_ref_dependences): Change max_vf from an int *
180         to an unsigned int *.
181         (vect_analyze_data_refs): Change min_vf from an int * to a
182         poly_uint64 *.
183         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
184         than an unsigned HOST_WIDE_INT.
185         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
186         (vect_analyze_data_ref_dependence): Change max_vf from an int *
187         to an unsigned int *.
188         (vect_analyze_data_ref_dependences): Likewise.
189         (vect_compute_data_ref_alignment): Handle polynomial vf.
190         (vect_enhance_data_refs_alignment): Likewise.
191         (vect_prune_runtime_alias_test_list): Likewise.
192         (vect_shift_permute_load_chain): Likewise.
193         (vect_supportable_dr_alignment): Likewise.
194         (dependence_distance_ge_vf): Take the vectorization factor as a
195         poly_uint64 rather than an unsigned HOST_WIDE_INT.
196         (vect_analyze_data_refs): Change min_vf from an int * to a
197         poly_uint64 *.
198         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
199         vfm1 as a poly_uint64 rather than an int.  Make the same change
200         for the returned bound_scalar.
201         (vect_gen_vector_loop_niters): Handle polynomial vf.
202         (vect_do_peeling): Likewise.  Update call to
203         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
204         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
205         be constant.
206         * tree-vect-loop.c (vect_determine_vectorization_factor)
207         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
208         (vect_get_known_peeling_cost): Likewise.
209         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
210         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
211         (vect_transform_loop): Likewise.  Use the lowest possible VF when
212         updating the upper bounds of the loop.
213         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
214         rather than an int.
215         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
216         polynomial unroll factors.
217         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
218         (vect_make_slp_decision): Likewise.
219         (vect_supported_load_permutation_p): Likewise, and polynomial
220         vf too.
221         (vect_analyze_slp_cost): Handle polynomial vf.
222         (vect_slp_analyze_node_operations): Likewise.
223         (vect_slp_analyze_bb_1): Likewise.
224         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
225         than an unsigned HOST_WIDE_INT.
226         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
227         (vectorizable_load): Handle polynomial vf.
228         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
229         a poly_uint64.
230         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
232 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
233             Alan Hayward  <alan.hayward@arm.com>
234             David Sherwood  <david.sherwood@arm.com>
236         * match.pd: Handle bit operations involving three constants
237         and try to fold one pair.
239 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
241         * tree-vect-loop-manip.c: Include gimple-fold.h.
242         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
243         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
244         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
245         Add a path that uses a step of VF instead of 1, but disable it
246         for now.
247         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
248         and niters_no_overflow parameters.  Update calls to
249         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
250         Create a new SSA name if the latter choses to use a ste other
251         than zero, and return it via niters_vector_mult_vf_var.
252         * tree-vect-loop.c (vect_transform_loop): Update calls to
253         vect_do_peeling, vect_gen_vector_loop_niters and
254         slpeel_make_loop_iterate_ntimes.
255         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
256         (vect_gen_vector_loop_niters): Update declarations after above changes.
258 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
260         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
261         128-bit round to integer instructions.
262         (ceil<mode>2): Likewise.
263         (btrunc<mode>2): Likewise.
264         (round<mode>2): Likewise.
266 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
268         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
269         unaligned VSX load/store on P8/P9.
270         (expand_block_clear): Allow the use of unaligned VSX
271         load/store on P8/P9.
273 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
275         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
276         New function.
277         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
278         swap associated with both a load and a store.
280 2018-01-02  Andrew Waterman  <andrew@sifive.com>
282         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
283         * config/riscv/riscv.md (clear_cache): Use it.
285 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
287         * web.c: Remove out-of-date comment.
289 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
291         * expr.c (fixup_args_size_notes): Check that any existing
292         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
293         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
294         (emit_single_push_insn): ...here.
296 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
298         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
299         (const_vector_encoded_nelts): New function.
300         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
301         (const_vector_int_elt, const_vector_elt): Declare.
302         * emit-rtl.c (const_vector_int_elt_1): New function.
303         (const_vector_elt): Likewise.
304         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
305         of CONST_VECTOR_ELT.
307 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
309         * expr.c: Include rtx-vector-builder.h.
310         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
311         directly on the tree encoding.
312         (const_vector_from_tree): Likewise.
313         * optabs.c: Include rtx-vector-builder.h.
314         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
315         sequence of "u" values.
316         * vec-perm-indices.c: Include rtx-vector-builder.h.
317         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
318         directly on the vec_perm_indices encoding.
320 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
322         * doc/rtl.texi (const_vector): Describe new encoding scheme.
323         * Makefile.in (OBJS): Add rtx-vector-builder.o.
324         * rtx-vector-builder.h: New file.
325         * rtx-vector-builder.c: Likewise.
326         * rtl.h (rtx_def::u2): Add a const_vector field.
327         (CONST_VECTOR_NPATTERNS): New macro.
328         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
329         (CONST_VECTOR_DUPLICATE_P): Likewise.
330         (CONST_VECTOR_STEPPED_P): Likewise.
331         (CONST_VECTOR_ENCODED_ELT): Likewise.
332         (const_vec_duplicate_p): Check for a duplicated vector encoding.
333         (unwrap_const_vec_duplicate): Likewise.
334         (const_vec_series_p): Check for a non-duplicated vector encoding.
335         Say that the function only returns true for integer vectors.
336         * emit-rtl.c: Include rtx-vector-builder.h.
337         (gen_const_vec_duplicate_1): Delete.
338         (gen_const_vector): Call gen_const_vec_duplicate instead of
339         gen_const_vec_duplicate_1.
340         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
341         (gen_const_vec_duplicate): Use rtx_vector_builder.
342         (gen_const_vec_series): Likewise.
343         (gen_rtx_CONST_VECTOR): Likewise.
344         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
345         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
346         Build a new vector rather than modifying a CONST_VECTOR in-place.
347         (handle_special_swappables): Update call accordingly.
348         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
349         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
350         Build a new vector rather than modifying a CONST_VECTOR in-place.
351         (handle_special_swappables): Update call accordingly.
353 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
355         * simplify-rtx.c (simplify_const_binary_operation): Use
356         CONST_VECTOR_ELT instead of XVECEXP.
358 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
360         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
361         the selector elements to be different from the data elements
362         if the selector is a VECTOR_CST.
363         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
364         ssizetype for the selector.
366 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
368         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
369         before testing each element individually.
370         * tree-vect-generic.c (lower_vec_perm): Likewise.
372 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
374         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
375         * selftest-run-tests.c (selftest::run_tests): Call it.
376         * vector-builder.h (vector_builder::operator ==): New function.
377         (vector_builder::operator !=): Likewise.
378         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
379         (vec_perm_indices::all_from_input_p): New function.
380         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
381         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
382         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
383         instead of reading the VECTOR_CST directly.  Detect whether both
384         vector inputs are the same before constructing the vec_perm_indices,
385         and update the number of inputs argument accordingly.  Use the
386         utility functions added above.  Only construct sel2 if we need to.
388 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
390         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
391         the broadcast of the low byte.
392         (expand_mult_highpart): Use an explicit encoding for the permutes.
393         * optabs-query.c (can_mult_highpart_p): Likewise.
394         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
395         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
396         (vectorizable_bswap): Likewise.
397         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
398         explicit encoding for the power-of-2 permutes.
399         (vect_permute_store_chain): Likewise.
400         (vect_grouped_load_supported): Likewise.
401         (vect_permute_load_chain): Likewise.
403 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
405         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
406         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
407         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
408         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
409         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
410         (vect_gen_perm_mask_any): Likewise.
412 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
414         * int-vector-builder.h: New file.
415         * vec-perm-indices.h: Include int-vector-builder.h.
416         (vec_perm_indices): Redefine as an int_vector_builder.
417         (auto_vec_perm_indices): Delete.
418         (vec_perm_builder): Redefine as a stand-alone class.
419         (vec_perm_indices::vec_perm_indices): New function.
420         (vec_perm_indices::clamp): Likewise.
421         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
422         (vec_perm_indices::new_vector): New function.
423         (vec_perm_indices::new_expanded_vector): Update for new
424         vec_perm_indices class.
425         (vec_perm_indices::rotate_inputs): New function.
426         (vec_perm_indices::all_in_range_p): Operate directly on the
427         encoded form, without computing elided elements.
428         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
429         encoding.  Update for new vec_perm_indices class.
430         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
431         the given vec_perm_builder.
432         (expand_vec_perm_var): Update vec_perm_builder constructor.
433         (expand_mult_highpart): Use vec_perm_builder instead of
434         auto_vec_perm_indices.
435         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
436         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
437         or double series encoding as appropriate.
438         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
439         vec_perm_indices instead of auto_vec_perm_indices.
440         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
441         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
442         (vect_permute_store_chain): Likewise.
443         (vect_grouped_load_supported): Likewise.
444         (vect_permute_load_chain): Likewise.
445         (vect_shift_permute_load_chain): Likewise.
446         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
447         (vect_transform_slp_perm_load): Likewise.
448         (vect_schedule_slp_instance): Likewise.
449         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
450         (vectorizable_mask_load_store): Likewise.
451         (vectorizable_bswap): Likewise.
452         (vectorizable_store): Likewise.
453         (vectorizable_load): Likewise.
454         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
455         vec_perm_indices instead of auto_vec_perm_indices.  Use
456         tree_to_vec_perm_builder to read the vector from a tree.
457         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
458         vec_perm_builder instead of a vec_perm_indices.
459         (have_whole_vector_shift): Use vec_perm_builder and
460         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
461         truncation to calc_vec_perm_mask_for_shift.
462         (vect_create_epilog_for_reduction): Likewise.
463         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
464         from auto_vec_perm_indices to vec_perm_indices.
465         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
466         instead of changing individual elements.
467         (aarch64_vectorize_vec_perm_const): Use new_vector to install
468         the vector in d.perm.
469         * config/arm/arm.c (expand_vec_perm_d::perm): Change
470         from auto_vec_perm_indices to vec_perm_indices.
471         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
472         instead of changing individual elements.
473         (arm_vectorize_vec_perm_const): Use new_vector to install
474         the vector in d.perm.
475         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
476         Update vec_perm_builder constructor.
477         (rs6000_expand_interleave): Likewise.
478         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
479         (rs6000_expand_interleave): Likewise.
481 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
483         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
484         to qimode could truncate the indices.
485         * optabs.c (expand_vec_perm_var): Likewise.
487 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
489         * Makefile.in (OBJS): Add vec-perm-indices.o.
490         * vec-perm-indices.h: New file.
491         * vec-perm-indices.c: Likewise.
492         * target.h (vec_perm_indices): Replace with a forward class
493         declaration.
494         (auto_vec_perm_indices): Move to vec-perm-indices.h.
495         * optabs.h: Include vec-perm-indices.h.
496         (expand_vec_perm): Delete.
497         (selector_fits_mode_p, expand_vec_perm_var): Declare.
498         (expand_vec_perm_const): Declare.
499         * target.def (vec_perm_const_ok): Replace with...
500         (vec_perm_const): ...this new hook.
501         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
502         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
503         * doc/tm.texi: Regenerate.
504         * optabs.def (vec_perm_const): Delete.
505         * doc/md.texi (vec_perm_const): Likewise.
506         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
507         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
508         expand_vec_perm for constant permutation vectors.  Assert that
509         the mode of variable permutation vectors is the integer equivalent
510         of the mode that is being permuted.
511         * optabs-query.h (selector_fits_mode_p): Declare.
512         * optabs-query.c: Include vec-perm-indices.h.
513         (selector_fits_mode_p): New function.
514         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
515         is defined, instead of checking whether the vec_perm_const_optab
516         exists.  Use targetm.vectorize.vec_perm_const instead of
517         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
518         fit in the vector mode before using a variable permute.
519         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
520         vec_perm_indices instead of an rtx.
521         (expand_vec_perm): Replace with...
522         (expand_vec_perm_const): ...this new function.  Take the selector
523         as a vec_perm_indices rather than an rtx.  Also take the mode of
524         the selector.  Update call to shift_amt_for_vec_perm_mask.
525         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
526         Use vec_perm_indices::new_expanded_vector to expand the original
527         selector into bytes.  Check whether the indices fit in the vector
528         mode before using a variable permute.
529         (expand_vec_perm_var): Make global.
530         (expand_mult_highpart): Use expand_vec_perm_const.
531         * fold-const.c: Includes vec-perm-indices.h.
532         * tree-ssa-forwprop.c: Likewise.
533         * tree-vect-data-refs.c: Likewise.
534         * tree-vect-generic.c: Likewise.
535         * tree-vect-loop.c: Likewise.
536         * tree-vect-slp.c: Likewise.
537         * tree-vect-stmts.c: Likewise.
538         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
539         Delete.
540         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
541         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
542         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
543         (aarch64_vectorize_vec_perm_const): ...this new function.
544         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
545         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
546         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
547         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
548         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
549         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
550         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
551         into...
552         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
553         check for NEON modes.
554         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
555         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
556         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
557         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
558         into...
559         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
560         the old VEC_PERM_CONST conditions.
561         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
562         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
563         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
564         (ia64_vectorize_vec_perm_const_ok): Merge into...
565         (ia64_vectorize_vec_perm_const): ...this new function.
566         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
567         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
568         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
569         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
570         * config/mips/mips.c (mips_expand_vec_perm_const)
571         (mips_vectorize_vec_perm_const_ok): Merge into...
572         (mips_vectorize_vec_perm_const): ...this new function.
573         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
574         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
575         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
576         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
577         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
578         (rs6000_expand_vec_perm_const): Delete.
579         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
580         Delete.
581         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
582         (altivec_expand_vec_perm_const_le): Take each operand individually.
583         Operate on constant selectors rather than rtxes.
584         (altivec_expand_vec_perm_const): Likewise.  Update call to
585         altivec_expand_vec_perm_const_le.
586         (rs6000_expand_vec_perm_const): Delete.
587         (rs6000_vectorize_vec_perm_const_ok): Delete.
588         (rs6000_vectorize_vec_perm_const): New function.
589         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
590         an element count and rtx array.
591         (rs6000_expand_extract_even): Update call accordingly.
592         (rs6000_expand_interleave): Likewise.
593         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
594         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
595         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
596         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
597         (rs6000_expand_vec_perm_const): Delete.
598         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
599         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
600         (altivec_expand_vec_perm_const_le): Take each operand individually.
601         Operate on constant selectors rather than rtxes.
602         (altivec_expand_vec_perm_const): Likewise.  Update call to
603         altivec_expand_vec_perm_const_le.
604         (rs6000_expand_vec_perm_const): Delete.
605         (rs6000_vectorize_vec_perm_const_ok): Delete.
606         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
607         reference to the SPE evmerge intructions.
608         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
609         an element count and rtx array.
610         (rs6000_expand_extract_even): Update call accordingly.
611         (rs6000_expand_interleave): Likewise.
612         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
613         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
614         new function.
615         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
617 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
619         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
620         vector mode and that that mode matches the mode of the data
621         being permuted.
622         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
623         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
624         directly using expand_vec_perm_1 when forcing selectors into
625         registers.
626         (expand_vec_perm_var): New function, split out from expand_vec_perm.
628 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
630         * optabs-query.h (can_vec_perm_p): Delete.
631         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
632         * optabs-query.c (can_vec_perm_p): Split into...
633         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
634         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
635         particular selector is valid.
636         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
637         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
638         (vect_grouped_load_supported): Likewise.
639         (vect_shift_permute_load_chain): Likewise.
640         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
641         (vect_transform_slp_perm_load): Likewise.
642         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
643         (vectorizable_bswap): Likewise.
644         (vect_gen_perm_mask_checked): Likewise.
645         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
646         implementations of variable permutation vectors into account
647         when deciding which selector to use.
648         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
649         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
650         with a false third argument.
651         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
652         to test whether the constant selector is valid and can_vec_perm_var_p
653         to test whether a variable selector is valid.
655 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
657         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
658         * optabs-query.c (can_vec_perm_p): Likewise.
659         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
660         instead of vec_perm_indices.
661         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
662         (vect_gen_perm_mask_checked): Likewise,
663         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
664         (vect_gen_perm_mask_checked): Likewise,
666 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
668         * optabs-query.h (qimode_for_vec_perm): Declare.
669         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
670         (qimode_for_vec_perm): ...this new function.
671         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
673 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
675         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
676         does not have a conditional at the top.
678 2018-01-02  Richard Biener  <rguenther@suse.de>
680         * ipa-inline.c (big_speedup_p): Fix expression.
682 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
684         PR target/81616
685         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
686         for generic 4->6.
688 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
690         PR target/81616
691         Generic tuning.
692         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
693         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
694         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
695         cond_taken_branch_cost 3->4.
697 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
699         PR tree-optimization/83581
700         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
701         TODO_cleanup_cfg if any changes have been made.
703         PR middle-end/83608
704         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
705         convert_modes if target mode has the right side, but different mode
706         class.
708         PR middle-end/83609
709         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
710         last argument when extracting from CONCAT.  If either from_real or
711         from_imag is NULL, use expansion through memory.  If result is not
712         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
713         the parts directly to inner mode, if even that fails, use expansion
714         through memory.
716         PR middle-end/83623
717         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
718         check for bswap in mode rather than HImode and use that in expand_unop
719         too.
721 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
723         PR target/83536
724         * config/i386/i386.c (ix86_attribute_table): Remove excess
725         initializer for "shared" attribute.
727 2017-12-30  Tom de Vries  <tom@codesourcery.com>
729         PR libgomp/83046
730         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
731         DECL_PRESERVE_P.
732         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
733         offload_funcs entries that no longer have a corresponding cgraph_node.
734         Mark the remaining ones as DECL_PRESERVE_P.
735         (output_lto): Call prune_offload_funcs.
737 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
739         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
740         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
741         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
742         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
743         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
744         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
745         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
746         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
747         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
748         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
749         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
750         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
751         vaesenc_<mode>, vpclmulqdq_<mode>,
752         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
754 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
756         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
757         _Float<N> and _Float<N>X variants for rounding built-in
758         functions.
759         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
760         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
761         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
762         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
763         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
764         * builtins.c (mathfn_built_in_2): Likewise.
765         * internal-fn.def (CEIL): Likewise.
766         (FLOOR): Likewise.
767         (NEARBYINT): Likewise.
768         (RINT): Likewise.
769         (ROUND): Likewise.
770         (TRUNC): Likewise.
771         * convert.c (convert_to_integer_1): Likewise.
772         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
773         (integer_valued_real_call_p): Likewise.
774         * fold-const-call.c (fold_const_call_ss): Likewise.
775         * gencfn-macros.c (print_case_cfn): Change CFN and operator
776         printers to take a const char * suffix instead of a bool.
777         (print_define_operator_list): Likewise.
778         (fltall_suffixes): New list of suffixes, that include the
779         traditional suffixes as well as all of the _Float<N> and
780         _Float<N>X suffixes.
781         (main): For _Float<N> and _Float<N>X functions, emit both
782         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
783         has the _Float<N> and _Float<N>X case names or operators.  The
784         <name>_ALL variant has both the traditional and the
785         _Float<N>/_Float<N>X case names or operators.
786         * match.pd (COPYSIGN optimizations): Provide optimizations for
787         _Float<N> and _Float<N>X types where possible.
788         (MIN/MAX optimizations): Likewise.
789         (sqrt optimizations): Likewise.
790         (rounding optimizations): Likewise.
792 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
794         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
795         gen_int_mode rather than GEN_INT.
796         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
798 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
800         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
801         instead of CONSTANT_P.
802         (gen_vec_series): Likewise.
803         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
805 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
807         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
808         CC_STATUS_INIT.
810 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
812         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
813         (aarch64_ld1x2<VDC:mode>): Likewise.
814         (aarch64_simd_ld1<mode>_x2): Likewise.
815         (aarch64_simd_ld1<mode>_x2): Likewise.
816         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
817         (vld1_s8_x2): Likewise.
818         (vld1_u16_x2): Likewise.
819         (vld1_s16_x2): Likewise.
820         (vld1_u32_x2): Likewise.
821         (vld1_s32_x2): Likewise.
822         (vld1_u64_x2): Likewise.
823         (vld1_s64_x2): Likewise.
824         (vld1_f16_x2): Likewise.
825         (vld1_f32_x2): Likewise.
826         (vld1_f64_x2): Likewise.
827         (vld1_p8_x2): Likewise.
828         (vld1_p16_x2): Likewise.
829         (vld1_p64_x2): Likewise.
830         (vld1q_u8_x2): Likewise.
831         (vld1q_s8_x2): Likewise.
832         (vld1q_u16_x2): Likewise.
833         (vld1q_s16_x2): Likewise.
834         (vld1q_u32_x2): Likewise.
835         (vld1q_s32_x2): Likewise.
836         (vld1q_u64_x2): Likewise.
837         (vld1q_s64_x2): Likewise.
838         (vld1q_f16_x2): Likewise.
839         (vld1q_f32_x2): Likewise.
840         (vld1q_f64_x2): Likewise.
841         (vld1q_p8_x2): Likewise.
842         (vld1q_p16_x2): Likewise.
843         (vld1q_p64_x2): Likewise.
845 2017-12-27  Martin Liska  <mliska@suse.cz>
847         PR tree-optimization/83552
848         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
849         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
851 2017-12-27  Tom de Vries  <tom@codesourcery.com>
853         * config/nvptx/nvptx.c (nvptx_option_override): Disable
854         -gstatement-frontiers.
856 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
858         PR rtl-optimization/83513
859         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
860         before priority comparison.
862 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
864         PR target/83488
865         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
866         ix86_isa_flags2 to ix86_isa_flags.
867         * config/i386/i386-c.c (ix86_target_macros_internal): Test
868         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
869         isa_flags rather than isa_flags2.
870         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
871         and -mavx512bitalg from isa2_opts to isa_opts.
872         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
873         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
874         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
875         Formatting fixes.
876         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
877         OPTION_MASK_ISA_AVX512F ored with another option similarly to
878         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
879         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
880         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
881         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
882         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
883         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
884         builtins from bdesc_args2 to bdesc_args section.
885         (__builtin_ia32_compressstoreuqi512_mask,
886         __builtin_ia32_compressstoreuhi512_mask,
887         __builtin_ia32_compressstoreuqi256_mask,
888         __builtin_ia32_expandloadqi512_mask,
889         __builtin_ia32_expandloadqi512_maskz,
890         __builtin_ia32_expandloadhi512_mask,
891         __builtin_ia32_expandloadhi512_maskz,
892         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
893         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
894         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
895         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
896         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
897         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
898         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
899         __builtin_ia32_vpshldv_v32hi_maskz,
900         __builtin_ia32_vpopcountb_v64qi_mask,
901         __builtin_ia32_vpopcountw_v32hi_mask,
902         __builtin_ia32_vpshufbitqmb512_mask,
903         __builtin_ia32_vpshufbitqmb256_mask): Add
904         " | OPTION_MASK_ISA_AVX512BW".
905         (__builtin_ia32_expandloadqi256_mask,
906         __builtin_ia32_expandloadqi256_maskz,
907         __builtin_ia32_vpopcountb_v32qi_mask): Add
908         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
909         (__builtin_ia32_expandloadhi256_mask,
910         __builtin_ia32_expandloadhi256_maskz,
911         __builtin_ia32_expandloadqi128_mask,
912         __builtin_ia32_expandloadqi128_maskz,
913         __builtin_ia32_expandloadhi128_mask,
914         __builtin_ia32_expandloadhi128_maskz,
915         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
916         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
917         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
918         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
919         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
920         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
921         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
922         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
923         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
924         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
925         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
926         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
927         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
928         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
929         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
930         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
931         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
932         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
933         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
934         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
935         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
936         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
937         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
938         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
939         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
940         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
941         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
942         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
943         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
944         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
945         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
946         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
947         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
948         __builtin_ia32_vpopcountw_v8hi_mask): Add
949         " | OPTION_MASK_ISA_AVX512VL".
950         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
951         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
952         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
953         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
954         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
955         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
956         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
957         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
958         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
959         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
960         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
961         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
962         _mm_mask_bitshuffle_epi64_mask): Likewise.
963         * common/config/i386/i386-common.c
964         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
965         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
966         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
967         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
968         OPTION_MASK_ISA_AVX512BITALG_UNSET.
969         (OPTION_MASK_ISA2_AVX512F_UNSET,
970         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
971         (ix86_handle_option): For -mno-general-regs-only, clear from
972         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
973         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
974         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
975         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
976         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
977         for moving from ix86_isa_flags2 to ix86_isa_flags.
979 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
981         PR c++/83553
982         * fold-const.c (struct contains_label_data): New type.
983         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
984         inside of a SWITCH_BODY seen during the walk.
985         (contains_label_p): Use walk_tree instead of
986         walk_tree_without_duplicates, prepare data for contains_label_1 and
987         provide own pset.
989 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
991         PR debug/83547
992         * tree-iterator.c (alloc_stmt_list): Start with cleared
993         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
994         or old one reused.
996         PR target/83488
997         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
998         enable avx512f explicitly in #pragma GCC target.
999         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
1000         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
1001         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
1002         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
1003         __builtin_ia32_vpdpbusds_v8si_mask,
1004         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
1005         __builtin_ia32_vpdpbusds_v4si_mask,
1006         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
1007         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
1008         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
1009         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
1010         __builtin_ia32_vpdpwssds_v8si_mask,
1011         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
1012         __builtin_ia32_vpdpwssds_v4si_mask,
1013         __builtin_ia32_vpdpwssds_v4si_maskz): Use
1014         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
1015         just OPTION_MASK_ISA_AVX512VNNI.
1017 2017-12-22  Martin Jambor  <mjambor@suse.cz>
1019         PR lto/82027
1020         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
1021         clones.
1023 2017-12-22  Julia Koval  <julia.koval@intel.com>
1024             Sebastian Peryt  <sebastian.peryt@intel.com>
1026         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1027         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1028         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1029         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1030         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1031         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1032         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1033         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1034         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1035         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1036         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1037         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1038         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1039         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1040         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1041         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1042         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1043         _mm256_maskz_popcnt_epi64): New intrinsics.
1044         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1045         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1046         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1047         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1048         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1049         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1050         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1051         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1052         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1053         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1054         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1055         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1056         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1057         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1058         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1059         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1060         __builtin_ia32_vpshufbitqmb256_mask,
1061         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1062         * config/i386/i386-c.c (__AVX512BITALG__): New.
1063         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1064         (ix86_valid_target_attribute_inner_p): Ditto.
1065         (ix86_expand_args_builtin): Handle new types.
1066         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1067         * config/i386/i386.opt: Add -mavx512bitalg.
1068         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1069         avx512bitalgintrin.h.
1070         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1071         (vpopcount<mode><mask_name>): Add more types.
1072         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1073         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1075 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1077         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1078         Or in OPTION_MASK_ISA_AVX512F_SET.
1079         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1080         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1081         (ix86_handle_option): Adjust for
1082         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1083         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1084         section to ARGS.
1085         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1086         isa_flag instead of isa_flag2.
1087         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1088         isa_opts2 to isa_opts.
1089         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1090         to ix86_isa_flags.
1092 2017-12-22  Mike Stump  <mikestump@comcast.net>
1093             Eric Botcazou  <ebotcazou@adacore.com>
1095         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1097 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1099         PR middle-end/83487
1100         * config/i386/i386.c (ix86_function_arg_boundary): Return
1101         PARM_BOUNDARY for TYPE_EMPTY_P types.
1103         PR c/83448
1104         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1105         if navail is >= dir.len.
1107 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1109         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1110         triplet for ilp32.
1112 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1114         PR rtl-optimization/80747
1115         PR rtl-optimization/83512
1116         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1117         succ edge from ENTRY, copy partition from e->dest to the newly
1118         created bb.
1119         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1120         ENTRY, use BB_PARTITION of its successor block as current_partition.
1121         Don't copy partition when splitting succ edge from ENTRY.
1123         PR tree-optimization/83523
1124         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1125         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1126         (convert_mult_to_fma): Likewise.
1128         PR tree-optimization/83521
1129         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1130         gimple_build_assign without code on result of
1131         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1132         a VIEW_CONVERT_EXPR.
1134 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1135             Steve Ellcey  <sellcey@cavium.com>
1137         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1138         multi-arch for ilp32.
1140 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1142         PR target/83467
1143         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1144         constraints to operand 2.
1145         (*ashl<mode>3_mask_1): Ditto.
1146         (*<shift_insn><mode>3_mask): Ditto.
1147         (*<shift_insn><mode>3_mask_1): Ditto.
1148         (*<rotate_insn><mode>3_mask): Ditto.
1149         (*<rotate_insn><mode>3_mask_1): Ditto.
1151 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1153         * reorg.c (make_return_insns): Reemit each insn with its own location.
1155 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1157         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1158         generation for cases where splatting a value is not useful.
1159         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1160         across a vec_duplicate and a paradoxical subreg forming a vector
1161         mode to a vec_concat.
1163 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1165         * combine.c (simplify_set): Do not transform subregs to zero_extends
1166         if the destination is not a scalar int mode.
1168 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1170         PR c++/82872
1171         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1172         shared zero if the input has overflowed.
1174 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1176         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1177         support for Cortex-A55 and Cortex-A75.
1179 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1181         * common/config/arm/arm-common.c (compare_opt_names): Add function
1182         comment.  Use strcmp instead of manual loop.
1184 2017-12-21  Martin Liska  <mliska@suse.cz>
1186         PR gcov-profile/83509
1187         * gcov-dump.c (dump_gcov_file): Do not read info about
1188         support_unexecuted_blocks for gcda files.
1190 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1192         PR rtl-optimization/82973
1193         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1194         (valid_for_const_vector_p): ... this.
1195         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1196         (valid_for_const_vector_p): ... this.  Adjust function comment.
1197         (gen_vec_duplicate): Adjust caller.
1198         * optabs.c (expand_vector_broadcast): Likewise.
1199         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1200         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1201         constant.
1202         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1203         instead of GET_CODE == CONST_FIXED.
1204         (simplify_subreg): Use CONST_FIXED_P macro instead of
1205         GET_CODE == CONST_FIXED.
1207         PR target/83488
1208         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1209         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1210         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1211         (ix86_option_override_internal): Adjust for
1212         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1213         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1214         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1215         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1216         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1217         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1218         def_builtin2 for CET builtins.
1219         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1220         up formatting in IX86_BUILTIN_RDPID code.
1221         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1222         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1223         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1224         ix86_isa_flags2 to ix86_isa_flags.
1225         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1226         ix86_isa_flags2.
1227         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1228         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1229         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1230         of isa_flag2.
1231         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1232         Or in OPTION_MASK_ISA_AVX512F_SET.
1233         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1234         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1235         (ix86_handle_option): Adjust for
1236         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1237         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1239 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1240             Alan Hayward  <alan.hayward@arm.com>
1241             David Sherwood  <david.sherwood@arm.com>
1243         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1244         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1245         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1246         Track polynomial offsets.
1248 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1249             Alan Hayward  <alan.hayward@arm.com>
1250             David Sherwood  <david.sherwood@arm.com>
1252         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1253         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1254         vector_alignment isn't known.
1256 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1257             Alan Hayward  <alan.hayward@arm.com>
1258             David Sherwood  <david.sherwood@arm.com>
1260         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1261         field.
1262         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1263         (vect_loop_versioning): Take the loop versioning threshold as a
1264         separate parameter.
1265         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1266         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1267         versioning_threshold.
1268         (vect_analyze_loop_2): Compute the loop versioning threshold
1269         whenever loop versioning is needed, and store it in the new
1270         field rather than combining it with the cost model threshold.
1271         (vect_transform_loop): Update call to vect_loop_versioning.
1272         Try to combine the loop versioning and cost thresholds here.
1274 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1275             Alan Hayward  <alan.hayward@arm.com>
1276             David Sherwood  <david.sherwood@arm.com>
1278         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1279         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1280         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1281         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1282         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1283         (group_compare_offset): Update accordingly.
1284         (split_small_address_groups_p): Likewise.
1285         (record_use): Take addr_offset as a poly_uint64 rather than
1286         an unsigned HOST_WIDE_INT.
1287         (strip_offset): Return the offset as a poly_uint64 rather than
1288         an unsigned HOST_WIDE_INT.
1289         (record_group_use, split_address_groups): Track polynomial offsets.
1290         (add_iv_candidate_for_use): Likewise.
1291         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1292         than a HOST_WIDE_INT.
1293         (strip_offset_1): Return the offset as a poly_int64 rather than
1294         a HOST_WIDE_INT.
1296 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1297             Alan Hayward  <alan.hayward@arm.com>
1298             David Sherwood  <david.sherwood@arm.com>
1300         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1301         rather than a HOST_WIDE_INT.
1302         * tree.c (get_binfo_at_offset): Likewise.
1304 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1305             Alan Hayward  <alan.hayward@arm.com>
1306             David Sherwood  <david.sherwood@arm.com>
1308         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1309         rather than a HOST_WIDE_INT.
1310         * tree-sra.c (build_ref_for_offset): Likewise.
1312 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1313             Alan Hayward  <alan.hayward@arm.com>
1314             David Sherwood  <david.sherwood@arm.com>
1316         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1317         than an offset_int.
1318         * tree.c (mem_ref_offset): Likewise.
1319         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1320         * builtins.c (get_object_alignment_2): Likewise.
1321         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1322         * gimple-fold.c (get_base_constructor): Likewise.
1323         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1324         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1325         Likewise.
1326         * ipa-polymorphic-call.c
1327         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1328         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1329         (get_ancestor_addr_info): Likewise.
1330         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1331         * match.pd: Likewise.
1332         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1333         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1334         * tree-eh.c (tree_could_trap_p): Likewise.
1335         * tree-object-size.c (addr_object_size): Likewise.
1336         * tree-ssa-address.c (copy_ref_info): Likewise.
1337         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1338         (indirect_refs_may_alias_p): Likewise.
1339         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1340         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1341         (non_rewritable_mem_ref_base): Likewise.
1342         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1343         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1344         * varasm.c (decode_addr_const): Likewise.
1346 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1347             Alan Hayward  <alan.hayward@arm.com>
1348             David Sherwood  <david.sherwood@arm.com>
1350         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1351         offsets for MEM_REFs.
1353 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1354             Alan Hayward  <alan.hayward@arm.com>
1355             David Sherwood  <david.sherwood@arm.com>
1357         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1358         as a poly_uint64 rather than an unsigned int.
1359         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1361 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1362             Alan Hayward  <alan.hayward@arm.com>
1363             David Sherwood  <david.sherwood@arm.com>
1365         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1366         to poly_int64.
1367         (decode_addr_const): Update accordingly.
1369 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1370             Alan Hayward  <alan.hayward@arm.com>
1371             David Sherwood  <david.sherwood@arm.com>
1373         * tree.h (bit_field_size, bit_field_offset): New functions.
1374         * hsa-gen.c (gen_hsa_addr): Use them.
1375         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1376         (simplify_vector_constructor): Likewise.
1377         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1378         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1379         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1380         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1381         tree_fits_uhwi_p.
1383 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1384             Alan Hayward  <alan.hayward@arm.com>
1385             David Sherwood  <david.sherwood@arm.com>
1387         * expr.h (emit_group_load, emit_group_load_into_temps)
1388         (emit_group_store): Take the size as a poly_int64 rather than an int.
1389         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1390         (emit_group_load_into_temp, emit_group_store): Likewise.
1392 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1393             Alan Hayward  <alan.hayward@arm.com>
1394             David Sherwood  <david.sherwood@arm.com>
1396         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1397         an unsigned int to a poly_uint64.
1398         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1399         sizes as poly_uint64s rather than unsigned ints.
1400         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1401         Likewise.
1403 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1404             Alan Hayward  <alan.hayward@arm.com>
1405             David Sherwood  <david.sherwood@arm.com>
1407         * lra-constraints.c (emit_inc): Change inc_amount from an int
1408         to a poly_int64.
1410 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1411             Alan Hayward  <alan.hayward@arm.com>
1412             David Sherwood  <david.sherwood@arm.com>
1414         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1415         to a poly_uint64.
1416         (add_stack_var, stack_var_cmp, partition_stack_vars)
1417         (dump_stack_var_partition): Update accordingly.
1418         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1419         than a HOST_WIDE_INT.
1420         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1421         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1422         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1423         rather than a HOST_WIDE_INT.
1425 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1426             Alan Hayward  <alan.hayward@arm.com>
1427             David Sherwood  <david.sherwood@arm.com>
1429         * target.def (return_pops_args): Treat both the input and output
1430         sizes as poly_int64s rather than HOST_WIDE_INTS.
1431         * targhooks.h (default_return_pops_args): Update accordingly.
1432         * targhooks.c (default_return_pops_args): Likewise.
1433         * doc/tm.texi: Regenerate.
1434         * emit-rtl.h (incoming_args): Change pops_args, size and
1435         outgoing_args_size from int to poly_int64_pod.
1436         * function.h (expr_status): Change x_pending_stack_adjust and
1437         x_stack_pointer_delta from int to poly_int64.
1438         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1439         (ARGS_SIZE_RTX): Update accordingly.
1440         * calls.c (highest_outgoing_arg_in_use): Change from int to
1441         unsigned int.
1442         (stack_usage_watermark, stored_args_watermark): New variables.
1443         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1444         (emit_call_1): Change the stack_size and rounded_stack_size
1445         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1446         as a poly_int64.
1447         (save_fixed_argument_area): Check stack_usage_watermark.
1448         (initialize_argument_information): Change old_pending_adj from
1449         a HOST_WIDE_INT * to a poly_int64_pod *.
1450         (compute_argument_block_size): Return the size as a poly_int64
1451         rather than an int.
1452         (finalize_must_preallocate): Track polynomial argument sizes.
1453         (compute_argument_addresses): Likewise.
1454         (internal_arg_pointer_based_exp): Track polynomial offsets.
1455         (mem_overlaps_already_clobbered_arg_p): Rename to...
1456         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1457         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1458         Check stored_args_used_watermark.
1459         (load_register_parameters): Update accordingly.
1460         (check_sibcall_argument_overlap_1): Likewise.
1461         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1462         args size as a poly_int64 rather than an int.  Return a bool
1463         indicating whether the optimization was possible and return
1464         the new adjustment by reference.
1465         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1466         Update stored_args_watermark.
1467         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1468         (expand_call): Likewise.  Maintain stack_usage_watermark and
1469         stored_args_watermark.  Update calls to
1470         combine_pending_stack_adjustment_and_call.
1471         (emit_library_call_value_1): Handle polynomial argument sizes.
1472         Call stack_region_maybe_used_p and mark_stack_region_used.
1473         Maintain stack_usage_watermark.
1474         (store_one_arg): Likewise.  Update call to
1475         mem_overlaps_already_clobbered_arg_p.
1476         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1477         HOST_WIDE_INT.
1478         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1479         * config/microblaze/microblaze.c (microblaze_function_prologue):
1480         Likewise.
1481         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1482         TARGET_RETURN_POPS_ARGS interface.
1483         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1484         to HOST_WIDE_INT.
1485         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1486         * config/i386/i386.c (ix86_return_pops_args): Update for new
1487         TARGET_RETURN_POPS_ARGS interface.
1488         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1489         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1490         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1491         TARGET_RETURN_POPS_ARGS interface.
1492         * config/vax/vax.c (vax_return_pops_args): Likewise.
1493         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1494         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1495         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1496         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1497         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1498         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1499         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1500         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1501         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1502         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1503         and x_stack_pointer_delta from int to poly_int64.
1504         * dojump.c (do_pending_stack_adjust): Update accordingly.
1505         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1506         stack_pointer_deltas.
1507         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1508         (pad_to_arg_alignment): Track polynomial offsets.
1509         (assign_parm_find_stack_rtl): Likewise.
1510         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1511         * toplev.c (output_stack_usage): Update reference to
1512         current_function_pushed_stack_size.
1514 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1515             Alan Hayward  <alan.hayward@arm.com>
1516             David Sherwood  <david.sherwood@arm.com>
1518         * function.c (in_arg_offset, var_offset, dynamic_offset)
1519         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1520         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1521         rather than a HOST_WIDE_INT.
1522         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1523         (instantiate_virtual_regs_in_insn): Likewise.
1525 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1526             Alan Hayward  <alan.hayward@arm.com>
1527             David Sherwood  <david.sherwood@arm.com>
1529         * rtl.h (get_args_size, add_args_size_note): New functions.
1530         (find_args_size_adjust): Return a poly_int64 rather than a
1531         HOST_WIDE_INT.
1532         (fixup_args_size_notes): Likewise.  Make the same change to the
1533         end_args_size parameter.
1534         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1535         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1536         * calls.c (emit_call_1): Likewise.
1537         * explow.c (adjust_stack_1): Likewise.
1538         * cfgcleanup.c (old_insns_match_p): Update use of
1539         find_args_size_adjust.
1540         * combine.c (distribute_notes): Track polynomial arg sizes.
1541         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1542         end_true_args_size, beg_delay_args_size and end_delay_args_size
1543         from HOST_WIDE_INT to poly_int64.
1544         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1545         than a HOST_WIDE_INT.
1546         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1547         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1548         polynomial arg sizes.
1549         * emit-rtl.c (try_split): Use get_args_size.
1550         * recog.c (peep2_attempt): Likewise.
1551         * reload1.c (reload_as_needed): Likewise.
1552         * expr.c (find_args_size_adjust): Return the adjustment as a
1553         poly_int64 rather than a HOST_WIDE_INT.
1554         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1555         to a poly_int64 and change the return type in the same way.
1556         (emit_single_push_insn): Track polynomial arg sizes.
1558 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1559             Alan Hayward  <alan.hayward@arm.com>
1560             David Sherwood  <david.sherwood@arm.com>
1562         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1563         from HOST_WIDE_INT to poly_int64.
1564         * expr.c (push_block, emit_push_insn): Likewise.
1566 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1567             Alan Hayward  <alan.hayward@arm.com>
1568             David Sherwood  <david.sherwood@arm.com>
1570         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1571         to poly_int64.
1572         (get_frame_size): Return the size as a poly_int64 rather than a
1573         HOST_WIDE_INT.
1574         (frame_offset_overflow): Take the offset as a poly_int64 rather
1575         than a HOST_WIDE_INT.
1576         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1577         (assign_stack_temp): Likewise for the size.
1578         * function.c (get_frame_size): Return a poly_int64 rather than
1579         a HOST_WIDE_INT.
1580         (frame_offset_overflow): Take the offset as a poly_int64 rather
1581         than a HOST_WIDE_INT.
1582         (try_fit_stack_local): Take the start, length and size as poly_int64s
1583         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1584         rather than a HOST_WIDE_INT.
1585         (add_frame_space): Take the start and end as poly_int64s rather than
1586         HOST_WIDE_INTs.
1587         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1588         (assign_stack_temp): Likewise for the size.
1589         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1590         to poly_int64.
1591         (find_temp_slot_from_address): Handle polynomial offsets.
1592         (combine_temp_slots): Likewise.
1593         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1594         to poly_int64.
1595         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1596         poly_int64 rather than a HOST_WIDE_INT.
1597         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1598         than a HOST_WIDE_INT.
1599         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1600         polynomial frame offsets.
1601         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1602         as a poly_int64 rather than an int.
1603         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1604         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1605         * config/v850/v850.c (compute_frame_size): Likewise.
1606         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1607         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1608         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1609         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1610         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1611         rather than a HOST_WIDE_INT.
1612         * explow.c (get_dynamic_stack_base): Likewise.
1613         * final.c (final_start_function): Use the constant lower bound
1614         of the frame size for -Wframe-larger-than.
1615         * ira.c (do_reload): Adjust for new get_frame_size return type.
1616         * lra.c (lra): Likewise.
1617         * reload1.c (reload): Likewise.
1618         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1619         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1620         * rtlanal.c (get_initial_register_offset): Return the offset as
1621         a poly_int64 rather than a HOST_WIDE_INT.
1623 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1624             Alan Hayward  <alan.hayward@arm.com>
1625             David Sherwood  <david.sherwood@arm.com>
1627         * reload1.c (elim_table): Change initial_offset, offset and
1628         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1629         (offsets_at): Change the target array's element type from
1630         HOST_WIDE_INT to poly_int64_pod.
1631         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1632         (elimination_costs_in_insn, update_eliminable_offsets)
1633         (verify_initial_elim_offsets, set_offsets_for_label)
1634         (init_eliminable_invariants): Update after above changes.
1636 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1637             Alan Hayward  <alan.hayward@arm.com>
1638             David Sherwood  <david.sherwood@arm.com>
1640         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1641         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1642         (decomposition): Change start and end from HOST_WIDE_INT
1643         to poly_int64_pod.
1644         (decompose, immune_p): Update accordingly.
1645         (find_inc_amount): Return a poly_int64 rather than an int.
1646         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1647         rather than an int.
1649 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1650             Alan Hayward  <alan.hayward@arm.com>
1651             David Sherwood  <david.sherwood@arm.com>
1653         * tree.h (get_inner_reference): Return the bitsize and bitpos
1654         as poly_int64_pods rather than HOST_WIDE_INT.
1655         * fold-const.h (ptr_difference_const): Return the pointer difference
1656         as a poly_int64_pod rather than a HOST_WIDE_INT.
1657         * expr.c (get_inner_reference): Return the bitsize and bitpos
1658         as poly_int64_pods rather than HOST_WIDE_INT.
1659         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1660         offsets and sizes.
1661         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1662         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1663         (optimize_bit_field_compare): Update call to get_inner_reference.
1664         (decode_field_reference): Likewise.
1665         (fold_unary_loc): Track polynomial offsets and sizes.
1666         (split_address_to_core_and_offset): Return the bitpos as a
1667         poly_int64_pod rather than a HOST_WIDE_INT.
1668         (ptr_difference_const): Likewise for the pointer difference.
1669         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1670         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1671         * dbxout.c (dbxout_expand_expr): Likewise.
1672         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1673         (loc_list_from_tree_1, fortran_common): Likewise.
1674         * gimple-laddress.c (pass_laddress::execute): Likewise.
1675         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1676         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1677         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1678         * tree-affine.c (tree_to_aff_combination): Likewise.
1679         (get_inner_reference_aff): Likewise.
1680         * tree-data-ref.c (split_constant_offset_1): Likewise.
1681         (dr_analyze_innermost): Likewise.
1682         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1683         * tree-sra.c (ipa_sra_check_caller): Likewise.
1684         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1685         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1686         (instrument_bool_enum_load, instrument_object_size): Likewise.
1687         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1688         to get_inner_reference.
1689         * hsa-gen.c (gen_hsa_addr): Likewise.
1690         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1691         * tsan.c (instrument_expr): Likewise.
1692         * match.pd: Update call to ptr_difference_const.
1694 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1695             Alan Hayward  <alan.hayward@arm.com>
1696             David Sherwood  <david.sherwood@arm.com>
1698         * fold-const.c (fold_comparison): Track sizes and offsets as
1699         poly_int64s rather than HOST_WIDE_INTs when folding address
1700         comparisons.
1702 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1703             Alan Hayward  <alan.hayward@arm.com>
1704             David Sherwood  <david.sherwood@arm.com>
1706         * expr.h (get_bit_range): Return the bitstart and bitend as
1707         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1708         as a poly_int64 rather than a HOST_WIDE_INT.
1709         * expr.c (get_bit_range): Likewise.
1710         (expand_assignment): Update call accordingly.
1711         * fold-const.c (optimize_bit_field_compare): Likewise.
1713 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1714             Alan Hayward  <alan.hayward@arm.com>
1715             David Sherwood  <david.sherwood@arm.com>
1717         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1718         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1719         as an optional final parameter.
1720         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1721         (aarch64_legitimate_address_p): Likewise.
1722         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1723         instead of an rtx code.
1724         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1725         (aarch64_legitimate_address_hook_p): Likewise.
1726         (aarch64_print_ldpstp_address): Likewise.
1727         (aarch64_print_operand_address): Likewise.
1728         (aarch64_address_cost): Likewise.
1729         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1730         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1731         (aarch64_mem_pair_lanes_operand): Likewise.
1733 2017-12-20  Richard Biener  <rguenther@suse.de>
1735         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1736         update_stmt_if_modified.
1738 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1740         PR tree-optimization/83491
1741         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1742         before walking uses.  Improve coding style and comments.
1744 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1746         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1748 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1750         PR ipa/83506
1751         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1752         todo_flags_finish.
1753         (pass_ipa_free_fn_summary): Add small_p private data member,
1754         initialize to false in the ctor.
1755         (pass_ipa_free_fn_summary::clone,
1756         pass_ipa_free_fn_summary::set_pass_param,
1757         pass_ipa_free_fn_summary::gate): New methods.
1758         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1759         | TODO_dump_symtab if small_p.
1760         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1761         entry and add another instance of the pass with false parm after
1762         ipa-pure-const.
1763         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1764         ipa_free_fn_summary here.
1766 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1768         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1770 2017-12-20  Martin Sebor  <msebor@redhat.com>
1772         PR testsuite/83131
1773         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1774         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1775         (expand_builtin_strncmp): Same.
1777 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1779         PR bootstrap/83396
1780         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1781         allowed debug stmts before labels.
1782         (expand_gimple_basic_block): Likewise.
1783         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1784         * gimple-iterator.h (gsi_after_labels): Likewise.
1785         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1786         rename reused variable, and simplify using gsi_move_before.
1787         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1788         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1789         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1790         (gimple_verify_flow_info, gimple_block_label): Likewise.
1791         (make_blocks): Move debug markers after adjacent labels.
1792         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1793         allowed debug insns outside blocks.
1794         * df-scan.c (df_insn_delete): Likewise.
1795         * lra-constraints.c (update_ebb_live_info): Likewise.
1796         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1797         (vt_initialize, delete_vta_debug_insns): Likewise.
1798         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1800 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1801             Alan Hayward  <alan.hayward@arm.com>
1802             David Sherwood  <david.sherwood@arm.com>
1804         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1805         (round_up_to_byte_boundary): Likewise.
1806         * expr.h (get_bit_range): Add temporary shim.
1807         * gimple-ssa-store-merging.c (store_operand_info): Change the
1808         bitsize, bitpos, bitregion_start and bitregion_end fields from
1809         unsigned HOST_WIDE_INT to poly_uint64.
1810         (merged_store_group): Likewise load_align_base.
1811         (compatible_load_p, compatible_load_p): Update accordingly.
1812         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1813         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1814         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1815         bitregion_start and bitregion_end as poly_uint64s rather than
1816         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1817         (handled_load): Take the bitsize, bitpos,
1818         bitregion_start and bitregion_end as poly_uint64s rather than
1819         unsigned HOST_WIDE_INTs.
1820         (pass_store_merging::process_store): Update call to
1821         mem_valid_for_store_merging.
1823 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1824             Alan Hayward  <alan.hayward@arm.com>
1825             David Sherwood  <david.sherwood@arm.com>
1827         * builtins.c (get_object_alignment_2): Track polynomial offsets
1828         and sizes.  Update the alignment handling.
1830 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1831             Alan Hayward  <alan.hayward@arm.com>
1832             David Sherwood  <david.sherwood@arm.com>
1834         * tree.h (get_inner_reference): Add a version that returns the
1835         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1836         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1837         the case in which bitpos is not associated with the first byte.
1839 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1840             Alan Hayward  <alan.hayward@arm.com>
1841             David Sherwood  <david.sherwood@arm.com>
1843         * tree-affine.h (get_inner_reference_aff): Return the size as a
1844         poly_widest_int.
1845         * tree-affine.c (get_inner_reference_aff): Likewise.
1846         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1847         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1849 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1850             Alan Hayward  <alan.hayward@arm.com>
1851             David Sherwood  <david.sherwood@arm.com>
1853         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1854         HOST_WIDE_INT rather than a poly_int64.
1856 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1857             Alan Hayward  <alan.hayward@arm.com>
1858             David Sherwood  <david.sherwood@arm.com>
1860         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1861         HOST_WIDE_INT to poly_int64_pod.
1862         (perform_symbolic_merge): Update accordingly.
1863         (bswap_replace): Likewise.
1865 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1866             Alan Hayward  <alan.hayward@arm.com>
1867             David Sherwood  <david.sherwood@arm.com>
1869         * tree-affine.h (aff_tree::offset): Change from widest_int
1870         to poly_widest_int.
1871         (wide_int_ext_for_comb): Delete.
1872         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1873         constants as poly_widest_int rather than widest_int.
1874         (aff_combination_constant_multiple_p): Return the multiplier
1875         as a poly_widest_int.
1876         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1877         polynomial offsets.
1878         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1879         version static and add an overload for poly_widest_int.
1880         (aff_combination_const, aff_combination_add_cst)
1881         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1882         the constants as poly_widest_int rather than widest_int.
1883         (tree_to_aff_combination): Generalize INTEGER_CST case to
1884         poly_int_tree_p.
1885         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1886         (aff_combination_add_product, aff_combination_mult): Handle
1887         polynomial offsets.
1888         (aff_combination_constant_multiple_p): Return the multiplier
1889         as a poly_widest_int.
1890         * tree-predcom.c (determine_offset): Return the offset as a
1891         poly_widest_int.
1892         (split_data_refs_to_components, suitable_component_p): Update
1893         accordingly.
1894         (valid_initializer_p): Update call to
1895         aff_combination_constant_multiple_p.
1896         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1897         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1898         as a poly_int64 rather than a HOST_WIDE_INT.
1899         (get_address_cost): Handle polynomial offsets.
1900         (iv_elimination_compare_lt): Likewise.
1901         (rewrite_use_nonlinear_expr): Likewise.
1903 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1904             Alan Hayward  <alan.hayward@arm.com>
1905             David Sherwood  <david.sherwood@arm.com>
1907         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1908         as a poly_int64_pod rather than a HOST_WIDE_INT.
1909         (get_addr_base_and_unit_offset): Likewise.
1910         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1911         (get_addr_base_and_unit_offset): Likewise.
1912         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1913         to poly_int64 in example.
1914         * fold-const.c (fold_binary_loc): Update call to
1915         get_addr_base_and_unit_offset.
1916         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1917         (maybe_canonicalize_mem_ref_addr): Likewise.
1918         (gimple_fold_stmt_to_constant_1): Likewise.
1919         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1920         Likewise.
1921         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1922         * match.pd: Likewise.
1923         * omp-low.c (lower_omp_target): Likewise.
1924         * tree-sra.c (build_ref_for_offset): Likewise.
1925         (build_debug_ref_for_model): Likewise.
1926         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1927         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1928         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1929         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1930         (constant_pointer_difference): Likewise.
1931         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1932         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1933         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1934         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1935         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1936         (set_ssa_val_to): Likewise.
1937         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1938         (maybe_diag_stxncpy_trunc): Likewise.
1939         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1940         * tree.c (build_simple_mem_ref_loc): Likewise.
1941         (array_at_struct_end_p): Likewise.
1943 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1944             Alan Hayward  <alan.hayward@arm.com>
1945             David Sherwood  <david.sherwood@arm.com>
1947         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
1948         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1949         (get_ref_base_and_extent_hwi): Declare.
1950         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
1951         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1952         (get_ref_base_and_extent_hwi): New function.
1953         * cfgexpand.c (expand_debug_expr): Update call to
1954         get_ref_base_and_extent.
1955         * dwarf2out.c (add_var_loc_to_decl): Likewise.
1956         * gimple-fold.c (get_base_constructor): Return the offset as a
1957         poly_int64_pod rather than a HOST_WIDE_INT.
1958         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
1959         * ipa-polymorphic-call.c
1960         (ipa_polymorphic_call_context::set_by_invariant)
1961         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
1962         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
1963         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
1964         rather than get_ref_base_and_extent.
1965         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
1966         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
1967         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
1968         Likewise.
1969         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
1970         call to get_ref_base_and_extent.
1971         * tree-sra.c (create_access, get_access_for_expr): Likewise.
1972         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
1973         (stmt_kills_ref_p): Likewise.
1974         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
1975         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
1976         Likewise.
1977         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
1978         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
1979         when calling native_encode_expr.
1980         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
1981         call to get_ref_base_and_extent.
1982         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
1983         get_ref_base_and_extent.
1984         * var-tracking.c (track_expr_p): Likewise.
1986 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1987             Alan Hayward  <alan.hayward@arm.com>
1988             David Sherwood  <david.sherwood@arm.com>
1990         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
1991         HOST_WIDE_INT to poly_int64_pod.
1992         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
1993         polynomail parameter offsets.
1995 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1996             Alan Hayward  <alan.hayward@arm.com>
1997             David Sherwood  <david.sherwood@arm.com>
1999         * gengtype.c (main): Handle poly_int64_pod.
2000         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
2001         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
2002         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
2003         from HOST_WIDE_INT to poly_int64_pod.
2004         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
2005         (copy_cfa): New function.
2006         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
2007         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
2008         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
2009         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
2010         (initial_return_save): Treat offsets as poly_ints.
2011         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
2012         offset is nonconstant.
2013         (reg_save): Take the offset as a poly_int64.  Fall back to
2014         DW_CFA_expression for nonconstant offsets.
2015         (queue_reg_save): Take the offset as a poly_int64.
2016         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
2018 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2019             Alan Hayward  <alan.hayward@arm.com>
2020             David Sherwood  <david.sherwood@arm.com>
2022         * rtl.h (operand_subword, operand_subword_force): Take the offset
2023         as a poly_uint64 an unsigned int.
2024         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
2026 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2027             Alan Hayward  <alan.hayward@arm.com>
2028             David Sherwood  <david.sherwood@arm.com>
2030         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2031         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2032         a code with an XINT and an XEXP.  Remove the implication that
2033         accessing an rtx field using XINT is expected to work.
2034         * rtl.def (SUBREG): Change format from "ei" to "ep".
2035         * rtl.h (rtunion::rt_subreg): New field.
2036         (XCSUBREG): New macro.
2037         (SUBREG_BYTE): Use it.
2038         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2039         Update constructor accordingly.
2040         (subreg_shape::operator ==): Update accordingly.
2041         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2042         than an unsigned int.
2043         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2044         a poly_uint64 rather than an unsigned int.
2045         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2046         than an unsigned int.
2047         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2048         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2049         an unsigned int.  Take the sizes as poly_uint64s.
2050         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2051         an unsigned int.  Take the shift as a poly_uint64 rather than
2052         an unsigned int.
2053         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2054         as a poly_uint64 rather than an unsigned int.
2055         (simplify_subreg_regno): Likewise.
2056         (byte_lowpart_offset): Return the memory offset as a poly_int64
2057         rather than an int.
2058         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2059         poly_uint64 rather than an unsigned int.
2060         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2061         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2062         poly_uint64 rather than an unsigned int.
2063         * rtl.c (rtx_format): Describe 'p' in comment.
2064         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2065         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2066         offset as a poly_uint64 rather than an unsigned int.
2067         (byte_lowpart_offset): Return the memory offset as a poly_int64
2068         rather than an int.
2069         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2070         poly_uint64 rather than an unsigned int.
2071         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2072         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2073         poly_uint64 rather than an unsigned int.
2074         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2075         (copy_insn_1): Handle 'p'.
2076         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2077         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2078         an unsigned int.  Return the shift in the same way.
2079         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2080         unsigned int.
2081         (subreg_size_offset_from_lsb): Take the sizes and shift as
2082         poly_uint64s rather than unsigned ints.  Return the offset as
2083         a poly_uint64.
2084         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2085         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2086         an unsigned int.
2087         * rtlhash.c (add_rtx): Handle 'p'.
2088         * genemit.c (gen_exp): Likewise.
2089         * gengenrtl.c (type_from_format, gendef): Likewise.
2090         * gensupport.c (subst_pattern_match, get_alternatives_number)
2091         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2092         (subst_dup): Likewise.
2093         * gengtype.c (adjust_field_rtx_def): Likewise.
2094         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2095         (match_pattern_2): Likewise.
2096         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2097         (rtx_test::subreg_field): New function.
2098         (operator ==, safe_to_hoist_p, transition_parameter_type)
2099         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2100         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2101         * genpeep.c (match_rtx): Likewise.
2102         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2103         (rtx_writer::print_rtx_operand): Handle 'p'.
2104         (print_value): Handle SUBREG.
2105         * read-rtl.c (apply_int_iterator): Likewise.
2106         (rtx_reader::read_rtx_operand): Handle 'p'.
2107         * alias.c (rtx_equal_for_memref_p): Likewise.
2108         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2109         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2110         as poly_ints.
2111         * calls.c (expand_call): Likewise.
2112         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2113         (make_extraction, gen_lowpart_for_combine): Likewise.
2114         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2115         Likewise.
2116         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2117         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2118         (exp_equiv_p): Handle 'p'.
2119         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2120         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2121         * dse.c (find_shift_sequence): Likewise.
2122         * dwarf2out.c (rtl_for_decl_location): Likewise.
2123         * expmed.c (extract_low_bits): Likewise.
2124         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2125         (expand_expr_real_2): Likewise.
2126         * final.c (alter_subreg): Likewise.
2127         (leaf_renumber_regs_insn): Handle 'p'.
2128         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2129         Treat subreg offsets as poly_ints.
2130         * fwprop.c (forward_propagate_and_simplify): Likewise.
2131         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2132         * ira.c (get_subreg_tracking_sizes): Likewise.
2133         * ira-conflicts.c (go_through_subreg): Likewise.
2134         * ira-lives.c (process_single_reg_class_operands): Likewise.
2135         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2136         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2137         as a poly_uint64 rather than an unsigned int.
2138         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2139         subreg offsets as poly_ints.
2140         * lra-constraints.c (operands_match_p): Handle 'p'.
2141         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2142         * lra-spills.c (assign_mem_slot): Likewise.
2143         * postreload.c (move2add_valid_value_p): Likewise.
2144         * recog.c (general_operand, indirect_operand): Likewise.
2145         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2146         (copyprop_hardreg_forward_1): Likewise.
2147         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2148         (record_subregs_of_mode): Likewise.
2149         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2150         * reload.c (operands_match_p): Handle 'p'.
2151         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2152         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2153         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2154         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2155         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2156         (simplify_const_poly_int_tests<N>::run): Likewise.
2157         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2158         a poly_uint64 rather than an unsigned int.
2159         * valtrack.c (debug_lowpart_subreg): Likewise.
2160         * var-tracking.c (var_lowpart): Likewise.
2161         (loc_cmp): Handle 'p'.
2163 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2164             Alan Hayward  <alan.hayward@arm.com>
2165             David Sherwood  <david.sherwood@arm.com>
2167         * ira.c (get_subreg_tracking_sizes): New function.
2168         (init_live_subregs): Take an integer size rather than a register.
2169         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2170         to init_live_subregs.
2172 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2173             Alan Hayward  <alan.hayward@arm.com>
2174             David Sherwood  <david.sherwood@arm.com>
2176         * expr.c (store_constructor_field): Change bitsize from a
2177         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2178         HOST_WIDE_INT to a poly_int64.
2179         (store_constructor): Change size from a HOST_WIDE_INT to
2180         a poly_int64.
2181         (store_field): Likewise bitsize and bitpos.
2183 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2184             Alan Hayward  <alan.hayward@arm.com>
2185             David Sherwood  <david.sherwood@arm.com>
2187         * expmed.h (store_bit_field): Change bitregion_start and
2188         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2189         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2190         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2191         (store_fixed_bit_field, store_split_bit_field): Likewise.
2192         * expr.c (store_constructor_field, store_field): Likewise.
2193         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2194         to bitsize and bitpos.
2195         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2196         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2197         same change in the constructor arguments.
2198         (get_best_mode): Change bitregion_start and bitregion_end from
2199         unsigned HOST_WIDE_INT to poly_uint64.
2200         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2201         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2202         poly_int64.
2203         (bit_field_mode_iterator::next_mode): Update for new types
2204         of m_bitregion_start and m_bitregion_end.
2205         (get_best_mode): Change bitregion_start and bitregion_end from
2206         unsigned HOST_WIDE_INT to poly_uint64.
2208 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2209             Alan Hayward  <alan.hayward@arm.com>
2210             David Sherwood  <david.sherwood@arm.com>
2212         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2213         accepts poly_uint64 offsets.
2214         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2215         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2216         * expmed.c (lowpart_bit_field_p): Likewise.
2217         (extract_bit_field_as_subreg): New function, split out from...
2218         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2219         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2220         extractions, check that BITSIZE matches the size of the extracted
2221         value and that BITNUM is an exact multiple of that size.
2222         If all else fails, try forcing the value into memory if
2223         BITNUM is variable, and adjusting the address so that the
2224         offset is constant.  Split the part that can only handle constant
2225         bitsize and bitnum out into...
2226         (extract_integral_bit_field): ...this new function.
2227         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2228         rather than unsigned HOST_WIDE_INTs.
2230 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2231             Alan Hayward  <alan.hayward@arm.com>
2232             David Sherwood  <david.sherwood@arm.com>
2234         * expmed.h (store_bit_field): Take bitsize and bitnum as
2235         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2236         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2237         that returns the byte size.
2238         (store_bit_field_1): Take bitsize and bitnum as
2239         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2240         to simple_mem_bitfield_p.  Split the part that can only handle
2241         constant bitsize and bitnum out into...
2242         (store_integral_bit_field): ...this new function.
2243         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2244         than unsigned HOST_WIDE_INTs.
2245         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2247 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2248             Alan Hayward  <alan.hayward@arm.com>
2249             David Sherwood  <david.sherwood@arm.com>
2251         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2252         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2253         to poly_int64.
2254         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2255         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2256         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2257         offset as a poly_int64 rather than an int.
2258         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2259         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2260         * lra-constraints.c (equiv_address_substitution): Track offsets
2261         as poly_int64s.
2262         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2263         (curr_insn_transform): Handle the new form of sp_offset.
2264         * lra-eliminations.c (lra_elim_table): Change previous_offset
2265         and offset from HOST_WIDE_INT to poly_int64.
2266         (print_elim_table, update_reg_eliminate): Update accordingly.
2267         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2268         (get_elimination): Update accordingly.
2269         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2270         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2271         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2272         poly_int64 offsets generally.
2273         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2274         (mark_not_eliminable, init_elimination): Update accordingly.
2275         (remove_reg_equal_offset_note): Return a bool and pass the new
2276         offset back by pointer as a poly_int64.
2277         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2278         rather than a HOST_WIDE_INT.
2279         (do_remat): Track offsets poly_int64s.
2280         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2282 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2283             Alan Hayward  <alan.hayward@arm.com>
2284             David Sherwood  <david.sherwood@arm.com>
2286         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2287         offset from HOST_WIDE_INT to poly_int64.
2288         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2289         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2290         (widen_memory_access): Take the sizes and offsets as poly_int64s
2291         rather than HOST_WIDE_INTs.
2292         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2293         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2294         and ints.
2295         (adjust_offset_for_component_ref): Change the offset from a
2296         HOST_WIDE_INT to a poly_int64.
2297         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2298         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2299         * dce.c (find_call_stack_args): Likewise.
2300         * dse.c (record_store): Likewise.
2301         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2302         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2303         * read-rtl-function.c (test_loading_mem): Likewise.
2304         * rtlanal.c (may_trap_p_1): Likewise.
2305         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2306         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2307         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2308         (mem_attrs::mem_attrs): New function.
2309         (set_mem_attributes_minus_bitpos): Change bitpos from a
2310         HOST_WIDE_INT to poly_int64.
2311         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2312         (clear_mem_offset, clear_mem_size, change_address)
2313         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2314         initialize mem_attrs.
2315         (set_mem_offset, set_mem_size, adjust_address_1)
2316         (adjust_automodify_address_1, offset_address, widen_memory_access):
2317         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2319 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2320             Alan Hayward  <alan.hayward@arm.com>
2321             David Sherwood  <david.sherwood@arm.com>
2323         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2324         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2325         rather than 0 to represent an unknown size.  Assert that the size
2326         is known when the mode isn't BLKmode.
2327         (may_trap_p_1): Use -1 for unknown sizes.
2328         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2330 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2331             Alan Hayward  <alan.hayward@arm.com>
2332             David Sherwood  <david.sherwood@arm.com>
2334         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2335         to poly_int64.  Update commentary for positions_needed.large.
2336         (read_info_type): Change offset and width from HOST_WIDE_INT
2337         to poly_int64.
2338         (set_usage_bits): Likewise.
2339         (canon_address): Return the offset as a poly_int64 rather than
2340         a HOST_WIDE_INT.  Use strip_offset_and_add.
2341         (set_all_positions_unneeded, any_positions_needed_p): Use
2342         positions_needed.large to track stores with non-constant widths.
2343         (all_positions_needed_p): Likewise.  Take the offset and width
2344         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2345         (record_store): Cope with non-constant offsets and widths.
2346         Nullify the rhs of an earlier store if we can't tell which bytes
2347         of it are needed.
2348         (find_shift_sequence): Take the access_size and shift as poly_int64s
2349         rather than ints.
2350         (get_stored_val): Take the read_offset and read_width as poly_int64s
2351         rather than HOST_WIDE_INTs.
2352         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2353         non-constant offsets and widths.
2355 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2356             Alan Hayward  <alan.hayward@arm.com>
2357             David Sherwood  <david.sherwood@arm.com>
2359         * inchash.h (inchash::hash::add_poly_int): New function.
2360         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2361         Use poly_int64 rather than HOST_WIDE_INT.
2362         (ao_ref::max_size_known_p): New function.
2363         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2364         rather than HOST_WIDE_INT.
2365         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2366         to temporaries until its interface is adjusted to match.
2367         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2368         (aliasing_component_refs_p, decl_refs_may_alias_p)
2369         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2370         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2371         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2372         ao_ref fields.
2373         * alias.c (ao_ref_from_mem): Likewise.
2374         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2375         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2376         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2377         (maybe_trim_complex_store, maybe_trim_constructor_store)
2378         (live_bytes_read, dse_classify_store): Likewise.
2379         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2380         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2381         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2382         (vn_reference_lookup_3): Likewise.
2383         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2385 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2386             Alan Hayward  <alan.hayward@arm.com>
2387             David Sherwood  <david.sherwood@arm.com>
2389         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2390         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2391         instead of ranges_overlap_p.
2393 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2394             Alan Hayward  <alan.hayward@arm.com>
2395             David Sherwood  <david.sherwood@arm.com>
2397         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2398         sizes as poly_int64s rather than HOST_WIDE_INTs.
2400 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2401             Alan Hayward  <alan.hayward@arm.com>
2402             David Sherwood  <david.sherwood@arm.com>
2404         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2405         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2406         * gimple-fold.c (fold_ctor_reference): Likewise.
2408 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2409             Alan Hayward  <alan.hayward@arm.com>
2410             David Sherwood  <david.sherwood@arm.com>
2412         * target.def (dwarf_poly_indeterminate_value): New hook.
2413         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2414         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2415         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2416         * doc/tm.texi: Regenerate.
2417         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2418         offset as a poly_int64.
2419         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2420         offset as a poly_int64.
2421         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2422         Take the offset as a poly_int64.
2423         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2424         (frame_pointer_fb_offset): Change to a poly_int64.
2425         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2426         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2427         (based_loc_descr): Take the offset as a poly_int64.
2428         Use strip_offset_and_add to handle (plus X (const)).
2429         Use new_reg_loc_descr instead of an open-coded version of the
2430         previous implementation.
2431         (mem_loc_descriptor): Handle CONST_POLY_INT.
2432         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2433         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2435 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2436             Alan Hayward  <alan.hayward@arm.com>
2437             David Sherwood  <david.sherwood@arm.com>
2439         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2440         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2441         * inchash.h (inchash::hash::add_poly_hwi): New function.
2442         * gengtype.c (main): Register poly_int64.
2443         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2444         offset as a poly_int.
2445         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2446         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2447         offset as a poly_int64.
2448         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2449         * print-rtl.c (print_poly_int): New function.
2450         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2451         a poly_int.
2452         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2453         functions.
2454         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2455         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2456         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2457         Enforce track_offset_p for parts of a PARALLEL.
2458         (vt_add_function_parameter): Use const_offset for the final
2459         offset to track.  Use get_tracked_reg_offset for the parts
2460         of a PARALLEL.
2462 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2463             Alan Hayward  <alan.hayward@arm.com>
2464             David Sherwood  <david.sherwood@arm.com>
2466         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2467         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2468         * doc/tm.texi: Regenerate.
2469         * hooks.h (hook_bool_uint_uint_true): Delete.
2470         (hook_bool_puint64_puint64_true): Declare.
2471         * hooks.c (hook_bool_uint_uint_true): Delete.
2472         (hook_bool_puint64_puint64_true): New function.
2473         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2474         instead of unsigned ints.
2475         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2476         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2478 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2479             Alan Hayward  <alan.hayward@arm.com>
2480             David Sherwood  <david.sherwood@arm.com>
2482         * optabs.h (expand_operand): Add an int_value field.
2483         (create_expand_operand): Add an int_value parameter and use it
2484         to initialize the new expand_operand field.
2485         (create_integer_operand): Replace with a declaration of a function
2486         that accepts poly_int64s.  Move the implementation to...
2487         * optabs.c (create_integer_operand): ...here.
2488         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2489         the mode preserves the value of int_value, instead of calling
2490         const_int_operand on the rtx.  Use gen_int_mode to generate
2491         the new rtx.
2493 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2494             Alan Hayward  <alan.hayward@arm.com>
2495             David Sherwood  <david.sherwood@arm.com>
2497         * dumpfile.h (dump_dec): Declare.
2498         * dumpfile.c (dump_dec): New function.
2499         * pretty-print.h (pp_wide_integer): Turn into a function and
2500         declare a poly_int version.
2501         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2503 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2504             Alan Hayward  <alan.hayward@arm.com>
2505             David Sherwood  <david.sherwood@arm.com>
2507         * doc/generic.texi (POLY_INT_CST): Document.
2508         * tree.def (POLY_INT_CST): New tree code.
2509         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2510         * tree-core.h (tree_poly_int_cst): New struct.
2511         (tree_node): Add a poly_int_cst field.
2512         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2513         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2514         instead of a wide_int_ref.
2515         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2516         of a HOST_WIDE_INT.
2517         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2518         instead of an unsigned HOST_WIDE_INT.
2519         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2520         (ptrdiff_tree_p): Declare.
2521         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2522         extern inline implementations if the target doesn't use POLY_INT_CST.
2523         (poly_int_tree_p): New function.
2524         (wi::unextended_tree): New class.
2525         (wi::int_traits <unextended_tree>): New override.
2526         (wi::extended_tree): Add a default constructor.
2527         (wi::extended_tree::get_tree): New function.
2528         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2529         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2530         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2531         (wi::tree_to_poly_wide_ref): New typedefs.
2532         (wi::ints_for): Provide overloads for extended_tree and
2533         unextended_tree.
2534         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2535         (wi::to_wide): New functions.
2536         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2537         * tree.c (poly_int_cst_hasher): New struct.
2538         (poly_int_cst_hash_table): New variable.
2539         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2540         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2541         POLY_INT_CST.
2542         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2543         (init_ttree): Initialize poly_int_cst_hash_table.
2544         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2545         a poly_int64 instead of a HOST_WIDE_INT.
2546         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2547         instead of an unsigned HOST_WIDE_INT.
2548         (wide_int_to_tree): Rename to...
2549         (wide_int_to_tree_1): ...this.
2550         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2551         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2552         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2553         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2554         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2555         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2556         TS_POLY_INT_CST.
2557         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2558         (streamer_read_tree_body): Likewise.
2559         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2560         (streamer_write_tree_body): Likewise.
2561         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2562         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2563         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2564         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2565         * gimple-expr.h (is_gimple_constant): Likewise.
2566         * gimplify.c (maybe_with_size_expr): Likewise.
2567         * print-tree.c (print_node): Likewise.
2568         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2569         * tree-pretty-print.c (dump_generic_node): Likewise.
2570         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2571         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2572         * tree-vrp.c (compare_values_warnv): Likewise.
2573         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2574         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2575         (force_expr_to_var_cost): Likewise.
2576         * tree-ssa-loop.c (for_each_index): Likewise.
2577         * fold-const.h (build_invariant_address, size_int_kind): Take a
2578         poly_int64 instead of a HOST_WIDE_INT.
2579         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2580         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2581         POLY_INT_CST.
2582         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2583         (int_const_binop_2): New function, split out from...
2584         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2585         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2586         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2587         * gimple-ssa-strength-reduction.c (slsr_process_add)
2588         (slsr_process_mul): Check for INTEGER_CSTs before using them
2589         as candidates.
2590         * stor-layout.c (bits_from_bytes): New function.
2591         (bit_from_pos): Use it.
2592         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2593         by BITS_PER_UNIT to get the TYPE_SIZE.
2594         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2595         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2597 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2598             Alan Hayward  <alan.hayward@arm.com>
2599             David Sherwood  <david.sherwood@arm.com>
2601         * doc/rtl.texi (const_poly_int): Document.  Also document the
2602         rtl sharing behavior.
2603         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2604         * rtl.h (const_poly_int_def): New struct.
2605         (rtx_def::u): Add a cpi field.
2606         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2607         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2608         (wi::rtx_to_poly_wide_ref): New typedef
2609         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2610         (poly_int_rtx_p): New functions.
2611         (trunc_int_for_mode): Declare a poly_int64 version.
2612         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2613         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2614         a wide_int_ref.
2615         (strip_offset): Declare.
2616         (strip_offset_and_add): New function.
2617         * rtl.def (CONST_POLY_INT): New rtx code.
2618         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2619         (shared_const_p): Use poly_int_rtx_p.
2620         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2621         HOST_WIDE_INT.
2622         (gen_int_shift_amount): Likewise.
2623         * emit-rtl.c (const_poly_int_hasher): New class.
2624         (const_poly_int_htab): New variable.
2625         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2626         (const_poly_int_hasher::hash): New function.
2627         (const_poly_int_hasher::equal): Likewise.
2628         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2629         (immed_wide_int_const): Rename to...
2630         (immed_wide_int_const_1): ...this and make static.
2631         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2632         instead of a wide_int_ref.
2633         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2634         (gen_lowpart_common): Handle CONST_POLY_INT.
2635         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2636         * cselib.c (cselib_hash_rtx): Likewise.
2637         * dwarf2out.c (const_ok_for_output_1): Likewise.
2638         * expr.c (convert_modes): Likewise.
2639         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2640         * rtlhash.c (add_rtx): Likewise.
2641         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2642         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2643         Handle existing CONST_POLY_INT rtxes.
2644         * expmed.h (expand_shift): Take a poly_int64 instead of a
2645         HOST_WIDE_INT.
2646         * expmed.c (expand_shift): Likewise.
2647         * rtlanal.c (strip_offset): New function.
2648         (commutative_operand_precedence): Give CONST_POLY_INT the same
2649         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2650         and CONST_INT.
2651         * rtl-tests.c (const_poly_int_tests): New struct.
2652         (rtl_tests_c_tests): Use it.
2653         * simplify-rtx.c (simplify_const_unary_operation): Handle
2654         CONST_POLY_INT.
2655         (simplify_const_binary_operation): Likewise.
2656         (simplify_binary_operation_1): Fold additions of symbolic constants
2657         and CONST_POLY_INTs.
2658         (simplify_subreg): Handle extensions and truncations of
2659         CONST_POLY_INTs.
2660         (simplify_const_poly_int_tests): New struct.
2661         (simplify_rtx_c_tests): Use it.
2662         * wide-int.h (storage_ref): Add default constructor.
2663         (wide_int_ref_storage): Likewise.
2664         (trailing_wide_ints): Use GTY((user)).
2665         (trailing_wide_ints::operator[]): Add a const version.
2666         (trailing_wide_ints::get_precision): New function.
2667         (trailing_wide_ints::extra_size): Likewise.
2669 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2670             Alan Hayward  <alan.hayward@arm.com>
2671             David Sherwood  <david.sherwood@arm.com>
2673         * emit-rtl.h (gen_int_shift_amount): Declare.
2674         * emit-rtl.c (gen_int_shift_amount): New function.
2675         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2676         instead of GEN_INT.
2677         * calls.c (shift_return_value): Likewise.
2678         * cse.c (fold_rtx): Likewise.
2679         * dse.c (find_shift_sequence): Likewise.
2680         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2681         (expand_shift, expand_smod_pow2): Likewise.
2682         * lower-subreg.c (shift_cost): Likewise.
2683         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2684         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2685         (expand_vec_perm_var): Likewise.
2686         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2687         (simplify_binary_operation_1): Likewise.
2688         * combine.c (try_combine, find_split_point, force_int_to_mode)
2689         (simplify_shift_const_1, simplify_shift_const): Likewise.
2690         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2692 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2694         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2696 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2698         * doc/invoke.texi (ARM Options): Document accepted extension options
2699         for -march=armv8.3-a.
2701 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2703         PR target/83105
2704         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2705         or --with-float=softfp, set the default CPU to arm10e.
2707 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2709         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2710         * config/visium/predicates.md (const_shift_operand): Likewise.
2711         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2712         (visium_legitimize_reload_address): Likewise.
2714 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2716         PR target/82975
2717         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2719 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2721         PR c++/83490
2722         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2724 2017-12-20  Julia Koval  <julia.koval@intel.com>
2726         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2727         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2728         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2729         * config.gcc: Include vpclmulqdqintrin.h.
2730         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2731         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2732         -mvpclmulqdq.
2733         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2734         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2735         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2736         * config/i386/i386.c (isa2_opts): Add -mcx16.
2737         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2738         (ix86_option_override_internal): Move mcx16 to flags2.
2739         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2740         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2741         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2742         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2743         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2744         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2745         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2746         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2747         * doc/invoke.texi: Add -mvpclmulqdq.
2749 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2751         PR middle-end/83423
2752         * config/i386/i386.c (ix86_static_chain): Move
2753         DECL_STATIC_CHAIN test ...
2754         * calls.c (rtx_for_static_chain): ... here.  New function.
2755         * calls.h (rtx_for_static_chain): Declare.
2756         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2757         instead of targetm.calls.static_chain.
2758         * df-scan.c (df_get_entry_block_def_set): Same.
2760 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2762         PR tree-optimization/83493
2763         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2764         and lb.
2766 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2768         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2769         inform with hardcoded english plural handling.
2771 2017-12-18  Jeff Law  <law@redhat.com>
2773         PR tree-optimization/83477
2774         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2775         a non-virtual PHI, always push a new range.
2777 2017-12-19  Martin Sebor  <msebor@redhat.com>
2779         PR middle-end/77608
2780         * builtins.c (compute_objsize): Handle non-constant offsets.
2782 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2784         PR tree-optimization/83444
2785         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2786         character load case, if get_stridx on MEM_REF's operand doesn't
2787         look usable, retry with get_addr_stridx.
2789 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2791         PR debug/83422
2792         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2793         (variable_tracking_main_1): Keep markers even when VTA fails.
2795         PR bootstrap/83396
2796         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2797         even if there are markers before it.
2798         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2800 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2802         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2803         typical order conditions.
2804         * sel-sched.c (extract_new_fences_from): Likewise.
2805         * config/visium/constraints.md (J, K, L): Likewise.
2806         * config/visium/predicates.md (const_shift_operand): Likewise.
2807         * config/visium/visium.c (visium_legitimize_address,
2808         visium_legitimize_reload_address): Likewise.
2809         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2810         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2811         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2812         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2813         SET_DUMP_DETAIL): Likewise.
2814         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2815         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2816         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2817         avr_set_core_architecture, avr_set_current_function,
2818         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2819         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2820         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2821         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2822         avr_map_decompose, avr_fold_builtin): Likewise.
2823         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2824         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2825         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2826         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2827         * config/m32c/m32c.c (m32c_conditional_register_usage,
2828         m32c_address_cost): Likewise.
2829         * config/m32c/predicates.md (shiftcount_operand,
2830         longshiftcount_operand): Likewise.
2831         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2832         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2833         can_use_cdx_ldstw): Likewise.
2834         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2835         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2836         Likewise.
2837         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2838         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2839         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2840         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2841         * config/vax/vax.c (adjacent_operands_p): Likewise.
2842         * config/ft32/constraints.md (L, b, KA): Likewise.
2843         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2844         Likewise.
2845         * cfgexpand.c (expand_stack_alignment): Likewise.
2846         * gcse.c (insert_expr_in_table): Likewise.
2847         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2848         * cgraphunit.c (cgraph_node::expand): Likewise.
2849         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2850         * emit-rtl.c (add_insn): Likewise.
2851         * input.c (dump_location_info): Likewise.
2852         * passes.c (NEXT_PASS): Likewise.
2853         * read-rtl-function.c (parse_note_insn_name,
2854         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2855         Likewise.
2856         * sched-rgn.c (sched_rgn_init): Likewise.
2857         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2858         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2859         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2860         Likewise.
2861         * explow.c (eliminate_constant_term): Likewise.
2862         * final.c (leaf_renumber_regs_insn): Likewise.
2863         * cfgrtl.c (print_rtl_with_bb): Likewise.
2864         * genhooks.c (emit_init_macros): Likewise.
2865         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2866         * tree-data-ref.c (conflict_fn): Likewise.
2867         * selftest.c (assert_streq): Likewise.
2868         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2869         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2870         fold_binary_loc, multiple_of_p): Likewise.
2871         * reload.c (push_reload, find_equiv_reg): Likewise.
2872         * et-forest.c (et_nca, et_below): Likewise.
2873         * dbxout.c (dbxout_symbol_location): Likewise.
2874         * reorg.c (relax_delay_slots): Likewise.
2875         * dojump.c (do_compare_rtx_and_jump): Likewise.
2876         * gengtype-parse.c (type): Likewise.
2877         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2878         simplify_const_relational_operation): Likewise.
2879         * reload1.c (do_output_reload): Likewise.
2880         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2881         * gengtype.c (type_for_name): Likewise.
2882         * gimple-ssa-sprintf.c (format_directive): Likewise.
2884 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2886         PR target/82975
2887         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2888         accessing it.  Adjust comment.
2890 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2892         PR middle-end/81914
2893         * predict.c (zero_one_minusone): New function.
2894         (apply_return_prediction): Avoid return prediction for functions
2895         returning only -1, 0 and 1 values, unless they only return -1 and 0
2896         or 0 and 1.
2898 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2900         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2901         fall-through warning.
2902         (arc_compute_frame_size): Remove unused variables.
2903         (arc_print_operand): Fix fprintif format.
2904         (arc_can_follow_jump): Clean fall-through warning.
2906 2017-12-19  Marek Polacek  <polacek@redhat.com>
2908         PR c++/83489
2909         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2910         on an error node.
2912 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2914         * config/arc/arc.c (overriderregs): New variable.
2915         (arc_override_options): Track fixed/call saved/call options.
2916         (arc_conditional_register_usage): Check against overrideregs
2917         variable whenever we change register properties.
2919 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2921         * opts.c (finish_options): Don't prefix dump_base_name if it
2922         already contains directories.
2924 2017-12-19  Martin Liska  <mliska@suse.cz>
2926         PR rtl-optimization/82675
2927         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2928         more element in sbitmap.
2930 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
2932         * gimplify.c (gimplify_expr): Use error_operand_p.
2934 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2936         PR target/83387
2937         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2938         multilib.
2940 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
2942         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
2943         the jump is to a label.
2945 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2947         PR tree-optimization/83444
2948         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
2949         character loads.
2951         PR ipa/82801
2952         PR ipa/83346
2953         * ipa-inline.c (flatten_remove_node_hook): New function.
2954         (ipa_inline): Keep only nodes with flatten attribute at the end of
2955         the array in the order from ipa_reverse_postorder, only walk that
2956         portion of array for flattening, if there is more than one such
2957         node, temporarily register a removal hook and ignore removed nodes.
2959         PR tree-optimization/80631
2960         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
2961         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
2962         IFN_REDUC_MAX or IFN_REDUC_MIN.
2964 2017-12-18  Jeff Law  <law@redhat.com>
2966         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
2967         record symbolic equivalences from backedges in the CFG.
2969         Revert
2970         2017-11-19  Jeff Law  <law@redhat.com>
2972         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
2973         of degenerates resulting from ignoring an edge.
2975 2017-12-18  Martin Sebor  <msebor@redhat.com>
2977         PR middle-end/83373
2978         PR tree-optimization/78450
2979         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
2980         (handle_builtin_strlen): Call it.
2982 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
2984         PR rtl-optimization/83424
2985         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
2987 2017-12-18  Marek Polacek  <polacek@redhat.com>
2989         PR middle-end/83463
2990         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2991         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
2992         values.
2994 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2996         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
2997         (umaddsidi4, umaddsidi_split): Likewise.
2999 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3001         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
3002         constants.
3004 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3006         PR target/83420
3007         * config/s390/s390.c (s390_option_override): Avoid strncpy.
3009 2017-12-18  Richard Biener  <rguenther@suse.de>
3011         PR tree-optimization/81877
3012         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
3013         (outermost_indep_loop): Adjust.
3014         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
3015         (can_sm_ref_p): Adjust.
3017 2017-12-18  Richard Biener  <rguenther@suse.de>
3019         PR middle-end/77291
3020         * tree.c (array_at_struct_end_p): Return true if the underlying
3021         object has space for at least one element in excess of what
3022         the array domain specifies.
3024 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3026         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3027         example to avoid overfull hbox.
3028         * doc/invoke.texi (Option Summary): Add missing @gol.
3029         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3030         overfull hbox.
3032 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3033             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3035         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3036         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3037         (AArch64 Options): Document -mverbose-cost-dump.
3038         (ARM Options): Likewise, plus -mflip-thumb.
3040 2017-12-17  Martin Sebor  <msebor@redhat.com>
3042         PR bootstrap/83446
3043         * gimple-ssa-warn-restrict.c
3044         (builtin_memref::offset_out_of_bounds): Correct the handling of
3045         anti-ranges.
3047 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3049         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3051 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3053         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3055 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3057         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3058         latencies.
3060 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3062         * doc/invoke.texi: Fix some typos.
3064 2017-12-16  Martin Sebor  <msebor@redhat.com>
3066         PR tree-optimization/78918
3067         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3068         * builtins.c (check_sizes): Rename...
3069         (check_access): ...to this.  Rename function arguments for clarity.
3070         (check_memop_sizes): Adjust names.
3071         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3072         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3073         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3074         (check_strncat_sizes, expand_builtin_strncat): Same.
3075         (expand_builtin_strncpy, expand_builtin_memset): Same.
3076         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3077         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3078         (maybe_emit_sprintf_chk_warning): Same.
3079         (expand_builtin_strcpy): Adjust.
3080         (expand_builtin_stpcpy): Same.
3081         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3082         in pointer-checking forms of memcpy, memmove, and mempcpy.
3083         (gcall_to_tree_minimal, max_object_size): Define new functions.
3084         * builtins.h (max_object_size): Declare.
3085         * calls.c (alloc_max_size): Call max_object_size instead of
3086         hardcoding ssizetype limit.
3087         (get_size_range): Handle new argument.
3088         * calls.h (get_size_range): Add a new argument.
3089         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3090         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3091         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3092         operations.
3093         (gimple_fold_builtin_memory_chk): Same.
3094         (gimple_fold_builtin_stxcpy_chk): New function.
3095         * gimple-ssa-warn-restrict.c: New source.
3096         * gimple-ssa-warn-restrict.h: New header.
3097         * gimple.c (gimple_build_call_from_tree): Propagate location.
3098         * passes.def (pass_warn_restrict): Add new pass.
3099         * tree-pass.h (make_pass_warn_restrict): Declare.
3100         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3101         operations.
3102         (handle_builtin_strcat): Same.
3103         (strlen_optimize_stmt): Rename...
3104         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3105         stpncpy, strncpy, and their checking forms.
3107 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3109         PR rtl-optimization/82849
3110         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3111         and get_max_loop_iterations_int.
3113 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3114             Alan Hayward  <alan.hayward@arm.com>
3115             David Sherwood  <david.sherwood@arm.com>
3117         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3118         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3119         size as a poly_uint64.
3120         (mode_for_vector, mode_for_int_vector): Take the number of vector
3121         elements as a poly_uint64.
3122         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3123         size as a poly_uint64.
3124         (mode_for_vector, mode_for_int_vector): Take the number of vector
3125         elements as a poly_uint64.
3127 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3128             Alan Hayward  <alan.hayward@arm.com>
3129             David Sherwood  <david.sherwood@arm.com>
3131         * machmode.h (MACRO_MODE): New macro.
3132         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3133         * targhooks.c (default_libcall_value, default_secondary_reload)
3134         (default_memory_move_cost, default_register_move_cost)
3135         (default_class_max_nregs): Likewise.
3137 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3138             Alan Hayward  <alan.hayward@arm.com>
3139             David Sherwood  <david.sherwood@arm.com>
3141         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3142         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3143         target C file.
3144         * genautomata.c (main): Likewise.
3145         * genconditions.c (write_header): Likewise.
3146         * genemit.c (main): Likewise.
3147         * genextract.c (print_header): Likewise.
3148         * genopinit.c (main): Likewise.
3149         * genoutput.c (output_prologue): Likewise.
3150         * genpeep.c (main): Likewise.
3151         * genpreds.c (write_insn_preds_c): Likewise.
3152         * genrecog.c (writer_header): Likewise.
3153         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3154         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3155         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3156         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3157         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3158         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3159         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3160         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3161         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3162         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3163         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3164         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3165         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3166         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3167         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3168         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3169         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3170         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3171         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3172         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3173         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3174         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3175         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3176         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3177         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3178         * config/darwin.c (IN_TARGET_CODE): Likewise.
3179         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3180         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3181         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3182         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3183         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3184         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3185         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3186         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3187         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3188         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3189         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3190         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3191         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3192         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3193         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3194         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3195         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3196         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3197         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3198         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3199         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3200         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3201         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3202         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3203         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3204         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3205         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3206         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3207         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3208         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3209         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3210         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3211         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3212         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3213         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3214         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3215         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3216         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3217         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3218         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3219         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3220         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3221         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3222         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3223         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3224         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3225         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3226         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3227         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3228         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3229         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3230         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3231         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3232         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3233         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3234         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3235         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3236         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3237         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3238         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3239         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3240         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3241         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3242         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3243         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3244         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3245         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3246         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3247         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3248         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3249         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3250         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3251         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3252         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3253         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3254         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3255         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3256         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3257         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3258         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3259         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3260         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3261         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3262         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3263         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3264         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3265         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3266         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3267         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3268         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3269         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3270         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3271         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3272         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3273         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3274         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3275         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3276         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3277         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3278         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3279         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3280         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3281         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3282         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3283         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3284         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3286 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3287             Alan Hayward  <alan.hayward@arm.com>
3288             David Sherwood  <david.sherwood@arm.com>
3290         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3291         checks for MEM_REF.
3293 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3294             Alan Hayward  <alan.hayward@arm.com>
3295             David Sherwood  <david.sherwood@arm.com>
3297         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3298         * doc/md.texi (vec_series@var{m}): Document.
3299         * tree.def (VEC_SERIES_EXPR): New tree code.
3300         * tree.h (build_vec_series): Declare.
3301         * tree.c (build_vec_series): New function.
3302         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3303         * tree-pretty-print.c (dump_generic_node): Likewise.
3304         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3305         * tree-inline.c (estimate_operator_cost): Likewise.
3306         * expr.c (expand_expr_real_2): Likewise.
3307         * optabs-tree.c (optab_for_tree_code): Likewise.
3308         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3309         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3310         * expmed.c (make_tree): Handle VEC_SERIES.
3311         * optabs.def (vec_series_optab): New optab.
3312         * optabs.h (expand_vec_series_expr): Declare.
3313         * optabs.c (expand_vec_series_expr): New function.
3314         * tree-vect-generic.c (expand_vector_operations_1): Check that
3315         the operands also have vector type.
3317 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3318             Alan Hayward  <alan.hawyard@arm.com>
3319             David Sherwood  <david.sherwood@arm.com>
3321         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3322         (VEC_COND_EXPR): Add missing @tindex.
3323         * doc/md.texi (vec_duplicate@var{m}): Document.
3324         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3325         * tree.c (build_vector_from_val): Add stubbed-out handling of
3326         variable-length vectors, using VEC_DUPLICATE_EXPR.
3327         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3328         * cfgexpand.c (expand_debug_expr): Likewise.
3329         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3330         * tree-inline.c (estimate_operator_cost): Likewise.
3331         * tree-pretty-print.c (dump_generic_node): Likewise.
3332         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3333         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3334         (test_vec_duplicate_folding): New function.
3335         (fold_const_c_tests): Call it.
3336         * optabs.def (vec_duplicate_optab): New optab.
3337         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3338         * optabs.h (expand_vector_broadcast): Declare.
3339         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3340         vec_duplicate_optab.
3341         * expr.c (store_constructor): Try using vec_duplicate_optab for
3342         uniform vectors.
3343         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3345 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3347         PR target/83358
3348         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3349         div/mod latencies a bit.
3351 2017-12-15  Jeff Law  <law@redhat.com>
3353         PR tree-optimization/36550
3354         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3355         (mark_threaded_blocks): Rewrite code to avoid block copying when
3356         optimizing for size.  Don't pessimize blocks which will be
3357         copied, but all the statements will be dead.
3359 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3361         PR tree-optimization/81165
3362         * tree-ssa-threadupdate.c (uses_in_bb): New.
3363         (estimate_threading_killed_stmts): New.
3364         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3365         * tree-ssa-threadedge.c
3366         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3367         when its hit.
3369 2017-12-15  Jeff Law  <law@redhat.com>
3371         PR tree-optimization/83410
3372         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3373         threads when parallelizing loops.
3375 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3377         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3378         handler fields.
3379         * config/alpha/alpha.c (vms_attribute_table): Swap
3380         affects_type_identity and handler fields, adjust comments.
3381         * config/mips/mips.c (mips_attribute_table): Likewise.
3382         * config/visium/visium.c (visium_attribute_table): Likewise.
3383         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3384         * config/microblaze/microblaze.c (microblaze_attribute_table):
3385         Likewise.
3386         * config/spu/spu.c (spu_attribute_table): Likewise.
3387         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3388         * config/arc/arc.c (arc_attribute_table): Likewise.
3389         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3390         * config/v850/v850.c (v850_handle_interrupt_attribute,
3391         v850_handle_data_area_attribute): Formatting fixes.
3392         (v850_attribute_table): Swap affects_type_identity and handler
3393         fields, adjust comments.
3394         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3395         * config/arm/arm.c (arm_attribute_table): Likewise.
3396         * config/avr/avr.c (avr_attribute_table): Likewise.
3397         * config/s390/s390.c (s390_attribute_table): Likewise.
3398         * config/sh/sh.c (sh_attribute_table): Likewise.
3399         * config/i386/i386.c (ix86_handle_cconv_attribute,
3400         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3401         (ix86_attribute_table): Swap affects_type_identity and handler
3402         fields, adjust comments.
3403         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3404         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3405         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3406         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3407         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3408         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3409         * config/rx/rx.c (rx_attribute_table): Likewise.
3410         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3411         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3412         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3413         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3414         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3415         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3416         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3417         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3418         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3419         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3420         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3421         exclude.
3422         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3423         handler fields.
3424         (register_scoped_attributes, decl_attributes): Formatting fixes.
3426         PR tree-optimization/83269
3427         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3428         subtraction in arg0's type if type is signed and arg0 is unsigned.
3429         Formatting fix.
3431         PR sanitizer/81281
3432         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3433         convert? on @0 instead of convert.  Check type of @1, not @0.
3434         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3435         convert? on @0 instead of convert.  Check type of @1, not @0.
3436         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3437         only optimize if either both @1 and @2 types are narrower
3438         precision, or both are wider or equal precision, and in the former
3439         case only if both have undefined overflow.
3441 2017-12-15  Richard Biener  <rguenther@suse.de>
3443         PR lto/83388
3444         * internal-fn.def (IFN_NOP): Add.
3445         * internal-fn.c (expand_NOP): Do nothing.
3446         * lto-streamer-in.c (input_function): Instead of removing
3447         sanitizer calls replace them with IFN_NOP calls.
3449 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3450             Alan Hayward  <alan.hayward@arm.com>
3451             David Sherwood  <david.sherwood@arm.com>
3453         * dse.c (store_info, read_info_type): Replace begin and end with
3454         offset and width.
3455         (print_range): New function.
3456         (set_all_positions_unneeded, any_positions_needed_p)
3457         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3458         accordingly.
3459         (record_store): Likewise.  Optimize the case in which all positions
3460         are unneeded.
3461         (get_stored_val): Replace read_begin and read_end with read_offset
3462         and read_width.
3463         (replace_read): Update call accordingly.
3465 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3467         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3468         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3469         (loop_cand::loop_cand): Initialize above members.
3470         (loop_cand::supported_operations): Delete.
3471         (loop_cand::can_interchange_p): Inline above function.
3472         (loop_cand::classify_simple_reduction): Record number of constant
3473         initialized simple reductions.
3474         (should_interchange_loops): New parameters.  Check stmt cost of loops
3475         to be interchange.
3476         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3477         Update call to should_interchange_loops.
3478         (should_interchange_loop_nest): Update call to
3479         should_interchange_loops.
3481 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3483         PR target/66488
3484         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3485         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3486         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3488 2017-12-15  Julia Koval  <julia.koval@intel.com>
3490         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3491         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3492         New.
3493         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3494         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3495         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3497 2017-12-15  Julia Koval  <julia.koval@intel.com>
3499         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3500         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3501         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3502         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3503         _mm_aesenc_epi128): New intrinsics.
3505 2017-12-15  Julia Koval  <julia.koval@intel.com>
3507         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3508         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3509         New.
3510         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3511         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3512         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3514 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3516         * gimple-ssa-strength-reduction.c (analyze_increments):
3517         Distinguish replacement costs for constant strides from those for
3518         unknown strides.
3520 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3522         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3524 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3526         * doc/invoke.texi: Document -Wcast-function-type.
3527         * recog.h (stored_funcptr): Change signature.
3528         * tree-dump.c (dump_node): Avoid warning.
3529         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3531 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3533         PR middle_end/79538
3534         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3535         array.
3537 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3539         PR tree-optimization/83312
3540         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3541         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3542         "val" param, and to cope with arbitrary basic blocks.
3543         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3544         handle NULL_TREE for "val", dropping "bb" param.
3545         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3546         drop "bb" param.  Handle NULL_TREE for "val".
3547         (find_case_label_for_value): Make "switch_stmt" param const.
3548         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3549         of dom_walker.
3550         (vrp_prop::check_all_array_refs): Reimplement as...
3551         (check_array_bounds_dom_walker::before_dom_children): ...this new
3552         vfunc.  Replace linear search through BB block list, excluding
3553         those with non-executable in-edges via dominator walk.
3555 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3557         * config/arm/arm.opt (mverbose-cost-dump): New option.
3558         * config/arm/arm.c (arm_rtx_costs): Use it.
3560 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3562         PR bootstrap/83396
3563         * reload1.c (emit_input_reload_insns): Skip debug markers.
3565 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3567         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3568         for its nop_endbr.
3570         PR bootstrap/83396
3571         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3572         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3573         * config/sh/sh.c: Skip debug insns besides notes.
3574         * config/sh/sh.md: Likewise.
3575         * config/sh/sh_treg_combine.cc: Likewise.
3576         * config/sh/sync.md: Likewise.
3578 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3580         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3581         Add item for weak_undefined.
3583 2017-12-14  Richard Biener  <rguenther@suse.de>
3585         PR tree-optimization/67842
3586         * sese.h (bb_in_region): Remove #if 0'ed code.
3588 2017-12-14  Richard Biener  <rguenther@suse.de>
3590         PR tree-optimization/83326
3591         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3592         may_be_zero parameter and handle it by not marking the first
3593         peeled copy as not exiting the loop.
3594         (try_peel_loop): Likewise.
3595         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3596         to handle the case of constant or zero iterations and perform
3597         loop header copying on-the-fly.
3599 2017-12-14  Richard Biener  <rguenther@suse.de>
3601         PR tree-optimization/83418
3602         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3603         Instead of asserting we don't get unfolded comparisons deal with them.
3605 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3607         PR bootstrap/83396
3608         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3609         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3610         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3611         bb.  Simplify.
3613         PR tree-optimization/83198
3614         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3615         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3616         value if arg is a REAL_CST with incompatible type.
3618 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3619             Bin Cheng  <bin.cheng@arm.com>
3621         PR target/81228
3622         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3623         to CCFPEmode.
3624         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3625         LTGT.
3627 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3629         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3630         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3632 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3633             Alan Hayward  <alan.hayward@arm.com>
3634             David Sherwood  <david.sherwood@arm.com>
3636         * poly-int.h: New file.
3637         * poly-int-types.h: Likewise.
3638         * coretypes.h: Include them.
3639         (POLY_INT_CONVERSION): Define.
3640         * target.def (estimated_poly_value): New hook.
3641         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3642         * doc/tm.texi: Regenerate.
3643         * doc/poly-int.texi: New file.
3644         * doc/gccint.texi: Include it.
3645         * doc/rtl.texi: Describe restrictions on subreg modes.
3646         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3647         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3648         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3649         * targhooks.h (default_estimated_poly_value): Declare.
3650         * targhooks.c (default_estimated_poly_value): New function.
3651         * target.h (estimated_poly_value): Likewise.
3652         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3653         (wi::unary_traits): Delete.
3654         (wi::binary_traits::signed_shift_result_type): Define for
3655         offset_int << HOST_WIDE_INT, etc.
3656         (generic_wide_int::operator <<=): Define for all types and use
3657         wi::lshift instead of <<.
3658         (wi::hwi_with_prec): Add a default constructor.
3659         (wi::ints_for): New class.
3660         (operator <<): Define for all wide-int types.
3661         (operator /): New function.
3662         (operator %): Likewise.
3663         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3664         (ASSERT_MAYBE_NE_AT): New macros.
3666 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3667             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3669         PR middle-end/78468
3670         * emit-rtl.c (init_emit): Remove ??? comment.
3671         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3672         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3673         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3674         alignment of 3 virtual registers to BITS_PER_WORD.
3676         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3678 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3680         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3681         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3683 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3685         PR bootstrap/83396
3686         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3688 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3690         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3691         with the exception of debug begin stmt markers appear before
3692         labels.
3694         PR bootstrap/83396
3695         * final.c (rest_of_handle_final): Call variable_tracking_main only
3696         if !flag_var_tracking.
3698 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3699             Jakub Jelinek  <jakub@redhat.com>
3701         PR bootstrap/83396
3702         PR debug/83391
3703         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3704         labels debug stmts that can only appear after labels.
3706 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3708         PR rtl-optimization/82398
3709         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3710         EXPR_USEFULNESS in priority comparison.
3712 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3714         PR rtl-optimization/83393
3715         * combine.c (move_deaths): If reg_stat points to a too new insn in
3716         last_death, do not use it: find the proper insn instead.
3718 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3720         PR target/82190
3721         * config/rs6000/rs6000-string.c (expand_block_compare,
3722         expand_strn_compare): Fix set_mem_size() calls.
3724 2017-12-12  Jeff Law  <law@redhat.com>
3726         PR tree-optimization/83298
3727         PR tree-optimization/83362
3728         PR tree-optimization/83383
3729         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3730         push_value_range a public interface.  Add new argument to
3731         record_ranges_from_stmt.
3732         * gimple-ssa-evrp-analyze.c
3733         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3734         Update comments.  Handle recording temporary equivalences.
3735         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3736         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3737         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3738         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3739         gimple-ssa-evrp-analyze.h.
3740         (record_temporary_equivalences_from_phis): Add new argument.  When
3741         the PHI arg is an SSA_NAME, set the result's range to the range
3742         of the PHI arg.
3743         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3744         from statements too.
3745         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3746         Pass it down to children as needed.
3747         (thread_outgoing_edges): Likewise.
3748         (thread_across_edge): Likewise.   Push/pop range state as needed.
3749         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3751 2017-12-12  Julia Koval  <julia.koval@intel.com>
3753         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3754         (PTA_CANNONLAKE): Remove PTA_CLWB.
3756 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3758         PR target/83332
3759         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3760         (vcondv2div2df): Likewise.
3761         (vconduv2dfv2di): Likewise.
3763 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3765         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3766         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3767         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3768         ATTR_MATHFN_FPROUNDING_ERRNO.
3770 2017-12-12  Richard Biener  <rguenther@suse.de>
3772         PR tree-optimization/83385
3773         * graphite-scop-detection.c (get_order, order): Remove.
3774         (bb_to_rpo): New global.
3775         (cmp_pbbs): Adjust.
3776         (build_scops): Sort pbbs in RPO order.
3778 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3780         * combine.c (simplify_set): Do not transform subregs to zero_extends
3781         if the destination mode is a vector mode.
3783 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3785         PR tree-optimization/83359
3786         * tree-cfg.h (fold_loop_internal_call): Declare.
3787         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3788         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3789         (find_loop_dist_alias): New function.
3790         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3791         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3792         calls.
3794         PR tree-optimization/80631
3795         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3796         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3797         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3798         hardcoding zero as the value if COND_EXPR is never true.  For
3799         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3800         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3801         hardcoding MAX_EXPR as the reduction operation.
3802         (is_nonwrapping_integer_induction): Allow negative step.
3803         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3804         vect_create_epilog_for_reduction, if no value is suitable, don't
3805         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3807 2017-12-12  Richard Biener  <rguenther@suse.de>
3809         PR tree-optimization/81889
3810         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3811         range info from the non-wrapping IV instead of just the range
3812         of the type.
3814 2017-12-12  Julia Koval  <julia.koval@intel.com>
3816         * config.gcc: Add vaesintrin.h.
3817         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3818         New type.
3819         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3820         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3821         New builtins.
3822         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3823         * config/i386/immintrin.h: Include vaesintrin.h.
3824         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3825         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3826         _mm_aesdec_epi128): New intrinsics.
3828 2017-12-12  Julia Koval  <julia.koval@intel.com>
3830         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3831         OPTION_MASK_ISA_VAES_UNSET): New.
3832         (ix86_handle_option): Handle -mvaes.
3833         * config/i386/cpuid.h: Define bit_VAES.
3834         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3835         * config/i386/i386-c.c (__VAES__): New.
3836         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3837         (ix86_valid_target_attribute_inner_p): Ditto.
3838         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3839         * config/i386/i386.opt: Add -mvaes.
3840         * doc/invoke.texi: Ditto.
3842 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3844         * debug.h (gcc_debug_hooks): Add inline_entry.
3845         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3846         * debug.c (do_nothing_debug_hooks): Likewise.
3847         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3848         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3849         (dwarf2_lineno_debug_hooks): Likewise.
3851         * common.opt (gstatement-frontiers): New, setting
3852         debug_nonbind_markers_p.
3853         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3854         * toplev.c (process_options): Autodetect value for debug statement
3855         frontiers option.
3856         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3857         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3859         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3860         markers.  Integrate source bind into debug stmt expand loop.
3861         (pass_expand::execute): Check debug marker limit.  Avoid deep
3862         TER and expand debug locations for debug bind insns only.
3863         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3864         followed by them.
3865         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3866         * final.c (reemit_insn_block_notes): Take current block from
3867         nonbind markers.  Declare note where it's first set.
3868         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3869         begin stmt markers if enabled.
3870         (notice_source_line): Handle nonbind markers.  Fail if their
3871         location is unknown or that of builtins.
3872         (rest_of_handle_final): Convert begin stmt markers to notes if
3873         var-tracking didn't run.
3874         (rest_of_clean_state): Skip begin stmt markers.
3875         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3876         markers.
3877         * function.c (allocate_struct_function): Set begin_stmt_markers.
3878         * function.h (struct function): Add debug_marker_count counter
3879         and debug_nonbind_markers flag.
3880         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3881         * gimple-low.c (lower_function_body): Adjust
3882         debug_nonbind_markers.
3883         (lower_stmt): Drop or skip gimple debug stmts.
3884         (lower_try_catch): Skip debug stmts.
3885         * gimple.c (gimple_build_debug_begin_stmt): New.
3886         (gimple_copy): Increment debug_marker_count if copying one.
3887         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3888         * gimplify.c (rexpr_location): New.
3889         (rexpr_has_location): New.
3890         (warn_switch_unreachable_r): Handle gimple debug stmts.
3891         (shortcut_cond_r): Call expr_location.
3892         (find_goto): New.
3893         (find_goto_label): New.
3894         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3895         find_goto_label.
3896         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3897         expr_location.
3898         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3899         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3900         (LANG_HOOKS_INITIALIZER): ... this.
3901         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3902         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3903         debug insns.
3904         (update_ebb_live_info): Skip debug insn markers.
3905         * lra.c (debug_insn_static_data): Rename to...
3906         (debug_bind_static_data): ... this.
3907         (debug_marker_static_data): New.
3908         (lra_set_insn_recog_data): Select one of the above depending
3909         on debug insn kind.
3910         (lra_update_isn_regno_info): Don't assume debug insns have
3911         freqs.
3912         (push_insns): Skip debug insns.
3913         * lto-streamer-in.c (input_function): Drop debug stmts
3914         depending on active options.  Adjust debug_nonbind_markers.
3915         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3916         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3917         begin stmt marker notes.
3918         (print_insn): Likewise.
3919         * recog.c (extract_insn): Recognize rtl for debug markers.
3920         * rtl.def (DEBUG_MARKER): New.
3921         * tree-inline.c: Include params.h.
3922         (remap_gimple_stmt): Handle nonbind markers.
3923         (maybe_move_debug_stmts_to_successors): Likewise.
3924         (copy_debug_stmt): Likewise.
3925         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3926         markers regardless of no side effects.
3927         (tsi_link_before): Don't update container's side effects when adding
3928         a begin stmt marker.
3929         (tsi_link_after): Likewise.
3930         (expr_first): Skip begin stmt markers.
3931         (expr_last): Likewise.
3932         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3933         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3934         Disregard nonbind markers.
3935         * tree.c (make_node_stat): Don't set side effects for begin stmt
3936         markers.
3937         (build1_stat): Likewise.
3938         * tree.def (DEBUG_BEGIN_STMT): New.
3939         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3940         * var-tracking.c (delete_debug_insns): Renamed to...
3941         (delete_vta_debug_insns): ... this.
3942         (reemit_marker_as_note): New.
3943         (vt_initialize): Reemit markers.
3944         (delete_vta_debug_insns): Likewise.
3945         (vt_debug_insns_local): Reemit or delete markers.
3946         (variable_tracking_main_1): Likewise.
3947         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
3948         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
3949         (gimple_debug_nonbind_marker_p): New.
3950         (gimple_build_debug_bind): Adjust.
3951         (gimple_build_debug_begin_stmt): New.
3952         * doc/invoke.texi (max-debug-marker-count): New param.
3953         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
3954         (debug_parameter_ref, debug_marker): New.
3955         (NOTE_INSN_BEGIN_STMT): New.
3956         (DEBUG_INSN): Describe begin stmt markers.
3958         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
3959         without debug insns, we wouldn't, but clean up debug insns
3960         after a control flow insn nevertheless.
3962         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
3963         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
3964         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
3965         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
3966         (rtl_tidy_fallthru_edge): Likewise.
3967         (rtl_verify_fallthru): Likewise.
3968         (rtl_verify_bb_layout): Likewise.
3969         (skip_insns_after_block): Likewise.
3970         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
3971         * dwarf2out.c: Include print-rtl.h.
3972         (dwarf2out_next_real_insn): New.
3973         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
3974         Dump debug binds in asm comments.
3975         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
3976         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
3977         callers to use gsi_start_nondebug_bb instead.
3978         (gsi_after_labels): Skip gimple debug stmts.
3979         (gsi_start_nondebug): New.
3980         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
3981         (proper_loop_form_for_interchange): Adjust.
3982         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
3983         * gimple.h (gimple_seq_last_nondebug_stmt): New.
3984         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
3985         (collect_fallthrough_labels): Likewise.
3986         (should_warn_for_implicit_fallthrough): Likewise.
3987         (warn_implicit_fallthrough_r): Likewise.
3988         (expand_FALLTHROUGH_r): Likewise.
3989         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
3990         (graphite_copy_stmts_from_block): Skip nonbind markers.
3991         * haifa-sched.c (sched_extend_bb): Skip debug insns.
3992         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
3993         * jump.c (clean_barriers): Skip debug insns.
3994         * omp-expand.c (expand_parallel_call): Skip debug insns.
3995         (expand_task_call): Likewise.
3996         (remove_exit_barrier): Likewise.
3997         (expand_omp_taskreg): Likewise.
3998         (expand_omp_for_init_counts): Likewise.
3999         (expand_omp_for_generic): Likewise.
4000         (expand_omp_for_static_nochunk): Likewise.
4001         (expand_omp_for_static_chunk): Likewise.
4002         (expand_omp_simd): Likewise.
4003         (expand_omp_taskloop_for_outer): Likewise.
4004         (expand_omp_taskloop_for_inner): Likewise.
4005         (expand_oacc_for): Likewise.
4006         (expand_omp_sections): Likewise.
4007         (expand_omp_single): Likewise.
4008         (expand_omp_synch): Likewise.
4009         (expand_omp_atomic_load): Likewise.
4010         (expand_omp_atomic_store): Likewise.
4011         (expand_omp_atomic_fetch_op): Likewise.
4012         (expand_omp_atomic_pipeline): Likewise.
4013         (expand_omp_atomic_mutex): Likewise.
4014         (expand_omp_target): Likewise.
4015         (grid_expand_omp_for_loop): Likewise.
4016         (grid_expand_target_grid_body): Likewise.
4017         (build_omp_regions_1): Likewise.
4018         * omp-low.c (check_combined_parallel): Skip debug stmts.
4019         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
4020         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
4021         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
4022         test.
4023         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4024         (make_edges): Likewise.
4025         (cleanup_dead_labels): Likewise.
4026         (gimple_can_merge_blocks_p): Likewise.
4027         (stmt_starts_bb_p): Likewise.
4028         (gimple_block_label): Likewise.
4029         (gimple_redirect_edge_and_branch): Likewise.
4030         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4031         of debug stmts.
4032         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4033         TDF_SLIM.
4034         * tree-pretty-print (print_declaration): Omit initializer in slim
4035         dumps.
4036         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4037         markers.
4038         (eliminate_unnecessary_stmts): Stabilize block removal order.
4039         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4040         * var-tracking.c (get_first_insn): New.
4041         (vt_emit_notes): Call it.
4042         (vt_initialize): Walk any insns before the first BB.
4043         (delete_debug_insns): Likewise.
4045         * gimple.h (enum gimple_debug_subcode): Add
4046         GIMPLE_DEBUG_BEGIN_STMT.
4047         (gimple_debug_begin_stmt_p): New.
4048         (gimple_debug_nonbind_marker_p): New.
4049         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4050         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4051         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4052         * insn-notes.def (BEGIN_STMT): New.
4053         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4054         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4055         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4056         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4057         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4058         (INSN_DEBUG_MARKER_KIND): New.
4059         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4060         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4061         (INSN_VAR_LOCATION_PTR): New.
4062         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4063         only.
4064         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4065         deps only if debug bind insns are enabled.
4066         (pass_expand::execute): Avoid deep TER and expand
4067         debug locations for debug bind insns only.
4068         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4069         debug stmts special handling down to debug bind stmts.
4070         * combine.c (try_combine): Narrow debug insns special handling
4071         down to debug bind insns.
4072         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4073         Narrow debug insns preexisting special handling down to debug
4074         bind insns.
4075         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4076         handling down to debug bind insns.
4077         * function.c (instantiate_virtual_regs): Skip debug markers,
4078         adjust handling of debug binds.
4079         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4080         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4081         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4082         insns to debug bind insns.
4083         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4084         special handling of debug stmts to debug bind stmts.
4085         * ipa-split.c (split_function): Likewise.
4086         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4087         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4088         debug insns.
4089         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4090         * regrename.c (build_def_use): Likewise.
4091         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4092         (pass_cprop_hardreg): Narrow special casing of debug insns to
4093         debug bind insns.
4094         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4095         * reload1.c (reload): Likewise.
4096         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4097         casing of debug stmts to debug bind stmts.
4098         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4099         * ssa-iterators.h (num_imm_uses): Likewise.
4100         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4101         debug stmts to debug bind stmts.
4102         * tree-inline.c (tree_function_versioning): Narrow special casing
4103         of debug stmts to debug bind stmts.
4104         * tree-loop-distribution.c (generate_loops_for_partition):
4105         Narrow special casing of debug stmts to debug bind stmts.
4106         * tree-sra.c (analyze_access_subtree): Narrow special casing
4107         of debug stmts to debug bind stmts.
4108         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4109         stmts to debug bind stmts.
4110         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4111         casing of debug stmts to debug bind stmts.
4112         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4113         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4114         casing of debug stmts to debug bind stmts.
4115         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4116         Likewise.
4117         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4118         debug stmts to debug bind stmts.
4119         (gimple_replace_ssa_lhs): Likewise.
4120         (insert_debug_temp_for_var_def): Likewise.
4121         (insert_debug_temps_for_defs): Likewise.
4122         (reset_debug_uses): Likewise.
4123         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4124         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4125         (adjust_debug_stmts): Likewise.
4126         (adjust_phi_and_debug_stmts): Likewise.
4127         (vect_do_peeling): Likewise.
4128         * tree-vect-loop.c (vect_transform_loop): Likewise.
4129         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4130         * var-tracking.c (adjust_mems): Narrow special casing of debug
4131         insns to debug bind insns.
4132         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4133         (compute_bb_dataflow, vt_find_locations): Likewise.
4134         (vt_expand_loc, emit_notes_for_changes): Likewise.
4135         (vt_init_cfa_base): Likewise.
4136         (vt_emit_notes): Likewise.
4137         (vt_initialize): Likewise.
4138         (vt_finalize): Likewise.
4140         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4141         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4142         (next_nonnote_nondebug_insn_bb): New.
4143         (prev_nonnote_nondebug_insn_bb): New.
4144         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4145         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4146         (prev_nonnote_nondebug_insn_bb): Declare.
4147         (next_nonnote_nondebug_insn_bb): Declare.
4148         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4149         * cfgrtl.c (get_last_bb_insn): Likewise.
4150         * lra.c (push_insns): Likewise.
4152 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4154         PR c/82050
4155         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4156         to...
4157         * selftest.c (selftest::test_runner::test_runner): New ctor.
4158         (selftest::test_runner::~test_runner): New dtor.
4159         * selftest.h (class selftest::test_runner): New class.
4161 2017-12-11  Carl Love  <cel@us.ibm.com>
4163         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4164         vec_extract_fp32_from_shortl]): Add #defines.
4165         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4166         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4167         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4168         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4169         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4170         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4171         * doc/extend.texi: Add documentation for the added builtins.
4173 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4175         PR rtl-optimization/80693
4176         PR rtl-optimization/81019
4177         PR rtl-optimization/81020
4178         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4179         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4180         possibly modified to REG_DEAD, if it did not originate in i3.
4182 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4184         * recog.c (store_data_bypass_p_1): New function.
4185         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4186         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4188 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4190         PR rtl-optimization/83361
4191         * ifcvt.c (if_convert): Call fixup_partitions.
4193 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4195         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4196         early folding of splat_u{8,16,32}.
4198 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4200         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4201         output_operand_lossage first argument with capital letter.
4202         (aarch64_override_options): Don't start error and sorry first argument
4203         with capital letter.
4205 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4207         PR gcov-profile/83355
4208         * auto-profile.c (string_table::get_index_by_decl): Don't
4209         recurse when abstract origin points to itself.
4211 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4213         PR tree-optimization/83320
4214         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4215         (prune_datarefs_not_in_loop): Ditto.
4217 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4219         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4221 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4223         PR tree-optimization/83337
4224         * gimple-loop-interchange.cc (compute_access_stride): Handle
4225         bitfield DRs properly.
4227 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4229         PR tree-optimization/83338
4230         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4231         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4232         vector of unsigned integers to vector of signed integers.
4234 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4236         PR rtl-optimization/83317
4237         * lra-constraints.c (process_address_1): Add insn code check.
4239 2017-12-08  Michael Matz  <matz@suse.de>
4241         Fix PR tree-optimization/83323
4242         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4243         head-controlled loops and loop BBs.
4244         * common.opt (funroll-and-jam): Remove, instead ...
4245         (floop-unroll-and-jam): ... reuse this option.
4246         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4247         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4248         (-floop-unroll-and-jam): ... this option.
4250 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4252         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4253         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4254         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4256 2017-12-08  Richard Biener  <rguenther@suse.de>
4258         PR middle-end/81782
4259         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4260         handle accesses outside of zero-sized vars.
4262 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4264         PR tree-optimization/83141
4265         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4266         test for MEM_REFs implicitely changing types with padding.  Remove
4267         inline keyword.
4268         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4270 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4272         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4273         the comment.
4274         * config/avr/avr.c (avr_attribute_table): Likewise.
4275         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4276         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4277         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4278         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4279         of struct attribute_spec.
4280         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4282 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4284         PR target/82960
4285         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4287 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4289         PR middle-end/83609
4290         * profile-count.c (profile_count::from_gcov_type): Move from
4291         profile-count.h; handle overflow.
4292         * profile-count.h (profile_count::from_gcov_type): Move offline.
4294 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4296         PR rtl-optimization/83304
4297         * combine.c (move_deaths): If we do not know where a register died,
4298         search for it.
4300 2017-12-08  Richard Biener  <rguenther@suse.de>
4302         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4303         Provide -fopt-info-loop feedback when we interchange in a nest.
4305 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4307         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4308         for armv6 ARM CPU IDs.
4310 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4312         * common/config/arm/arm-common.c: Include <algorithm>.
4313         (INCLUDE_VECTOR): Define.
4314         (compare_opt_names): New function.
4315         (arm_rewrite_selected_arch): Only strip out extensions that can be
4316         expressed through -mfpu.  Sort the remaining extensions
4317         alphabetically.
4319 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4321         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4322         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4323         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4324         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4326 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4328         PR ada/81470
4329         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4330         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4332 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4334         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4335         of struct attribute_spec.
4337 2017-12-08  Julia Koval  <julia.koval@intel.com>
4339         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4340         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4341         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4342         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4343         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4344         _mm_maskz_dpwssds_epi32): Ditto.
4346 2017-12-08  Richard Biener  <rguenther@suse.de>
4348         PR tree-optimization/81303
4349         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4350         conditions try to create a comparison vector type matching
4351         the data vector type.
4352         (vectorizable_condition): Adjust.
4353         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4354         Leave invariant conditions alone in case we can vectorize those.
4356 2017-12-08  Julia Koval  <julia.koval@intel.com>
4358         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4359         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4360         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4361         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4362         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4364 2017-12-08  Julia Koval  <julia.koval@intel.com>
4366         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4367         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4368         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4369         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4370         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4371         _mm_maskz_dpbusds_epi32): New intrinsics.
4373 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4375         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4376         operand.
4377         (ld<bh>io_signed): Likewise.
4378         (st<bhw>io): Likewise.
4379         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4380         SMALL_INT12 constant integer operand.
4382 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4384         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4385         Rename to...
4386         (nios2_large_constant_allowed): ...this.  Adjust uses.
4387         (nios2_plus_symbolic_constant_p): Rename to...
4388         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4389         (nios2_legitimate_address_p): Correct CONST_INT handling.
4390         (nios2_symbolic_memory_operand_p): Rename to...
4391         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4392         (nios2_large_constant_p): Check for large constant integers too.
4393         (nios2_split_large_constant): Handle constant integers.
4394         (nios2_split_symbolic_memory_operand): Rename to...
4395         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4396         (nios2_legitimize_constant_address): Handle constant integers.
4397         (r0rel_constant_p): Handle small constant integers.
4398         (nios2_print_operand_address): Handle r0-relative integer addresses.
4399         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4400         * config/nios2/nios2.md: Adjust for renamed functions.
4402 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4404         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4405         (TARGET_HAVE_SRODATA_SECTION): New define.
4406         (riscv_select_section): New function.
4408 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4409             Alexander Monakov  <amonakov@ispras.ru>
4410             Jakub Jelinek  <jakub@redhat.com>
4412         PR target/81906
4413         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4415 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4417         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4418         a bool success value.  Don't call output_operand_lossage here.
4419         (aarch64_print_ldpstp_address): Return a bool success value.
4420         (aarch64_print_operand_address): Call output_addr_const if
4421         aarch64_print_address_internal fails.
4422         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4423         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4424         if aarch64_print_ldpstp_address fails.
4426 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4428         * tree-vector-builder.h
4429         (tree_vector_builder::binary_encoded_nelts): Declare.
4430         * tree-vector-builder.c
4431         (tree_vector_builder::binary_encoded_nelts): New function.
4432         * fold-const.c (negate_expr_p): Likewise.
4433         (operand_equal_p, fold_checksum_tree): Likewise.
4434         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4435         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4436         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4437         (uniform_vector_p): Likewise.
4438         * varasm.c (const_hash_1, compare_constant): Likewise.
4439         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4440         (valid_lattice_transition): Operate directly on the VECTOR_CST
4441         encoding.
4442         * ipa-icf.c: Include tree-vector-builder.h.
4443         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4444         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4446 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4448         * tree.c (build_vector): Delete.
4449         * tree.h (build_vector): Make static and move into the self-testing
4450         block.
4452 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4454         * vector-builder.h (vector_builder::derived): New const overload.
4455         (vector_builder::elt): New function.
4456         * tree-vector-builder.h (tree_vector_builder::type): New function.
4457         (tree_vector_builder::apply_step): Declare.
4458         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4459         function.
4460         * gimple-fold.h (tree_vector_builder): Declare.
4461         (gimple_build_vector): Take a tree_vector_builder instead of a
4462         type and vector of elements.
4463         * gimple-fold.c (gimple_build_vector): Likewise.
4464         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4465         accordingly.
4466         (get_initial_defs_for_reduction): Likewise.
4467         (vectorizable_induction): Likewise.
4469 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4471         * tree-vector-builder.h
4472         (tree_vector_builder::new_binary_operation): Declare.
4473         * tree-vector-builder.c
4474         (tree_vector_builder::new_binary_operation): New function.
4475         * fold-const.c (fold_relational_const): Use it.
4476         (const_binop): Likewise.  Check that both input vectors have
4477         the same number of elements, thus excluding things like WIDEN_SUM.
4478         Check whether it is possible to operate directly on the encodings
4479         of stepped inputs.
4481 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4483         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4484         new_unary_operation, operating only on the encoded elements.
4485         (const_unop): Likewise.
4486         (exact_inverse): Likewise.
4487         (distributes_over_addition_p): New function.
4488         (const_binop): Use tree_vector_builder and new_unary_operation
4489         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4490         on the encoded elements unless the encoding is strided and the
4491         operation does not distribute over addition.
4492         (fold_convert_const):  Use tree_vector_builder and
4493         new_unary_operation.  Operate only on the encoded elements
4494         for truncating integer conversions, or for non-stepped encodings.
4496 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4498         * config/sparc/sparc.c: Include tree-vector-builder.h.
4499         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4500         * expmed.c: Include tree-vector-builder.h.
4501         (make_tree): Use tree_vector_builder instead of build_vector.
4502         * fold-const.c: Include tree-vector-builder.h.
4503         (const_binop): Use tree_vector_builder instead of build_vector.
4504         (const_unop): Likewise.
4505         (native_interpret_vector): Likewise.
4506         (fold_vec_perm): Likewise.
4507         (fold_ternary_loc): Likewise.
4508         * gimple-fold.c: Include tree-vector-builder.h.
4509         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4510         of build_vector.
4511         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4512         (simplify_vector_constructor): Use tree_vector_builder instead
4513         of build_vector.
4514         * tree-vect-generic.c: Include tree-vector-builder.h.
4515         (add_rshift): Use tree_vector_builder instead of build_vector.
4516         (expand_vector_divmod): Likewise.
4517         (optimize_vector_constructor): Likewise.
4518         * tree-vect-loop.c: Include tree-vector-builder.h.
4519         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4520         of build_vector.  Explicitly use a stepped encoding for
4521         { 1, 2, 3, ... }.
4522         * tree-vect-slp.c: Include tree-vector-builder.h.
4523         (vect_get_constant_vectors): Use tree_vector_builder instead
4524         of build_vector.
4525         (vect_transform_slp_perm_load): Likewise.
4526         (vect_schedule_slp_instance): Likewise.
4527         * tree-vect-stmts.c: Include tree-vector-builder.h.
4528         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4529         (vect_gen_perm_mask_any): Likewise.
4530         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4531         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4532         of build_vector.
4533         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4534         encoding.
4536 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4538         * doc/generic.texi (VECTOR_CST): Describe new representation of
4539         vector constants.
4540         * vector-builder.h: New file.
4541         * tree-vector-builder.h: Likewise.
4542         * tree-vector-builder.c: Likewise.
4543         * Makefile.in (OBJS): Add tree-vector-builder.o.
4544         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4545         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4546         (tree_vector): Change the number of elements to
4547         vector_cst_encoded_nelts.
4548         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4549         (VECTOR_CST_ELTS): Delete.
4550         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4551         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4552         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4553         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4554         (VECTOR_CST_ENCODED_ELT): Likewise.
4555         (vector_cst_encoded_nelts): New function.
4556         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4557         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4558         (vector_cst_int_elt, vector_cst_elt): Declare.
4559         * tree.c: Include tree-vector-builder.h.
4560         (tree_code_size): Abort if passed VECTOR_CST.
4561         (tree_size): Update for new VECTOR_CST layout.
4562         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4563         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4564         (build_vector): Use tree_vector_builder.
4565         (vector_cst_int_elt, vector_cst_elt): New functions.
4566         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4567         encoded elements and then create the vector in the canonical form.
4568         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4569         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4570         (tree_c_tests): Call test_vector_cst_patterns.
4571         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4572         VECTOR_CST fields.
4573         (hash_tree): Likewise.
4574         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4575         (streamer_write_tree_header): Likewise.
4576         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4577         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4578         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4580 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4582         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4583         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4584         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4586 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4587             Richard Biener  <rguenther@suse.de>
4589         PR tree-optimization/81303
4590         * Makefile.in (gimple-loop-interchange.o): New object file.
4591         * common.opt (floop-interchange): Reuse the option from graphite.
4592         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4593         -floop-interchange and mention it for -O3.
4594         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4595         * gimple-loop-interchange.cc: New file.
4596         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4597         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4598         * passes.def (pass_linterchange): New pass.
4599         * timevar.def (TV_LINTERCHANGE): New time var.
4600         * tree-pass.h (make_pass_linterchange): New declaration.
4601         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4602         interchange.  Record IV before/after increment in new parameters.
4603         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4604         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4605         path check into...
4606         (check_reduction_path): ...New function here.
4607         * tree-vectorizer.h (check_reduction_path): New declaration.
4609 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4611         PR target/83252
4612         PR rtl-optimization/80818
4613         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4614         always early clobbered.
4615         * lra-lives.c (process_bb_lives): Check input hard regs for early
4616         clobbered non-operand hard reg.
4618 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4620         PR middle-end/83164
4621         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4622         types_compatible_p, just that TYPE_MODE is the same.
4624 2017-12-07  Martin Sebor  <msebor@redhat.com>
4626         PR c/81544
4627         * attribs.c (empty_attribute_table): Initialize new member of
4628         struct attribute_spec.
4629         (decl_attributes): Add argument.  Handle mutually exclusive
4630         combinations of attributes.
4631         (selftests::test_attribute_exclusions): New function.
4632         (selftests::attribute_c_tests): Ditto.
4633         * attribs.h (decl_attributes): Add default argument.
4634         * selftest.h (attribute_c_tests): Declare.
4635         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4636         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4637         member.
4638         * doc/extend.texi (Common Function Attributes): Update const and pure.
4639         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4640         of struct attribute_spec.
4641         * config/arc/arc.c (arc_attribute_table): Same.
4642         * config/arm/arm.c (arm_attribute_table): Same.
4643         * config/avr/avr.c ( avr_attribute_table): Same.
4644         * config/bfin/bfin.c (bfin_attribute_table): Same.
4645         * config/cr16/cr16.c (cr16_attribute_table): Same.
4646         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4647         * config/h8300/h8300.c (h8300_attribute_table): Same.
4648         * config/i386/i386.c (ix86_attribute_table): Same.
4649         * config/ia64/ia64.c (ia64_attribute_table): Same.
4650         * config/m32c/m32c.c (m32c_attribute_table): Same.
4651         * config/m32r/m32r.c (m32r_attribute_table): Same.
4652         * config/m68k/m68k.c (m68k_attribute_table): Same.
4653         * config/mcore/mcore.c (mcore_attribute_table): Same.
4654         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4655         * config/mips/mips.c (mips_attribute_table): Same.
4656         * config/msp430/msp430.c (msp430_attribute_table): Same.
4657         * config/nds32/nds32.c (nds32_attribute_table): Same.
4658         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4659         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4660         * config/rl78/rl78.c (rl78__attribute_table): Same.
4661         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4662         * onfig/rx/rx.c (rx_attribute_table): Same.
4663         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4664         * config/sh/sh.c (sh_attribute_table): Same.
4665         * config/sparc/sparc.c (sparc_attribute_table): Same.
4666         * config/spu/spu.c (spu_attribute_table): Same.
4667         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4668         * config/v850/v850.c (v850_attribute_table): Same.
4669         * config/visium/visium.c (visium_attribute_table): Same.
4671 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4673         PR target/82641
4674         * config/arm/arm.c (INCLUDE_STRING): Define.
4675         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4676         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4677         and .fpu.
4679 2017-12-07  Michael Matz  <matz@suse.de>
4681         Add unroll and jam pass
4683         * gimple-loop-jam.c: New file.
4684         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4685         * common.opt (funroll-and-jam): New option.
4686         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4687         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4688         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4689         * passes.def: Add pass_loop_jam.
4690         * timevar.def (TV_LOOP_JAM): Add.
4691         * tree-pass.h (make_pass_loop_jam): Declare.
4692         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4693         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4694         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4695         to flow_loop_tree_node_add.
4696         (duplicate_subloops, copy_loops_to): Append to sibling list.
4697         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4698         * doc/invoke.texi (-funroll-and-jam): Document new option.
4699         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4701 2017-12-07  Richard Biener  <rguenther@suse.de>
4703         PR tree-optimization/83296
4704         PR tree-optimization/67769
4705         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4706         flow sensitive info in an unrelated BB.
4707         (value_replacement): Use reset_flow_sensitive_info.
4708         (minmax_replacement): Reset flow sensitive info on the def
4709         we move.  Do not reset flow sensitive info in the whole BB
4710         we move the stmt to.
4711         (abs_replacement): Likewise.
4713 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4715         PR target/43871
4716         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4717         rs6000_cpu to the given -mcpu=, or to the default processor.
4719 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4721         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4722         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4723         instead of rs6000_cpu_attr.
4724         (rs6000_adjust_cost): Ditto.
4725         (is_microcoded_insn): Ditto.
4726         (rs6000_adjust_priority): Ditto.
4727         (rs6000_issue_rate): Ditto.
4728         (rs6000_use_sched_lookahead): Ditto.
4729         (rs6000_use_sched_lookahead_guard): Ditto.
4730         (rs6000_sched_reorder): Ditto.
4731         (force_new_group): Ditto.
4732         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4733         (group_ending_nop): Ditto.
4735 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4737         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4738         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4739         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4740         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4741         appropriate.
4742         (rs6000_reassociation_width): Ditto.
4743         (rs6000_emit_epilogue): Ditto.
4744         (rs6000_adjust_cost): Ditto.
4745         (is_microcoded_insn): Ditto.
4746         (is_cracked_insn): Ditto.
4747         (rs6000_adjust_priority): Ditto.
4748         (rs6000_sched_reorder): Ditto.
4749         (rs6000_sched_reorder2): Ditto.
4750         (insn_must_be_first_in_group): Ditto.
4751         (insn_must_be_last_in_group): Ditto.
4752         (rs6000_register_move_cost): Ditto.
4753         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4754         rs6000_cpu.
4756 2017-12-07  Julia Koval  <julia.koval@intel.com>
4758         * config.gcc: Add vaesintrin.h.
4759         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4760         New type.
4761         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4762         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4763         New builtins.
4764         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4765         * config/i386/immintrin.h: Include vaesintrin.h.
4766         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4767         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4768         _mm_aesdec_epi128): New intrinsics.
4770 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4772         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4773         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4774         c-family/c-spellcheck.cc.
4775         (best_macro_match::best_macro_match): Likewise.
4776         * spellcheck-tree.h
4777         (struct edit_distance_traits<cpp_hashnode *>): Move to
4778         c-family/c-spellcheck.h.
4779         (class best_macro_match): Likewise.
4781 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4783         PR tree-optimization/83293
4784         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4785         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4786         might insert into empty bb.
4788         PR sanitizer/81281
4789         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4790         simplify for plus with :c added, and pointer_plus without that.
4791         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4792         with undefined overflow and the conversion is not widening,
4793         perform negation in utype and only convert to type afterwards.
4794         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4795         simplify for plus with :c added, and pointer_plus without that.
4796         If type is integral with undefined overflow and the conversion is
4797         not widening, perform minus in utype and only convert to type
4798         afterwards.  Move the last pointer_diff_expr simplify into the
4799         two outermost ifs.
4801 2017-12-06  Martin Sebor  <msebor@redhat.com>
4803         PR tree-optimization/82646
4804         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4805         strncpy, not maxlen.
4807 2017-12-06  Martin Sebor  <msebor@redhat.com>
4809         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4810         nonstring.
4812         PR tree-optimization/83075
4813         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4814         strncat/strncpy don't change length of source string.
4816 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4818         Revert
4819         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4821         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4822         to prevent b2bst errata sequence.
4823         (sqrtdf2_fix): Likewise.
4825 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4827         PR tree-optimization/81945
4828         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4829         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4830         to dest_cfun has orig_loop_num set, either remap it to the new
4831         loop number if the loop got moved too, or clear it.
4833 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4835         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4836         to reservation.
4837         (thunderx2t99_nothing): New insn reservation.
4838         (thunderx2t99_mrs): New insn reservation.
4839         (thunderx2t99_multiple): New insn reservation.
4840         (thunderx2t99_alu_basi): Add bfx to reservation.
4841         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4843 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4845         PR target/82248
4846         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4848 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4850         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4851         tree-ssa-pre.c::remove_dead_inserted_code.
4852         * tree-ssa-dce.h: New file.
4853         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4854         (remove_dead_inserted_code): Move and rename to function
4855         tree-ssa-dce.c::simple_dce_from_worklist.
4856         (pass_pre::execute): Update use.
4858 2017-12-05  Richard Biener  <rguenther@suse.de>
4860         PR tree-optimization/83277
4861         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4862         to code-gen liveout vars.
4864 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4866         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4867         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4868         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4869         split condition.
4871 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4873         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4874         function.
4875         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4876         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4877         ASAN is enabled.
4879 2017-12-05  Richard Biener  <rguenther@suse.de>
4881         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4882         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4883         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4884         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4886 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4888         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4890 2017-12-05  Martin Liska  <mliska@suse.cz>
4891             Jakub Jelinek  <jakub@redhat.com>
4893         * doc/invoke.texi: Document the options.
4894         * flag-types.h (enum sanitize_code): Add
4895         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4896         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4897         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4898         * opts.c: Define new sanitizer options.
4899         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4900         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4902 2017-12-05  Julia Koval  <julia.koval@intel.com>
4904         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4905         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4906         (ix86_handle_option): Handle -mavx512vnni.
4907         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4908         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4909         * config/i386/i386-c (__AVX512VNNI__): New.
4910         * config/i386/i386.c (ix86_target_string): Handle new option.
4911         (ix86_valid_target_attribute_inner_p): Handle new option.
4912         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4913         * config/i386/i386.opt (mavx512vnni): New option.
4915 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4917         PR target/81616
4918         * athlon.md: Disable for generic.
4919         * haswell.md: Enable for generic.
4920         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4921         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4922         to 4.
4923         (ix86_adjust_cost): Move generic to haswell path.
4925 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
4927         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4928         instead of MEM_P in a couple more places.  Fix formatting issues.
4930 2017-12-04  Jim Wilson  <jimw@sifive.com>
4932         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4933         instead of GP_REG_LAST-1.
4934         (riscv_adjust_libcall_cfi_prologue): Likewise.
4935         (riscv_adjust_libcall_cri_epilogue): Likewise.
4936         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4937         comment.
4939 2017-12-04  Luis Machado  <luis.machado@linaro.org>
4941         * ipa-pure-const.c (check_decl): Add missing newline.
4942         (state_from_flags): Likewise.
4944 2017-12-04  Jeff Law  <law@redhat.com>
4946         PR tree-optimizatin/78496
4947         * gimple-ssa-evrp-analyze.h
4948         (evrp_range_analyzer::get_vr_values): Simplify.
4949         * gimple-ssa-evrp-analyze.c: Corresponding changes.
4950         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
4951         and gimple-ssa-evrp-analyze.h.
4952         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
4953         (simplify_stmt_for_jump_threading): Copy a blob of code from
4954         tree-vrp.c to use ranges to simplify statements.
4955         (dom_opt_dom_walker::before_dom_children): Call
4956         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
4957         (dom_opt_dom_walker::after_dom_children): Similarly for
4958         evrp_range_analyzer::leave.
4959         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
4960         conditionals.
4962         * gimple-ssa-evrp-analyze.c
4963         (evrp_range_analyzer::extract_range_from_stmt):  Always use
4964         vr_values::update_value_range so preexisting range info is
4965         medged with new range info, even if the new range is VR_VARYING.
4967 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4969         * combine.c: Adjust comment.
4970         (use_crosses_set_p): Delete.
4971         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
4972         (try_combine): Ditto.
4974 2017-12-04  Richard Biener  <rguenther@suse.de>
4976         PR tree-optimization/83255
4977         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
4978         Re-add zero-iteration check.
4980 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4982         PR rtl-optimization/83245
4983         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
4984         hard registers as earlyclobber, also if not in an asm.
4986 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4988         PR bootstrap/83265
4989         Revert
4990         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4992         PR target/43871
4993         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4994         rs6000_cpu based on cpu_index, not tune_index.
4996 2017-12-04  Richard Biener  <rguenther@suse.de>
4998         PR tree-optimization/83238
4999         * graphite-scop-detection.c (scop_detection::merge_sese): Make
5000         code match comment, rejecting invalid SESE regions.
5002 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
5004         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
5005         require base operand is a REG_POINTER prior to reload on targets
5006         with non-equivalent space registers.
5008 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5010         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
5011         (update_specialized_profile): Fix updating of counts.
5012         (perhaps_add_new_callers): Likewise.
5014 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5016         PR target/81616
5017         * x86-tune.def: Remove obsolette FIXMEs.
5018         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
5019         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
5020         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
5021         Enable for generic.
5022         (X86_TUNE_PAD_RETURNS): Disable for generic.
5024 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5026         PR tree-optimization/83170
5027         PR tree-optimization/83241
5028         * gimple-ssa-store-merging.c
5029         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5030         gimple_vuse (ins_stmt) in case it has changed.
5031         (imm_store_chain_info::output_merged_store): Likewise.
5033         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5034         POINTER_DIFF_EXPR.
5036         PR c++/81212
5037         * tree-cfg.c (pass_warn_function_return::execute): Handle
5038         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5039         with BUILTINS_LOCATION.
5041         PR target/78643
5042         PR target/80583
5043         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5044         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5045         instead of DECL_MODE.
5047         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5048         last argument to rtx pointer.
5049         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5050         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5051         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5052         depending on the chosen ISAs.
5053         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5054         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5055         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5056         callers.
5057         * config/i386/sse.md (mov<mode>_internal): Likewise.
5058         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5060 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5062         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5064 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5066         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5067         parameters from prototype.
5068         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5069         parameters.  Don't print a comment.
5070         (emit_fusion_gpr_load): Adjust.
5071         (emit_fusion_load_store): Adjust.
5072         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5073         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5074         comment on the second line.
5076 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5078         PR target/43871
5079         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5080         rs6000_cpu based on cpu_index, not tune_index.
5082 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5084         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5085         which_alternative instead of which_alternative + 1.
5086         (output_asm_insn): Print an extra tab if the template is short.
5088 2017-12-01  Jim Wilson  <jimw@sifive.com>
5090         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5091         comment.
5092         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5093         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5094         reference.
5095         * doc/tm.texi: Regenerate.
5097 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5099         PR target/81959
5100         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5101         whether we can allocate pseudos before trying to fix an address.
5102         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5103         memory address is indexed or indirect.
5104         (floatuns_<mode>si2_hw2): Likewise.
5106 2017-12-01  Jason Merrill  <jason@redhat.com>
5108         * Makefile.in (TAGS): Add c-family/*.cc.
5110 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5112         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5113         (call_value_insn): Likewise.
5114         (sibcall_insn): Likewise.
5115         (sibcall_value_insn): Likewise.
5116         (movsi_aarch64): Likewise.
5117         (movdi_aarch64): Likewise.
5118         (add_losym_): Likewise.
5119         (ldr_got_small_): Likewise.
5120         (ldr_got_small_sidi): Likewise.
5121         (ldr_got_small_28k_): Likewise.
5122         (ldr_got_small_28k_sidi): Likewise.
5123         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5124         Move output_addr_const to symbolic case. Add error check.
5126 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5128         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5129         * config/s390/s390.md: Disable MVC merging peephole if it would
5130         disable operand forwarding.
5131         (new peephole2): Split MVCs if it would turn them into up to 2
5132         forwardable MVCs.
5134 2017-12-01  Richard Biener  <rguenther@suse.de>
5136         PR tree-optimization/83232
5137         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5138         detection of same access. Instead of breaking the group here
5139         do not consider the duplicate.  Add comment explaining real fix.
5141 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5143         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5145 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5147         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5149 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5151         * function.h (struct function): Remove cilk_frame_decl,
5152         is_cilk_function and calls_cilk_spawn fields.
5153         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5154         field.
5155         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5156         cilk_elemental field.
5157         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5158         * target.def: Adjust comment.
5159         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5160         Don't test cilk_elemental.
5162         PR tree-optimization/83233
5163         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5164         bswap_stat name for the struct.
5166         PR c/79153
5167         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5168         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5169         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5170         SWITCH_BREAK_LABEL_P set on the label.
5171         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5172         added for default case if it was missing and not all cases covered.
5173         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5174         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5175         set on the label.
5176         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5177         to avoid -Wimplicit-fallthrough warning.
5178         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5179         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5181         PR tree-optimization/83221
5182         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5183         down by 16.
5184         (init_reassoc): Formatting fix.
5186         PR sanitizer/81275
5187         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5188         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5189         is diagnosed.
5191         PR sanitizer/83219
5192         * tree-cfg.c: Include asan.h.
5193         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5195 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5197         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5199 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5201         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5202         vpcmpeqd instruction.
5204 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5206         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5207         registers type for 128bit mode.
5209 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5211         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5212         of ;;.
5213         * gengtype-state.c (read_state_pair): Likewise.
5214         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5215         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5216         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5217         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5218         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5219         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5220         * builtins.c (fold_builtin_3): Likewise.
5221         * graphite-scop-detection.c
5222         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5223         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5224         Likewise.
5226 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5228         PR sanitizer/81697
5229         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5230         parameter. Return true if ignore_decl_rtl_set_p is true and other
5231         conditions are satisfied.
5232         * asan.h (asan_protect_global): Add new parameter.
5233         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5234         to asan_protect_global calls.
5236 2017-11-30  Jim Wilson  <jimw@sifive.com>
5238         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5239         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5240         -mstrict-align, add info on default value.  Delete redundant lines for
5241         -mabi.  Add missing -mexplicit-relocs docs.
5243 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5245         * config/arc/arc.md (trap): New pattern.
5247 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5249         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5250         instruction to end into a delay slot.
5251         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5252         can be placed into a delay slot against reg_note.
5254 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5256         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5257         labels number of usages.
5259 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5261         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5262         function.
5263         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5265 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5267         PR target/83210
5268         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5269         multiplication by power of 2 constant into two shifts + comparison.
5271 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5273         PR target/81616
5274         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5276 2017-11-30  Richard Biener  <rguenther@suse.de>
5278         PR tree-optimization/83202
5279         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5280         allow_peel argument and guard peeling.
5281         (canonicalize_loop_induction_variables): Likewise.
5282         (canonicalize_induction_variables): Pass false.
5283         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5284         peeling from cunrolli.
5286 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5288         * combine.c (try_combine): Print a message to dump file whenever
5289         I0, I1, or I2 cannot be combined into I3.
5291 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5293         PR rtl-optimization/83156
5294         PR rtl-optimization/82621
5295         * combine.c (try_combine): Don't split an I2 if one of the dests is
5296         set again before I3.  Allow unused dests.
5298 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5300         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5302 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5304         PR rtl-optimization/80818
5305         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5306         recursively.  Use insn code for clobber.
5307         (lra_set_insn_recog_data): Pass the new arg to
5308         collect_non_operand_hard_regs.
5309         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5310         code for clobber.
5311         (lra_update_insn_regno_info): Pass insn to
5312         add_regs_to_insn_regno_info.
5314 2017-11-29  Jim Wilson  <jimw@sifive.com>
5315             Andrew Waterman  <andrew@sifive.com>
5317         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5318         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5319         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5320         use SINGLE_SHIFT_COST.
5321         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5322         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5324 2017-11-29  Julia Koval  <julia.koval@intel.com>
5326         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5327         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5328         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5329         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5330         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5331         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5332         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5333         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5334         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5335         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5336         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5337         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5338         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5339         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5340         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5341         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5342         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5343         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5344         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5345         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5346         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5347         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5348         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5349         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5350         __builtin_ia32_vpshldv_v2di_mask,
5351         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5352         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5353         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5355 2017-11-29  Julia Koval  <julia.koval@intel.com>
5357         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5358         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5359         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5360         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5361         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5362         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5363         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5364         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5365         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5366         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5367         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5368         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5369         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5370         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5371         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5372         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5373         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5374         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5375         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5376         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5377         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5379 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5381         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5382         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5383         workaround.
5385 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5387         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5388         to prevent b2bst errata sequence.
5389         (sqrtdf2_fix): Likewise.
5391 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5393         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5394         (div_sqrt_insn_p): New function.
5395         (sparc_do_work_around_errata): Insert NOP instructions to
5396         prevent sequences that could trigger the TN-0013 errata for
5397         certain LEON3 processors.
5398         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5399         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5400         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5401         (in_branch_delay): Prevent div and sqrt in delay slot if
5402         fix_lost_divsqrt.
5403         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5405 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5407         * config/sparc/sparc.c (atomic_insn_p): New function.
5408         (sparc_do_work_around_errata): Insert NOP instructions to
5409         prevent sequences that could trigger the TN-0010 errata for
5410         UT700.
5411         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5412         instruction referable in atomic_insns_p.
5414 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5416         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5417         (atomic_compare_and_swap_leon3_1): Likewise.
5418         (ldstub): Likewise.
5420 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5422         * config/sparc/sparc.c (fpop_insn_p): New function.
5423         (sparc_do_work_around_errata): Insert NOP instructions to
5424         prevent sequences that could trigger the TN-0012 errata for
5425         GR712RC.
5426         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5427         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5428         (in_branch_annul_delay): Prevent floating-point instructions
5429         in delay slot of annulled integer branch.
5431 2017-11-29  Richard Biener  <rguenther@suse.de>
5433         PR tree-optimization/83202
5434         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5435         (bst_fail): Use it.
5436         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5437         nodes vectorized to the same stmts multiple times.
5438         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5439         (vect_analyze_slp_instance): Adjust.
5440         (scalar_stmts_to_slp_tree_map_t): New typedef.
5441         (vect_schedule_slp_instance): Add a map recording the SLP node
5442         representing the vectorized stmts for a set of scalar stmts.
5443         Avoid code-generating redundancies.
5444         (vect_schedule_slp): Allocate map and pass it down.
5446 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5448         PR c++/83187
5449         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5450         type if it is new.
5452 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5454         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5455         cases for printing LDP/STP memory addresses.
5456         (aarch64_print_address_internal): Renamed from
5457         aarch64_print_operand_address, added parameter, add Pmode check.
5458         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5459         (aarch64_print_operand_address): Indirect to
5460         aarch64_print_address_internal.
5461         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5462         'y' operand output specifier.
5464 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5466         PR middle-end/83185
5467         * tree.c (build_simple_mem_ref_loc): Handle
5468         get_addr_base_and_unit_offset returning a MEM_REF.
5470         PR middle-end/80929
5471         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5473         PR target/80819
5474         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5475         alternative.
5477 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5479         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5480         m_x_offset.
5481         (layout::move_to_column): Likewise.
5483 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5485         PR sanitizer/81275
5486         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5487         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5489 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5490             Martin Jambor  <mjambor@suse.cz>
5492         PR ipa/82808
5493         * tree.h (expr_type_first_operand_type_p): Declare
5494         * tree.c (expr_type_first_operand_type_p): New function.
5495         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5496         (ipa_value_from_jfunc): Adjust declaration.
5497         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5498         Use it as result type for arithmetics, unless it is NULL in which case
5499         be more conservative.
5500         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5501         ipa_get_jf_pass_through_result.
5502         (propagate_vals_across_pass_through): Likewise.
5503         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5504         is to propagate_vals_across_pass_through.
5505         (propagate_constants_across_call): Pass PARM_TYPE to
5506         propagate_scalar_across_jump_function.
5507         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5508         ipa_value_from_jfunc.
5509         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5510         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5511         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5512         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5513         target_type, pass it to ipa_value_from_jfunc.
5514         (update_indirect_edges_after_inlining): Pass parameter type to
5515         try_make_edge_direct_simple_call.
5517 2017-11-28  Jeff Law  <law@redhat.com>
5519         * gimple-ssa-evrp-analyze.c
5520         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5521         refine ranges if scev_initialized_p returns true.
5522         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5524 2017-11-28  Julia Koval  <julia.koval@intel.com>
5526         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5527         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5528         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5529         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5530         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5531         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5532         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5533         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5534         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5535         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5536         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5537         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5538         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5539         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5540         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5541         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5542         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5543         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5544         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5545         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5546         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5547         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5549 2017-11-28  Julia Koval  <julia.koval@intel.com>
5551         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5552         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5553         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5554         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5555         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5556         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5557         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5558         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5559         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5560         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5561         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5562         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5563         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5564         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5565         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5566         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5567         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5568         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5569         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5570         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5571         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5572         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5573         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5574         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5575         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5576         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5577         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5578         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5579         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5580         __builtin_ia32_vpshld_v2di_mask): New builtins.
5581         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5582         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5584 2017-11-28  Richard Biener  <rguenther@suse.de>
5586         PR tree-optimization/80776
5587         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5588         Declare.
5589         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5590         New function.
5591         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5592         If the incoming edge is an effective fallthru because the other
5593         edge only reaches a __builtin_unreachable () then record ranges
5594         derived from the controlling condition in SSA info.
5595         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5596         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5598 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5600         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5601         on mingw build hosts.
5603 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5605         PR debug/81307
5606         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5608 2017-11-28  Richard Biener  <rguenther@suse.de>
5610         PR middle-end/83141
5611         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5612         copies generated from memcpy use a character array as reference
5613         type.
5615 2017-11-28  Julia Koval  <julia.koval@intel.com>
5616             Sebastian Peryt  <sebastian.peryt@intel.com>
5618         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5619         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5620         cilk-common.o, cilk.h, cilk-common.c): Remove.
5621         * builtin-types.def
5622         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5623         * builtins.c (is_builtin_name): Remove cilkplus condition.
5624         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5625         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5626         cilk-builtins.def, cilkplus.def): Remove.
5627         * cif-code.def (CILK_SPAWN): Remove.
5628         * cilk-builtins.def: Delete.
5629         * cilk-common.c: Ditto.
5630         * cilk.h: Ditto.
5631         * cilkplus.def: Ditto.
5632         * config/darwin.h (fcilkplus): Delete.
5633         * cppbuiltin.c: Ditto.
5634         * doc/extend.texi: Remove cilkplus doc.
5635         * doc/generic.texi: Ditto.
5636         * doc/invoke.texi: Ditto.
5637         * doc/passes.texi: Ditto.
5638         * gcc.c (fcilkplus): Remove.
5639         * gengtype.c (cilk.h): Remove.
5640         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5641         support.
5642         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5643         Remove.
5644         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5645         gimplify_call_expr,
5646         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5647         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5648         cilkplus conditions.
5649         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5650         inline_read_section): Ditto.
5651         * ipa-inline-analysis.c (cilk.h): Remove.
5652         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5653         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5654         append_linker_options): Remove condition for fcilkplus.
5655         * lto/lto-lang.c (cilk.h): Remove.
5656         (lto_init): Remove condition for fcilkplus.
5657         * omp-expand.c (expand_cilk_for_call): Delete.
5658         (expand_omp_taskreg, expand_omp_for_static_chunk,
5659         expand_omp_for): Remove cilkplus
5660         conditions.
5661         (expand_cilk_for): Delete.
5662         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5663         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5664         execute_lower_omp, diagnose_sb_0): Ditto.
5665         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5666         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5667         * tree-nested.c: Ditto.
5668         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5669         (dump_generic_node): Ditto.
5670         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5671         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5672         Delete.
5673         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5675 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5677         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5678         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5679         for vector mode and !TARGET_SIMD.
5681 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5683         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5684         Adjust comment.
5685         * tree.h (SWITCH_LABELS): Remove.
5686         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5687         assert SWITCH_BODY is non-NULL.
5688         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5689         handling.
5690         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5692         PR tree-optimization/80788
5693         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5694         has TREE_OVERFLOW set, call drop_tree_overflow.
5696 2017-11-28  Richard Biener  <rguenther@suse.de>
5698         PR tree-optimization/83158
5699         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5701 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5703         PR 81288/target
5704         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5705         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5707 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5709         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5710         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5712 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5714         PR middle_end/82333
5715         * varasm.c (compare_constant): Take the mode of the constants into
5716         account when comparing floating point constants.
5718 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5720         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5721         from explicit instantiation of debug_helper.
5722         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5724 2017-11-27  Richard Biener  <rguenther@suse.de>
5726         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5727         refactor a bit.
5729 2017-11-27  Richard Biener  <rguenther@suse.de>
5731         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5732         (type_hash_canon): Also clear int_cst_hash_table entry for
5733         TYPE_MIN/MAX_VALUE.
5734         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5736 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5738         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5739         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5740         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5742 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5744         * hash-map.h (gt_cleare_cache): Avoid UB.
5746 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5748         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5749         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5750         (decide_unroll_runtime_iterations): Likewise.
5751         (decide_unroll_stupid): Likewise.
5753 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5755         PR target/83109
5756         * config/i386/i386.md: Add a loop with incssp.
5758 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5760         PR tree-optimization/81248
5761         * tree-sra.c (splice_param_accesses): Remove size check.
5762         (decide_one_param_reduction): Fix size check.
5763         * gimple-pretty-print.c (dump_profile): Silence warning.
5764         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5766 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5768         PR debug/81307
5769         * dbxout.c (lastlineno): New variable.
5770         (dbx_debug_hooks): Use dbxout_switch_text_section as
5771         switch_text_section debug hook.
5772         (dbxout_function_end): Switch to current_function_section
5773         rather than function_section.  If crtl->has_bb_partition,
5774         output just one N_FUN, depending on in_cold_section_p.
5775         (dbxout_source_line): Remember last lineno in lastlineno.
5776         (dbxout_switch_text_section): New function.
5777         (dbxout_function_decl): Adjust dbxout_block caller.
5778         (dbx_block_with_cold_children): New function.
5779         (dbxout_block): Return true if any LBRAC/RBRAC have been
5780         emitted.  Use dbx_block_with_cold_children at depth == 0
5781         in second partition.  Add PARENT_BLOCKNUM argument, pass
5782         it optionally adjusted to children.  Output LBRAC/RBRAC
5783         around recursive call only if the block is in the current
5784         partition, if not and anything was output, emit empty
5785         range LBRAC/RBRAC.
5786         * final.c (final_scan_insn): Compute cold_function_name
5787         before calling switch_text_section debug hook.  Call
5788         that hook even if dwarf2out_do_frame if not emitting
5789         dwarf debug info.
5791         PR target/83100
5792         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5793         TREE_READONLY decls.
5795 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5797         PR rtl-optimization/82488
5798         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5800 2017-11-26  Julia Koval  <julia.koval@intel.com>
5802         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5803         skylake-avx512.
5804         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5805         skylake_cost): New.
5807 2017-11-26  Julia Koval  <julia.koval@intel.com>
5809         * config/i386/driver-i386.c (host_detect_local_cpu):
5810         Detect skylake-avx512.
5812 2017-11-26  Julia Koval  <julia.koval@intel.com>
5814         * config.gcc: Add -march=cannonlake.
5815         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5816         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5817         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5818         (PTA_CANNONLAKE): New.
5819         (processor_target_table): Add cannonlake.
5820         (ix86_option_override_internal): Ditto.
5821         (fold_builtin_cpu): Ditto.
5822         (get_builtin_code_for_version): Handle cannonlake.
5823         (M_INTEL_COREI7_CANNONLAKE): New.
5824         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5825         * doc/invoke.texi: Add -march=cannonlake.
5827 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5829         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5830         (try_init_one_plugin): Alternative implementation for MinGW.
5831         * Makefile.in (plugin_implib): New.
5832         (gengtype-lex.c): Fix broken AIX workaround.
5833         * configure: Regenerate.
5834         * doc/plugins.texi: Document support for MinGW.
5836 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5838         PR rtl-optimization/81553
5839         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5840         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5841         is a shift where C1 has different mode than the whole shift, use C1's
5842         mode for MULT rather than the shift's mode.
5844         PR target/82848
5845         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5846         builtins not enabled in the currently selected ISA.
5848 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5850         PR tree-optimization/71026
5851         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5852         (insert_reciprocals): Change to insert reciprocals before a division
5853         by a square and to insert the square of a reciprocal.
5854         (execute_cse_reciprocals_1): Change to consider division by a square.
5855         (register_division_in): Add importance parameter.
5857 2017-11-24  Richard Biener  <rguenther@suse.de>
5859         PR tree-optimization/82402
5860         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5861         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5863 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5865         * match.pd (0-ptr): New transformation.
5867 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5869         PR bootstrap/83015
5870         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5871         when skipping checking.
5873 2017-11-24  Richard Biener  <rguenther@suse.de>
5875         PR tree-optimization/83128
5876         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5877         (vn_reference_lookup_3): Likewise.
5879 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5881         PR sanitizer/83014
5882         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5883         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5884         tree_to_uhwi twice.
5886         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5887         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5888         check for the latter with gimple_call_builtin_p.  Do not handle
5889         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5891 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5893         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5894         vdot_u32.
5896 2017-11-23  Julia Koval  <julia.koval@intel.com>
5898         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5899         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5900         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5901         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5902         _mm512_maskz_expandloadu_epi16): New intrinsics.
5903         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5904         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5905         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5906         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5907         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5908         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5909         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5910         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5911         _mm256_maskz_expandloadu_epi8): New intrinsics.
5912         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5913         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5914         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5915         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5916         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5917         * config/i386/sse.md (VI248_VLBW): New iterator.
5918         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5920 2017-11-23  Julia Koval  <julia.koval@intel.com>
5922         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5923         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5924         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5925         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5926         _mm512_mask_compressstoreu_epi16): New.
5927         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5928         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5929         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5930         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5931         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5932         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5933         _mm256_mask_compressstoreu_epi8): New.
5934         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5935         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5936         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5937         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5938         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5939         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5940         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
5941         __builtin_ia32_compresshi128_mask,
5942         __builtin_ia32_compressstoreuqi512_mask,
5943         __builtin_ia32_compressstoreuhi512_mask,
5944         __builtin_ia32_compressstoreuqi256_mask,
5945         __builtin_ia32_compressstoreuqi128_mask,
5946         __builtin_ia32_compressstoreuhi256_mask,
5947         __builtin_ia32_compressstoreuhi128_mask): New builtins.
5948         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
5949         array for flags2.
5950         (ix86_expand_special_args_builtin): Handle new types.
5951         (s4fma_expand): Handle new builtin array.
5952         * config/i386/immintrin.h: Include new headers.
5953         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
5954         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
5956 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
5958         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
5959         fixes.  Declare temp and g variables at the top in order to avoid
5960         {} in most of the cases.
5962 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
5964         * match.pd (ptr-0): New transformation.
5966 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5968         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
5969         (struct addr_mode_cost_table): New.
5970         (struct tune_params): Add field addr_mode_costs.
5971         * config/arm/arm.c (generic_addr_mode_costs): New.
5972         (arm_slowmul_tune): Initialise addr_mode_costs field.
5973         (arm_fastmul_tune): Likewise.
5974         (arm_strongarm_tune): Likewise.
5975         (arm_xscale_tune): Likewise.
5976         (arm_9e_tune): Likewise.
5977         (arm_marvell_pj4_tune): Likewise.
5978         (arm_v6t2_tune): Likewise.
5979         (arm_cortex_tune): Likewise.
5980         (arm_cortex_a8_tune): Likewise.
5981         (arm_cortex_a7_tune): Likewise.
5982         (arm_cortex_a15_tune): Likewise.
5983         (arm_cortex_a35_tune): Likewise.
5984         (arm_cortex_a53_tune): Likewise.
5985         (arm_cortex_a57_tune): Likewise.
5986         (arm_exynosm1_tune): Likewise.
5987         (arm_xgene1_tune): Likewise.
5988         (arm_cortex_a5_tune): Likewise.
5989         (arm_cortex_a9_tune): Likewise.
5990         (arm_cortex_a12_tune): Likewise.
5991         (arm_cortex_a73_tune): Likewise.
5992         (arm_v7m_tune): Likewise.
5993         (arm_cortex_m7_tune): Likewise.
5994         (arm_v6m_tune): Likewise.
5995         (arm_fa726te_tune): Likewise.
5996         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
5998 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6000         * config/arm/arm.c (arm_mem_costs): New function.
6001         (arm_rtx_costs_internal): Use arm_mem_costs.
6003 2017-11-23  Mark Wielaard  <mark@klomp.org>
6005         * dwarf2out.c (init_sections_and_labels): Use generation to create
6006         unique ranges_section_label and ranges_base_label. Return generation.
6007         (output_rnglists): Add generation argument. Use generation to create
6008         unique ranges labels.
6009         (dwarf2out_finish): Get generation from init_sections_and_labels
6010         and pass generation to output_rnglists.
6012 2017-11-23  Mike Stump  <mikestump@comcast.net>
6013             Eric Botcazou  <ebotcazou@adacore.com>
6015         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
6016         * cfgloop.h (struct loop): Add unroll field.
6017         * function.h (struct function): Add has_unroll bitfield.
6018         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
6019         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
6020         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
6021         (pass_rtl_unroll_loops::gate): Likewise.
6022         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
6023         for which loop->unroll==1.
6024         (decide_unroll_constant_iterations): Use note for consistency and
6025         take loop->unroll into account.  Return early if loop->unroll is set.
6026         Fix thinko in existing test.
6027         (decide_unroll_runtime_iterations): Use note for consistency and
6028         take loop->unroll into account.
6029         (decide_unroll_stupid): Likewise.
6030         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6031         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6032         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6033         New case.
6034         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6035         (print_loop): Print loop->unroll if set.
6036         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6037         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6038         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6039         New case.
6040         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6041         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6042         entirely the heuristics if loop->unroll is set.  Remove dead note.
6043         Fix off-by-one bug in other note.
6044         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6045         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6046         is greater than 1.
6047         (tree_unroll_loops_completely): Make static.
6048         (pass_complete_unroll::execute): Use correct type for variable.
6049         (pass_complete_unrolli::execute): Fix formatting.
6050         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6052 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6054         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6055         enable when TARGET_PREFER_AVX128 is set.
6057 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6059         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6060         frequency.
6061         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6062         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6064 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6066         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6067         frequencies. Fix estimation of aggregate parameters.
6069 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6071         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6072         when inlining.
6074 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6076         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6078 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6080         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6081         macro.
6083 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6085         PR target/83111
6086         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6087         sibcall_value_pcrel_fdpic): Use local variable instead of
6088         operands[3].
6089         (calli_tbr_rel): Add missing operand 2.
6090         (call_valuei_tbr_rel): Add missing operand 3.
6092 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6094         PR middle-end/82253
6095         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6096         bitpos/bitsize covering the whole destination, use store_expr only if
6097         the complex mode is the same.  Otherwise, use expand_normal and if
6098         it returns CONCAT, subreg each part separately instead of trying to
6099         subreg the whole result.
6101 2017-11-23  Richard Biener  <rguenther@suse.de>
6103         PR tree-optimization/23094
6104         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6105         come here from walking over backedges in the first iteration.
6106         (vn_reference_lookup_3): Skip clobbers that store the same value.
6108 2017-11-23  Richard Biener  <rguenther@suse.de>
6110         PR tree-optimization/81403
6111         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6112         a block we need a leader relative to.
6113         (phi_translate_1): For nary processing require a leader from
6114         get_representative_for given we run expression simplification
6115         using match-and-simplify.  Remove previous fix.
6117 2017-11-22  Jeff Law  <law@redhat.com>
6119         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6120         Use new method allocate_value_range rather than accessing the
6121         vrp_value_range_pool data member directly.
6122         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6123         to use extract_range_from_stmt method to avoid need for
6124         extract_range_from_assignment method.
6125         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6126         method rather than setting values_propgated data member directly.
6127         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6128         and values propagated data members and extract_range_from_assignment
6129         method.  Reorder private data members to conform to standards.
6130         Add new methods set_lattice_propagation_complete and
6131         allocate_value_range.
6133 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6135         PR rtl-optimization/83030
6136         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6137         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6138         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6139         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6140         CROSSING_JUMP_P flag is already set before setting it.
6141         * cfgrtl.c (fixup_partition_crossing): Likewise.
6142         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6143         insn as useless.
6145 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6147         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6148         Handle the case where both arguments are using gen_const_vec_series.
6150 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6152         PR c++/62170
6153         * pretty-print.c (pp_format): Move quoting implementation to
6154         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6155         to pass address of "quote" local.
6156         (pp_begin_quote): New function.
6157         (pp_end_quote): New function.
6158         * pretty-print.h (printer_fn): Convert penultimate param from bool
6159         to bool *.
6160         (pp_begin_quote): New decl.
6161         (pp_end_quote): New decl.
6162         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6163         param from bool to bool *.
6164         * tree-diagnostic.h (default_tree_printer): Likewise.
6166 2017-11-22  Jeff Law  <law@redhat.com>
6168         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6169         Initialize vr_values.
6170         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6171         vr_values via vr_values class instance rather than delegators.
6172         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6173         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6174         (evrp_range_analyzer::push_value_range): Likewise.
6175         (evrp_range_analyzer::pop_value_range): Likewise.
6176         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6177         most delegators.  Those remaining are exposed as public interfaces.
6178         Make vr_values a pointer and private.
6179         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6180         vr_values.
6181         (evrp_range_analyzer::get_vr_value): New method.
6182         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6183         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6184         (class evrp_dom_walker): Attach evrp_folder class, initialize
6185         it in the ctor.  Remove temporary delegators.
6186         (evrp_dom_walker::before_dom_children): Call methods in attached
6187         evrp_range_analyzer class via class instance pointer.  Use
6188         free value_range_constant_singleton to remove need for
6189         op_with_constant_singleton_value delegator method.  Do not
6190         create a vrp_prop class instance for every call!  Narrow
6191         scope of a couple variables.
6192         (evrp_dom_walker::cleanup): Call methods in attached
6193         evrp_range_analyzer class via class instance pointer.
6194         * vr-values.h (class vr_values): Privatize many methods and
6195         data members.
6197 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6199         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6200         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6201         * expr.c (expand_expr_real_2): Likewise.
6202         * fold-const.c (const_unop): Likewise.
6203         * optabs-tree.c (optab_for_tree_code): Likewise.
6204         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6205         * tree-inline.c (estimate_operator_cost): Likewise.
6206         * tree-pretty-print.c (dump_generic_node): Likewise.
6207         (op_code_prio): Likewise.
6208         (op_symbol_code): Likewise.
6209         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6210         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6211         * internal-fn.c (direct_internal_fn_optab): New function.
6212         (direct_internal_fn_array, direct_internal_fn_supported_p
6213         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6214         * fold-const-call.c (fold_const_reduction): New function.
6215         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6216         CFN_REDUC_MIN.
6217         * tree-vect-loop.c: Include internal-fn.h.
6218         (reduction_code_for_scalar_code): Rename to...
6219         (reduction_fn_for_scalar_code): ...this and return an internal
6220         function.
6221         (vect_model_reduction_cost): Take an internal_fn rather than
6222         a tree_code.
6223         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6224         than assignments.
6225         (vectorizable_reduction): Use internal functions rather than tree
6226         codes for the reduction operation.  Update calls to the functions
6227         above.
6228         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6229         Use calls to internal functions rather than REDUC tree codes.
6230         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6232 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6234         * config/vxworks.c (vxworks_override_options): Pick default
6235         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6236         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6237         DWARF_GNAT_ENCODINGS_DEFAULT.
6238         * config/vxworksae.h: Likewise.
6240 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6242         PR tree-optimization/83104
6243         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6244         not result.
6246 2017-11-22  Marek Polacek  <polacek@redhat.com>
6247             H.J. Lu  <hongjiu.lu@intel.com>
6248             Jason Merrill  <jason@redhat.com>
6250         PR c++/60336
6251         PR middle-end/67239
6252         PR target/68355
6253         * calls.c (initialize_argument_information): Call
6254         warn_parameter_passing_abi target hook.
6255         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6256         argument onto stack.
6257         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6258         * common.opt: Update -fabi-version description.
6259         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6260         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6261         int_size_in_bytes.
6262         (ix86_is_empty_record): New function.
6263         (ix86_warn_parameter_passing_abi): New function.
6264         (TARGET_EMPTY_RECORD_P): Redefine.
6265         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6266         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6267         * doc/tm.texi: Regenerated.
6268         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6269         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6270         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6271         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6272         instead of int_size_in_bytes.
6273         * expr.c (copy_blkmode_to_reg): Likewise.
6274         * function.c (aggregate_value_p): Return 0 for empty types.
6275         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6276         target hook.
6277         (locate_and_pad_parm): Call arg size_in_bytes instead
6278         size_in_bytes.
6279         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6280         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6281         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6282         hooks.
6283         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6284         (std_gimplify_va_arg_expr): Skip empty records.  Call
6285         arg_size_in_bytes instead size_in_bytes.
6286         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6287         * tree-core.h (tree_type_common): Add empty_flag.
6288         (tree_decl_common): Update comments.
6289         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6290         DECL_PADDING_P.
6291         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6292         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6293         DECL_PADDING_P.
6294         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6295         * tree.c (default_is_empty_type): New function.
6296         (default_is_empty_record): New function.
6297         (arg_int_size_in_bytes): New function.
6298         (arg_size_in_bytes): New function.
6299         (get_ultimate_context): New function.
6300         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6301         TRANSLATION_UNIT_WARN_EMPTY_P.
6302         (default_is_empty_record, arg_int_size_in_bytes,
6303         arg_size_in_bytes, get_ultimate_context): Declare.
6305 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6307         * config/arm/arm.c (cmse_clear_registers): New function.
6308         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6309         code by call to cmse_clear_registers.
6310         (cmse_nonsecure_entry_clear_before_return): Likewise.
6312 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6314         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6315         (vdot_s32, vdotq_s32): New.
6316         (vdot_lane_u32, vdotq_lane_u32): New.
6317         (vdot_lane_s32, vdotq_lane_s32): New.
6320 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6322         PR middle-end/82547
6323         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6324         for unsigned values with fewer HWIs than the precision.
6325         (test_overflow): New function.
6326         (wide_int_cc_tests): Call it.
6328 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6330         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6331         has the same mode class as Pmode.
6333 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6335         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6336         padding_bits_to_clear_ptr.
6337         (cmse_nonsecure_entry_clear_before_return): Likewise.
6339 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6341         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6342         auto_sbitap instead of integer bitfield to control register needing
6343         clearing.
6345 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6347         PR tree-optimization/83044
6348         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6349         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6350         that need the upper bound.  Subtract offset from
6351         get_addr_base_and_unit_offset only if positive and subtract it
6352         before division by eltsize rather than after it.
6354         PR debug/83084
6355         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6356         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6357         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6358         asm.
6360         PR middle-end/82875
6361         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6362         expand_binop with *mul_widen_optab, make sure at least one of the
6363         operands doesn't have VOIDmode.
6365         PR debug/83034
6366         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6368         PR rtl-optimization/82044
6369         PR tree-optimization/82042
6370         * dse.c (record_store): Check for overflow.
6371         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6372         add_wild_read instead of clear_rhs_from_active_local_stores on
6373         overflow and log it into dump_file.
6375 2017-11-22  Richard Biener  <rguenther@suse.de>
6377         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6378         fake edges to exit when looking for a place to insert.
6379         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6380         and remove.
6381         (insert_into_preds_of_block): Commit edge insertion immediately,
6382         assert that doesn't require new BBs.
6383         (fini_pre): Release expressions.
6384         (pass_pre::execute): Shuffle things around a bit, if the fn
6385         is too large do not compute AVAIL either as this is really the
6386         quadratic bit.
6388 2017-11-22  Richard Biener  <rguenther@suse.de>
6390         PR tree-optimization/83089
6391         * tree-if-conv.c (pass_if_conversion::execute): If anything
6392         changed reset SCEV and free the number of iteration estimates.
6394 2017-11-21  Martin Sebor  <msebor@redhat.com>
6396         PR tree-optimization/82945
6397         * calls.h (warn_nonstring_bound): Remove unused function.
6399 2017-11-21  Martin Sebor  <msebor@redhat.com>
6401         PR tree-optimization/82945
6402         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6403         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6404         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6405         functions.
6406         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6407         * calls.h (get_attr_nonstring_decl): Declare new function.
6408         * doc/extend.texi (attribute nonstring): Update.
6409         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6410         get_attr_nonstring_decl and handle it.
6411         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6412         detection of nul-termination.
6413         (strlen_to_stridx): Change to a pointer.
6414         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6415         (pass_strlen::execute): Same.
6417 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6419         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6420         for the new option -mprefer-vector-width=[none|128|256|512].
6421         * config/i386/i386.c (ix86_target_string): remove old style options
6422         -mprefer-avx256 and make -mprefer-avx128 as alias.
6423         (ix86_option_override_internal):  Apply defaults for the
6424         -mprefer-vector-width=[128|256] option.
6425         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6426         Implement macros to work with -mprefer-vector-width=.
6427         * config/i386/i386.opt: Implemented option
6428         -mprefer-vector-width=[none|128|256|512].
6429         * doc/invoke.texi: Documentation for
6430         -mprefer-vector-width=[none|128|256|512].
6432 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6434         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6436 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6438         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6440 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6442         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6443         POINTER_PLUS_EXPR.
6444         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6445         * expr.c (expand_expr_real_2): Likewise.
6446         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6447         fold_binary_loc): Likewise.
6448         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6449         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6450         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6451         MINUS_EXPR transformations.
6452         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6453         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6454         * tree-inline.c (estimate_operator_cost): Likewise.
6455         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6456         op_symbol_code): Likewise.
6457         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6458         * vr-values.c (extract_range_from_binary_expr): Likewise.
6459         * varasm.c (initializer_constant_valid_p_1): Likewise.
6460         * tree.def: New tree code POINTER_DIFF_EXPR.
6462 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6464         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6465         integer suffix to movbe mnemonic.
6466         (*bswaphi2_movbe): Ditto.
6467         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6469 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6471         PR c++/83045
6472         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6473         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6474         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6476 2017-11-21  Martin Liska  <mliska@suse.cz>
6478         * tree-inline.c (expand_call_inline): Remove not needed
6479         xstrdup_for_dump.
6481 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6482             Jakub Jelinek  <jakub@redhat.com>
6484         PR target/82880
6485         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6486         Remove static keyword from f variable.
6488 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6490         PR tree-optimization/83086
6491         * gimple-ssa-store-merging.c
6492         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6493         rather than n.base_addr.
6495 2017-11-21  Martin Liska  <mliska@suse.cz>
6497         PR rtl-optimization/82044
6498         PR tree-optimization/82042
6499         * dse.c (check_mem_read_rtx): Check for overflow.
6501 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6503         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6504         typo in comment.
6506 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6508         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6509         and its corresponding call arg location note.
6511 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6512             Andrew Burgess  <andrew.burgess@embecosm.com>
6514         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6515         declaration.
6516         (arc_return_slot_offset): Likewise.
6517         (arc_eh_return_address_location): New declaration.
6518         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6519         (MUST_SAVE_REGISTER): Add exception handler case.
6520         (MUST_SAVE_RETURN_ADDR): Likewise.
6521         (arc_frame_pointer_required): Likewise.
6522         (arc_frame_pointer_needed): New function.
6523         (arc_compute_frame_size): Changed.
6524         (arc_expand_prologue): Likewise.
6525         (arc_expand_epilogue): Likewise.
6526         (arc_initial_elimination_offset): Likewise.
6527         (arc_return_slot_offset): Delete.
6528         (arc_eh_return_address_location): New function.
6529         (arc_builtin_setjmp_frame_value): Likewise.
6530         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6531         (EH_RETURN_STACKADJ_RTX): Define.
6532         (EH_RETURN_HANDLER_RTX): Likewise.
6533         * config/arc/arc.md (eh_return): Delete.
6535 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6537         * print-tree.h (debug_vec_tree): Remove prototype.
6538         * gdbinit.in (pvt): Remove macro.
6540 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6542         PR tree-optimization/83047
6543         * gimple-ssa-store-merging.c
6544         (imm_store_chain_info::output_merged_store): If the loads with the
6545         same vuse are in different basic blocks, for load_gsi pick a load
6546         location that is dominated by the other loads.
6548         PR c++/83059
6549         * config/i386/i386.c (ix86_memmodel_check): Start
6550         -Winvalid-memory-model diagnostics with lowercase letter.
6552         PR debug/82718
6553         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6554         set in_cold_section_p to the partition containing loc_list->first.
6555         When seeing loc_list->last_before_switch node, update secname and
6556         perform range_across_switch second partition handling only after that.
6558         PR debug/82933
6559         * run-rtl-passes.c: Include debug.h.
6560         (run_rtl_passes): Call debug_hooks->assembly_start.
6561         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6562         multiple times.
6564         PR target/82981
6565         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6566         OPTAB_DIRECT in calls to expand_simple_binop.
6568 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6570         PR c/81404
6571         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6573 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6575         PR c++/72786
6576         * spellcheck.h (best_match::blithely_get_best_candidate): New
6577         accessor.
6579 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6581         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6582         with lower case letter.
6584 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6586         * config/i386/i386.md (bswaphi2): New expander.
6587         (*bswaphi2_movbe): New insn pattern.
6588         (bswaphi -> rorhi pepehole2): New peephole pattern.
6590 2017-11-20  Jeff Law  <law@redhat.com>
6592         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6593         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6594         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6595         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6596         gimple-ssa-evrp-analyze.h.
6598         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6599         set BB_VISITED here.
6600         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6602 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6604         * tree-predcom.c: Add general comment on Store-Store chains.
6605         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6606         flag in component.
6607         (get_chain_last_ref_at): Rename into...
6608         (get_chain_last_write_at): ...this.
6609         (get_chain_last_write_before_load): New function.
6610         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6611         CT_STORE_STORE when write reference is added.
6612         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6613         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6614         (initialize_root_vars_store_elim_1): Ditto.
6615         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6616         definition is created.
6617         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6618         chain by replacing it with dominant stored value.
6620 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6622         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6624 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6626         * vr-values.c (extract_range_from_binary_expr): Use a full range
6627         for VR_VARYING.
6629 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6631         * config/arm/arm.md (R4_REGNUM): Define constant.
6632         (nonsecure_call_internal): Remove r4 clobber.
6633         (nonsecure_call_value_internal): Likewise.
6634         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6635         clobber and resequence match_operands.
6636         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6637         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6638         (nonsecure_call_value_reg_thumb2): Likewise.
6640 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6642         PR tree-optimization/78821
6643         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6644         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6645         to the address of the base rather than the base itself.
6646         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6647         (find_bswap_or_nop_finalize): New function.
6648         (find_bswap_or_nop): Use it.
6649         (bswap_replace): Return a tree rather than bool, change first
6650         argument from gimple * to gimple_stmt_iterator, allow inserting
6651         into an empty sequence, allow ins_stmt to be NULL - then emit
6652         all stmts into gsi.  Fix up MEM_REF address gimplification.
6653         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6654         (struct store_immediate_info): Add N and INS_STMT non-static
6655         data members.
6656         (store_immediate_info::store_immediate_info): Initialize them
6657         from newly added ctor args.
6658         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6659         bitpos at the end.
6660         (stmts_may_clobber_ref_p): For stores call also
6661         refs_anti_dependent_p.
6662         (gather_bswap_load_refs): New function.
6663         (imm_store_chain_info::try_coalesce_bswap): New method.
6664         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6665         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6666         (imm_store_chain_info::output_merged_store): Fail if number of
6667         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6668         and NOP_EXPR rhs_code.
6669         (pass_store_merging::process_store): Compute n and ins_stmt, if
6670         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6671         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6672         ctor.
6673         (pass_store_merging::execute): Calculate dominators.
6675         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6676         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6677         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6678         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6679         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6680         class pass_optimize_bswap, bswap_replace,
6681         pass_optimize_bswap::execute): Moved to ...
6682         * gimple-ssa-store-merging.c: ... this file.
6683         Include optabs-tree.h.
6684         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6685         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6686         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6687         anonymous namespace, remove static keywords.
6688         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6689         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6691 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6693         PR bootstrap/83062
6694         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6696 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6698         * vec.h (debug_helper): New function.
6699         (DEFINE_DEBUG_VEC): New macro.
6700         * hash-set.h (debug_helper): New function.
6701         (DEFINE_DEBUG_HASH_SET): New macro.
6702         * cfg.c (debug_slim (edge)): New function.
6703         Call DEFINE_DEBUG_VEC for edges.
6704         Call DEFINE_DEBUG_HASH_SET for edges.
6705         * cfghooks.c (debug_slim (basic_block)): New function.
6706         Call DEFINE_DEBUG_VEC for basic blocks.
6707         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6708         * print-tree.c (debug_slim): New function to handle trees.
6709         Call DEFINE_DEBUG_VEC for trees.
6710         Call DEFINE_DEBUG_HASH_SET for trees.
6711         (debug (vec<tree, va_gc>) &): Remove.
6712         (debug (<vec<tree, va_gc>) *): Remove.
6713         * print-rtl.c (debug_slim): New function to handle const_rtx.
6714         Call DEFINE_DEBUG_VEC for rtx_def.
6715         Call DEFINE_DEBUG_VEC for rtx_insn.
6716         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6717         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6718         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6719         (debug (vec<rtx_insn *> *ptr): Remove.
6720         (debug_insn_vector): Remove.
6721         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6723 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6725         PR rtl-optimization/82020
6726         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6727         IF_THEN_ELSE condition.
6729 2017-11-19  Jeff Law  <law@redhat.com>
6731         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6732         of degenerates resulting from ignoring an edge.
6734 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6736         PR ipa/81360
6737         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6739 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6741         PR ipa/83001
6742         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6743         for uninitialied counts.
6745 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6747         PR ipa/60243
6748         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6750 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6752         PR target/82713
6753         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6755 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6757         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6758         "do while (0)".
6760 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6762         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6763         "do {} while (0)".
6765 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6767         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6768         macro body.
6770 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6772         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6773         "do {} while (0)".
6774         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6776 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6778         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6779         "do {} while (0)".
6780         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6781         semicolon after MCORE_EXPORT_NAME call.
6783 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6785         PR target/82961
6786         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6787         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6789 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6791         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6792         (cgraph_edge::clone): Cleanup updating of profile.
6793         * ipa-cp.c (update_profiling_info): Likewise.
6794         * ipa-inline-transform.c (inline_transform): Likewise.
6795         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6796         * ipa-split.c (execute_split_functions): Do not split when function
6797         is cold.
6798         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6799         * profile-count.c (profile_count::dump): Add global0.
6800         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6801         undefined.
6802         (profile_count::to_sreal_scale): Likewise.
6803         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6804         (profile_count::combine_with_ipa_count): New function.
6805         * profile-count.h (profile_guessed_global0adjusted): New.
6806         (profile_count::adjusted_zero): New.
6807         (profile_count::global0adjusted): New.
6808         (profile_count::combine_with_ipa_count): New.
6809         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6810         correct profile of return block of split functions.
6811         (copy_cfg_body): Remove unused profile_count.
6812         (copy_body): Likewise.
6813         (expand_call_inline): Update.
6814         (tree_function_versioning): Update.
6816 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6818         * hash-set.h (hash_set::empty): New.
6819         * tree-ssa-threadbackward.h: Delete.
6820         * tree-ssa-threadbackward.c (class thread_jumps): New.
6821         Move max_threaded_paths into class.
6822         (fsm_find_thread_path): Remove arguments that are now in class.
6823         (profitable_jump_thread_path): Rename to...
6824         (thread_jumps::profitable_jump_thread_path): ...this.
6825         (convert_and_register_jump_thread_path): Rename to...
6826         (thread_jumps::convert_and_register_current_path): ...this.
6827         (check_subpath_and_update_thread_path): Rename to...
6828         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6829         (register_jump_thread_path_if_profitable): Rename to...
6830         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6831         (handle_phi): Rename to...
6832         (thread_jumps::handle_phi): ...this.
6833         (handle_assignment): Rename to...
6834         (thread_jumps::handle_assignment): ...this.
6835         (fsm_find_control_statement_thread_paths): Rename to...
6836         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6837         (find_jump_threads_backwards): Rename to...
6838         (thread_jumps::find_jump_threads_backwards): ...this.
6839         Initialize path local data.
6840         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6841         from within thread_jumps class.
6842         (pass_early_thread_jumps::execute): Same.
6844 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6846         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6848 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6850         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6851         to canonical location.
6853 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6855         PR target/81356
6856         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6857         Remove.
6858         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6860 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6862         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6863         rcrt1.o%s/grcrt1.o%s for -static-pie.
6865 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6867         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6868         ix86_shift_rotate_cost): Break out from ...
6869         (ix86_rtx_costs): ... here.
6870         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6871         vector operations.
6873 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6875         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6876         when entry block was promoted unlikely.
6877         (estimate_bb_frequencies): Increase frequency scale.
6878         * profile-count.h (profile_count): Export precision info.
6880 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6882         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6883         disturbin profile of entry block.
6885 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6887         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6888         roundoff errors.
6890 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6892         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6893         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6895 2017-11-17  Jeff Law  <law@redhat.com>
6897         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6898         from evrp_dom_walker class.  Various methods moved into new class.
6899         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6900         (evrp_range_analyzer::enter): New method.
6901         (evrp_range_analyzer::leave): New method.
6902         (evrp_dom_walker): Remove delegators no longer needed by this class.
6903         Replace vr_values data member with evrp_range_analyzer
6905         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6906         method extracted from evrp_dom_walker::before_dom_children.
6907         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6908         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6910         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6911         Add private copy constructor and move assignment operators.
6912         Privatize methods and class data where trivially possible.
6913         (evrp_dom_walker::cleanup): New function, extracted from
6914         execute_early_vrp.  Simplify access to class data.
6916         * vr-values.h (get_output_for_vrp): Prototype.
6917         * vr-values.c (get_output_for_vrp): New function extracted from
6918         vrp_visit_assignment_or_call and extract_range_from_stmt.
6919         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6921 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6923         * config/aarch64/aarch64.c
6924         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6925         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6926         tune_params::AUTOPREFETCHER_WEAK.
6928 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6930         PR target/82641
6931         * config/arm/arm.c (arm_valid_target_attribute_rec):
6932         Parse "arch=" and "+<ext>".
6933         (arm_valid_target_attribute_tree): Re-init global options.
6934         (arm_option_override): Make non-static.
6935         (arm_options_perform_arch_sanity_checks): Make errors fatal.
6936         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6937         (__ARM_FEATURE_CRC32): Support undef.
6938         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6939         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
6941 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
6943         * gdbinit.in (break-on-diagnostic): New command.
6945 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6947         * config/i386/i386.c (ix86_expand_epilogue): Change simple
6948         return to indirect jump for EH return if control-flow
6949         protection is enabled. Change explicit 'false' argument in
6950         pro_epilogue_adjust_stack with a value of
6951         flag_cf_protection.
6952         * config/i386/i386.md (simple_return_indirect_internal):
6953         Remove SImode restriction to support 64-bit.
6955 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6957         * combine.c (added_notes_insn): New.
6958         (try_combine): Handle added_notes_insn like added_links_insn.
6959         Rewrite return value code.
6960         (distribute_notes): Set added_notes_insn to the earliest insn we added
6961         a note to.
6963 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6965         PR rtl-optimization/82621
6966         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
6967         dest of one of those SETs is unused.
6969 2017-11-17  Richard Biener  <rguenther@suse.de>
6971         PR fortran/83017
6972         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
6973         * tree-pretty-print.c (dump_generic_node): Handle
6974         annot_expr_parallel_kind.
6975         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
6976         * gimplify.c (gimple_boolify): Likewise.
6978 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6980         * config.gcc (extra_headers): Add cet.h for x86 targets.
6981         * config/i386/cet.h: New file.
6982         * doc/install.texi: Add --enable-cet/--disable-cet.
6984 2017-11-17  Richard Biener  <rguenther@suse.de>
6986         PR tree-optimization/83017
6987         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
6988         (gen_parallel_loop): Properly count iterations.
6989         (parallelize_loops): Handle loop->can_be_parallel independent
6990         of flag_loop_parallelize_all.  Make static profitability test match
6991         the runtime one.
6992         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
6993         * invoke.texi (parloops-min-per-thread): Document.
6995 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
6997         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
6998         upstreaming review comments.
7000 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7002         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
7003         with fast unaligned access.
7004         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
7006 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7008         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
7009         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
7011 2017-11-17  Richard Biener  <rguenther@suse.de>
7013         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
7014         folding of references.
7016 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
7018         PR middle-end/78809
7019         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
7020         of replacing call to strncmp with corresponding call to strcmp when
7021         meeting conditions.
7023 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7025         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7026         option prefer-avx256 for skylake-avx512 configuration.
7027         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7028         (get_builtin_code_for_version): Ditto.
7030 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7031             Monk Chiang  <sh.chiang04@gmail.com>
7033         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7034         (FIXED_REGISTERS): Reserve more register numbers.
7035         (CALL_USED_REGISTERS): Likewise.
7036         (REG_ALLOC_ORDER): Likewise.
7037         (REG_CLASS_CONTENTS): Likewise.
7038         (REGISTER_NAMES): Likewise.
7040 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7041             Kito Cheng  <kito.cheng@gmail.com>
7043         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7044         V2SI.
7045         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7047 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7049         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7050         parameter list for vec_splats.
7052 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7054         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7055         date of C17.
7056         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7058 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7060         PR bootstrap/82856
7061         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7062         with the required version of automake 1.11.6.
7064 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7066         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7067         for it.
7068         (DU_C2_3_power9): Correct reservation combinations.
7069         (FP_DIV_power9, VEC_DIV_power9): New.
7070         (power9-alu): Split out rotate/shift...
7071         (power9-rot): ...to here, correct dispatch resource.
7072         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7073         resource.
7074         (power9-fp): Correct latency.
7075         (power9-sdiv): Add div/sqrt resource.
7076         (power9-ddiv): Correct latency, add div/sqrt resource.
7077         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7078         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7079         resource.
7080         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7082 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7084         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7085         switch statement mapping KF built-ins to TF built-ins if we don't
7086         have the proper ISA 3.0 assembler support.
7088 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7090         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7091         (lower_emutls_function_body): Do not compute it.
7093 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7095         * ipa-split.c (split_bb_info): Turn time to sreal.
7096         (split_point): Likewise.
7097         (dump_split_point): Likewise.
7098         (fine_split_points): Likewise.
7099         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7101 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7103         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7104         in sreal.
7106 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7108         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7109         edges.
7110         (expensive_function_p): Remove useless assert.
7111         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7113 2017-11-16  Martin Sebor  <msebor@redhat.com>
7115         PR tree-optimization/82588
7116         PR tree-optimization/82583
7117         * tree-vrp.c (check_array_ref): Handle flexible array members,
7118         string literals, and inner indices.
7119         (search_for_addr_array): Add detail to diagnostics.
7121 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7123         PR c++/82836
7124         PR c++/82737
7125         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7126         (SET_DECL_ASSEMBLER_NAME): Forward to
7127         overwrite_decl_assembler_name.
7128         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7129         (overwrite_decl_assembler_name): Declare.
7130         * tree.c (overwrite_decl_assembler_name): New.
7131         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7132         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7133         (LANG_HOOKS_INITIALIZER): Add it.
7134         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7135         * langhooks.c (lhd_set_decl_assembler_name): Use
7136         SET_DECL_ASSEMBLER_NAME.
7137         (lhd_overwrite_decl_assembler_name): Default implementation.
7139 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7140             Jackson Woodruff  <jackson.woodruff@arm.com>
7142         PR tree-optimization/71026
7143         * match.pd: Canonicalize constant multiplies in division.
7145 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7147         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7148         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7149         Remove OPT_fomit_frame_pointer entry.
7150         * common/config/arc/arc-common.c: Likewise.
7151         * common/config/arm/arm-common.c: Likewise.
7152         * common/config/avr/avr-common.c: Likewise.
7153         * common/config/c6x/c6x-common.c: Likewise.
7154         * common/config/cr16/cr16-common.c: Likewise.
7155         * common/config/cris/cris-common.c: Likewise.
7156         * common/config/epiphany/epiphany-common.c: Likewise.
7157         * common/config/fr30/fr30-common.c: Likewise.
7158         * common/config/frv/frv-common.c: Likewise.
7159         * common/config/ia64/ia64-common.c: Likewise.
7160         * common/config/iq2000/iq2000-common.c: Likewise.
7161         * common/config/lm32/lm32-common.c: Likewise.
7162         * common/config/m32r/m32r-common.c: Likewise.
7163         * common/config/mcore/mcore-common.c: Likewise.
7164         * common/config/microblaze/microblaze-common.c: Likewise.
7165         * common/config/mips/mips-common.c: Likewise.
7166         * common/config/mmix/mmix-common.c: Likewise.
7167         * common/config/mn10300/mn10300-common.c: Likewise.
7168         * common/config/nios2/nios2-common.c: Likewise.
7169         * common/config/pa/pa-common.c: Likewise.
7170         * common/config/pdp11/pdp11-common.c: Likewise.
7171         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7172         * common/config/riscv/riscv-common.c: Likewise.
7173         * common/config/rs6000/rs6000-common.c: Likewise.
7174         * common/config/rx/rx-common.c: Likewise.
7175         * common/config/s390/s390-common.c: Likewise.
7176         * common/config/sh/sh-common.c: Likewise.
7177         * common/config/sparc/sparc-common.c: Likewise.
7178         * common/config/tilegx/tilegx-common.c: Likewise.
7179         * common/config/tilepro/tilepro-common.c: Likewise.
7180         * common/config/v850/v850-common.c: Likewise.
7181         * common/config/visium/visium-common.c: Likewise.
7182         * common/config/xstormy16/xstormy16-common.c: Likewise.
7183         * common/config/xtensa/xtensa-common.c: Likewise.
7184         * invoke.texi (-fomit-frame-pointer): Update documentation.
7186 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7188         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7190 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7192         * cfg.c (scale_bbs_frequencies_int,
7193         cale_bbs_frequencies_gcov_type): Remove.
7194         * cfg.h (scale_bbs_frequencies_int,
7195         cale_bbs_frequencies_gcov_type): Remove.
7197 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7199         * tree-ssa-loop-manip.c
7200         (scale_dominated_blocks_in_loop): Update to profile counts.
7201         (tree_transform_and_unroll_loop): Likewise.
7203 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7205         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7206         scale_bbs_frequencies_int.
7208 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7210         * final.c (compute_alignments): Use counts rather than frequencies.
7212 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7214         * cfgloopanal.c: Include sreal.h
7215         (average_num_loop_insns): Use counts and sreal for accounting.
7217 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7219         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7220         manipulation.
7222 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7223             Kito Cheng  <kito.cheng@gmail.com>
7225         * config/nds32/constraints.md: Provide more constraints.
7226         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7227         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7228         support constraints usage.
7230 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7232         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7234 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7235             Kito Cheng  <kito.cheng@gmail.com>
7237         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7238         * config/nds32/nds32.opt: Refine the layout.
7239         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7240         TARGET_EXT_STRING): Support new options.
7241         * config/nds32/nds32.h: Likewise.
7242         * config/nds32/nds32.md: Likewise.
7243         * config/nds32/nds32-predicates.c: Likewise.
7244         * config/nds32/constraints.md: Likewise.
7245         * common/config/nds32/nds32-common.c: Likewise.
7247 2017-11-16  Julia Koval  <julia.koval@intel.com>
7249         PR target/82983
7250         * config/i386/gfniintrin.h: Add sse check.
7251         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7253 2017-11-16  Julia Koval  <julia.koval@intel.com>
7255         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7256         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7257         (ix86_handle_option): Handle -mavx512vbmi2.
7258         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7259         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7260         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7261         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7262         (ix86_valid_target_attribute_inner_p): Ditto.
7263         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7264         * config/i386/i386.opt (mavx512vbmi2): New option.
7265         * doc/invoke.texi: Add new option.
7267 2017-11-16  Julia Koval  <julia.koval@intel.com>
7269         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7270         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7271         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7272         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7273         _mm512_gf2p8mul_epi8): New intrinsics.
7274         * config/i386/i386-builtin-types.def
7275         (V64QI_FTYPE_V64QI_V64QI): New type.
7276         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7277         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7278         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7279         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7280         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7281         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7283 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7285         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7286         explicitly as a stream of bytes.
7288 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7290         * config/rs6000/altivec.h (vec_xst_be): New #define.
7291         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7292         and externalize from *altivec_vperm_<mode>_internal.
7293         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7294         instantiation.
7295         (XL_BE_V8HI): Likewise.
7296         (XL_BE_V4SI): Likewise.
7297         (XL_BE_V4SI): Likewise.
7298         (XL_BE_V2DI): Likewise.
7299         (XL_BE_V4SF): Likewise.
7300         (XL_BE_V2DF): Likewise.
7301         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7302         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7303         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7304         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7305         VSX_BUILTIN_VEC_XST_BE.
7306         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7307         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7308         built-ins.
7309         (altivec_init_builtins): Replace conditional calls to def_builtin
7310         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7311         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7312         with unconditional calls.  Remove calls to def_builtin for
7313         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7314         __builtin_vec_xst_be.
7315         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7316         to define_expand, and add alternate RTL generation for P8.
7317         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7318         vsx_ld_elemrev_v8hi.
7319         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7320         add alternate RTL generation for P8.
7321         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7322         vsx_ld_elemrev_v16qi.
7323         (vsx_st_elemrev_v8hi): Convert define_insn
7324         to define_expand, and add alternate RTL generation for P8.
7325         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7326         vsx_st_elemrev_v8hi.
7327         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7328         add alternate RTL generation for P8.
7329         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7330         vsx_st_elemrev_v16qi.
7332 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7334         PR target/82990
7335         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7336         TARGET_AVX512ER check.
7337         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7338         neither -mzeroupper nor -mno-zeroupper is used and
7339         TARGET_EMIT_VZEROUPPER is set.
7340         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7341         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7343 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7345         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7346         folding of vector compares.
7347         (fold_build_vec_cmp): New helper function.
7348         (fold_compare_helper): New helper function.
7349         (builtin_function_type): Add compare builtins to the list of functions
7350         having unsigned arguments.  Cosmetic updates to comment indentation.
7351         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7352         the not+eq combination.
7354 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7356         PR tree-optimization/82726
7357         PR tree-optimization/70754
7358         * tree-predcom.c (order_drefs_by_pos): New function.
7359         (combine_chains): Move code setting has_max_use_after to...
7360         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7361         according to position information.
7362         (tree_predictive_commoning_loop): Update call to above function.
7363         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7365 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7367         PR tree-optimization/82726
7368         Revert
7369         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7371         PR tree-optimization/70754
7372         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7373         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7374         combined stmt before it if not NULL.
7375         (combine_chains): Process refs reversely and compute dominance point
7376         for root ref.
7378         Revert
7379         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7381         PR tree-optimization/79663
7382         * tree-predcom.c (combine_chains): Process refs in reverse order
7383         only for ZERO length chains, and add explaining comment.
7385 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7387         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7388         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7389         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7391 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7393         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7395 2017-11-15  Martin Liska  <mliska@suse.cz>
7397         * tree-cfg.c (pass_warn_function_return::execute):
7398         Compare warn_return_type for greater than zero.
7400 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7402         PR target/82941
7403         PR target/82942
7404         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7405         to return true on Xeon and not on Xeon Phi.
7406         (ix86_check_avx256_register): Changed to ...
7407         (ix86_check_avx_upper_register): ... this. Add extra check for
7408         VALID_AVX512F_REG_OR_XI_MODE.
7409         (ix86_avx_u128_mode_needed): Changed
7410         ix86_check_avx256_register to ix86_check_avx_upper_register.
7411         (ix86_check_avx256_stores): Changed to ...
7412         (ix86_check_avx_upper_stores): ... this. Changed
7413         ix86_check_avx256_register to ix86_check_avx_upper_register.
7414         (ix86_avx_u128_mode_after): Changed
7415         avx_reg256_found to avx_upper_reg_found. Changed
7416         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7417         (ix86_avx_u128_mode_entry): Changed
7418         ix86_check_avx256_register to ix86_check_avx_upper_register.
7419         (ix86_avx_u128_mode_exit): Ditto.
7420         * config/i386/i386.h: (host_detect_local_cpu): New define.
7422 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7424         * config/arm/xgene1.md (xgene1): Split into automatons
7425         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7426         (xgene1_f_load): Adjust reservations and/or types.
7427         (xgene1_f_store): Likewise.
7428         (xgene1_load_pair): Likewise.
7429         (xgene1_store_pair): Likewise.
7430         (xgene1_fp_load1): Likewise.
7431         (xgene1_load1): Likewise.
7432         (xgene1_store1): Likewise.
7433         (xgene1_move): Likewise.
7434         (xgene1_alu): Likewise.
7435         (xgene1_simd): Likewise.
7436         (xgene1_bfm): Likewise.
7437         (xgene1_neon_load1): Likewise.
7438         (xgene1_neon_store1): Likewise.
7439         (xgene1_neon_logic): Likewise.
7440         (xgene1_neon_st1): Likewise.
7441         (xgene1_neon_ld1r): Likewise.
7442         (xgene1_alu_cond): Added.
7443         (xgene1_shift_reg): Likwise.
7444         (xgene1_bfx): Likewise.
7445         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7447 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7449         PR target/82981
7450         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7451         ssa-iterators.h.
7452         (can_widen_mult_without_libcall): New function.
7453         (expand_mul_overflow): If only checking unsigned mul overflow,
7454         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7455         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7456         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7457         (expand_DIVMOD): Formatting fix.
7458         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7459         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7460         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7462         PR tree-optimization/82977
7463         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7464         constructed temporary to strlen_to_stridx.put.
7466 2017-11-15  Martin Liska  <mliska@suse.cz>
7468         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7469         * configure: Regenerate.
7471 2017-11-15  Martin Liska  <mliska@suse.cz>
7473         PR target/82927
7474         * config/sh/sh-mem.cc: Use proper probability for
7475         REG_BR_PROB_NOTE.
7477 2017-11-14  Jeff Law  <law@redhat.com>
7479         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7480         the red zone for stack_clash_protection_final_dynamic_probe targets
7481         when the total dynamic stack size is zero bytes.
7483         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7484         blocks is post order.
7486 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7488         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7489         * final.c (rest_of_clean_state): Set it for the
7490         -fcompare-debug dump.
7491         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7492         class when TDF_COMPARE_DEBUG is set.
7494         * dwarf2out.c (gen_producer_string): Discard
7495         OPT_fcompare_debug.
7497 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7499         PR c/81156
7500         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7501         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7502         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7503         (__TGMATH_REAL_2_3): Remove macros.
7504         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7505         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7506         __builtin_tgmath.
7507         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7508         __TGMATH_REAL_2.
7509         (remquo): Define using __TGMATH_REAL_3.
7511 2017-11-14  Jeff Law  <law@redhat.com>
7513         * vr-values.c: New file with contents extracted from tree-vrp.c.
7514         * Makefile.in (OBJS): Add vr-values.o
7515         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7516         (set_value_range, set_and_canonicalize_value_range): Likewise.
7517         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7518         (value_range_constant_singleton, symbolic_range_p): Likewise.
7519         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7520         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7521         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7522         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7523         (find_case_label_range, find_case_label_index): Likewise.
7524         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7525         (range_int_cst_singleton_p, value_inside_range): Likewise.
7526         (get_single_symbol): Likewise.
7527         (switch_update): Move structure definition here.
7528         (to_remove_edges, to_update_switch_stmts): Provide externs.
7529         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7530         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7531         (vrp_val_is_min, set_value_range): Likewise.
7532         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7533         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7534         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7535         (range_is_nonnull, range_int_cst_p): Likewwise.
7536         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7537         (get_single_symbol, operand_less_p): Likewise
7538         (compare_values_warnv, compare_values): Likewise.
7539         (value_inside_range, value_range_constant_singleton): Likewise.
7540         (zero_nonzero_bitgs_from_vr): Likewise.
7541         (extract_range_from_binary_expr_1): Likewise.
7542         (overflow_comparison_p): Likewise.
7543         (to_remove_edges, to_update_switch_stmts): Likewise.
7544         (find_case_label-index, find_case_label_range): Likewise.
7545         (switch_update, set_value_range_to_nonnegative): Remove.
7546         (set_value_range_to_truthvalue): Likewise.
7547         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7548         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7549         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7550         (find_case_label_ranges, test_for_singularity): Likewise.
7551         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7552         (x_vr_values): Move to its remaining use site.
7554 2017-11-10  Jeff Law  <law@redhat.com>
7556         * vr-values.h (VR_INITIALIZER): Move #define here.
7557         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7558         * Makefile.in (OBJS): Add tree-evrp.o
7559         * tree-vrp.h (assert_info): Move structure definition here.
7560         (set_value_range_to_varying): Prototype.
7561         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7562         (infer_value_range, register_edge_assert_for): Likewise.
7563         (stmt_interesting_for_vrp): Likewise.
7564         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7565         (set_value_range_to_varying): No longer static.
7566         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7567         (infer_value_range, register_edge_assert_for): Likewise.
7569 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7571         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7572         generate the XXBRD instruction.
7574         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7575         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7576         as being compatible if -mabi=ieeelongdouble.
7577         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7578         to setup float128 built-ins with hardware support.
7579         (BU_FLOAT128_HW_2): Likewise.
7580         (BU_FLOAT128_HW_3): Likewise.
7581         (BU_FLOAT128_HW_VSX_1): Likewise.
7582         (BU_FLOAT128_HW_VSX_2): Likewise.
7583         (scalar_extract_expq): Change float128 built-in functions to
7584         accommodate having both KFmode and TFmode functions.  Use the
7585         KFmode variant as the default.
7586         (scalar_extract_sigq): Likewise.
7587         (scalar_test_neg_qp): Likewise.
7588         (scalar_insert_exp_q): Likewise.
7589         (scalar_insert_exp_qp): Likewise.
7590         (scalar_test_data_class_qp): Likewise.
7591         (sqrtf128_round_to_odd): Delete processing the round to odd
7592         built-in functions as special built-in functions, and define them
7593         as float128 built-ins.  Use the KFmode variant as the default.
7594         (truncf128_round_to_odd): Likewise.
7595         (addf128_round_to_odd): Likewise.
7596         (subf128_round_to_odd): Likewise.
7597         (mulf128_round_to_odd): Likewise.
7598         (divf128_round_to_odd): Likewise.
7599         (fmaf128_round_to_odd): Likewise.
7600         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7601         support for KFmode and TFmode xststdcqp calls.
7602         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7603         point, switch the built-in handlers for the get/set float128
7604         exponent, get float128 mantissa, float128 test built-ins, and the
7605         float128 round to odd built-in functions.  Eliminate creating the
7606         float128 round to odd built-in functions as special built-ins.
7607         (rs6000_init_builtins): Eliminate special creation of the float128
7608         round to odd built-in functions.
7609         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7610         function insns to support both TFmode and KFmode variants.
7611         (xsxsigqp_<mode>): Likewise.
7612         (xsiexpqpf_<mode>): Likewise.
7613         (xsiexpqp_<mode>): Likewise.
7614         (xststdcqp_<mode>): Likewise.
7615         (xststdcnegqp_<mode>): Likewise.
7616         (xststdcqp_<mode>): Likewise.
7618 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7620         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7621         unused path_in_freq_ptr parameter.
7622         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7624 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7626         * ipa-inline.c (edge_badness): Dump sreal frequency.
7627         (compute_inlined_call_time): Match natural implementaiton ...
7628         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7629         forgotten division by CGRAPH_FREQ_BASE.
7631 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7633         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7634         Solaris 11.  Update comment.
7635         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7636         renaming.
7637         * config/sol2.h (STARTFILE_SPEC): Likewise.
7638         * configure: Regenerate.
7640 2017-11-14  Carl Love  <cel@us.ibm.com>
7642         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7643         le_ and be_ prefixes to swap* variables.  Remove
7644         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7645         statements.
7647 2017-11-14  Jason Merrill  <jason@redhat.com>
7649         Support GTY((cache)) on hash_map.
7650         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7651         (ggc_cache_remove): Override it instead of ggc_mx.
7652         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7653         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7654         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7655         (simple_cache_map_traits): Override maybe_mx.
7656         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7657         (hash_map): Friend gt_cleare_cache.
7658         (gt_cleare_cache): New.
7659         * tree.h (tree_cache_traits): New hash_map traits class.
7660         (tree_cache_map): New typedef.
7662 2017-11-14  Richard Biener  <rguenther@suse.de>
7664         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7665         paramter and handling.
7666         (cleanup_control_flow_bb): Likewise.
7667         (cleanup_control_flow_pre): New helper performing a DFS walk
7668         to call cleanup_control_flow_bb in PRE order.
7669         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7670         via cleanup_control_flow_pre.
7672 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7674         * config/aarch64/aarch64-simd.md
7675         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7676         (*aarch64_simd_bsl<mode>_alt): Likewise.
7677         (aarch64_simd_bsldi_internal): New.
7678         (aarch64_simd_bsldi_alt): Likewise.
7680 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7682         * tracer.c (better_p): Do not compare frequencies.
7683         * reg-stack.c (better_edge): Likewise.
7684         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7685         and back.
7687 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7689         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7690         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7691         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7692         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7693         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7694         * predict.c (maybe_hot_frequency_p): Inline to ...
7695         (maybe_hot_count_p): ... here; rewrite to counts.
7696         (counts_to_freqs): Rename to ...
7697         (update_max_bb_count): ... this one.
7698         (expensive_function_p): Use counts.
7699         (estimate_bb_frequencies): Update.
7700         (rebuild_frequencies): Update.
7701         * predict.h (counts_to_freqs): Rename to ...
7702         (update_max_bb_count): ... this one.
7703         * profile.c (compute_branch_probabilities): Add debug info
7704         * tree-inline.c (expand_call_inline): Update debug info.
7705         (optimize_inline_calls): Use update_max_bb_count..
7706         (tree_function_versioning): Use update_max_bb_count..
7707         * value-prof.c (gimple_value_profile_transformations):
7708         Do not use update_max_bb_count.
7710 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7712         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7713         always use frequencies.
7715 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7717         * bb-reorder.c: Remove frequencies from comments.
7718         (better_edge_p): Use profile counts.
7719         (find_traces): Dump profile counts.
7720         (rotate_loop): Use profile counts.
7721         (find_traces_1_round): Likewise.
7722         (connect_better_edge_p): Use counts instead of probabilities for
7723         reverse walk.
7724         (copy_bb_p): Drop early check for non-0 frequency.
7725         (sanitize_hot_paths): Update comments.
7727 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7729         * ipa-split.c (struct split_point): Add count.
7730         (consider_split): Do not compute incoming frequency; compute incoming
7731         count and store it to split_point.
7732         (split_function): Set count of the call to split part correctly.
7734 2017-11-13  Carl Love  <cel@us.ibm.com>
7736         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7738 2017-11-13  Tom Tromey  <tom@tromey.com>
7740         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7742 2017-11-13  Carl Love  <cel@us.ibm.com>
7744         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7745         Add support for builtins:
7746         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7747         vector {un,}signed {char,int,short},
7748         vector {un,}signed {char,int,short}) arguments.
7749         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7750         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7751         Add BU_P9V_AV_2 expansions for the builtins.
7752         * config/rs6000/altivec.h (vec_first_match_index,
7753         vec_first_mismatch_index, vec_first_match_or_eos_index,
7754         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7755         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7756         new extern declaration.
7757         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7758         * config/rs6000/vsx.md (first_match_index_<mode>,
7759         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7760         first_mismatch_or_eos_index_<mode>): Add define expand.
7761         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7762         * doc/extend.texi: Update the built-in documenation file for the new
7763         built-in functions.
7765 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7767         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7768         into the min/max operations for _Float<N> and _Float<N>X types.
7770 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7772         PR lto/81351
7773         * dwarf2out.c (do_eh_frame): New static variable.
7774         (dwarf2out_begin_prologue): Set it.
7775         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7777 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7779         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7781         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7783 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7785         PR lto/81351
7786         * debug.h (dwarf2out_do_eh_frame): Declare.
7787         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7788         (dwarf2out_do_frame): Use it.
7789         (dwarf2out_do_cfi_asm): Likewise.
7790         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7791         (dwarf2out_assembly_start): Likewise.
7792         (dwarf2out_begin_prologue): Fix comment.
7793         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7794         if the target needs either debug or unwind DWARF2 info.
7795         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7796         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7797         -ftrapping-math, -ftrapv and -fwrapv.
7799 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7801         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7802         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7803         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7804         (estimate_edge_size_and_time): Likewise.
7805         (ipa_merge_fn_summary_after_inlining): Likewise.
7806         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7807         (compute_uninlined_call_time): Use sreal_frequency.
7808         (compute_inlined_call_time): Likewise.
7809         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7810         * profile-count.c: Include sreal.h.
7811         (profile_count::to_sreal_scale): New.
7812         * profile-count.h: Forward declare sreal.
7813         (profile_count::to_sreal_scale): Declare.
7815 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7817         * diagnostic.c (maybe_line_and_column): New.
7818         (diagnostic_get_location_text): Use it.
7819         (diagnostic_report_current_module): Likewise.
7820         (test_diagnostic_get_location_text): Add tests.
7822 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7824         * doc/md.texi (Specifying processor pipeline description): Fix
7825         incorrect latency for the div instruction example.
7827 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7829         PR tree-optimization/78821
7830         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7831         that bit_not_p is the same.
7832         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7833         (split_group): Count precisely bit_not_p bits in each statement.
7834         (invert_op): New function.
7835         (imm_store_chain_info::output_merged_store): Use invert_op to
7836         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7837         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7839 2017-11-13  Martin Liska  <mliska@suse.cz>
7841         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7842         (struct source_info): Likewise.
7843         (add_branch_counts): Likewise.
7844         (add_line_counts): Likewise.
7845         (function_summary): Likewise.
7846         (output_intermediate_line): Likewise.
7847         (generate_results): Likewise.
7849 2017-11-13  Martin Liska  <mliska@suse.cz>
7851         * gcov.c (struct block_info): Remove typedef for block_t.
7852         (struct line_info): Likewise.
7853         (line_info::has_block): Likewise.
7854         (EXIT_BLOCK): Likewise.
7855         (unblock): Likewise.
7856         (circuit): Likewise.
7857         (get_cycles_count): Likewise.
7858         (process_file): Likewise.
7859         (read_graph_file): Likewise.
7860         (solve_flow_graph): Likewise.
7861         (find_exception_blocks): Likewise.
7862         (add_line_counts): Likewise.
7863         (accumulate_line_info): Likewise.
7864         (output_line_details): Likewise.
7866 2017-11-13  Martin Liska  <mliska@suse.cz>
7868         * gcov.c (struct arc_info): Remove typedef for arc_t.
7869         (struct line_info): Likewise.
7870         (add_branch_counts): Likewise.
7871         (output_branch_count): Likewise.
7872         (function_info::~function_info): Likewise.
7873         (circuit): Likewise.
7874         (output_intermediate_line): Likewise.
7875         (read_graph_file): Likewise.
7876         (solve_flow_graph): Likewise.
7877         (find_exception_blocks): Likewise.
7878         (add_line_counts): Likewise.
7879         (accumulate_line_info): Likewise.
7880         (output_line_details): Likewise.
7881         (output_function_details): Likewise.
7883 2017-11-13  Martin Liska  <mliska@suse.cz>
7885         * gcov.c (struct function_info): Remove typedef for function_t.
7886         (struct source_info): Likewise.
7887         (source_info::get_functions_at_location): Likewise.
7888         (solve_flow_graph): Likewise.
7889         (find_exception_blocks): Likewise.
7890         (add_line_counts): Likewise.
7891         (output_intermediate_file): Likewise.
7892         (process_file): Likewise.
7893         (generate_results): Likewise.
7894         (release_structures): Likewise.
7895         (read_graph_file): Likewise.
7896         (read_count_file): Likewise.
7897         (accumulate_line_counts): Likewise.
7898         (output_lines): Likewise.
7900 2017-11-13  Martin Liska  <mliska@suse.cz>
7902         * gcov.c (function_info::function_info): Remove num_counts
7903         and add vector<gcov_type>.
7904         (function_info::~function_info): Use the vector.
7905         (process_file): Likewise.
7906         (read_graph_file): Likewise.
7907         (read_count_file): Likewise.
7908         (solve_flow_graph): Likewise.
7910 2017-11-13  Martin Liska  <mliska@suse.cz>
7912         * gcov.c (function_info::is_artificial): New function.
7913         (process_file): Erase all artificial early.
7914         (generate_results): Skip as all artificial are already
7915         removed.
7917 2017-11-13  Martin Liska  <mliska@suse.cz>
7919         * gcov.c (read_graph_file): Store to global vector of functions.
7920         (read_count_file): Iterate the vector.
7921         (process_file): Likewise.
7922         (generate_results): Likewise.
7923         (release_structures): Likewise.
7925 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7927         PR tree-optimization/82954
7928         * gimple-ssa-store-merging.c
7929         (imm_store_chain_info::coalesce_immediate_stores): If
7930         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7932 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
7934         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7935         Upddate call to ENDIAN_LANE_N.
7936         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7937         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7938         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7939         and use aarch64_endian_lane_rtx.
7940         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
7942 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7944         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
7946 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7948         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
7950 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7952         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
7953         body.
7954         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7955         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7956         * defaults.h (ASM_OUTPUT_LABELREF): Same.
7958 2017-11-11  Martin Sebor  <msebor@redhat.com>
7960         PR c/81117
7961         * doc/extend.texi (attribute nonstring): Remove spurious argument.
7963         PR bootstrap/82948
7964         * prefic.c (translate_name): Replace strncpy with memcpy to
7965         avoid -Wstringop-truncation.
7967 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7969         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
7971 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7973         * predict.c (maybe_hot_frequency_p): Do not use cfun.
7975 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7977         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
7978         merging.
7980 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
7982         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
7983         enable generating XXBRH if the value is in a vector register.
7984         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
7985         value is in a vector register.
7986         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
7987         register to register bswap64's instead of doing the GPR sequence
7988         used on previous machines.
7989         (bswapdi2_xxbrd): New insn.
7990         (bswapdi2_reg): Disallow on ISA 3.0.
7991         (register to register bswap64 splitter): Do not split the insn on
7992         ISA 3.0 systems that use XXBRD.
7994 2017-11-10  Martin Sebor  <msebor@redhat.com>
7996         PR c/81117
7997         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
7998         with memcpy.
7999         (find_subframework_file): Same.
8001 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8003         * auto-profile.c (afdo_indirect_call): Drop frequency.
8004         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
8005         (cgraph_node::create_edge): Drop frequency argument.
8006         (cgraph_node::create_indirect_edge): Drop frequency argument.
8007         (cgraph_edge::make_speculative): Drop frequency arguments.
8008         (cgraph_edge::resolve_speculation): Do not update frequencies
8009         (cgraph_edge::dump_edge_flags): Do not dump frequency.
8010         (cgraph_node::dump): Check consistency in IPA mode.
8011         (cgraph_edge::maybe_hot_p): Use IPA counter.
8012         (cgraph_edge::verify_count_and_frequency): Rename to ...
8013         (cgraph_edge::verify_count): ... this one; drop frequency checking.
8014         (cgraph_node::verify_node): Update.
8015         * cgraph.h (struct cgraph_edge): Drop frequency.
8016         (cgraph_edge::frequency): New function.
8017         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
8018         frequencies.
8019         (cgraph_edge::rebuild_edges): Likewise.
8020         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
8021         (duplicate_thunk_for_node): Do not pass frequency.
8022         (cgraph_node::create_clone): Scale only counts.
8023         (cgraph_node::create_virtual_clone): Do not pass frequency.
8024         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8025         (cgraph_node::create_version_clone): Do not pass frequency.
8026         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8027         (cgraph_node::expand_thunk): Do not pass frequency.
8028         (cgraph_node::create_wrapper): Do not pass frequency.
8029         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8030         frequency.
8031         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8032         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8033         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8034         (gather_caller_stats): Use frequency function.
8035         (ipcp_cloning_candidate_p): Use frequency function.
8036         (ipcp_propagate_stage): Use frequency function.
8037         (get_info_about_necessary_edges): Use frequency function.
8038         (update_profiling_info): Update only IPA profile.
8039         (update_specialized_profile): Use frequency functoin.
8040         (perhaps_add_new_callers): Update only IPA profile.
8041         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8042         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8043         (dump_ipa_call_summary): Use frequency function.
8044         (estimate_edge_size_and_time): Use frequency function.
8045         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8046         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8047         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8048         (update_noncloned_counts): ... ths one; scale counts only.
8049         (clone_inlined_nodes): Do not scale frequency.
8050         (inline_call): Do not pass frequency.
8051         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8052         (compute_inlined_call_time): Use IPA profile.
8053         (want_inline_small_function_p): Use IPA profile.
8054         (want_inline_self_recursive_call_p): Use IPA profile.
8055         (edge_badness): Use IPA profile.
8056         (lookup_recursive_calls): Use IPA profile.
8057         (recursive_inlining): Do not pass frequency.
8058         (resolve_noninline_speculation): Do not update frequency.
8059         (inline_small_functions): Collect max of IPA profile.
8060         (dump_overall_stats): Dump IPA porfile.
8061         (dump_inline_stats): Dump IPA porfile.
8062         (ipa_inline): Collect IPA stats.
8063         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8064         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8065         (ipa_propagate_frequency): Use frequency function.
8066         (ipa_profile): Cleanup.
8067         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8068         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8069         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8070         (input_node): Do not stream frequency.
8071         (input_edge): Do not stream frequency.
8072         (merge_profile_summaries): Scale only IPA profiles.
8073         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8074         * predict.c (drop_profile): Do not recompute frequency.
8075         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8076         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8077         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8078         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8079         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8080         * tree-inline.c (copy_bb): Do not scale frequency.
8081         (expand_call_inline): Do not scale frequency.
8082         (tree_function_versioning): Do not scale frequency.
8083         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8085 2017-11-10  Julia Koval  <julia.koval@intel.com>
8087         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8088         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8089         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8090         (_mm256_maskz_gf2p8affine_epi64_epi8)
8091         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8092         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8093         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8094         (__builtin_ia32_vgf2p8affineqb_v32qi)
8095         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8096         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8098 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8100         PR target/82641
8101         * config/arm/arm.c
8102         (arm_option_override): Refactor.
8103         (arm_option_reconfigure_globals): New.
8104         (arm_options_perform_arch_sanity_checks): New.
8105         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8106         New prototype.
8107         (arm_options_perform_arch_sanity_checks): Likewise
8109 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8111         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8112         (power9-qpmul): New.
8113         * rs6000/rs6000.md ("type" attr): Add qmul.
8114         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8115         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8116         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8118 2017-11-10  Martin Sebor  <msebor@redhat.com>
8120         PR c/81117
8121         * builtins.c (compute_objsize): Handle arrays that
8122         compute_builtin_object_size likes to fail for.  Make extern.
8123         * builtins.h (compute_objsize): Declare.
8124         (check_strncpy_sizes): New function.
8125         (expand_builtin_strncpy): Call check_strncpy_sizes.
8126         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8127         -Wstringop-truncation.
8128         (gimple_fold_builtin_strncat): Same.
8129         * gimple.c (gimple_build_call_from_tree): Set call location.
8130         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8131         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8132         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8133         (handle_builtin_strlen): Use strlen_to_stridx.
8134         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8135         stpncpy.
8136         Use strlen_to_stridx.
8137         (pass_strlen::execute): Release strlen_to_stridx.
8138         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8139         (-Wstringop-truncation): Document new option.
8141 2017-11-10  Martin Liska  <mliska@suse.cz>
8143         PR gcov-profile/82702
8144         * gcov.c (main): Handle intermediate files in a different
8145         way.
8146         (get_gcov_intermediate_filename): New function.
8147         (output_gcov_file): Remove support of intermediate files.
8148         (generate_results): Allocate intermediate file.
8149         (release_structures): Clean-up properly fn_end.
8150         (output_intermediate_file): Start iterating with line 1.
8152 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8154         PR tree-optimization/82929
8155         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8156         ops_swapped_p non-static data member.
8157         (store_immediate_info::store_immediate_info): Clear it.
8158         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8159         ops set ops_swapped_p.
8160         (count_multiple_uses): Handle ops_swapped_p.
8162 2017-11-10  Martin Liska  <mliska@suse.cz>
8164         * coverage.c (coverage_init): Stream information about
8165         support of has_unexecuted_blocks.
8166         * doc/gcov.texi: Document that.
8167         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8168         * gcov.c (read_graph_file): Likewise.
8169         (output_line_beginning): Fix a small issue with
8170         color output.
8172 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8174         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8175         reference of trivial component.
8177 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8179         PR bootstrap/82916
8180         * gimple-ssa-store-merging.c
8181         (pass_store_merging::terminate_all_aliasing_chains): For
8182         gimple_store_p stmts also call refs_output_dependent_p.
8184         PR rtl-optimization/82913
8185         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8186         single set.
8188 2017-11-09  Jeff Law  <law@redhat.com>
8190         * vr-values.h: New file with vr_values class.
8191         * tree-vrp.c: Include vr-values.h
8192         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8193         data objects into the vr_values class.
8194         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8195         (get_value_range): Make it a member function within vr_values class.
8196         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8197         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8198         (op_with_constant_singleton_value_range): Likewise.
8199         (extract_range_for_var_from_comparison_expr): Likewise.
8200         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8201         (extract_range_from_binary_expr): Likewise.
8202         (extract_range_from_unary_expr): Likewise.
8203         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8204         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8205         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8206         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8207         (compare_name_with_value, compare_names): Likewise.
8208         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8209         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8210         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8211         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8212         (extract_range_from_phi_node): Likewise.
8213         (simplify_truth_ops_using_ranges): Likewise.
8214         (simplify_div_or_mod_using_ranges): Likewise.
8215         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8216         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8217         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8218         (simplify_float_conversion_using_ranges): Likewise.
8219         (simplify_internal_call_using_ranges): Likewise.
8220         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8221         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8222         poitner via x_vr_values for calls into gimple folder.
8223         (vrp_initialize_lattice): Make this the vr_values ctor.
8224         (vrp_free_lattice): Make this the vr_values dtor.
8225         (set_vr_value): New function.
8226         (class vrp_prop): Add vr_values data member.  Add various member
8227         functions as well as member functions that delegate to vr_values.
8228         (check_array_ref): Make a member function within vrp_prop class.
8229         (search_for_addr_array, vrp_initialize): Likewise.
8230         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8231         vr_value, values_propagated, etc.
8232         (check_array_bounds): Extract vrp_prop class instance pointer from
8233         walk info structure.  Use it to call member functions.
8234         (check_all_array_refs): Make a member function within vrp_prop class.
8235         Smuggle class instance pointer via walk info structure.
8236         (x_vr_values): New local static.
8237         (vrp_valueize): Use x_vr_values to get class instance.
8238         (vr_valueize_1): Likewise.
8239         (class vrp_folder): Add vr_values data member.  Add various member
8240         functions as well as member functions that delegate to vr_values.
8241         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8242         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8243         class instance from vr_values.  Use it to call member functions.
8244         (vrp_dom_walker): Add vr_values data member.
8245         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8246         instance via x_vr_values.
8247         (identify_jump_threads): Accept vr_values as argument.  Store
8248         it into the walker structure.
8249         (evrp_dom_walker): Add vr_values class data member.  Add various
8250         delegators.
8251         (evrp_dom_walker::try_find_new_range): Use vr_values data
8252         member to access the memory allocator.
8253         (evrp_dom_walker::before_dom_children): Store vr_values class
8254         instance into the vrp_folder class.
8255         (evrp_dom_walker::push_value_range): Rework to avoid direct
8256         access to num_vr_values and vr_value.
8257         (evrp_dom_walker::pop_value_range): Likewise.
8258         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8259         Use vr_values to get to dump_all_value_ranges member function.
8260         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8261         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8262         Pass vr_values class instance down to identify_jump_threads.
8263         Remove call to vrp_free_lattice.
8264         (debug_all_value_ranges): Remove.
8266         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8267         (vrp_folder): Likewise.
8269         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8270         Get it from the existing bitmap instead.
8271         (vrp_intersect_ranges_1): Likewise.
8273 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8275         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8276         bit_not_p field.
8277         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8278         set bit_not_p to it.
8279         (imm_store_chain_info::coalesce_immediate_stores): Break group
8280         if bit_not_p is different.
8281         (count_multiple_uses, split_group,
8282         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8283         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8284         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8285         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8286         to store_immediate_info ctor.
8288 2017-11-09  Jim Wilson  <jimw@sifive.com>
8290         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8291         (scan_prog_file): Likewise.
8293 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8295         * bb-reorder.c (max_entry_frequency): Remove.
8296         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8297         connect_traces, push_to_next_round_p): Remove prototypes.
8298         (find_traces_1_round): Use counts only.
8299         (push_to_next_round_p): Likewise.
8300         (find_traces): Likewise.
8301         (rotate_loop): Likewise.
8302         (find_traces_1_round): Likewise.
8303         (connect_traces): Likewise.
8304         (edge_order): Likewise.
8306 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8308         * config/arm/arm.c (output_return_instruction): Add comments to
8309         indicate requirement for cmse_nonsecure_entry return to account
8310         for the size of clearing instruction output here.
8311         (thumb_exit): Likewise.
8312         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8313         return in hardfloat mode.
8315 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8317         * config/rs6000/rs6000.c (machine_function): Add a bool,
8318         "toc_is_wrapped_separately".
8319         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8320         if it wasn't explicitly set or unset, we are optimizing for speed, and
8321         doing separate shrink-wrapping.
8322         (rs6000_get_separate_components): Enable the TOC component if
8323         saving the TOC register in the prologue.
8324         (rs6000_components_for_bb): Handle the TOC component.
8325         (rs6000_emit_prologue_components): Store the TOC register where needed.
8326         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8327         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8329 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8331         * ipa-param-manipulation.c: New file.
8332         * ipa-param-manipulation.h: Likewise.
8333         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8334         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8335         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8336         (ipa_parm_adjustment): Likewise.
8337         (ipa_parm_adjustment_vec): Likewise.
8338         (ipa_get_vector_of_formal_parms): Moved declaration to
8339         ipa-param-manipulation.h.
8340         (ipa_get_vector_of_formal_parm_types): Likewise.
8341         (ipa_modify_formal_parameters): Likewise.
8342         (ipa_modify_call_arguments): Likewise.
8343         (ipa_combine_adjustments): Likewise.
8344         (ipa_dump_param_adjustments): Likewise.
8345         (ipa_modify_expr): Likewise.
8346         (ipa_get_adjustment_candidate): Likewise.
8347         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8348         ipa-param-manipulation.c.
8349         (ipa_get_vector_of_formal_parm_types): Likewise.
8350         (ipa_modify_formal_parameters): Likewise.
8351         (ipa_modify_call_arguments): Likewise.
8352         (ipa_modify_expr): Likewise.
8353         (get_ssa_base_param): Likewise.
8354         (ipa_get_adjustment_candidate): Likewise.
8355         (index_in_adjustments_multiple_times_p): Likewise.
8356         (ipa_combine_adjustments): Likewise.
8357         (ipa_dump_param_adjustments): Likewise.
8358         * tree-sra.c: Also include ipa-param-manipulation.h
8359         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8360         ipa-param.h.
8362 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8363             Alan Hayward  <alan.hayward@arm.com>
8364             David Sherwood  <david.sherwood@arm.com>
8366         * doc/sourcebuild.texi (vect_masked_store): Document.
8368 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8369             Alan Hayward  <alan.hayward@arm.com>
8370             David Sherwood  <david.sherwood@arm.com>
8372         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8374 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8375             Alan Hayward  <alan.hayward@arm.com>
8376             David Sherwood  <david.sherwood@arm.com>
8378         * doc/sourcebuild.texi (vect_variable_length): Document.
8380 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8381             Alan Hayward  <alan.hayward@arm.com>
8382             David Sherwood  <david.sherwood@arm.com>
8384         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8386 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8387             Alan Hayward  <alan.hayward@arm.com>
8388             David Sherwood  <david.sherwood@arm.com>
8390         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8392 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8393             Alan Hayward  <alan.hayward@arm.com>
8394             David Sherwood  <david.sherwood@arm.com>
8396         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8397         previously undocumented selectors.
8398         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8400 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8402         * doc/rtl.texi (const_vector): Say that elements can be
8403         const_wide_ints too.
8404         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8405         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8406         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8407         * optabs.c (expand_vector_broadcast): Likewise.
8409 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8410             Alan Hayward  <alan.hayward@arm.com>
8411             David Sherwood  <david.sherwood@arm.com>
8413         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8414         scaled index even if the unscaled address was invalid.
8415         Don't increase the complexity of using a scale in that case.
8417 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8418             Alan Hayward  <alan.hayward@arm.com>
8419             David Sherwood  <david.sherwood@arm.com>
8421         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8422         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8423         UNITS_PER_WORD bytes.
8424         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8425         based on REGMODE_NATURAL_SIZE of the inner mode.
8426         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8427         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8428         * expmed.c (lowpart_bit_field_p): Divide the value up into
8429         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8430         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8431         whether something is likely to occupy more than one register.
8433 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8435         PR ipa/82879
8436         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8437         profile_count::adjust_for_ipa_scaling.
8438         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8439         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8440         function.
8441         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8443 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8445         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8446         (split_group): Add total_orig and total_new arguments, estimate the
8447         number of statements related to the store group without store merging
8448         and with store merging.
8449         (imm_store_chain_info::output_merged_store): Adjust split_group
8450         callers, punt if estimated number of statements with store merging
8451         is not smaller than estimated number of statements without it.
8452         Formatting fix.
8453         (handled_load): Remove has_single_use checks.
8454         (pass_store_merging::process_store): Likewise.
8456 2017-11-09  Richard Biener  <rguenther@suse.de>
8458         PR tree-optimization/82902
8459         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8461 2017-11-09  Martin Liska  <mliska@suse.cz>
8463         PR target/82863
8464         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8465         uninitialized.
8467 2017-11-09  Martin Liska  <mliska@suse.cz>
8469         PR tree-optimization/82669
8470         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8472 2017-11-09  Martin Liska  <mliska@suse.cz>
8474         PR gcov-profile/48463
8475         * coverage.c (coverage_begin_function): Output also end locus
8476         of a function and information whether the function is
8477         artificial.
8478         * gcov-dump.c (tag_function): Parse and print the information.
8479         * gcov.c (INCLUDE_MAP): Add include.
8480         (INCLUDE_SET): Likewise.
8481         (struct line_info): Move earlier in the source file because
8482         of vector<line_info> in function_info structure.
8483         (line_info::line_info): Likewise.
8484         (line_info::has_block): Likewise.
8485         (struct source_info): Add new member index.
8486         (source_info::get_functions_at_location): New function.
8487         (function_info::group_line_p): New function.
8488         (output_intermediate_line): New function.
8489         (output_intermediate_file): Use the mentioned function.
8490         (struct function_start): New.
8491         (struct function_start_pair_hash): Likewise.
8492         (process_file): Add code that identifies group functions.
8493         Assign lines either to global or function scope.
8494         (generate_results): Skip artificial functions.
8495         (find_source): Assign index for each source file.
8496         (read_graph_file): Read new flag artificial and end_line.
8497         (add_line_counts): Assign it either to global of function scope.
8498         (accumulate_line_counts): Isolate core of the function to
8499         accumulate_line_info and call it for both function and global
8500         scope lines.
8501         (accumulate_line_info): New function.
8502         (output_line_beginning): Fix GNU coding style.
8503         (print_source_line): New function.
8504         (output_line_details): Likewise.
8505         (output_function_details): Likewise.
8506         (output_lines): Iterate both source (global) scope and function
8507         scope.
8508         (struct function_line_start_cmp): New class.
8509         * doc/gcov.texi: Reflect changes in documentation.
8511 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8513         PR debug/82837
8514         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8515         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8516         and similarly for not instead of neg.
8518 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8520         * config/i386/i386.opt: Add -mforce-indirect-call.
8521         * config/i386/predicates.md: Check for flag_force_indirect_call.
8522         * doc/invoke.texi: Document -mforce-indirect-call
8524 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8526         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8527         New extern.
8528         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8529         config/riscv/riscv.c (predict.h): New include.
8530         (riscv_slow_unaligned_access_p): No longer static.
8531         (riscv_block_move_straight): Add require.
8532         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8534 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8536         PR target/82855
8537         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8538         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8539         nonimmediate_operand predicate for operand 1 instead of
8540         register_operand.
8542 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8544         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8545         New pattern.
8546         * config/aarch64/constraints.md (Uml): New constraint.
8547         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8548         predicate.
8550 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8552         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8553         of two vec_duplicates into a vec_concat.
8555 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8557         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8558         Simplify vec_merge of vec_duplicate and vec_concat.
8559         * config/aarch64/constraints.md (Utq): New constraint.
8560         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8561         define_insn.
8563 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8565         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8566         Simplify vec_merge of vec_duplicate and const_vector.
8567         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8568         New predicate.
8569         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8570         mode iterator.  Update predicate on operand 1 to
8571         handle non-const_vec constants.  Delete constraints.
8572         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8574 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8576         PR tree-optimization/78821
8577         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8578         data member.
8579         (store_operand_info::store_operand_info): Initialize it to false.
8580         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8581         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8582         store in the group, and if chain_info is non-NULL, to ignore altogether
8583         that chain.
8584         (compatible_load_p): Fail if bit_not_p does not match.
8585         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8586         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8587         (pass_store_merging::process_store): Adjust
8588         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8589         call terminate_all_aliasing_chains newly when adding a store into
8590         a chain with non-NULL chain_info.
8592 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8594         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8596 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8598         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8599         Remove.
8600         (aarch64_layout_frame): Initialise emit_frame_chain.
8601         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8602         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8604 2017-11-08  Martin Liska  <mliska@suse.cz>
8606         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8607         of gimple_call_internal_p.
8609 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8611         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8613 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8615         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8616         "do {} while (0)".
8618 2017-11-08  Martin Liska  <mliska@suse.cz>
8620         PR sanitizer/82792
8621         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8623 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8625         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8627 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8629         PR target/82855
8630         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8631         (*cmp<mode>_ccz_1): New insn with $k alternative.
8633         PR target/82855
8634         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8635         RTX_COMM_COMPARE as commutative as well.
8636         (ix86_binary_operator_ok): Formatting fix.
8637         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8638         *<code><mode>3<mask_name><round_saeonly_name>,
8639         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8640         *<s>mul<mode>3_highpart<mask_name>,
8641         *vec_widen_umult_even_v16si<mask_name>,
8642         *vec_widen_umult_even_v8si<mask_name>,
8643         *vec_widen_umult_even_v4si<mask_name>,
8644         *vec_widen_smult_even_v16si<mask_name>,
8645         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8646         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8647         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8648         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8649         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8650         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8651         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8652         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8653         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8654         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8655         ix86_binary_operator_ok.  Formatting fixes.
8656         (*<plusminus_insn><mode>3<mask_name><round_name>,
8657         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8658         fixes.
8660 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8662         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8663         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8664         eq and ne if TARGET_ISEL.
8665         (cmp): New code_iterator.
8666         (UNS, UNSU_, UNSIK): New code_attrs.
8667         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8668         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8669         to...
8670         ("eq<mode>3"): ... this.
8671         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8672         to...
8673         ("ne<mode>3"): ... this.
8675 2017-11-07  Julia Koval  <julia.koval@intel.com>
8677         PR target/82812
8678         * common/config/i386/i386-common.c
8679         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8680         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8681         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8682         * config/i386/i386.opt: Ditto.
8683         * config/i386/i386.c (ix86_target_string): Ditto.
8684         (ix86_option_override_internal): Ditto.
8685         (ix86_init_mpx_builtins): Move MPX to args2.
8686         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8687         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8688         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8689         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8690         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8691         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8692         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8694 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8696         PR target/80425
8697         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8698         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8699         (zero-extendsidi peephole2): Remove peephole.
8701 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8703         PR c/53037
8704         * stor-layout.c: Include attribs.h.
8705         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8706         explicit lookup of "aligned" attribute.
8708 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8710         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8711         (riscv_expand_block_move): Likewise.
8712         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8713         implementation.
8714         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8715         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8716         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8717         (riscv_adjust_block_mem): Likewise.
8718         (riscv_block_move_loop): Likewise.
8719         (riscv_expand_block_move): Likewise.
8720         * config/riscv/riscv.md (movmemsi): New pattern.
8722 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8724         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8725         (MUSL_DYNAMIC_LINKER): Likewise.
8727 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8729         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8730         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8731         nonmmory_operand.
8733 2017-11-07  Richard Biener  <rguenther@suse.de>
8735         * match.pd: Fix build.
8737 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8738             Jackson Woodruff  <jackson.woodruff@arm.com>
8740         PR tree-optimization/71026
8741         * match.pd: Canonicalize negate in division.
8743 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8745         PR middle-end/80131
8746         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8748 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8750         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8751         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8753 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8755         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8756         non-scalar integral types.
8757         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8758         (-(A-B), -(~A)): New transformations.
8760 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8762         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8763         semicolon after "do {} while (0)".
8764         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8765         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8766         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8767         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8768         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8770 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8772         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8773         after "do {} while (0)".
8774         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8775         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8776         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8777         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8778         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8780 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8782         PR other/82784
8783         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8784         "while {} do (0)".
8785         (arm_rtx_costs_internal): Add missing semicolon after
8786         HANDLE_NARROW_SHIFT_ARITH call.
8788 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8790         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8791         disable isel if it was not set explicitly.
8793 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8795         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8796         (add_type_attribute) likewise.
8798 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8800         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8801         of struct ix86_frame.
8802         (ix86_initial_elimination_offset): Likewise.
8803         (ix86_expand_split_stack_prologue): Likewise.
8805 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8807         * tree-vrp.h (enum value_range_type): Update stale comment.
8809 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8810             Alan Hayward  <alan.hayward@arm.com>
8811             David Sherwood  <david.sherwood@arm.com>
8813         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8814         (aarch64_expand_vec_perm_const): Take the number of units too.
8815         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8816         (aarch64_expand_vec_perm_const): Likewise.
8817         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8818         (vec_perm<mode>): Update accordingly.
8820 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8821             Alan Hayward  <alan.hayward@arm.com>
8822             David Sherwood  <david.sherwood@arm.com>
8824         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8825         Take the number of units too.
8826         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8827         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8828         but check for a vector mode before rather than after the call.
8829         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8830         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8831         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8832         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8833         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8834         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8835         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8836         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8837         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8838         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8839         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8840         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8841         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8842         (aarch64_sqdmull2_n<mode>): Update accordingly.
8844 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8845             Alan Hayward  <alan.hayward@arm.com>
8846             David Sherwood  <david.sherwood@arm.com>
8848         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8849         the number of units too.
8850         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8851         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8852         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8853         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8854         (vec_store_lanesxi<mode>): Update accordingly.
8856 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8857             Alan Hayward  <alan.hayward@arm.com>
8858             David Sherwood  <david.sherwood@arm.com>
8860         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8861         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8862         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8863         of units rather than the mode.
8864         * config/aarch64/iterators.md (nunits): New mode attribute.
8865         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8866         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8867         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8868         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8869         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8870         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8871         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8872         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8873         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8874         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8875         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8876         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8877         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8878         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8879         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8880         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8881         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8882         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8883         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8884         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8885         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8886         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8887         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8888         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8889         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8890         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8891         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8892         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8893         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8894         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8895         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8896         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8897         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8898         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8899         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8900         (aarch64_simd_vec_setv2di): Likewise.
8902 2017-11-06  Carl Love  <cel@us.ibm.com>
8904         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8905         definitions.
8906         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8907         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8908         to power 8.
8909         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8910         extern declaration.
8911         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8912         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8913         Add power 8 macro expansions.
8914         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8915         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8916         power 8 instructions.  (VSX_XXBR): Add iterator.
8918 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8920         * config/arm/arm.md (predicable_short_it): Change default to "no",
8921         improve documentation, remove uses that are identical to the default.
8922         (enabled_for_depr_it): Rename to enabled_for_short_it.
8923         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8924         Remove default uses.
8925         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8926         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8927         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8928         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8930 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8932         PR target/82748
8933         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8934         float128 helper macros, which are no longer used after deleting
8935         the old 'q' built-in functions, and moving the round to odd
8936         built-in functions to being special built-in functions.
8937         (BU_FLOAT128_2): Likewise.
8938         (BU_FLOAT128_1_HW): Likewise.
8939         (BU_FLOAT128_2_HW): Likewise.
8940         (BU_FLOAT128_3_HW): Likewise.
8941         (FABSQ): Delete old 'q' built-in functions.
8942         (COPYSIGNQ): Likewise.
8943         (SQRTF128_ODD): Move round to odd built-in functions to be
8944         special built-in functions, so that we can handle
8945         -mabi=ieeelongdouble.
8946         (TRUNCF128_ODD): Likewise.
8947         (ADDF128_ODD): Likewise.
8948         (SUBF128_ODD): Likewise.
8949         (MULF128_ODD): Likewise.
8950         (DIVF128_ODD): Likewise.
8951         (FMAF128_ODD): Likewise.
8952         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
8953         built-in names to 'f128'.
8954         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
8955         old 'q' built-in functions, as the machine independent code for
8956         'f128' built-in functions handles this.
8957         (rs6000_expand_builtin): Add expansion for float128 round to odd
8958         functions, keying off on -mabi=ieeelongdouble of whether to use
8959         the KFmode or TFmode variant.
8960         (rs6000_init_builtins): Initialize the _Float128 round to odd
8961         built-in functions.
8962         * doc/extend.texi (PowerPC Built-in Functions): Document the old
8963         _Float128 'q' built-in functions are now mapped into the new
8964         'f128' built-in functions.
8966 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
8968         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
8969         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
8971 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
8973         PR jit/82826
8974         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
8975         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
8976         * toplev.c: Include "ipa-fnsummary.h".
8977         (toplev::finalize): Call ipa_fnsummary_c_finalize.
8979 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
8981         PR tree-optimization/82838
8982         * gimple-ssa-store-merging.c
8983         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
8984         on a separate gimple_seq which is then appended to seq.
8986 2017-11-06  Jeff Law  <law@redhat.com>
8988         PR target/82788
8989         * config/i386/i386.c (PROBE_INTERVAL): Remove.
8990         (get_probe_interval): New functions.
8991         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
8992         (ix86_adjust_stack_and_probe): Likewise.
8993         (output_adjust_stack_and_probe): Likewise.
8994         (ix86_emit_probe_stack_range): Likewise.
8995         (ix86_expand_prologue): Likewise.
8997 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8999         PR tree-optimization/82816
9000         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
9001         if the modes of the two types are the same.
9002         (convert_plusminus_to_widen): Likewise.
9004 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9006         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
9007         p9_vadu<mode>3.
9008         (usadv16qi): New define_expand.
9009         (usadv8hi): New define_expand.
9011 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
9013         PR bootstrap/82832
9014         * ipa-inline-transform.c (update_noncloned_frequencies): Always
9015         scale.
9016         (inline_transform): Likewise.
9017         * predict.c (counts_to_freqs): Remove useless conditional.
9018         * profile-count.h (profile_count::apply_scale): Move sanity check.
9019         * tree-inline.c (copy_bb): Always scale.
9020         (copy_cfg_body): Likewise.
9022 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9024         PR target/67591
9025         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9026         attribute.
9027         (*cmp_ite0): Add enabled_for_depr_it attribute.
9028         (*cmp_ite1): Likewise.
9030 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9032         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9033         TYPE_MFCRF.
9035 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9037         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9038         return the value by pointer.
9039         (extract_range_from_multiplicative_op_1): Update accordingly.
9040         Return as soon as an operation fails.
9042 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9044         PR other/82784
9045         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9046         (DEF_SANITIZER_BUILTIN): ... here.
9047         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9048         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9050 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9052         PR other/82784
9053         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9054         macro body.
9055         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9056         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9057         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9058         after macro body.
9059         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9060         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9061         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9063 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9065         PR other/82784
9066         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9067         "do {} while (0)".
9069 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9071         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9072         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9073         (adddi3): Likewise.
9074         (*addsi3_extended): Likewise.
9075         (*addsi3_extended2): Likewise.
9076         (<optab>si3): Likewise.
9077         (<optab>di3): Likewise.
9078         (<optab><mode>3): Likewise.
9079         (<*optabe>si3_internal): Likewise.
9080         (zero_extendqi<SUPERQI:mode>2): Likewise.
9081         (*add<mode>hi3): Likewise.
9082         (*xor<mode>hi3): Likewise.
9083         (<optab>di3): Likewise.
9084         (*<optab>si3_extend): Likewise.
9085         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9086         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9087         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9089 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9091         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9092         TARGET_STRICT_ALIGN based upon -mtune argument.
9094 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9096         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9098 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9100         * config/i386/i386.c (choose_basereg): Use optional scratch
9101         register and add assertion.
9102         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9103         needed, and don't allocate stack.
9104         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9105         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9106         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9108 2017-11-03  Jeff Law  <law@redhat.com>
9110         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9111         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9112         to probe at the start of a noreturn function.
9114 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9116         PR tree-optimization/78821
9117         * gimple-ssa-store-merging.c: Update the file comment.
9118         (MAX_STORE_ALIAS_CHECKS): Define.
9119         (struct store_operand_info): New type.
9120         (store_operand_info::store_operand_info): New constructor.
9121         (struct store_immediate_info): Add rhs_code and ops data members.
9122         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9123         and op1r arguments to the ctor, initialize corresponding data members.
9124         (struct merged_store_group): Add load_align_base and load_align
9125         data members.
9126         (merged_store_group::merged_store_group): Initialize them.
9127         (merged_store_group::do_merge): Update them.
9128         (merged_store_group::apply_stores): Pick the constant for
9129         encode_tree_to_bitpos from one of the two operands, or skip
9130         encode_tree_to_bitpos if neither operand is a constant.
9131         (class pass_store_merging): Add process_store method decl.  Remove
9132         bool argument from terminate_all_aliasing_chains method decl.
9133         (pass_store_merging::terminate_all_aliasing_chains): Remove
9134         var_offset_p argument and corresponding handling.
9135         (stmts_may_clobber_ref_p): New function.
9136         (compatible_load_p): New function.
9137         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9138         if there is overlap and rhs_code is not INTEGER_CST.  For
9139         non-overlapping stores terminate group if rhs is not mergeable.
9140         (get_alias_type_for_stmts): Change first argument from
9141         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9142         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9143         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9144         alias type.
9145         (get_location_for_stmts): Change first argument from
9146         auto_vec<gimple *> & to vec<gimple *> &.
9147         (struct split_store): Remove orig_stmts data member, add orig_stores.
9148         (split_store::split_store): Create orig_stores rather than orig_stmts.
9149         (find_constituent_stmts): Renamed to ...
9150         (find_constituent_stores): ... this.  Change second argument from
9151         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9152         to info structures rather than the statements.
9153         (split_group): Rename ALLOW_UNALIGNED argument to
9154         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9155         it.  Adjust find_constituent_stores caller.
9156         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9157         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9158         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9159         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9160         (mem_valid_for_store_merging): New function.
9161         (handled_load): New function.
9162         (pass_store_merging::process_store): New method.
9163         (pass_store_merging::execute): Use process_store method.  Adjust
9164         terminate_all_aliasing_chains caller.
9166 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9168         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9169         Return true for more constants, symbols and label references.
9170         (aarch64_valid_floating_const): Remove unused function.
9172 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9174         PR target/82786
9175         * config/aarch64/aarch64.c (aarch64_layout_frame):
9176         Undo forcing of LR at bottom of frame.
9178 2017-11-03  Jeff Law  <law@redhat.com>
9180         PR target/82823
9181         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9182         for int_registers_saved.
9184         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9185         extracted from tree-ssa-dom.c.
9186         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9187         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9188         (record_equivalences_from_incoming_edge): Add additional argument
9189         to single_pred_edge_ignoring_loop_edges call.
9190         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9191         (uncprop_dom_walker::before_dom_children): Add additional argument
9192         to single_pred_edge_ignoring_loop_edges call.
9193         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9194         single_pred_edge_ignoring_loop_edges rather than open coding.
9195         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9197 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9199         * match.pd (-(-A)): Rewrite.
9201 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9203         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9204         (rs6000_emit_int_cmove): New declaration.
9205         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9206         (rs6000_emit_sISEL): Delete.
9207         (rs6000_emit_int_cmove): Make non-static.
9208         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9209         instead of rs6000_emit_sISEL.
9211 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9213         * asan.c (create_cond_insert_point): Maintain profile.
9214         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9215         merged.
9216         * basic-block.h (struct basic_block_def): Remove frequency.
9217         (EDGE_FREQUENCY): Use to_frequency
9218         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9219         heuristics.
9220         (find_traces): Update to use to_frequency.
9221         (find_traces_1_round): Likewise; use only IPA counts.
9222         (bb_to_key): Likewise.
9223         (connect_traces): Use IPA counts only.
9224         (copy_bb_p): Update to use to_frequency.
9225         (fix_up_crossing_landing_pad): Likewise.
9226         (sanitize_hot_paths): Likewise.
9227         * bt-load.c (basic_block_freq): Likewise.
9228         * cfg.c (init_flow): Set count_max to uninitialized.
9229         (check_bb_profile): Remove frequencies; check counts.
9230         (dump_bb_info): Do not dump frequencies.
9231         (update_bb_profile_for_threading): Update counts only.
9232         (scale_bbs_frequencies_int): Likewise.
9233         (MAX_SAFE_MULTIPLIER): Remove.
9234         (scale_bbs_frequencies_gcov_type): Update counts only.
9235         (scale_bbs_frequencies_profile_count): Update counts only.
9236         (scale_bbs_frequencies): Update counts only.
9237         * cfg.h (struct control_flow_graph): Add count-max.
9238         (update_bb_profile_for_threading): Update prototype.
9239         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9240         (find_many_sub_basic_blocks): Likewise.
9241         * cfgcleanup.c (try_forward_edges): Likewise.
9242         (try_crossjump_to_edge): Likewise.
9243         * cfgexpand.c (expand_gimple_cond): Likewise.
9244         (expand_gimple_tailcall): Likewise.
9245         (construct_init_block): Likewise.
9246         (construct_exit_block): Likewise.
9247         * cfghooks.c (verify_flow_info): Check consistency of counts.
9248         (dump_bb_for_graph): Do not dump frequencies.
9249         (split_block_1): Do not update frequencies.
9250         (split_edge): Do not update frequencies.
9251         (make_forwarder_block): Do not update frequencies.
9252         (duplicate_block): Do not update frequencies.
9253         (account_profile_record): Do not update frequencies.
9254         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9255         for global heuristics.
9256         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9257         (expected_loop_iterations_unbounded): Use counts only.
9258         * cfgloopmanip.c (scale_loop_profile): Simplify.
9259         (create_empty_loop_on_edge): Simplify
9260         (loopify): Simplify
9261         (duplicate_loop_to_header_edge): Simplify
9262         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9263         (update_br_prob_note): Take care of removing note when profile
9264         becomes undefined.
9265         (relink_block_chain): Do not dump frequency.
9266         (rtl_account_profile_record): Use to_frequency.
9267         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9268         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9269         (cgraph_update_edges_for_call_stmt_node): Likewise.
9270         (cgraph_edge::verify_count_and_frequency): Update.
9271         (cgraph_node::verify_node): Temporarily disable frequency verification.
9272         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9273         to_cgraph_frequency.
9274         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9275         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9276         frequencies.
9277         (cgraph_node::expand_thunk): Update profile.
9278         * except.c (dw2_build_landing_pads): Do not update frequency.
9279         * final.c (compute_alignments): Use to_frequency.
9280         (dump_basic_block_info): Do not dump frequency.
9281         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9282         (dump_gimple_bb_header): Do not dump frequency.
9283         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9284         do update count.
9285         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9286         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9287         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9288         (init_before_recovery): Do not update frequency.
9289         (sched_create_recovery_edges): Do not update frequency.
9290         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9291         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9292         (ipa_cp_c_finalize): Set max_count to uninitialized.
9293         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9294         (param_change_prob): Use counts.
9295         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9296         local profiles.
9297         * ipa-split.c (consider_split): Use to_frequency.
9298         (split_function): Use to_frequency.
9299         * ira-build.c (loop_compare_func): Likewise.
9300         (mark_loops_for_removal): Likewise.
9301         (mark_all_loops_for_removal): Likewise.
9302         * loop-doloop.c (doloop_modify): Do not update frequency.
9303         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9304         frequency.
9305         * lto-streamer-in.c (input_function): Update count_max.
9306         * omp-expand.c (expand_omp_taskreg): Update count_max.
9307         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9308         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9309         (maybe_hot_count_p): Use ipa counts only.
9310         (maybe_hot_bb_p): Simplify.
9311         (maybe_hot_edge_p): Simplify.
9312         (probably_never_executed): Do not take frequency argument.
9313         (probably_never_executed_bb_p): Do not pass frequency.
9314         (probably_never_executed_edge_p): Likewise.
9315         (combine_predictions_for_bb): Check that profile is nonzero.
9316         (propagate_freq): Do not set frequency.
9317         (drop_profile): Simplify.
9318         (counts_to_freqs): Simplify.
9319         (expensive_function_p): Use to_frequency.
9320         (propagate_unlikely_bbs_forward): Simplify.
9321         (determine_unlikely_bbs): Simplify.
9322         (estimate_bb_frequencies): Add hack to silence graphite issues.
9323         (compute_function_frequency): Use ipa counts.
9324         (pass_profile::execute): Update.
9325         (rebuild_frequencies): Use counts only.
9326         (force_edge_cold): Use counts only.
9327         * profile-count.c (profile_count::dump): Dump new count types.
9328         (profile_count::differs_from_p): Check compatiblity.
9329         (profile_count::to_frequency): New function.
9330         (profile_count::to_cgraph_frequency): New function.
9331         * profile-count.h (struct function): Declare.
9332         (enum profile_quality): Add profile_guessed_local and
9333         profile_guessed_global0.
9334         (class profile_proability): Decrease number of bits to 29;
9335         update from_reg_br_prob_note and to_reg_br_prob_note.
9336         (class profile_count: Update comment; decrease number of bits
9337         to 61. Check compatibility.
9338         (profile_count::compatible_p): New private member function.
9339         (profile_count::ipa_p): New member function.
9340         (profile_count::operator<): Handle global zero correctly.
9341         (profile_count::operator>): Handle global zero correctly.
9342         (profile_count::operator<=): Handle global zero correctly.
9343         (profile_count::operator>=): Handle global zero correctly.
9344         (profile_count::nonzero_p): New member function.
9345         (profile_count::force_nonzero): New member function.
9346         (profile_count::max): New member function.
9347         (profile_count::apply_scale): Handle IPA scalling.
9348         (profile_count::guessed_local): New member function.
9349         (profile_count::global0): New member function.
9350         (profile_count::ipa): New member function.
9351         (profile_count::to_frequency): Declare.
9352         (profile_count::to_cgraph_frequency): Declare.
9353         * profile.c (OVERLAP_BASE): Delete.
9354         (compute_frequency_overlap): Delete.
9355         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9356         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9357         * sched-ebb.c (rank): Use counts only.
9358         * shrink-wrap.c (handle_simple_exit): Use counts only.
9359         (try_shrink_wrapping): Use counts only.
9360         (place_prologue_for_one_component): Use counts only.
9361         * tracer.c (find_best_predecessor): Use to_frequency.
9362         (find_trace): Use to_frequency.
9363         (tail_duplicate): Use to_frequency.
9364         * trans-mem.c (expand_transaction): Do not update frequency.
9365         * tree-call-cdce.c: Do not update frequency.
9366         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9367         (gimple_merge_blocks): Likewise.
9368         (gimple_split_edge): Likewise.
9369         (gimple_duplicate_sese_region): Likewise.
9370         (gimple_duplicate_sese_tail): Likewise.
9371         (move_sese_region_to_fn): Likewise.
9372         (gimple_account_profile_record): Likewise.
9373         (insert_cond_bb): Likewise.
9374         * tree-complex.c (expand_complex_div_wide): Likewise.
9375         * tree-eh.c (lower_resx): Update profile.
9376         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9377         frequencies.
9378         (initialize_cfun): Do not initialize frequencies
9379         (freqs_to_counts): Delete.
9380         (copy_cfg_body): Ignore count parameter.
9381         (copy_body): Update.
9382         (expand_call_inline): Update count_max.
9383         (optimize_inline_calls): Update count_max.
9384         (tree_function_versioning): Update count_max.
9385         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9386         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9387         frequency.
9388         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9389         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9390         (try_peel_loop): Likewise.
9391         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9392         to_frequency.
9393         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9394         (tree_transform_and_unroll_loop): Do not use frequencies
9395         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9396         Use reliable prediction only.
9397         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9398         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9399         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9400         probability scaling.
9401         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9402         not update frequency
9403         (any_remaining_duplicated_blocks): Likewise.
9404         (update_profile): Likewise.
9405         (estimated_freqs_path): Delete.
9406         (freqs_to_counts_path): Delete.
9407         (clear_counts_path): Delete.
9408         (ssa_fix_duplicate_block_edges): Likewise.
9409         (duplicate_thread_path): Likewise.
9410         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9411         * tree-tailcall.c (decrease_profile): Do not update frequency.
9412         (eliminate_tail_call): Likewise.
9413         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9414         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9415         (optimize_mask_stores): Likewise.
9416         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9417         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9418         (ubsan_expand_ptr_ifn): Update profile.
9419         * value-prof.c (gimple_ic): Simplify.
9420         * value-prof.h (gimple_ic): Update prototype.
9421         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9422         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9423         counts are nonzero.
9424         (want_inline_self_recursive_call_p): Likewise.
9425         (resolve_noninline_speculation): Only cummulate defined counts.
9426         (inline_small_functions): Use nonzero_p.
9427         (ipa_inline): Do not access freed node.
9429 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9431         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9432         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9434 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9436         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9437         non-legitimate address.
9439 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9441         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9442         (*lt0_<mode>di, *lt0_<mode>si): New.
9444 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9446         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9447         TARGET_PAIRED_FLOAT.
9449 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9450             Jim Wilson  <jim.wilson@linaro.org>
9452         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9453         * config/aarch64/aarch64-tune.md: Regenerated.
9454         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9455         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9457 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9459         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9460         (arc_expand_prologue): Restore blink for millicode.
9461         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9463 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9465         PR target/82809
9466         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9467         gen_vec_duplicate after forcing the scalar into a register.
9469 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9471         * combine (try_combine): Print the insns input to try_combine to the
9472         dump file.
9474 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9476         PR target/79868
9477         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9478         Remove second argument from aarch64_process_target_attr call.
9479         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9480         Ditto.
9481         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9482         field type.
9483         (aarch64_handle_attr_arch): Remove second argument.
9484         (aarch64_handle_attr_cpu): Ditto.
9485         (aarch64_handle_attr_tune): Ditto.
9486         (aarch64_handle_attr_isa_flags): Ditto.
9487         (aarch64_process_one_target_attr): Ditto.
9488         (aarch64_process_target_attr): Ditto.
9489         (aarch64_option_valid_attribute_p): Remove second argument.
9490         on aarch64_process_target_attr call.
9492 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9494         * diagnostic.c: Include "selftest-diagnostic.h".
9495         (selftest::assert_location_text): New function.
9496         (selftest::test_diagnostic_get_location_text): New function.
9497         (selftest::diagnostic_c_tests): Call it.
9499 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9501         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9502         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9503         (class selftest::test_diagnostic_context): Move to...
9504         * selftest-diagnostic.c: New file.
9505         * selftest-diagnostic.h: New file.
9507 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9509         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9510         offset range for FT32B.
9511         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9512         * config/ft32/ft32.md: Add TARGET_NOPM.
9513         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9514         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9516 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9518         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9520 2017-11-02  Jeff Law  <law@redhat.com>
9522         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9523         virtual keyword on FINAL OVERRIDE members.
9525         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9526         virtuals together.  Add virtual destructor.
9527         (substitute_and_fold_engine): Similarly.
9529 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9531         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9533 2017-11-02  Richard Biener  <rguenther@suse.de>
9535         PR tree-optimization/82795
9536         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9538 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9540         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9541         gcc_SUN_LD_VERSION.
9542         (gcc_GAS_CHECK_FEATURE): Remove.
9543         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9544         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9545         * configure: Regenerate.
9547 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9549         * config/arc/arc.c (hwloop_optimize): Account for empty
9550         body loops.
9552 2017-11-02  Richard Biener  <rguenther@suse.de>
9554         PR middle-end/82765
9555         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9556         Truncate ARRAY_REF index and element size.
9558 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9560         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9562 2017-11-01  Jeff Law  <law@redhat.com>
9564         * tree-ssa-ccp.c (ccp_folder): New class derived from
9565         substitute_and_fold_engine.
9566         (ccp_folder::get_value): New member function.
9567         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9568         (ccp_fold_stmt): Remove prototype.
9569         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9570         * tree-ssa-copy.c (copy_folder): New class derived from
9571         substitute_and_fold_engine.
9572         (copy_folder::get_value): Renamed from get_value.
9573         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9574         * tree-vrp.c (vrp_folder): New class derived from
9575         substitute_and_fold_engine.
9576         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9577         (vrp_folder::get_value): New member function.
9578         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9579         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9580         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9581         provide a class interface to folder/substitute routines.
9582         (ssa_prop_fold_stmt_fn): Remove typedef.
9583         (ssa_prop_get_value_fn): Likewise.
9584         (subsitute_and_fold): Remove prototype.
9585         (replace_uses_in): Likewise.
9586         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9587         Renamed from replace_uses_in.  Call the virtual member function
9588         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9589         (substitute_and_fold_dom_walker): Remove initialization of
9590         data member entries for calbacks.  Add substitute_and_fold_engine
9591         member and initialize it.
9592         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9593         member functions for get_value, replace_phi_args_in c
9594         replace_uses_in, and fold_stmt calls.
9595         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9596         substitute_and_fold.  Remove assert.   Update ctor call.
9598         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9599         (ssa_prop_visit_phi_fn): Likewise.
9600         (class ssa_propagation_engine): New class to provide an interface
9601         into ssa_propagate.
9602         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9603         variable.
9604         (ssa_prop_visit_phi): Likewise.
9605         (ssa_propagation_engine::simulate_stmt): Moved into class.
9606         Call visit_phi/visit_stmt from the class rather than via
9607         file scoped static variables.
9608         (ssa_propagation_engine::simulate_block): Moved into class.
9609         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9610         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9611         set file scoped statics for the visit_stmt/visit_phi callbacks.
9612         * tree-complex.c (complex_propagate): New class derived from
9613         ssa_propagation_engine.
9614         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9615         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9616         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9617         class.
9618         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9619         ssa_propagation_engine.
9620         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9621         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9622         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9623         * tree-ssa-copy.c (copy_prop): New class derived from
9624         ssa_propagation_engine.
9625         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9626         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9627         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9628         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9629         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9630         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9631         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9633 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9635         PR rtl-optimization/82778
9636         PR rtl-optimization/82597
9637         * compare-elim.c (struct comparison): Add in_a_setter field.
9638         (find_comparison_dom_walker::before_dom_children): Remove killed
9639         bitmap and df_simulate_find_defs call, instead walk the defs.
9640         Compute last_setter and initialize in_a_setter.  Merge definitions
9641         with first initialization for a few variables.
9642         (try_validate_parallel): Use insn_invalid_p instead of
9643         recog_memoized.  Return insn rather than just the pattern.
9644         (try_merge_compare): Fix up comment.  Don't uselessly test if
9645         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9646         chains.
9647         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9648         call.
9650 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9651             Alan Hayward  <alan.hayward@arm.com>
9652             David Sherwood  <david.sherwood@arm.com>
9654         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9655         aarch64_hard_regno_nregs to get the number of registers
9656         in a mode.
9658 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9659             Alan Hayward  <alan.hayward@arm.com>
9660             David Sherwood  <david.sherwood@arm.com>
9662         * config/aarch64/constraints.md (Upl): Rename to...
9663         (Uaa): ...this.
9664         * config/aarch64/aarch64.md
9665         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9666         Update accordingly.
9668 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9669             Alan Hayward  <alan.hayward@arm.com>
9670             David Sherwood  <david.sherwood@arm.com>
9672         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9673         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9674         earlier in file.
9676 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9677             Alan Hayward  <alan.hayward@arm.com>
9678             David Sherwood  <david.sherwood@arm.com>
9680         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9681         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9682         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9683         named expanders.
9684         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9685         of a single element.
9686         * config/aarch64/iterators.md: Add a comment above the permute
9687         unspecs to say that they are generated directly by
9688         aarch64_expand_vec_perm_const.
9689         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9691 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9693         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9695 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9697         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9698         and medany code models, and describe what they do.
9700 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9702         Revert accidental duplicate:
9704         * combine.c (can_change_dest_mode): Reject changes in
9705         REGMODE_NATURAL_SIZE.
9707 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9709         PR rtl-optimization/64682
9710         PR rtl-optimization/69567
9711         PR rtl-optimization/69737
9712         PR rtl-optimization/82683
9713         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9714         register mentioned in the note, drop the note, unless it came from I3,
9715         in which case it should go to I3 again.
9717 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9719         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9720         and return false if not.
9721         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9723 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9725         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9726         range is known to be empty.
9728 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9729             Alan Hayward  <alan.hayward@arm.com>
9730             David Sherwood  <david.sherwood@arm.com>
9732         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9733         and CONST_VECTOR_NUNITS instead of computing the number of units from
9734         the byte sizes of the vector and element.
9735         (simplify_binary_operation_1): Likewise.
9736         (simplify_const_binary_operation): Likewise.
9737         (simplify_ternary_operation): Likewise.
9739 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9740             Alan Hayward  <alan.hayward@arm.com>
9741             David Sherwood  <david.sherwood@arm.com>
9743         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9744         (int_mem_offset): ...this new function.
9745         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9746         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9747         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9748         Update accordingly.
9750 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9751             Alan Hayward  <alan.hayward@arm.com>
9752             David Sherwood  <david.sherwood@arm.com>
9754         * lower-subreg.c (interesting_mode_p): New function.
9755         (compute_costs, find_decomposable_subregs, decompose_register)
9756         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9757         (resolve_clobber, dump_choices): Use it.
9759 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9760             Alan Hayward  <alan.hayward@arm.com>
9761             David Sherwood  <david.sherwood@arm.com>
9763         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9765 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9766             Alan Hayward  <alan.hayward@arm.com>
9767             David Sherwood  <david.sherwood@arm.com>
9769         * alias.c (find_base_value, find_base_term): Only process integer
9770         truncations.  Check the precision rather than the size.
9772 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9773             Alan Hayward  <alan.hayward@arm.com>
9774             David Sherwood  <david.sherwood@arm.com>
9776         * machmode.h (is_narrower_int_mode): New function
9777         * optabs.c (expand_float, expand_fix): Use it.
9778         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9780 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9781             Alan Hayward  <alan.hayward@arm.com>
9782             David Sherwood  <david.sherwood@arm.com>
9784         * rtl.h (narrower_subreg_mode): New function.
9785         * ira-color.c (update_costs_from_allocno): Use it.
9787 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9788             Alan Hayward  <alan.hayward@arm.com>
9789             David Sherwood  <david.sherwood@arm.com>
9791         * optabs-query.h (convert_optab_p): New function, split out from...
9792         (convert_optab_handler): ...here.
9793         (widening_optab_handler): Delete.
9794         (find_widening_optab_handler): Remove permit_non_widening parameter.
9795         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9796         override that operates on mode class wrappers.
9797         * optabs-query.c (widening_optab_handler): Delete.
9798         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9799         parameter.  Assert that the two modes are the same class and that
9800         the "from" mode is narrower than the "to" mode.  Use
9801         convert_optab_handler instead of widening_optab_handler.
9802         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9803         instead of widening_optab_handler.
9804         * expr.c (expand_expr_real_2): Update calls to
9805         find_widening_optab_handler.
9806         * optabs.c (expand_widen_pattern_expr): Likewise.
9807         (expand_binop_directly): Take the insn_code as a parameter.
9808         (expand_binop): Only call find_widening_optab_handler for
9809         conversion optabs; use optab_handler otherwise.  Update calls
9810         to find_widening_optab_handler and expand_binop_directly.
9811         Use convert_optab_handler instead of widening_optab_handler.
9812         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9813         find_widening_optab_handler and use scalar_mode rather than
9814         machine_mode.
9815         (convert_plusminus_to_widen): Likewise.
9817 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9818             Alan Hayward  <alan.hayward@arm.com>
9819             David Sherwood  <david.sherwood@arm.com>
9821         * machmode.h (fixed_size_mode): New class.
9822         * rtl.h (get_pool_mode): Return fixed_size_mode.
9823         * gengtype.c (main): Add fixed_size_mode.
9824         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9825         (get_raw_arg_mode): Likewise.
9826         * doc/tm.texi: Regenerate.
9827         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9828         * targhooks.c (default_get_reg_raw_mode): Likewise.
9829         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9830         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9831         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9832         (msp430_get_raw_result_mode): Likewise.
9833         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9834         * dbxout.c (dbxout_parms): Require fixed-size modes.
9835         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9836         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9837         * omp-low.c (lower_oacc_reductions): Likewise.
9838         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9839         (simplify_subreg): Update accordingly.
9840         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9841         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9842         that aren't fixed-size.
9843         (get_pool_mode): Return a fixed_size_mode.
9844         (output_constant_pool_2): Take a fixed_size_mode.
9846 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9847             Alan Hayward  <alan.hayward@arm.com>
9848             David Sherwood  <david.sherwood@arm.com>
9850         * doc/rtl.texi (vec_series): Document.
9851         (const): Say that the operand can be a vec_series.
9852         * rtl.def (VEC_SERIES): New rtx code.
9853         * rtl.h (const_vec_series_p_1): Declare.
9854         (const_vec_series_p): New function.
9855         * emit-rtl.h (gen_const_vec_series): Declare.
9856         (gen_vec_series): Likewise.
9857         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9858         (gen_vec_series): Likewise.
9859         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9860         * simplify-rtx.c (simplify_unary_operation): Handle negations
9861         of vector series.
9862         (simplify_binary_operation_series): New function.
9863         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9864         (test_vector_ops_series): New function.
9865         (test_vector_ops): Call it.
9866         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9867         gen_const_vec_series.
9868         (altivec_lvsr): Likewise.
9869         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9871 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9872             Alan Hayward  <alan.hayward@arm.com>
9873             David Sherwood  <david.sherwood@arm.com>
9875         * doc/rtl.texi (const): Update description of address constants.
9876         Say that vector constants are allowed too.
9877         * common.md (E, F): Use CONSTANT_P instead of checking for
9878         CONST_VECTOR.
9879         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9880         checking for CONST_VECTOR.
9881         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9882         VEC_DUPLICATE.
9883         * expr.c (expand_expr_real_2): Check for vector modes instead
9884         of checking for CONST_VECTOR.
9885         * rtl.h (const_vec_p): New function.
9886         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9887         (unwrap_const_vec_duplicate): Handle them here too.
9889 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9890             David Malcolm  <dmalcolm@redhat.com>
9891             Alan Hayward  <alan.hayward@arm.com>
9892             David Sherwood  <david.sherwood@arm.com>
9894         * rtl.h (vec_duplicate_p): New function.
9895         * selftest-rtl.c (assert_rtx_eq_at): New function.
9896         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9897         (assert_rtx_eq_at): Declare.
9898         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9899         * selftest-run-tests.c (selftest::run_tests): Call it.
9900         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9901         (simplify_unary_operation_1): Recursively handle vector duplicates.
9902         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9903         vector duplicates.
9904         (simplify_subreg): Handle subregs of vector duplicates.
9905         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9906         (selftest::simplify_rtx_c_tests): New functions.
9908 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9909             Alan Hayward  <alan.hayward@arm.com>
9910             David Sherwood  <david.sherwood@arm.com>
9912         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9913         (gen_vec_duplicate): Likewise.
9914         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9915         out from...
9916         (gen_const_vector): ...here.
9917         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9918         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9919         whose elements are all equal.
9920         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9921         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9922         (simplify_relational_operation): Likewise.
9923         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9924         Likewise.
9925         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9926         (aarch64_expand_vector_init): Likewise.
9927         * config/arm/arm.c (neon_vdup_constant): Likewise.
9928         (neon_expand_vector_init): Likewise.
9929         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9930         (arm_block_set_unaligned_vect): Likewise.
9931         (arm_block_set_aligned_vect): Likewise.
9932         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9933         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9934         (expand_vec_perm_even_odd_pack): Likewise.
9935         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9936         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9937         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9938         gen_const_vec_duplicate.
9939         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9940         * config/mips/mips.c (mips_gen_const_int_vector): Use
9941         gen_const_vec_duplicate.
9942         (mips_expand_vector_init): Use CONST0_RTX.
9943         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
9944         (define_split): Use gen_const_vec_duplicate.
9945         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
9946         (define_split): Use gen_const_vec_duplicate.
9947         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
9948         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
9949         * config/spu/spu.c (spu_const): Likewise.
9951 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9952             Alan Hayward  <alan.hayward@arm.com>
9953             David Sherwood  <david.sherwood@arm.com>
9955         * combine.c (can_change_dest_mode): Reject changes in
9956         REGMODE_NATURAL_SIZE.
9958 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
9960         * configure.ac (--enable-libssp): New.
9961         (gcc_cv_libc_provides_ssp): Check for explicit setting before
9962         trying to determine target-specific default.  Adjust indentation.
9963         * configure: Regenerated.
9964         * doc/install.texi (Configuration): Expand --disable-libssp
9965         documentation.
9967 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
9969         config/i386/i386.c (ix86_expand_epilogue): Correct stack
9970         calculation.
9972 2017-10-31  Martin Jambor  <mjambor@suse.cz>
9974         PR c++/81702
9975         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
9977 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
9979         * auto-profile.c (autofdo_source_profile::read): Use
9980         UNKNOWN_LOCATION rather than 0.
9981         * diagnostic-core.h (warning_at_rich_loc): Rename to...
9982         (warning_at): ...this overload.
9983         (warning_at_rich_loc_n): Rename to...
9984         (warning_n): ...this overload.
9985         (error_at_rich_loc): Rename to...
9986         (error_at): ...this overload.
9987         (pedwarn_at_rich_loc): Rename to...
9988         (pedwarn): ...this overload.
9989         (permerror_at_rich_loc): Rename to...
9990         (permerror): ...this overload.
9991         (inform_at_rich_loc): Rename to...
9992         (inform): ...this overload.
9993         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
9994         (diagnostic_n_impl_richloc): Rename to...
9995         (diagnostic_n_impl): ...this rich_location *-based decl.
9996         (inform_at_rich_loc): Rename to...
9997         (inform): ...this, and add an assertion.
9998         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
9999         (warning_at_rich_loc): Rename to...
10000         (warning_at): ...this, and add an assertion.
10001         (warning_at_rich_loc_n): Rename to...
10002         (warning_n): ...this, and add an assertion.
10003         (warning_n): Update location_t-based implementation for removal of
10004         location_t-based diagnostic_n_impl.
10005         (pedwarn_at_rich_loc): Rename to...
10006         (pedwarn): ...this, and add an assertion.
10007         (permerror_at_rich_loc): Rename to...
10008         (permerror): ...this, and add an assertion.
10009         (error_n): Update for removal of location_t-based diagnostic_n_impl.
10010         (error_at_rich_loc): Rename to...
10011         (error_at): ...this, and add an assertion.
10012         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
10013         (driver::do_spec_on_infiles): Likewise.
10014         * substring-locations.c (format_warning_va): Update for renaming
10015         of inform_at_rich_loc.
10017 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
10019         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
10020         _Float<N>X built-in functions so that the variant without the
10021         "__builtin_" prefix is only enabled for the GNU C and Objective C
10022         languages when they are in non-strict ANSI/ISO mode.
10023         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10024         * target.def (floatn_builtin_p): Add a target hook to control
10025         whether _Float<N> and _Float<N>X built-in functions without the
10026         "__builtin_" prefix are enabled, and return true for C and
10027         Objective C in the default hook.  Include langhooks.h in
10028         targhooks.c.
10029         * targhooks.h (default_floatn_builtin_p): Likewise.
10030         * targhooks.c (default_floatn_builtin_p): Likewise.
10031         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10032         floatn_builtin_p target hook.
10033         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10035 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10036             Eric Botcazou  <ebotcazou@adacore.com>
10038         PR rtl-optimization/81803
10039         * lra-constraints.c (curr_insn_transform): Also reload the whole
10040         register for a strict subreg no wider than a word if this is for
10041         a WORD_REGISTER_OPERATIONS target.
10043 2017-10-31  Jason Merrill  <jason@redhat.com>
10045         * gdbinit.in: Skip over inlines from timevar.h.
10047 2017-10-31  Martin Liska  <mliska@suse.cz>
10049         * doc/gcov.texi: Document new option.
10050         * gcov.c (print_usage): Likewise print it.
10051         (process_args): Support the argument.
10052         (format_count): New function.
10053         (format_gcov): Use the function.
10055 2017-10-31  Martin Liska  <mliska@suse.cz>
10057         * gcov.c (struct name_map): do not use typedef.
10058         Define operator== and operator<.
10059         (name_search): Remove.
10060         (name_sort): Remove.
10061         (main): Do not allocate names.
10062         (process_file): Add vertical space.
10063         (generate_results): Use std::find.
10064         (release_structures): Do not release memory.
10065         (find_source): Use std::find.
10067 2017-10-31  Martin Liska  <mliska@suse.cz>
10069         * gcov.c (struct line_info): Remove it's typedef.
10070         (line_info::line_info): Add proper ctor.
10071         (line_info::has_block): Do not use a typedef.
10072         (struct source_info): Do not use typedef.
10073         (circuit): Likewise.
10074         (get_cycles_count): Likewise.
10075         (output_intermediate_file): Iterate via vector iterator.
10076         (add_line_counts): Use std::vector methods.
10077         (accumulate_line_counts): Likewise.
10078         (output_lines): Likewise.
10080 2017-10-31  Martin Liska  <mliska@suse.cz>
10082         * gcov.c (struct source_info): Remove typedef.
10083         (source_info::source_info): Add proper ctor.
10084         (accumulate_line_counts): Use struct, not it's typedef.
10085         (output_gcov_file): Likewise.
10086         (output_lines): Likewise.
10087         (main): Do not allocate an array.
10088         (output_intermediate_file): Use size of vector container.
10089         (process_file): Resize the vector.
10090         (generate_results): Do not preallocate, use newly added vector
10091         lines.
10092         (release_structures): Do not release sources.
10093         (find_source): Use vector methods.
10094         (add_line_counts): Do not use typedef.
10096 2017-10-31  Martin Liska  <mliska@suse.cz>
10098         * doc/gcov.texi: Document that.
10099         * gcov.c (add_line_counts): Mark lines with a non-executed
10100         statement.
10101         (output_line_beginning): Handle such lines.
10102         (output_lines): Pass new argument.
10103         (output_intermediate_file): Print it in intermediate format.
10105 2017-10-31  Martin Liska  <mliska@suse.cz>
10107         * color-macros.h: New file.
10108         * diagnostic-color.c: Factor out color related to macros to
10109         color-macros.h.
10110         * doc/gcov.texi: Document -k option.
10111         * gcov.c (INCLUDE_STRING): Include string.h.
10112         (print_usage): Add -k option.
10113         (process_args): Parse it.
10114         (pad_count_string): New function.
10115         (output_line_beginning): Likewise.
10116         (DEFAULT_LINE_START): New macro.
10117         (output_lines): Support color output.
10119 2017-10-31  Martin Liska  <mliska@suse.cz>
10121         PR gcov-profile/82633
10122         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10123         their interaction with GCOV infrastructure.
10124         * configure.ac: Add -fkeep-{inline,static}-functions to
10125         coverage_flags.
10126         * configure: Regenerate.
10128 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10130         PR target/82772
10131         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10133 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10135         PR target/82674
10136         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10137         into a register if it does not fit into an immediate offset field.
10139 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10141         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10143 2017-10-31  Julia Koval  <julia.koval@intel.com>
10145         * config.gcc: Add gfniintrin.h.
10146         * config/i386/gfniintrin.h: New.
10147         * config/i386/i386-builtin-types.def
10148         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10149         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10150         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10151         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10152         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10153         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10154         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10155         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10156         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10157         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10158         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10159         * config/i386/immintrin.h: Include gfniintrin.h.
10160         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10162 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10164         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10166 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10168         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10169         (arm_ashldi3_1bit): Remove pattern.
10170         (ashrdi3): Remove shift by 1 expansion.
10171         (arm_ashrdi3_1bit): Remove pattern.
10172         (lshrdi3): Remove shift by 1 expansion.
10173         (arm_lshrdi3_1bit): Remove pattern.
10174         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10175         cost of ashldi3 by 1.
10176         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10177         (<shift>di3_neon): Likewise.
10179 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10181         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10182         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10183         and (*aarch64_simd_mov<VQ:mode>).
10184         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10185         pattern alternative.
10186         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10187         attributes to match pattern alternative.
10189 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10191         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10192         * config/rs6000/emmintrin.h: New file.
10193         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10195 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10197         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10198         * (movdi_vfp_cortexa8): Remove pattern.
10200 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10202         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10203         etc. reference.
10204         (Specific, alpha*-dec-osf5.1): Remove.
10205         (Specific, mips-sgi-irix5): Remove.
10206         (Specific, mips-sgi-irix6): Remove.
10208 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10210         PR middle-end/22141
10211         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10212         arguments to clear_bit_region_be.
10214 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10216         * gimplify.c: Include memmodel.h.
10218 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10220         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10221         condition whether loop should be executed at all.
10223 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10225         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10226         gimple folding of vec_madd() intrinsics.
10227         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10228         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10229         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10231 2017-10-30  Richard Biener  <rguenther@suse.de>
10233         PR tree-optimization/82762
10234         Revert
10235         2017-10-23  Richard Biener  <rguenther@suse.de>
10237         PR tree-optimization/82129
10238         Revert
10239         2017-08-01  Richard Biener  <rguenther@suse.de>
10241         PR tree-optimization/81181
10242         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10243         (compute_antic): ... end of iteration here.
10245 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10247         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10248         -std=iso9899:2017 and -std=gnu17.
10249         * doc/standards.texi (C Language): Document C17 support.
10250         * doc/cpp.texi (Overview): Mention -std=c17.
10251         (Standard Predefined Macros): Document C11 and C17 values of
10252         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10253         * doc/extend.texi (Inline): Do not list individual options for
10254         standards newer than C99.
10255         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10256         "GNU C17".
10257         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10258         language name.
10260 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10262         * asan.c (asan_finish_file): Align asan globals array by shadow
10263         granularity.
10265 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10267         PR middle-end/22141
10268         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10269         (struct store_immediate_info): Add bitregion_start and bitregion_end
10270         fields.
10271         (store_immediate_info::store_immediate_info): Add brs and bre
10272         arguments and initialize bitregion_{start,end} from those.
10273         (struct merged_store_group): Add bitregion_start, bitregion_end,
10274         align_base and mask fields.  Drop unnecessary struct keyword from
10275         struct store_immediate_info.  Add do_merge method.
10276         (clear_bit_region_be): Use memset instead of loop storing zeros.
10277         (merged_store_group::do_merge): New method.
10278         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10279         stores as long as the surrounding bitregions have no gaps.
10280         (merged_store_group::merge_overlapping): Use do_merge.
10281         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10282         is byte aligned, rather than requiring that start and width are
10283         byte aligned.  Drop unnecessary struct keyword from
10284         struct store_immediate_info.  Allocate and populate also mask array.
10285         Make start of the arrays relative to bitregion_start rather than
10286         start and size them according to bitregion_{end,start} difference.
10287         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10288         struct store_immediate_info.
10289         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10290         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10291         struct keyword from struct store_immediate_info.
10292         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10293         between stores as long as the surrounding bitregions have no gaps.
10294         Formatting fixes.
10295         (struct split_store): Add orig non-static data member.
10296         (split_store::split_store): Initialize orig to false.
10297         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10298         if there is exactly a single original stmt.  Change stmts argument
10299         to pointer from reference, if NULL, don't push anything to it.  Add
10300         first argument, use it to optimize skipping over orig stmts that
10301         are known to be before bitpos already.  Simplify.
10302         (split_group): Return unsigned int count how many stores are or
10303         would be needed rather than a bool.  Add allow_unaligned argument.
10304         Change split_stores argument from reference to pointer, if NULL,
10305         only do a dry run computing how many stores would be produced.
10306         Rewritten algorithm to use both alignment and misalign if
10307         !allow_unaligned and handle bitfield stores with gaps.
10308         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10309         from bitregion_start instead of start.  Compute allow_unaligned
10310         here, if true, do 2 split_group dry runs to compute which one
10311         produces fewer stores and prefer aligned if equal.  Punt if
10312         new count is bigger or equal than original before emitting any
10313         statements, rather than during that.  Remove no longer needed
10314         new_ssa_names tracking.  Replace num_stmts with
10315         split_stores.length ().  Use 32-bit stack allocated entries
10316         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10317         if possible.  Handle bitfields with gaps.
10318         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10319         Compute bitregion_{start,end} for the stores and construct
10320         store_immediate_info with that.  Formatting fixes.
10322 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10324         PR target/82725
10325         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10326         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10328 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10330         * gimplify.c: Include tm_p.h.
10332         * common.opt (gcoff): Re-add as ignored option.
10333         (gcoff1, gcoff2, gcoff3): Likewise.
10335         * Makefile.in (OBJS): Delete sdbout.o.
10336         (GTFILES): Delete $(srcdir)/sdbout.c.
10337         * debug.h: Delete sdb_debug_hooks.
10338         * final.c: Delete sdbout.h include.
10339         (final_scan_insn): Delete SDB_DEBUG check.
10340         (rest_of_clean_state): Likewise.
10341         * output.h: Delete sdb_begin_function_line.
10342         * sdbout.c: Delete.
10343         * sdbout.h: Delete.
10344         * toplev.c: Delete sdbout.h include.
10345         (process_options): Delete SDB_DEBUG check.
10346         * tree-core.h (tree_type_common): Delete pointer field of
10347         tree_type_symtab.
10348         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10349         TYPE_SYMTAB_POINTER.
10350         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10351         (TYPE_SYMTAB_IS_POINTER): Delete.
10352         (TYPE_SYMTAB_IS_DIE): Renumber.
10353         * xcoffout.c: Refer to former sdbout.c file.
10354         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10356         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10357         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10358         (Debugging Options): Delete -gcoff.
10359         (-gstabs): Delete SDB reference.
10360         (-gcoff): Delete.
10361         (-gcoff@var{level}): Delete.
10362         * doc/passes.texi (Debugging information output): Delete SDB and
10363         sdbout.c references.
10364         * doc/tm.texi: Regenerate.
10365         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10366         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10367         references.
10368         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10369         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10370         (SDB_DEBUGGING_INFO): Delete.
10371         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10372         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10373         * target.def (output_source_filename): Delete COFF reference.
10375         * common.opt (gcoff): Delete.
10376         (gxcoff+): Update Negative chain.
10377         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10378         SDB_DEBUG.
10379         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10380         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10381         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10382         and SDB references.
10383         (expand_function_start): Change sdb reference to past tense.
10384         (expand_function_end): Change sdb reference to past tense.
10385         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10386         * opts.c (debug_type_names): Delete coff entry.
10387         (common_handle_option): Delete OPT_gcoff case.
10388         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10390         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10391         * config/cris/cris.h: Delete SDB reference in comment.
10392         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10393         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10394         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10395         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10396         to past tense.
10397         (ix86_expand_prologue): Likewise.
10398         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10399         * config/ia64/ia64.h: Likewise.
10400         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10401         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10402         support.
10403         * config/mmix/mmix.h: Likewise.
10404         * config/nds32/nds32.c: Likewise.
10405         * config/stormy/storym16.h: Likewise.
10406         * config/visium/visium.h: Likewise.
10407         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10409 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10411         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10412         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10413         FRAME_POINTER_REGNUM point at high end of local var area.
10415 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10417         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10418         Move comment around.  Do not reset best_edge for a copiable
10419         destination if the copy would cause a partition change.
10420         (better_edge_p): Remove redundant check.
10422 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10424         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10426 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10428         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10429         for math functions that have _Float<N> and _Float<N>X variants.
10430         (mathfn_built_in_2): Add support for math functions that have
10431         _Float<N> and _Float<N>X variants.
10432         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10433         (expand_builtin_mathfn_ternary): Add support for fma with
10434         _Float<N> and _Float<N>X variants.
10435         (expand_builtin): Likewise.
10436         (fold_builtin_3): Likewise.
10437         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10438         create math function _Float<N> and _Float<N>X variants as external
10439         library builtins.
10440         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10441         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10442         the __builtin_ prefix and if not strict ansi, without the prefix.
10443         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10444         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10445         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10446         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10447         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10448         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10449         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10450         function signatures for fma _Float<N> and _Float<N>X variants.
10451         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10452         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10453         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10454         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10455         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10456         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10457         * gencfn-macros.c (print_case_cfn): Add support for math functions
10458         that have _Float<N> and _Float<N>X variants.
10459         (print_define_operator_list): Likewise.
10460         (fltfn_suffixes): Likewise.
10461         (main): Likewise.
10462         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10463         for math functions that have _Float<N> and _Float<N>X variants.
10464         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10465         and _Float<N>X variants.
10466         (COPYSIGN): Likewise.
10467         (FMIN): Likewise.
10468         (FMAX): Likewise.
10469         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10470         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10471         variants.
10472         (integer_valued_read_call_p): Likewise.
10473         * fold-const-call.c (fold_const_call_ss): Likewise.
10474         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10475         _Float<N> and _Float<N>X variants.
10476         (fold_const_call_ssss): Add support for fma _Float<N> and
10477         _Float<N>X variants.
10478         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10479         support for copysign and fma _Float<N> and _Float<N>X variants.
10480         (backprop::process_builtin_call_use): Likewise.
10481         * tree-call-cdce.c (can_test_argument_range); Add support for
10482         sqrt _Float<N> and _Float<N>X variants.
10483         (edom_only_function): Likewise.
10484         (get_no_error_domain): Likewise.
10485         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10486         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10487         copysign _Float<N> and _Float<N>X variants.
10488         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10489         handled by machine independent code.
10490         (FMAF128): Likewise.
10491         * doc/cpp.texi (Common Predefined Macros): Document defining
10492         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10493         fma _Float<N> and _Float<N>X variants.
10495 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10497         PR target/82692
10498         * config/i386/i386-modes.def (CCFPU): Remove definition.
10499         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10500         (ix86_cc_modes_compatible): Ditto.
10501         (ix86_expand_carry_flag_compare): Ditto.
10502         (ix86_expand_int_movcc): Ditto.
10503         (ix86_expand_int_addcc): Ditto.
10504         (ix86_reverse_condition): Ditto.
10505         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10506         Return true/false for unordered/ordered fp comparisons.
10507         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10508         (ix86_prepare_fp_compare_args): Update for rename.
10509         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10510         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10511         (ix86_expand_sse_compare_and_jump): Ditto.
10512         * config/i386/predicates.md (fcmov_comparison_operator):
10513         Remove CCFPU mode handling.
10514         (ix86_comparison_operator): Ditto.
10515         (ix86_carry_flag_operator): Ditto.
10516         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10517         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10518         (*cmpu<mode>_cc_i387): Ditto.
10519         (FPCMP): Remove mode iterator.
10520         (unord): Remove mode attribute.
10521         (unord_subst): New define_subst transformation
10522         (unord): New define_subst attribute.
10523         (unordered): Ditto.
10524         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10525         (*cmpi<unord>xf_i387): Ditto.
10526         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10527         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10528         using unord_subst transformation.
10529         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10530         (round_saeonly): Also handle CCFP mode.
10531         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10532         Remove UNSPEC_SAHF unspec handling.
10534 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10536         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10538 2017-10-27  Jeff Law  <law@redhat.com>
10540         * gimple-ssa-sprintf.c: Include domwalk.h.
10541         (class sprintf_dom_walker): New class, derived from dom_walker.
10542         (sprintf_dom_walker::before_dom_children): New function.
10543         (struct call_info): Moved into sprintf_dom_walker class
10544         (compute_formath_length, handle_gimple_call): Likewise.
10545         (sprintf_length::execute): Call the dominator walker rather
10546         than walking the statements.
10548         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10549         gimple statement locations.
10550         (check_array_bounds): Corresponding changes.  Get the statement's
10551         location directly from wi->stmt.
10553 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10555         PR target/82717
10556         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10558 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10560         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10561         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10563 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10565         PR target/82703
10566         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10567         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10568         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10569         maybe_get_pool_constant.
10570         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10571         Likewise.
10573 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10575         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10576         2.26 caveat.  Update gas and gld versions.
10577         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10578         reference.
10580 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10582         * cgraph.h (set_malloc_flag): Declare.
10583         * cgraph.c (set_malloc_flag_1): New function.
10584         (set_malloc_flag): Likewise.
10585         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10586         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10587         false.
10588         (read_ipa_call_summary): Add support for reading is_return_callee.
10589         (write_ipa_call_summary): Stream is_return_callee.
10590         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10591         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10592         ipa-prop.h, ipa-fnsummary.h.
10593         (pure_const_names): Change to static.
10594         (malloc_state_e): Define.
10595         (malloc_state_names): Define.
10596         (funct_state_d): Add field malloc_state.
10597         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10598         (check_retval_uses): New function.
10599         (malloc_candidate_p): Likewise.
10600         (analyze_function): Add support for malloc attribute.
10601         (pure_const_write_summary): Stream malloc_state.
10602         (pure_const_read_summary): Add support for reading malloc_state.
10603         (dump_malloc_lattice): New function.
10604         (propagate_malloc): New function.
10605         (warn_function_malloc): New function.
10606         (ipa_pure_const::execute): Call propagate_malloc and
10607         ipa_free_fn_summary.
10608         (pass_local_pure_const::execute): Add support for malloc attribute.
10609         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10610         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10612 2017-10-27  Martin Liska  <mliska@suse.cz>
10614         PR gcov-profile/82457
10615         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10616         for fork-like functions to be properly instrumented.
10618 2017-10-27  Richard Biener  <rguenther@suse.de>
10620         PR middle-end/81659
10621         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10622         info when we redirected EH.
10624 2017-10-26  Michael Collison  <michael.collison@arm.com>
10626         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10627         New pattern.
10628         (<optab>_trunchf<GPI:mode>2: New pattern.
10629         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10630         * config/aarch64/iterators.md (wv): New mode attribute.
10631         (vf, VF): New mode attributes.
10632         (vgp, VGP): New mode attributes.
10633         (s): Update attribute with SImode and DImode prefixes.
10635 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10637         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10638         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10639         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10640         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10641         addressing with PIC.
10642         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10643         (nios2_symbolic_constant_p): Likewise.
10644         (nios2_legitimate_address_p): Likewise.
10645         (nios2_r0rel_section_name_p): New.
10646         (nios2_symbol_ref_in_r0rel_data_p): New.
10647         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10648         (r0rel_constant_p): New.
10649         (nios2_print_operand_address): Handle r0rel_constant_p.
10650         (nios2_cdx_narrow_form_p): Likewise.
10651         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10652         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10653         (Nios II Options): Document -mr0rel-sec.
10655 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10657         * config/nios2/nios2.c: Include xregex.h.
10658         (nios2_gprel_sec_regex): New.
10659         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10660         addressing with PIC.
10661         (nios2_small_section_name_p): Check for regex match.
10662         * config/nios2/nios2.opt (mgprel-sec=): New option.
10663         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10664         (Nios II Options): Document -mgprel-sec.
10666 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10668         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10670 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10672         PR tree-optimization/82707
10673         * gimple.c (gimple_copy): Fix unsharing of
10674         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10676 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10678         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10679         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10680         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10681         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10682         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10683         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10684         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10685         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10686         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10687         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10688         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10689         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10690         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10691         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10692         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10693         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10694         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10696 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10698         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10699         default to IBM.
10700         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10701         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10702         warning to rs6000.c.  Remove the Undocumented flag, since it has
10703         been documented.
10704         (-mabi=ibmlongdouble): Likewise.
10705         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10706         already set, set the default format for long double.
10707         (rs6000_debug_reg_global): Print whether long double is IBM or
10708         IEEE.
10709         (rs6000_option_override_internal): Rework setting long double
10710         format.  Only warn if the user is changing the long double default
10711         and they did not use -Wno-psabi.
10712         * doc/invoke.texi (PowerPC options): Update the documentation for
10713         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10715 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10716             Alan Hayward  <alan.hayward@arm.com>
10717             David Sherwood  <david.sherwood@arm.com>
10719         * rtl.h (wider_subreg_mode): New function.
10720         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10721         rather than an unsigned int *.
10722         * ira-color.c (regno_max_ref_width): Replace with...
10723         (regno_max_ref_mode): ...this new variable.
10724         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10725         Use wider_subreg_mode.
10726         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10727         rather than an unsigned int *.
10728         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10729         (process_alt_operands): Likewise.
10730         (invariant_p): Likewise.
10731         * lra-spills.c (assign_mem_slot): Likewise.
10732         (add_pseudo_to_slot): Likewise.
10733         * lra.c (collect_non_operand_hard_regs): Likewise.
10734         (add_regs_to_insn_regno_info): Likewise.
10735         * reload1.c (regno_max_ref_width): Replace with...
10736         (regno_max_ref_mode): ...this new variable.
10737         (reload): Update accordingly.  Update call to
10738         ira_sort_regnos_for_alter_reg.
10739         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10740         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10741         (scan_paradoxical_subregs): Likewise.
10743 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10745         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10746         (aarch64_frame): Add emit_frame_chain boolean.
10747         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10748         Move eh_return case to aarch64_layout_frame.
10749         (aarch64_layout_frame): Initialize emit_frame_chain.
10750         (aarch64_expand_prologue): Use emit_frame_chain.
10752 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10754         * config/aarch64/aarch64.c (aarch64_layout_frame):
10755         Ensure LR is always stored at the bottom of the callee-saves.
10756         Remove rarely used frame layout which saves callee-saves at top of
10757         frame, so the store of LR can be used as a valid probe in all cases.
10759 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10761         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10762         Improve unaligned TImode/TFmode base/offset split.
10764 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10765             Alan Hayward  <alan.hayward@arm.com>
10766             David Sherwood  <david.sherwood@arm.com>
10768         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10769         * combine.c (find_single_use_1): Likewise.
10770         (expand_field_assignment): Likewise.
10771         (move_deaths): Likewise.
10772         * lra-constraints.c (simplify_operand_subreg): Likewise.
10773         (curr_insn_transform): Likewise.
10774         * lra.c (collect_non_operand_hard_regs): Likewise.
10775         (add_regs_to_insn_regno_info): Likewise.
10776         * rtlanal.c (reg_referenced_p): Likewise.
10777         (covers_regno_no_parallel_p): Likewise.
10779 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10781         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10782         Don't print any bits outside the precision of the value.
10783         * wide-int.cc (test_printing): Add some new tests.
10785 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10787         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10788         supports -xbrace_comment option.
10789         * configure: Regenerate.
10790         * config.in: Regenerate.
10791         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10792         (ASM_CPU_SPEC): Use it.
10794 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10796         * target.def (static_rtx_alignment): New hook.
10797         * targhooks.h (default_static_rtx_alignment): Declare.
10798         * targhooks.c (default_static_rtx_alignment): New function.
10799         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10800         * doc/tm.texi: Regenerate.
10801         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10802         instead of targetm.constant_alignment.  Remove call to
10803         set_mem_attributes.
10804         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10805         (cris_preferred_mininum_alignment): New function, split out from...
10806         (cris_constant_alignment): ...here.
10807         (cris_static_rtx_alignment): New function.
10808         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10809         split out from...
10810         (ix86_constant_alignment): ...here.
10811         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10812         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10813         (mmix_static_rtx_alignment): New function.
10814         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10815         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10817 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10819         PR target/81800
10820         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10821         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10823 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10825         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10826         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10827         mark as a sign-extending load.
10828         (local_pic_load_u): Define.
10830 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10832         PR middle-end/82062
10833         * fold-const.c (operand_equal_for_comparison_p): Also return true
10834         if ARG0 is a simple variant of ARG1 with narrower precision.
10835         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10837 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10839         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10840         cost correctly.
10841         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10842         scatter_static, scatter_per_elt.
10843         * x86-tune-costs.h: Add new cost entries.
10845 2017-10-25  Richard Biener  <rguenther@suse.de>
10847         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10848         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10849         class pass_fre): Move to ...
10850         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10851         class pass_fre): ... here and adjust for statistics.
10853 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10855         PR libstdc++/81706
10856         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10857         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10858         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10859         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10860         declarations.
10862 2017-10-25  Richard Biener  <rguenther@suse.de>
10864         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10865         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10866         eliminate_push_avail, eliminate_insert): Move inside...
10867         (class eliminate_dom_walker): ... this class in preparation
10868         of move.
10869         (fini_eliminate): Remove by merging with ...
10870         (eliminate): ... this function.  Adjust for class changes.
10871         (pass_pre::execute): Remove fini_eliminate call.
10872         (pass_fre::execute): Likewise.
10874 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10876         PR target/82460
10877         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10878         (VPERMI2, VPERMI2I): New mode iterators.
10879         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10880         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10881         patterns.
10882         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10883         mode iterator.  Remove 3 old define_insn patterns.
10884         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10885         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10886         VPERMI2 mode iterator, remove the other two expanders.
10887         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10888         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10889         instructions, remove the other two patterns.
10890         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10891         mode iterator, remove the other two patterns.
10892         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10893         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10894         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10895         and adjust argument order accordingly.
10896         (ix86_expand_vec_perm): Adjust caller.
10897         (expand_vec_perm_1): Likewise.
10898         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10899         (expand_vec_perm_vpermt2_vpshub2): ... this.
10900         (ix86_expand_vec_perm_const_1): Adjust caller.
10901         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10903         PR target/82370
10904         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10905         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10906         (vec_shl_<mode>): Remove unused expander.
10907         (avx512bw_<shift_insn><mode>3): New define_insn.
10908         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10909         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10911 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10913         PR c++/82466
10914         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10915         description.
10917 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10919         PR rtl-optimization/82396
10920         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10921         (autopref_multipass_init): Simplify initialization.
10922         (autopref_rank_data): Simplify sort order.
10923         * gcc/sched-int.h (autopref_multipass_data_): Remove
10924         multi_mem_insn_p, min_offset and max_offset.
10926 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10928         PR middle-end/60580
10929         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10930         Check special value of flag_omit_frame_pointer.
10931         (aarch64_can_eliminate): Likewise.
10932         (aarch64_override_options_after_change_1): Simplify handling of
10933         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10935 2017-10-24  Richard Biener  <rguenther@suse.de>
10937         PR tree-optimization/82697
10938         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10939         zero for conditional load and unconditional store.
10941 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10943         * doc/install.texi: Document bootstrap-cet.
10945 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10947         PR target/82659
10948         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
10949         ENDBR instruction at function entrance if function is only
10950         called directly.
10952 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10954         PR target/82628
10955         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
10956         patterns to better describe from which operation the CF is computed.
10957         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
10958         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
10959         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
10960         is 0, use _0 suffixed expanders instead of emitting a comparison
10961         before it.
10963 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
10965         * config/i386/i386.md(*movsf_internal, *movdf_internal):
10966         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
10968 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
10970         PR middle-end/82569
10971         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
10972         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
10973         * loop-iv.c (iv_get_reaching_def): Likewise.
10974         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
10975         variable is promoted and the partition contains undefined values.
10977 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10979         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
10980         reflect reality.
10981         (nios2_address_cost): Define.
10982         (nios2_legitimize_address): Recognize (exp + constant) directly.
10983         (TARGET_ADDRESS_COST): Define.
10985 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10987         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
10988         (nios2_symbolic_memory_operand_p): Declare.
10989         (nios2_split_large_constant): Declare.
10990         (nios2_split_symbolic_memory_operand): Declare.
10991         * config/nios2/nios2.c: Adjust includes.
10992         (nios2_symbolic_constant_allowed): New.
10993         (nios2_symbolic_constant_p): New.
10994         (nios2_plus_symbolic_constant_p): New.
10995         (nios2_valid_addr_expr_p): Recognize addresses involving
10996         symbolic constants.
10997         (nios2_legitimate_address_p): Likewise, also LO_SUM.
10998         (nios2_symbolic_memory_operand_p): New.
10999         (nios2_large_constant_p): New.
11000         (nios2_split_large_constant): New.
11001         (nios2_split_plus_large_constant): New.
11002         (nios2_split_symbolic_memory_operand): New.
11003         (nios2_legitimize_address): Code refactoring.  Handle addresses
11004         involving symbolic constants.
11005         (nios2_emit_move_sequence): Likewise.
11006         (nios2_print_operand): Improve error output.
11007         (nios2_print_operand_address): Handle LO_SUM.
11008         (nios2_cdx_narrow_form_p): Likewise.
11009         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
11010         operands involving symbolic constants.
11011         (movhi_internal, movsi_internal): Likewise.
11012         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
11013         (extendhisi2, extendqi<mode>2): Likewise.
11015 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11017         * tree-pass.h (PROP_rtl_split_insns): Define.
11018         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
11020 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11022         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11024 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11026         PR debug/82630
11027         * target.def (const_not_ok_for_debug_p): Default to
11028         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11029         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11030         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11031         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11032         which targetm.const_not_ok_for_debug_p returned true.
11033         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11034         for UNSPECs.
11035         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11036         Likewise.
11037         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11038         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11039         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11040         if !base_term_p.
11041         (ix86_const_not_ok_for_debug_p): New function.
11042         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11043         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11045 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11047         PR bootstrap/82610
11048         * system.h: Conditionally include "unique-ptr.h" if
11049         INCLUDE_UNIQUE_PTR is defined.
11050         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11051         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11053 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11055         * config/rl78/rl78.md: New define_expand "subdi3".
11057 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11059         PR target/82673
11060         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11061         DF_REF_INSN if DF_REF_INSN_INFO is false.
11063 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11065         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11066         xmm_move instead of sse_move.
11067         (sse_store_index): New function.
11068         (ix86_register_move_cost): Be more sensible about mismatch stall;
11069         model AVX moves correctly; make difference between sse->integer and
11070         integer->sse.
11071         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11072         moves; make difference between SSE and AVX.
11073         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11074         and zmm_move. Increase size of sse load and store tables;
11075         add unaligned load and store tables; add ssemmx_to_integer.
11076         * x86-tune-costs.h: Update all entries according to real
11077         move latencies from Agner Fog's manual and chip documentation.
11079 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11081         PR target/82628
11082         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11083         * config/i386/constraints.md (Wf): New constraint.
11084         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11085         (cmp<dwi>_doubleword): Removed.
11086         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11087         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11088         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11089         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11090         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11092         * common.opt (gcolumn-info): Enable by default.
11093         * doc/invoke.texi (gcolumn-info): Document new default.
11095 2017-10-23  Richard Biener  <rguenther@suse.de>
11097         PR tree-optimization/82672
11098         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11099         Fold the stmt if we propagated into it.
11101 2017-10-23  Richard Biener  <rguenther@suse.de>
11103         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11104         (bitmap_remove_expr_from_set): ... this.  All callers call this
11105         for non-constant values.
11106         (bitmap_set_subtract): Rename to...
11107         (bitmap_set_subtract_expressions): ... this.  Adjust and
11108         optimize.
11109         (bitmap_set_contains_value): Remove superfluous check.
11110         (bitmap_set_replace_value): Inline into single caller ...
11111         (bitmap_value_replace_in_set): ... here and simplify.
11112         (dependent_clean): Merge into ...
11113         (clean): ... this using an overload.  Adjust.
11114         (prune_clobbered_mems): Adjust.
11115         (compute_antic_aux): Likewise.
11116         (compute_partial_antic_aux): Likewise.
11118 2017-10-23  Richard Biener  <rguenther@suse.de>
11120         PR tree-optimization/82129
11121         Revert
11122         2017-08-01  Richard Biener  <rguenther@suse.de>
11124         PR tree-optimization/81181
11125         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11126         (compute_antic): ... end of iteration here.
11128 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11130         * target.def (starting_frame_offset): New hook.
11131         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11132         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11133         * doc/tm.texi.in: Regenerate.
11134         * hooks.h (hook_hwi_void_0): Declare.
11135         * hooks.c (hook_hwi_void_0): New function.
11136         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11137         STARTING_FRAME_OFFSET.
11138         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11139         * reload1.c (reload): Likewise.
11140         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11141         instead of STARTING_FRAME_OFFSET.
11142         * function.c (try_fit_stack_local): Likewise.
11143         (assign_stack_local_1): Likewise
11144         (instantiate_virtual_regs): Likewise.
11145         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11146         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11147         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11148         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11149         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11150         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11151         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11152         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11153         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11154         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11155         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11156         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11157         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11158         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11159         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11160         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11161         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11162         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11163         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11164         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11165         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11166         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11167         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11168         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11169         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11170         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11171         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11172         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11173         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11174         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11175         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11176         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11177         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11178         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11179         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11180         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11181         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11182         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11183         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11184         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11185         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11186         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11187         return a HOST_WIDE_INT.
11188         (avr_builtin_setjmp_frame_value): Use it instead of
11189         STARTING_FRAME_OFFSET.
11190         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11191         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11192         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11193         New function.
11194         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11195         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11196         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11197         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11198         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11199         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11200         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11201         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11202         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11203         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11204         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11205         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11206         New function.
11207         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11208         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11209         * config/mips/mips.c (mips_compute_frame_info): Refer to
11210         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11211         (mips_starting_frame_offset): New function.
11212         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11213         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11214         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11215         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11216         and return a HOST_WIDE_INT.
11217         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11218         (mmix_initial_elimination_offset): Refer to
11219         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11220         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11221         * config/pa/pa.c (pa_starting_frame_offset): New function.
11222         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11223         (pa_expand_prologue): Likewise.
11224         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11225         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11226         !FRAME_GROWS_DOWNWARD handling to...
11227         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11228         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11229         !FRAME_GROWS_DOWNWARD handling to...
11230         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11231         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11232         !FRAME_GROWS_DOWNWARD handling to...
11233         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11234         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11235         Redefine.
11236         (rs6000_starting_frame_offset): New function.
11237         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11238         !FRAME_GROWS_DOWNWARD handling to...
11239         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11240         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11241         !FRAME_GROWS_DOWNWARD handling to...
11242         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11243         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11244         !FRAME_GROWS_DOWNWARD handling to...
11245         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11246         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11247         (rs6000_starting_frame_offset): New function.
11248         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11249         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11250         * config/vax/vax.c (vax_starting_frame_offset): New function.
11251         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11252         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11253         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11254         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11255         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11256         * system.h (STARTING_FRAME_OFFSET): Poison.
11258 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11260         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11261         SCALAR_TYPE_MODE instead of TYPE_MODE.
11263 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11264             Alan Hayward  <alan.hayward@arm.com>
11265             David Sherwood  <david.sherwood@arm.com>
11267         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11269 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11270             Alan Hayward  <alan.hayward@arm.com>
11271             David Sherwood  <david.sherwood@arm.com>
11273         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11275 2017-10-23  Richard Biener  <rguenther@suse.de>
11277         PR tree-optimization/82129
11278         * tree-ssa-pre.c (bitmap_set_and): Remove.
11279         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11280         canonicalizing expressions in the set to those with lowest
11281         ID rather than taking that from the first edge.
11283 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11285         * combine.c (rtx_equal_for_field_assignment_p): Use
11286         byte_lowpart_offset.
11288 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11289             Alan Hayward  <alan.hayward@arm.com>
11290             David Sherwood  <david.sherwood@arm.com>
11292         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11293         to a SUBREG_PROMOTED_VAR.
11295 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11296             Alan Hayward  <alan.hayward@arm.com>
11297             David Sherwood  <david.sherwood@arm.com>
11299         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11300         (expand_debug_source_expr): Likewise.
11301         * combine.c (combine_simplify_rtx): Likewise.
11302         * cse.c (fold_rtx): Likewise.
11303         * optabs.c (expand_float): Likewise.
11304         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11305         (simplify_binary_operation_1): Likewise.
11307 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11308             Alan Hayward  <alan.hayward@arm.com>
11309             David Sherwood  <david.sherwood@arm.com>
11311         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11312         (record_promoted_value): Likewise.
11313         * expr.c (expand_expr_real_2): Likewise.
11314         * ree.c (update_reg_equal_equiv_notes): Likewise.
11315         (combine_set_extension): Likewise.
11316         * rtlanal.c (low_bitmask_len): Likewise.
11317         * simplify-rtx.c (neg_const_int): Likewise.
11318         (simplify_binary_operation_1): Likewise.
11320 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11321             Alan Hayward  <alan.hayward@arm.com>
11322             David Sherwood  <david.sherwood@arm.com>
11324         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11325         * regcprop.c (maybe_mode_change): Likewise.
11326         * reload1.c (alter_reg): Likewise.
11328 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11330         * inchash.h (inchash::hash::add_wide_int): New function.
11331         * lto-streamer-out.c (hash_tree): Use it.
11333 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11335         * inchash.h (inchash::hash::add_wide_int): Rename to...
11336         (inchash::hash::add_hwi): ...this.
11337         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11338         (polymorphic_call_target_hasher::hash): Likewise.
11339         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11340         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11341         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11342         * lto-streamer-out.c (hash_tree): Likewise.
11343         * optc-save-gen.awk: Likewise.
11344         * tree.c (add_expr): Likewise.
11346 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11348         PR target/52451
11349         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11350         for ordered inequality comparisons even with TARGET_IEEE_FP.
11352 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11354         PR target/82628
11355         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11356         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11357         Expand with cmp<dwi>_doubleword.
11359 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11361         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11362         List CET intrinsics.
11363         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11364         specific to -fcf-protection option.
11366 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11368         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11369         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11370         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11371         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11372         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11373         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11374         (extra_objs): Add cet.o for Linux/x86 targets.
11375         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11376         * config/i386/cet.c: New file.
11377         * config/i386/cetintrin.h: Likewise.
11378         * config/i386/t-cet: Likewise.
11379         * config/i386/cpuid.h (bit_SHSTK): New.
11380         (bit_IBT): Likewise.
11381         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11382         pass IBT and SHSTK bits.
11383         * config/i386/i386-builtin-types.def
11384         (VOID_FTYPE_UNSIGNED_PVOID): New.
11385         (VOID_FTYPE_UINT64_PVOID): Likewise.
11386         * config/i386/i386-builtin.def: Add CET intrinsics.
11387         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11388         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11389         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11390         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11391         prototype.
11392         * config/i386/i386.c (rest_of_insert_endbranch): New.
11393         (pass_data_insert_endbranch): Likewise.
11394         (pass_insert_endbranch): Likewise.
11395         (make_pass_insert_endbranch): Likewise.
11396         (ix86_notrack_prefixed_insn_p): Likewise.
11397         (ix86_target_string): Add -mibt, -mshstk flags.
11398         (ix86_option_override_internal): Add flag_cf_protection
11399         processing.
11400         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11401         (ix86_print_operand): Add 'notrack' prefix output.
11402         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11403         (ix86_expand_builtin): Expand CET intrinsics.
11404         (x86_output_mi_thunk): Add 'endbranch' instruction.
11405         * config/i386/i386.h (TARGET_IBT): New.
11406         (TARGET_IBT_P): Likewise.
11407         (TARGET_SHSTK): Likewise.
11408         (TARGET_SHSTK_P): Likewise.
11409         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11410         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11411         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11412         (builtin_setjmp_setup): New pattern.
11413         (builtin_longjmp): Likewise.
11414         (rdssp<mode>): Likewise.
11415         (incssp<mode>): Likewise.
11416         (saveprevssp): Likewise.
11417         (rstorssp): Likewise.
11418         (wrss<mode>): Likewise.
11419         (wruss<mode>): Likewise.
11420         (setssbsy): Likewise.
11421         (clrssbsy): Likewise.
11422         (nop_endbr): Likewise.
11423         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11424         options.
11425         * config/i386/immintrin.h: Include <cetintrin.h>.
11426         * config/i386/linux-common.h
11427         (file_end_indicate_exec_stack_and_cet): New prototype.
11428         (TARGET_ASM_FILE_END): New.
11430 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11432         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11433         latencies instead of having separate table; make difference between
11434         integer and float costs.
11435         * i386.h (processor_costs): Remove scalar_stmt_cost,
11436         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11437         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11438         vec_store_cost.
11439         * x86-tune-costs.h: Remove entries which has been removed in
11440         procesor_costs from all tables; make cond_taken_branch_cost
11441         and cond_not_taken_branch_cost COST_N_INSNS based.
11443 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11445         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11447 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11449         * config/i386/i386.md (isa): Remove fma_avx512f.
11450         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11451         <avx512>_fmadd_<mode>_mask3<round_name>,
11452         <avx512>_fmsub_<mode>_mask<round_name>,
11453         <avx512>_fmsub_<mode>_mask3<round_name>,
11454         <avx512>_fnmadd_<mode>_mask<round_name>,
11455         <avx512>_fnmadd_<mode>_mask3<round_name>,
11456         <avx512>_fnmsub_<mode>_mask<round_name>,
11457         <avx512>_fnmsub_<mode>_mask3<round_name>,
11458         <avx512>_fmaddsub_<mode>_mask<round_name>,
11459         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11460         <avx512>_fmsubadd_<mode>_mask<round_name>,
11461         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11462         (*vec_widen_umult_even_v16si<mask_name>,
11463         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11464         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11466 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11468         * extend.texi: Add 'nocf_check' documentation.
11469         * gimple.texi: Add second parameter to
11470         gimple_build_call_from_tree.
11471         * invoke.texi: Add -fcf-protection documentation.
11472         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11474 2017-10-20  Richard Biener  <rguenther@suse.de>
11476         PR tree-optimization/82473
11477         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11478         the largest input type.
11480 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11482         * c-attribs.c (handle_nocf_check_attribute): New function.
11483         (c_common_attribute_table): Add 'nocf_check' handling.
11484         * gimple-parser.c: Add second argument NULL to
11485         gimple_build_call_from_tree.
11486         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11487         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11488         call insn.
11489         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11490         * common.opt: Add fcf-protection flag.
11491         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11492         * flag-types.h: Add enum cf_protection_level.
11493         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11494         Add 'nocf_check' attribute propagation to gimple call.
11495         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11496         (gimple_build_call_from_tree): Update prototype.
11497         (gimple_call_nocf_check_p): New function.
11498         (gimple_call_set_nocf_check): Likewise.
11499         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11500         * ipa-icf.c: Add nocf_check attribute in statement hash.
11501         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11502         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11503         * toplev.c (process_options): Add flag_cf_protection handling.
11505 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11507         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11509 2017-10-20  Richard Biener  <rguenther@suse.de>
11511         PR tree-optimization/82603
11512         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11513         remove false predicated stores.
11515 2017-10-20  Richard Biener  <rguenther@suse.de>
11517         * graphite-isl-ast-to-gimple.c
11518         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11519         Remove return value and simplify, dump copied stmt after lhs
11520         adjustment.
11521         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11522         Reduce dump verbosity.
11523         (gsi_insert_earliest): Likewise.
11524         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11525         * graphite.c (print_global_statistics): Adjust dumping.
11526         (print_graphite_scop_statistics): Likewise.
11527         (print_graphite_statistics): Do not dump loops here.
11528         (graphite_transform_loops): But here.
11530 2017-10-20  Nicolas Roche  <roche@adacore.com>
11532         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11533         * configure: Regenerate.
11535 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11537         PR target/82158
11538         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11539         functions when optimizing replace GIMPLE_RETURN stmts with
11540         calls to __builtin_unreachable ().
11542         PR sanitizer/82595
11543         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11544         for -fsanitize=thread link of executables.
11545         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11546         link of executables.
11548         PR target/82370
11549         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11550         New mode iterators.
11551         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11552         define_insns for logical vector shifts to use VI248_AVX512BW
11553         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11554         condition, useless isa and prefix attributes.  Change the first
11555         2 of these define_insns to ...
11556         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11557         define_insn for avx512vl.
11558         (<shift_insn><mode>3): ... and this, new define_insn without
11559         masking for non-avx512vl.
11561         PR target/82370
11562         * config/i386/sse.md (*andnot<mode>3,
11563         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11564         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11565         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11566         not applied use empty suffix even for TARGET_AVX512VL.
11567         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11568         is applied, supply evex,evex or evex,evex,evex instead of just
11569         evex.
11571 2017-10-20  Julia Koval  <julia.koval@intel.com>
11573         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11574         (OPTION_MASK_ISA_GFNI_UNSET): New.
11575         (ix86_handle_option): Handle OPT_mgfni.
11576         * config/i386/cpuid.h (bit_GFNI): New.
11577         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11578         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11579         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11580         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11581         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11582         * config/i386/i386.opt: Add mgfni.
11584 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11586         * config/msp430/msp430.c (msp430_option_override): Disable
11587         -fdelete-null-pointer-checks.
11588         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11590 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11592         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11593         of x87 and SSE instructions.
11595 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11597         * asan.c (create_cond_insert_point): Do not update edge count.
11598         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11599         (afdo_propagate_circuit): Likewise.
11600         (afdo_calculate_branch_prob): Likewise.
11601         (afdo_annotate_cfg): Likewise.
11602         * basic-block.h (struct edge_def): Remove count.
11603         (edge_def::count): New accessor.
11604         * bb-reorder.c (rotate_loop): Update.
11605         (find_traces_1_round): Update.
11606         (connect_traces): Update.
11607         (sanitize_hot_paths): Update.
11608         * cfg.c (unchecked_make_edge): Update.
11609         (make_single_succ_edge): Update.
11610         (check_bb_profile): Update.
11611         (dump_edge_info): Update.
11612         (update_bb_profile_for_threading): Update.
11613         (scale_bbs_frequencies_int): Update.
11614         (scale_bbs_frequencies_gcov_type): Update.
11615         (scale_bbs_frequencies_profile_count): Update.
11616         (scale_bbs_frequencies): Update.
11617         * cfganal.c (connect_infinite_loops_to_exit): Update.
11618         * cfgbuild.c (compute_outgoing_frequencies): Update.
11619         (find_many_sub_basic_blocks): Update.
11620         * cfgcleanup.c (try_forward_edges): Update.
11621         (try_crossjump_to_edge): Update
11622         * cfgexpand.c (expand_gimple_cond): Update
11623         (expand_gimple_tailcall): Update
11624         (construct_exit_block): Update
11625         * cfghooks.c (verify_flow_info): Update
11626         (redirect_edge_succ_nodup): Update
11627         (split_edge): Update
11628         (make_forwarder_block): Update
11629         (duplicate_block): Update
11630         (account_profile_record): Update
11631         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11632         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11633         * cfgloopmanip.c (scale_loop_profile): Update.
11634         (loopify): Update.
11635         (lv_adjust_loop_entry_edge): Update.
11636         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11637         (force_nonfallthru_and_redirect): Update.
11638         (purge_dead_edges): Update.
11639         (rtl_flow_call_edges_add): Update.
11640         * cgraphunit.c (init_lowered_empty_function): Update.
11641         (cgraph_node::expand_thunk): Update.
11642         * gimple-pretty-print.c (dump_probability): Update.
11643         (dump_edge_probability): Update.
11644         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11645         * haifa-sched.c (sched_create_recovery_edges): Update.
11646         * hsa-gen.c (convert_switch_statements): Update.
11647         * ifcvt.c (dead_or_predicable): Update.
11648         * ipa-inline-transform.c (inline_transform): Update.
11649         * ipa-split.c (split_function): Update.
11650         * ipa-utils.c (ipa_merge_profiles): Update.
11651         * loop-doloop.c (add_test): Update.
11652         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11653         * lto-streamer-in.c (input_cfg): Update.
11654         (input_function): Update.
11655         * lto-streamer-out.c (output_cfg): Update.
11656         * modulo-sched.c (sms_schedule): Update.
11657         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11658         * predict.c (maybe_hot_edge_p): Update.
11659         (unlikely_executed_edge_p): Update.
11660         (probably_never_executed_edge_p): Update.
11661         (dump_prediction): Update.
11662         (drop_profile): Update.
11663         (propagate_unlikely_bbs_forward): Update.
11664         (determine_unlikely_bbs): Update.
11665         (force_edge_cold): Update.
11666         * profile.c (compute_branch_probabilities): Update.
11667         * reg-stack.c (better_edge): Update.
11668         * shrink-wrap.c (handle_simple_exit): Update.
11669         * tracer.c (better_p): Update.
11670         * trans-mem.c (expand_transaction): Update.
11671         (split_bb_make_tm_edge): Update.
11672         * tree-call-cdce.c: Update.
11673         * tree-cfg.c (gimple_find_sub_bbs): Update.
11674         (gimple_split_edge): Update.
11675         (gimple_duplicate_sese_region): Update.
11676         (gimple_duplicate_sese_tail): Update.
11677         (gimple_flow_call_edges_add): Update.
11678         (insert_cond_bb): Update.
11679         (execute_fixup_cfg): Update.
11680         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11681         * tree-complex.c (expand_complex_div_wide): Update.
11682         * tree-eh.c (lower_resx): Update.
11683         (unsplit_eh): Update.
11684         (cleanup_empty_eh_move_lp): Update.
11685         * tree-inline.c (copy_edges_for_bb): Update.
11686         (freqs_to_counts): Update.
11687         (copy_cfg_body): Update.
11688         * tree-ssa-dce.c (remove_dead_stmt): Update.
11689         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11690         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11691         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11692         (unloop_loops): Update.
11693         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11694         * tree-ssa-loop-split.c (connect_loops): Update.
11695         (split_loop): Update.
11696         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11697         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11698         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11699         * tree-ssa-reassoc.c (branch_fixup): Update.
11700         * tree-ssa-tail-merge.c (replace_block_by): Update.
11701         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11702         (compute_path_counts): Update.
11703         (update_profile): Update.
11704         (recompute_probabilities): Update.
11705         (update_joiner_offpath_counts): Update.
11706         (estimated_freqs_path): Update.
11707         (freqs_to_counts_path): Update.
11708         (clear_counts_path): Update.
11709         (ssa_fix_duplicate_block_edges): Update.
11710         (duplicate_thread_path): Update.
11711         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11712         (case_bit_test_cmp): Update.
11713         (collect_switch_conv_info): Update.
11714         (gen_inbound_check): Update.
11715         (do_jump_if_equal): Update.
11716         (emit_cmp_and_jump_insns): Update.
11717         * tree-tailcall.c (decrease_profile): Update.
11718         (eliminate_tail_call): Update.
11719         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11720         (vect_do_peeling): Update.
11721         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11722         * ubsan.c (ubsan_expand_null_ifn): Update.
11723         (ubsan_expand_ptr_ifn): Update.
11724         * value-prof.c (gimple_divmod_fixed_value): Update.
11725         (gimple_mod_pow2): Update.
11726         (gimple_mod_subtract): Update.
11727         (gimple_ic): Update.
11728         (gimple_stringop_fixed_value): Update.
11730 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11732         PR target/82618
11733         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11735 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11737         PR rtl-optimization/82395
11738         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11739         based on non_spilled_static_chain_regno_p.
11741 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11743         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11744         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11745         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11747 2017-10-19  Martin Sebor  <msebor@redhat.com>
11749         PR tree-optimization/82596
11750         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11752 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11754         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11755         (get_mem_refs_of_builtin_call): Likewise.
11756         * builtins.c (expand_builtin_apply): Adjust call to
11757         allocate_dynamic_stack_space.
11758         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11759         the third argument to allocate_dynamic_stack_space, otherwise -1.
11760         (expand_builtin): Deal with all alloca variants.
11761         (is_inexpensive_builtin): Likewise.
11762         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11763         * calls.c (special_function_p): Deal with all alloca variants.
11764         (initialize_argument_information): Adjust call to
11765         allocate_dynamic_stack_space.
11766         (expand_call): Likewise.
11767         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11768         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11769         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11770         use it for the stack usage computation.
11771         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11772         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11773         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11774         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11775         (in_loop_p): Remove first argument and useless check.
11776         (pass_walloca::execute): Remove useless test and adjust call to above.
11777         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11778         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11779         (gimplify_call_expr): Deal with all alloca variants.
11780         * hsa-gen.c (gen_hsa_alloca): Likewise.
11781         (gen_hsa_insns_for_call): Likewise.
11782         * ipa-pure-const.c (special_builtin_state): Likewise.
11783         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11784         * tree-object-size.c (alloc_object_size): Likewise.
11785         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11786         (call_may_clobber_ref_p_1): Likewise.
11787         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11788         (ccp_fold_stmt): Likewise.
11789         (optimize_stack_restore): Likewise.
11790         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11791         (mark_all_reaching_defs_necessary_1): Likewise.
11792         (propagate_necessity): Likewise.
11793         (eliminate_unnecessary_stmts): Likewise.
11794         * tree.c (build_common_builtin_nodes): Build
11795         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11796         (build_alloca_call_expr): New function.
11797         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11798         (CASE_BUILT_IN_ALLOCA): Likewise.
11799         (build_alloca_call_expr): Declare.
11800         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11802 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11804         PR debug/82509
11805         * dwarf2out.c (new_die_raw): New static inline function.
11806         (new_die): Use it to create the DIE.
11807         (add_AT_external_die_ref): Likewise.
11808         (clone_die): Likewise.
11809         (clone_as_declaration): Likewise.
11810         (dwarf2out_vms_debug_main_pointer): Likewise.
11811         (base_type_die): Likewise.  Remove early return for corner cases.
11812         Do not call add_pubtype on the DIE here.
11813         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11814         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11815         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11816         native order exists for base types, attach the DIE manually and call
11817         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11819 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11821         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11822         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11823         the mem into it.
11824         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11826 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11828         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11829         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11830         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11831         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11832         * opts.c (sanitizer_opts): Add builtin.
11833         * ubsan.c (instrument_builtin): New function.
11834         (pass_ubsan::execute): Call it.
11835         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11836         * doc/invoke.texi: Document -fsanitize=builtin.
11838         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11839         builtins, store max (log2 (align), 0) into uchar field instead of
11840         align into uptr field.
11841         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11842         store uchar 0 field instead of uptr 0 field.
11843         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11844         instead of passing one address of struct with 2 locations pass
11845         two addresses of structs with 1 location each.
11846         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11847         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11848         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11849         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11850         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11851         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11852         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11853         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11855 2017-10-19  Martin Liska  <mliska@suse.cz>
11857         PR driver/81829
11858         * file-find.c (remove_prefix): Remove.
11859         * file-find.h (remove_prefix): Likewise.
11860         * gcc-ar.c: Remove smartness of lookup.
11862 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11864         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11865         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11866         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11868 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11870         PR target/82580
11871         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11872         (setcc + and to xor + setcc): New peephole2.
11874 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11876         * doc/sourcebuild.texi (Test Directives, Variants of
11877         dg-require-support): Add dg-require-stack-size.
11879 2017-10-19  Martin Liska  <mliska@suse.cz>
11881         PR sanitizer/82517
11882         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11883         that have a large alignment.
11884         (gimplify_target_expr): Likewise.
11886 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11888         PR rtl-optimization/82602
11889         * ira.c (rtx_moveable_p): Return false for volatile asm.
11891 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11893         PR target/82580
11894         * config/i386/i386-modes.def (CCGZ): New CC mode.
11895         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11896         * config/i386/predicates.md (ix86_comparison_operator):
11897         Handle CCGZmode.
11898         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11899         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11900         with double-word subtraction.
11901         (put_condition_code): Handle CCGZmode.
11903 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11905         * wide-int.cc (debug (const wide_int &)): New.
11906         (debug (const wide_int *)): New.
11907         (debug (const widest_int &)): New.
11908         (debug (const widest_int *)): New.
11910 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11912         PR middle-end/82556
11913         * lra-constraints.c (curr_insn_transform): Use non-input operand
11914         instead of output one for matched reload.
11916 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11918         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11919         (tree-ssa-loop-ivopts.h): New header file.
11920         (struct builtin_info): New fields.
11921         (classify_builtin_1): Compute and record base and offset parts for
11922         memset builtin partition by calling strip_offset.
11923         (offset_cmp, fuse_memset_builtins): New functions.
11924         (finalize_partitions): Fuse adjacent memset partitions by calling
11925         above function.
11926         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11927         Expose the interface.
11928         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11930 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11932         PR tree-optimization/82574
11933         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
11934         that data reference must be executed exactly once per iteration
11935         against the outermost loop in nest.
11936         (classify_partition): Update call to above function.
11938 2017-10-18  Richard Biener  <rguenther@suse.de>
11940         PR tree-optimization/82591
11941         * graphite.c (graphite_transform_loops): Move code gen message
11942         printing ...
11943         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
11944         Here.  Handle scop_to_isl_ast failing.
11945         (scop_to_isl_ast): Limit the number of ISL operations.
11947 2017-10-18  Richard Biener  <rguenther@suse.de>
11949         * graphite-isl-ast-to-gimple.c
11950         (translate_isl_ast_to_gimple::set_rename): Simplify.
11951         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
11952         (graphite_copy_stmts_from_block): ... here.
11953         (copy_bb_and_scalar_dependences): Simplify.
11954         (add_parameters_to_ivs_params): Canonicalize.
11955         (generate_entry_out_of_ssa_copies): Simplify.
11956         * graphite-sese-to-poly.c (extract_affine_name): Simplify
11957         by passing in ISL dimension.
11958         (parameter_index_in_region_1): Rename to ...
11959         (parameter_index_in_region): ... this.
11960         (extract_affine): Adjust assert, pass down parameter index.
11961         (add_param_constraints): Use range-info when available.
11962         (build_scop_context): Adjust.
11963         * sese.c (new_sese_info): Adjust.
11964         (free_sese_info): Likewise.
11965         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
11966         Remove unused typedefs.
11967         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
11969 2017-10-18  Martin Liska  <mliska@suse.cz>
11971         * combine.c (simplify_compare_const): Add gcc_fallthrough.
11973 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11975         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
11976         (s390_sched_init): Do not reset s390_sched_state if we entered the
11977         current basic block via a fallthru edge and all others are unlikely.
11979 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11981         * config/s390/s390.c (NUM_SIDES): New variable.
11982         (LONGRUNNING_THRESHOLD): New variable.
11983         (LATENCY_FACTOR): New variable.
11984         (s390_sched_score): Decrease score for long-running instructions on
11985         wrong side.
11986         (s390_sched_variable_issue): Perform bookkeeping for long-running
11987         instructions.
11989 2017-10-18  Richard Biener  <rguenther@suse.de>
11991         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
11992         Simplify with removal of the parameter rename map.
11993         (set_rename): Likewise.
11994         (should_copy_to_new_region): Likewise.
11995         (graphite_copy_stmts_from_block): Likewise.
11996         (copy_bb_and_scalar_dependences): Remove initialization of
11997         unused copied_bb_map.
11998         (copy_def): Remove.
11999         (copy_internal_parameters): Likewise.
12000         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
12001         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12002         Use INTEGRAL_TYPE_P.
12003         (parameter_index_in_region_1): Rename to ...
12004         (assign_parameter_index_in_region): ... this.  Assert we have
12005         a parameter we handle.
12006         (scan_tree_for_params): Adjust.
12007         * sese.h (parameter_rename_map_t): Remove.
12008         (struct sese_info_t): Remove unused parameter_rename_map and
12009         copied_bb_map members.
12010         * sese.c (new_sese_info): Adjust.
12011         (free_sese_info): Likewise.
12013 2017-10-18  Martin Liska  <mliska@suse.cz>
12015         PR sanitizer/82545
12016         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
12017         on an abnormal edge.
12019 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12021         * doc/invoke.texi (ffunction-sections and fdata-sections):
12022         Update.
12024 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12026         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12027         the use statement can throw internally.
12029 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12031         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12032         any RTX present on the RHS of a SET.
12033         * compare-elim.c (try_eliminate_compare): Restore comment.
12035 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12037         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12038         may be also called on tcc_type nodes.
12039         * langhooks.c (lhd_tree_size): Likewise.
12041 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12043         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12044         format_warning_at_substring.
12045         (maybe_warn): Convert source_range * param to a location_t.  Pass
12046         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12047         (format_directive): Remove code to extract source_ranges and
12048         source_range * in favor of just a location_t.
12049         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12050         fmtwarn.
12051         * substring-locations.c (format_warning_va): Convert
12052         source_range * param to a location_t.
12053         (format_warning_at_substring): Likewise.
12054         * substring-locations.h (format_warning_va): Likewise.
12055         (format_warning_at_substring): Likewise.
12057 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12059         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12060         vec_scatter_store
12061         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12062         and scatter/gather ops.
12064         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12065         vec_gather_load and vec_scatter_store.
12066         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12067         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12068         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12069         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12070         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12071         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12073 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12075         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12076         Detect FTST instruction and handle its register pops.  Only pop
12077         second operand if can_pop_second_op is true.
12078         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12079         set can_pop_second_op to false in the compare_for_stack_reg call.
12081         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12082         output_fp_compare for stack register operands.
12083         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12084         instructions here.  Do not emit stack register pops here.  Assert
12085         that FCOMPP pops next to top stack register.  Rewrite function.
12087 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12089         PR middle-end/82577
12090         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12091         use DECL_ASSEMBLER_NAME_RAW.
12093         PR middle-end/82546
12094         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12095         TYPE nodes.
12097 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12098             Wilco Dijkstra <wilco.dijkstra@arm.com>
12100         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12101         converstion to Pmode from the buf_addr.
12103 2017-10-17  Richard Biener  <rguenther@suse.de>
12105         * graphite-dependences.c (scop_get_reads_and_writes): Change
12106         output parameters to references.
12108 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12110         PR 71026/tree-optimization
12111         * fold-const.c (distribute_real_division): Removed.
12112         (fold_binary_loc): Remove calls to distribute_real_divison.
12114 2017-10-17  Richard Biener  <rguenther@suse.de>
12116         * graphite-scop-detection.c
12117         (scop_detection::stmt_has_simple_data_refs_p): Always use
12118         the full nest as region.
12119         (try_generate_gimple_bb): Likewise.
12120         * sese.c (scalar_evolution_in_region): Simplify now that
12121         SCEV can handle instantiation in regions.
12122         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12123         in the non-loop part of a function if requested.
12125 2017-10-17  Richard Biener  <rguenther@suse.de>
12127         PR tree-optimization/82563
12128         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12129         New function.
12130         (graphite_regenerate_ast_isl): Call it.
12131         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12133 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12135         PR tree-optimization/82549
12136         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12137         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12138         bitpos return 0.
12140 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12142         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12143         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12144         _mm_maskz_reduce_ss): New.
12145         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12146         __builtin_ia32_reducess_mask): Ditto..
12147         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12148         * config/i386/sse.md (reduces<mode>): Renamed to ...
12149         (reduces<mode><mask_scalar_name>): ... this.
12150         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12151         Changed to ...
12152         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12153         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12155 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12157         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12158         * selftest-run-tests.c (selftest::run_tests): Call
12159         selftest::unique_ptr_tests_cc_tests.
12160         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12161         * unique-ptr-tests.cc: New file.
12163 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12165         PR sanitizer/82353
12166         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12167         locations.
12168         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12169         (make_hard_regno_born, make_hard_regno_dead): Update
12170         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12172 2017-10-16  Jeff Law  <law@redhat.com>
12174         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12176 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12178         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12180 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12182         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12183         with_cpu if we were configured for an e500v2 target cpu name.
12185 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12187         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12188         * doc/invoke.texi: Document +nodsp as a valid extension for
12189         -mcpu=cortex-m33.
12191 2017-10-16  Martin Liska  <mliska@suse.cz>
12193         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12194         (test_set_range): Likewise.
12195         (test_range_functions): Rename to ...
12196         (test_bit_in_range): ... this.
12197         (sbitmap_c_tests): Add new test.
12199 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12201         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12202         New.
12203         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12204         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12206 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12208         * config/aarch64/aarch64-builtins.c
12209         (aarch64_types_quadopu_lane_qualifiers): New.
12210         (TYPES_QUADOPU_LANE): New.
12211         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12212         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12213         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12214         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12215         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12216         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12217         (Vdottype, DOTPROD): New.
12218         (sur): Add SDOT and UDOT.
12220 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12222         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12223         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12224         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12225         Add TARGET_DOTPROD.
12226         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12227         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12228         Enable TARGET_DOTPROD.
12229         (cortex-a75.cortex-a55): Likewise.
12230         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12232 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12234         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12235         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12236         New.
12237         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12238         New.
12239         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12240         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12241         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12242         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12243         * config/arm/types.md (neon_dot, neon_dot_q): New.
12244         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12246 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12248         * config/arm/arm.h (TARGET_DOTPROD): New.
12249         * config/arm/arm.c (arm_arch_dotprod): New.
12250         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12251         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12252         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12253         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12254         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12255         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12256         * doc/invoke.texi (armv8.2-a): Document dotprod
12258 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12260         * i386.c (ix86_vec_cost): New function.
12261         (ix86_rtx_costs): Handle vector operations better.
12262         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12263         * x86-tune-costs.h: Add new costs to all tables.
12265 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12267         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12268         operations.
12269         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12270         divsd, sqrtss and sqrtsd
12271         * x86-tune-costs.h: Add new entries to all costs.
12272         (znver1_cost): Fix to match real instruction latencies.
12274 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12275             Michael Collison <michael.collison@arm.com>
12277         * compare-elim.c: Include emit-rtl.h.
12278         (can_merge_compare_into_arith): New function.
12279         (try_validate_parallel): Likewise.
12280         (try_merge_compare): Likewise.
12281         (try_eliminate_compare): Call the above when no previous clobber
12282         is available.
12283         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12284         dataflow problems.
12286 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12288         PR middle-end/62263
12289         PR middle-end/82498
12290         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12291         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12293         PR middle-end/62263
12294         PR middle-end/82498
12295         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12296         to be any operand_equal_p operands.  For & (B - 1) require
12297         B to be power of 2.  Recognize
12298         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12300 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12302         PR bootstrap/82553
12303         * optabs.c (expand_memory_blockage): Fix call of
12304         targetm.have_memory_blockage.
12306 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12308         PR bootstrap/82548
12309         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12310         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12311         objects to extra_objs instead of overwriting it.
12313 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12315         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12316         Use any_fp_register_operand as operand[3] predicate.  Simplify
12317         equality test for operands[2] and operands[4] memory location.
12318         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12319         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12320         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12321         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12322         any_fp_register_operand as operand[1] predicate.  Simplify
12323         equality test for operands[0] and operands[3] memory location.
12324         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12325         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12326         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12328 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12330         * target-insns.def: Add memory_blockage.
12331         * optabs.c (expand_memory_blockage): New function.
12332         (expand_asm_memory_barrier): Rename ...
12333         (expand_asm_memory_blockage): ... to this.
12334         (expand_mem_thread_fence): Call expand_memory_blockage
12335         instead of expand_asm_memory_barrier.
12336         (expand_mem_singnal_fence): Ditto.
12337         (expand_atomic_load): Ditto.
12338         (expand_atomic_store): Ditto.
12339         * doc/md.texi (Standard Pattern Names For Generation):
12340         Document memory_blockage instruction pattern.
12342 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12344         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12345         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12346         * config/rl78/rl78.md: New define_expand "adddi3".
12348 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12350         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12351         are set correctly.
12353 2017-10-13  Jeff Law  <law@redhat.com>
12355         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12357 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12359         PR target/82274
12360         * internal-fn.c (expand_mul_overflow): If both operands have
12361         the same highpart of -1 or 0 and the topmost bit of lowpart
12362         is different, overflow is if res <= 0 rather than res < 0.
12364 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12366         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12367         TARGET_P9_VECTOR code for unaligned_load case.
12369 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12371         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12373 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12375         * tree-core.h (tree_contains_struct): Make bool.
12376         * tree.c (tree_contains_struct): Likewise.
12377         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12378         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12379         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12380         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12381         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12383 2017-10-13  Richard Biener  <rguenther@suse.de>
12385         * graphite-isl-ast-to-gimple.c
12386         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12387         parameters and dominance check.
12388         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12389         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12390         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12391         Do not update SSA form here or do intermediate IL verification.
12392         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12393         (graphite_initialize): Remove check on the number of loops in
12394         the function and inline into graphite_transform_loops.
12395         (graphite_finalize): Inline into graphite_transform_loops.
12396         (graphite_transform_loops): Perform SSA update and IL verification
12397         here.
12398         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12400 2017-10-13  Richard Biener  <rguenther@suse.de>
12402         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12403         graphite_expression_type_precision): Avoid global constructor
12404         by moving ...
12405         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12406         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12407         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12408         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12409         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12410         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12412 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12414         PR target/82499
12415         * config/i386/i386.h (ix86_red_zone_size): New.
12416         * config/i386/i386.md (push peephole2s): Replace
12417         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12419 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12420             Alan Hayward  <alan.hayward@arm.com>
12421             David Sherwood  <david.sherwood@arm.com>
12423         * combine.c (can_change_dest_mode): Reject changes in
12424         REGMODE_NATURAL_SIZE.
12426 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12427             Alan Hayward  <alan.hayward@arm.com>
12428             David Sherwood  <david.sherwood@arm.com>
12430         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12431         (expand_debug_source_expr): Likewise.
12432         * combine.c (combine_simplify_rtx): Likewise.
12433         * cse.c (fold_rtx): Likewise.
12434         * fwprop.c (canonicalize_address): Likewise.
12435         * targhooks.c (default_shift_truncation_mask): Likewise.
12437 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12438             Alan Hayward  <alan.hayward@arm.com>
12439             David Sherwood  <david.sherwood@arm.com>
12441         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12442         (widened_mode): Likewise.
12443         (expand_unop): Likewise.
12444         * ree.c (transform_ifelse): Likewise.
12445         (merge_def_and_ext): Likewise.
12446         (combine_reaching_defs): Likewise.
12447         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12449 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12450             Alan Hayward  <alan.hayward@arm.com>
12451             David Sherwood  <david.sherwood@arm.com>
12453         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12454         * combine.c (gen_lowpart_for_combine): Likewise.
12455         * dwarf2out.c (rtl_for_decl_location): Likewise.
12456         * final.c (alter_subreg): Likewise.
12457         * rtlhooks.c (gen_lowpart_general): Likewise.
12458         (gen_lowpart_if_possible): Likewise.
12460 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12461             Alan Hayward  <alan.hayward@arm.com>
12462             David Sherwood  <david.sherwood@arm.com>
12464         * calls.c (expand_call): Use subreg_lowpart_offset.
12465         * cse.c (cse_insn): Likewise.
12466         * regcprop.c (copy_value): Likewise.
12467         (copyprop_hardreg_forward_1): Likewise.
12469 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12471         PR target/82524
12472         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12473         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12474         =Q constraints to +Q and into insn condition add check
12475         that operands[0] and operands[1] are equal.
12476         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12477         =Q constraints to +Q and into insn condition add check
12478         that operands[0] is equal to either operands[1] or operands[2].
12480         PR target/82498
12481         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12482         instead of handling MINUS_EXPR twice (once for each argument),
12483         canonicalize operand order and handle just once, use rtype where
12484         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12486         PR target/82498
12487         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12488         any values of __C while still being pattern recognizable as a simple
12489         rotate instruction.
12491 2017-10-13  Richard Biener  <rguenther@suse.de>
12493         PR tree-optimization/82451
12494         Revert
12495         2017-10-02  Richard Biener  <rguenther@suse.de>
12497         PR tree-optimization/82355
12498         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12499         a mapping for the enclosing loop but avoid generating one for
12500         the loop tree root.
12501         (copy_bb_and_scalar_dependences): Remove premature codegen
12502         error on PHIs in blocks duplicated into multiple places.
12503         * graphite-scop-detection.c
12504         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12505         in the region use it as loop and nest to analyze the DR in.
12506         (try_generate_gimple_bb): Likewise.
12507         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12508         (add_loop_constraints): For blocks in a loop not in the region
12509         create a dimension with a single iteration.
12510         * sese.h (gbb_loop_at_index): Remove assert.
12512         * cfgloop.c (loop_preheader_edge): For the loop tree root
12513         return the single successor of the entry block.
12514         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12515         Reset the SCEV hashtable and niters.
12516         * graphite-scop-detection.c
12517         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12518         assert that we only have POLYNOMIAL_CHREC that vary in loops
12519         contained in the region.
12520         (scop_detection::graphite_can_represent_expr): Adjust.
12521         (scop_detection::stmt_has_simple_data_refs_p): For loops
12522         not in the region set loop to NULL.  The nest is now the
12523         entry edge to the region.
12524         (try_generate_gimple_bb): Likewise.
12525         * sese.c (scalar_evolution_in_region): Adjust for
12526         instantiate_scev change.
12527         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12528         Make nest parameter the edge into the region.
12529         (create_data_ref): Likewise.
12530         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12531         entry edge into a region and adjust instantiate_scev calls.
12532         (create_data_ref): Likewise.
12533         (graphite_find_data_references_in_stmt): Likewise.
12534         (find_data_references_in_stmt): Pass the loop preheader edge
12535         from the nest argument.
12536         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12537         parameter the edge into the region.
12538         (instantiate_parameters): Use the loop preheader edge as entry.
12539         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12540         NULL loop.
12541         (get_instantiated_value_entry): Make instantiate_below parameter
12542         the edge into the region.
12543         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12544         when we cannot use loop-based instantiation instantiate by
12545         walking use-def chains.
12546         (instantiate_scev_poly): Adjust.
12547         (instantiate_scev_binary): Likewise.
12548         (instantiate_scev_convert): Likewise.
12549         (instantiate_scev_not): Likewise.
12550         (instantiate_array_ref): Remove.
12551         (instantiate_scev_3): Likewise.
12552         (instantiate_scev_2): Likewise.
12553         (instantiate_scev_1): Likewise.
12554         (instantiate_scev_r): Do not blindly handle N-operand trees.
12555         Do not instantiate array-refs.  Handle all constants and invariants.
12556         (instantiate_scev): Make instantiate_below parameter
12557         the edge into the region.
12558         (resolve_mixers): Use the loop preheader edge for the region
12559         parameter to instantiate_scev_r.
12560         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12562 2017-10-13  Richard Biener  <rguenther@suse.de>
12564         PR tree-optimization/82525
12565         * graphite-isl-ast-to-gimple.c
12566         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12567         out from ...
12568         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12569         Fail code generation when we cannot represent the isl integer.
12570         (binary_op_to_tree): Elide modulo operations that are no-ops
12571         in the type we code generate.  Remove now superfluous code
12572         generation errors.
12574 2017-10-13  Richard Biener  <rguenther@suse.de>
12576         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12577         (scop_detection::harmful_loop_in_region): Remove premature
12578         IV type restriction.
12579         (scop_detection::graphite_can_represent_scev): We can handle
12580         pointer IVs just fine.
12582 2017-10-13  Alan Modra  <amodra@gmail.com>
12584         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12585         "Clobbers and Scratch Registers".  Add paragraph on
12586         alternative to clobbers for scratch registers and OpenBLAS
12587         example.
12589 2017-10-13  Alan Modra  <amodra@gmail.com>
12591         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12592         example of a memory input for a string of known length.  Move
12593         commentary out of table.  Add a number of new examples
12594         covering array memory inputs.
12596 2017-10-12  Martin Liska  <mliska@suse.cz>
12598         PR tree-optimization/82493
12599         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12600         (test_range_functions): New function.
12601         (sbitmap_c_tests): Likewise.
12602         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12603         * selftest.h (sbitmap_c_tests): New function.
12605         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12607 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12609         * config/rs6000/amo.h: Fix spacing issue.
12611 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12613         PR target/82498
12614         * config/i386/i386.md (*ashl<mode>3_mask_1,
12615         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12616         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12617         patterns.
12619 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12621         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12622         (profile_probability): Set max_probability
12623         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12624         in temporaries.
12625         * profile-count.c (profile_probability::differs_from_p): Do not
12626         rely on max_probaiblity == 10000
12628 2017-10-12  Jeff Law  <law@redhat.com>
12630         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12631         negative offsets.
12633 2017-10-12  Martin Sebor  <msebor@redhat.com>
12635         PR other/82301
12636         PR c/82435
12637         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12638         (handle_alias_pairs): Call it.
12639         * common.opt (-Wattribute-alias): New option.
12640         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12641         * doc/invoke.texi (-Wattribute-alias): Document.
12643 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12645         Revert
12646         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12647         PR sanitizer/82353
12648         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12649         locations.
12650         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12651         (make_hard_regno_born, make_hard_regno_dead): Update
12652         bb_killed_pseudos and bb_gen_pseudos.
12654 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12656         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12658 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12660         * config/alpha/alpha.c (alpha_split_conditional_move):
12661         Use std::swap instead of manually swapping.
12662         (alpha_stdarg_optimize_hook): Ditto.
12663         (alpha_canonicalize_comparison): Ditto.
12665 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12667         * tree-loop-distribution.c (struct builtin_info): New struct.
12668         (struct partition): Refactor fields into struct builtin_info.
12669         (partition_free): Free struct builtin_info.
12670         (build_size_arg_loc, build_addr_arg_loc): Delete.
12671         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12672         information from struct builtin_info.
12673         (find_single_drs): New function refactored from classify_partition.
12674         Also moved builtin validity checks to this function.
12675         (compute_access_range, alloc_builtin): New functions.
12676         (classify_builtin_st, classify_builtin_ldst): New functions.
12677         (classify_partition): Refactor code into functions find_single_drs,
12678         classify_builtin_st and classify_builtin_ldst.
12679         (distribute_loop): Don't do runtime alias check when distributing
12680         loop nest.
12681         (find_seed_stmts_for_distribution): New function.
12682         (pass_loop_distribution::execute): Refactor code finding seed
12683         stmts into above function.  Support distribution for the innermost
12684         two-level loop nest.  Adjust dump information.
12686 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12688         * tree-loop-distribution.c: Adjust the general comment.
12689         (NUM_PARTITION_THRESHOLD): New macro.
12690         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12691         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12692         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12693         in call to build_partition_graph.
12694         (finalize_partitions): New parameter.  Make loop distribution more
12695         conservative by fusing more partitions.
12696         (distribute_loop): Don't do runtime alias check in case of loop nest
12697         distribution.
12698         (find_seed_stmts_for_distribution): New function.
12699         (prepare_perfect_loop_nest): New function.
12700         (pass_loop_distribution::execute): Refactor code finding seed stmts
12701         and loop nest into above functions.  Support loop nest distribution.
12702         Adjust dump information accordingly.
12704 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12706         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12707         and set PTYPE_SEQUENTIAL for merged partition.
12709 2017-10-12  Richard Biener  <rguenther@suse.de>
12711         PR tree-optimization/69728
12712         Revert
12713         2017-09-19  Richard Biener  <rguenther@suse.de>
12715         PR tree-optimization/69728
12716         * graphite-sese-to-poly.c (schedule_error): New global.
12717         (add_loop_schedule): Handle empty domain by failing the
12718         schedule.
12719         (build_original_schedule): Handle schedule_error.
12721         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12722         domain by returning an unchanged schedule.
12724 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12726         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12727         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12729 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12731         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12732         Handle params.def.
12734 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12736         PR c++/82159
12737         * expr.c (store_field): Don't optimize away bitsize == 0 store
12738         from CALL_EXPR with addressable return type.
12740 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12742         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12743         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12744         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12745         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12746         TARGET_ISEL instead of TARGET_ISEL<sel>.
12748 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12750         * config/rs6000/rs6000.c
12751         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12753 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12755         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12756         Move up in file.
12757         (reg_or_cint_operand): Fix comment.
12758         (reg_or_zero_operand): New predicate.
12759         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12760         * config/rs6000/rs6000.c (output_isel): Delete.
12761         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12762         instead of reg_or_cint_operand.  Output instruction directly (not via
12763         output_isel).
12764         (isel_unsigned_<mode>): Ditto.
12765         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12766         gpc_reg_operand.  Add an instruction alternative for this.  Output
12767         instruction directly.
12768         (*isel_reversed_unsigned_<mode>): Ditto.
12770 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12772         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12773         (TARGET_CANONICALIZE_COMPARISON): Define.
12775 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12777         PR target/81422
12778         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12779         Check whether the dest is REG before adding REG_EQUIV note.
12781 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12783         PR sanitizer/82353
12784         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12785         locations.
12786         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12787         (make_hard_regno_born, make_hard_regno_dead): Update
12788         bb_killed_pseudos and bb_gen_pseudos.
12790 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12792         * incpath.h (enum incpath_kind): Name enum, prefix values.
12793         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12794         * incpath.c (heads, tails): Use INC_MAX.
12795         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12796         (merge_include_chains, split_quote_chain,
12797         register_include_chains): Update incpath_kind names.
12798         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12799         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12800         names.
12801         (add_framework_path, darwin_register_objc_includes): Likewise.
12802         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12804 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12806         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12807         Do not use float_operator operator predicate.
12808         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12809         * config/i386/predicates.md (float_operator): Remove predicate.
12811 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12813         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12814         (*jccxf_i387): Ditto.
12815         (*jcc<mode>_i387): Ditto.
12816         (*jccu<mode>_i387): Ditto.
12817         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12818         (*jcc_*_i387 splitters): Remove.
12819         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12820         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12821         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12822         Remove predicate.
12824 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12826         * profile-count.h (slow_safe_scale_64bit): New function.
12827         (safe_scale_64bit): New inline.
12828         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12829         * profile-count.c: Include wide-int.h
12830         (slow_safe_scale_64bit): New.
12832 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12834         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12835         HAS_DECL_ASSEMBLER_NAME_P.
12836         * gimple-expr.c (gimple_decl_printable_name: Check
12837         HAS_DECL_ASSEMBLER_NAME_P too.
12838         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12839         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12840         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12841         * passes.c (rest_of_decl_compilation): Check
12842         HAS_DECL_ASSEMBLER_NAME_P too.
12843         * recog.c (verify_changes): Likewise.
12844         * tree-pretty-print.c (dump_decl_name): Likewise.
12845         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12847         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12848         (SET_DECL_ASSEMBLER_NAME): Use it.
12849         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12850         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12851         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12853 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12855         * config.gcc (i386, x86_64): Add extra objects.
12856         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12857         (ix86_min_insn_size): Declare.
12858         (ix86_issue_rate): Declare.
12859         (ix86_adjust_cost): Declare.
12860         (ia32_multipass_dfa_lookahead): Declare.
12861         (ix86_macro_fusion_p): Declare.
12862         (ix86_macro_fusion_pair_p): Declare.
12863         (ix86_bd_has_dispatch): Declare.
12864         (ix86_bd_do_dispatch): Declare.
12865         (ix86_core2i7_init_hooks): Declare.
12866         (ix86_atom_sched_reorder): Declare.
12867         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12868         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12869         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12870         (rip_relative_addr_p): Rename to ...
12871         (ix86_rip_relative_addr_p): ... this one; export.
12872         (memory_address_length): Update.
12873         (ix86_issue_rate): Move to x86-tune-sched.c.
12874         (ix86_flags_dependent): Move to x86-tune-sched.c.
12875         (ix86_agi_dependent): Move to x86-tune-sched.c.
12876         (exact_dependency_1): Move to x86-tune-sched.c.
12877         (exact_store_load_dependency): Move to x86-tune-sched.c.
12878         (ix86_adjust_cost): Move to x86-tune-sched.c.
12879         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12880         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12881         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12882         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12883         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12884         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12885         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12886         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12887         (min_insn_size): Rename to ...
12888         (ix86_min_insn_size): ... this one; export.
12889         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12890         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12891         (core2i7_first_cycle_multipass_backtrack): Move to
12892         x86-tune-sched-core.c.
12893         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12894         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12895         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12896         (ix86_avoid_jump_mispredicts): Update.
12897         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12898         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12899         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12900         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12901         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12902         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12903         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12904         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12905         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12906         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12907         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12908         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12909         (BIG): Move to ix86-tune-sched-bd.c.
12910         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12911         (enum insn_path): Move to ix86-tune-sched-bd.c.
12912         (get_mem_group): Move to ix86-tune-sched-bd.c.
12913         (is_cmp): Move to ix86-tune-sched-bd.c.
12914         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12915         (is_branch): Move to ix86-tune-sched-bd.c.
12916         (is_prefetch): Move to ix86-tune-sched-bd.c.
12917         (init_window): Move to ix86-tune-sched-bd.c.
12918         (allocate_window): Move to ix86-tune-sched-bd.c.
12919         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12920         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12921         (process_end_window): Move to ix86-tune-sched-bd.c.
12922         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12923         (find_constant): Move to ix86-tune-sched-bd.c.
12924         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12925         (has_immediate): Move to ix86-tune-sched-bd.c.
12926         (get_insn_path): Move to ix86-tune-sched-bd.c.
12927         (get_insn_group): Move to ix86-tune-sched-bd.c.
12928         (count_num_restricted): Move to ix86-tune-sched-bd.c.
12929         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12930         (add_insn_window): Move to ix86-tune-sched-bd.c.
12931         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12932         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12933         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12934         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12935         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12936         (do_dispatch): Move to ix86-tune-sched-bd.c.
12937         (has_dispatch): Move to ix86-tune-sched-bd.c.
12938         * config/i386/t-i386: Add new object files.
12939         * config/i386/x86-tune-costs.h: New file.
12940         * config/i386/x86-tune-sched-atom.c: New file.
12941         * config/i386/x86-tune-sched-bd.c: New file.
12942         * config/i386/x86-tune-sched-core.c: New file.
12943         * config/i386/x86-tune-sched.c: New file.
12945 2017-10-11  Liu Hao  <lh_mouse@126.com>
12947         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
12948         the generic version below instead.
12949         (should_colorize): Recognize Windows consoles as terminals
12950         for MinGW targets.
12951         * pretty-print.c [__MINGW32__] (write_all): New function.
12952         [__MINGW32__] (find_esc_head): Likewise.
12953         [__MINGW32__] (find_esc_terminator): Likewise.
12954         [__MINGW32__] (eat_esc_sequence): Likewise.
12955         [__MINGW32__] (mingw_ansi_fputs): New function that handles
12956         ANSI escape codes.
12957         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
12958         for MinGW targets.
12960 2017-10-11  Richard Biener  <rguenther@suse.de>
12962         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
12963         Properly call analyze_scalar_evolution with the loop of the stmt.
12965 2017-10-11  Richard Biener  <rguenther@suse.de>
12967         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
12968         * tree-core.h (tree_base): Add chrec_var union member.
12969         * tree.h (CHREC_VAR): Remove.
12970         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
12971         * tree-chrec.h (build_polynomial_chrec): Adjust.
12972         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
12973         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
12975 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
12977         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
12978         * match.pd: ... here.
12979         ((T) X == (T) Y): Relax condition.
12981 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
12983         PR tree-optimization/82472
12984         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
12985         comment.
12986         (break_alias_scc_partitions): Update postorder number.
12988 2017-10-11  Martin Liska  <mliska@suse.cz>
12990         PR sanitizer/82490
12991         * opts.c (parse_no_sanitize_attribute): Do not use error_value
12992         variable.
12993         * opts.h (parse_no_sanitize_attribute): Remove last argument.
12995 2017-10-11  Martin Liska  <mliska@suse.cz>
12997         * print-rtl.c (print_insn): Move declaration of idbuf
12998         to same scope as name.
13000 2017-10-11  Martin Liska  <mliska@suse.cz>
13002         Revert r253637:
13004         PR sanitizer/82484
13005         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13006         volatile arguments.
13008 2017-10-11  Martin Liska  <mliska@suse.cz>
13010         PR sanitizer/82484
13011         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13012         volatile arguments.
13014 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
13016         * config.gcc (default_gnu_indirect_function): Default to yes for
13017         arm*-*-linux* with glibc.
13019 2017-10-11  Richard Biener  <rguenther@suse.de>
13021         * tree-scalar-evolution.c (get_scalar_evolution): Handle
13022         default-defs and types we do not want to analyze.
13023         (interpret_loop_phi): Replace unreachable code with an assert.
13024         (compute_scalar_evolution_in_loop): Remove and inline ...
13025         (analyze_scalar_evolution_1): ... here, replacing condition with
13026         what makes the intent clearer.  Remove handling of cases
13027         get_scalar_evolution now handles.
13029 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13031         PR rtl-optimization/81434
13032         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13033         comment for main loop.  In sched_group_found if, also add checks for
13034         pass and min_cost_group.
13036 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13038         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13039         (rs6000_insn_cost): New function.
13040         * config/rs6000/rs6000.md (cost): New attribute.
13042 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13043             H.J. Lu  <hongjiu.lu@intel.com>
13045         PR target/79565
13046         PR target/82483
13047         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13048         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13049         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13050         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13051         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13052         OPTION_MASK_ISA_MMX and some other bit set require both
13053         mmx and the ISAs without the mmx bit.
13054         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13055         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13056         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13057         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13058         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13059         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13060         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13061         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13062         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13063         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13064         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13065         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13066         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13067         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13068         Add OPTION_MASK_ISA_MMX.
13070 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13072         * config.gcc (armv7*-*-freebsd*): New target.
13073         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13075 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13077         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13078         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13079         spot in the file.
13081 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13083         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13084         a template parameter.
13085         (WIDE_INT_REF_FOR): Update accordingly.
13086         * tree.h (wi::int_traits <const_tree>): Delete.
13087         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13088         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13089         (wi::tree_to_wide_ref): New typedef.
13090         (wi::to_wide): New function.
13091         * calls.c (get_size_range): Use wi::to_wide when operating on
13092         trees as wide_ints.
13093         * cgraph.c (cgraph_node::create_thunk): Likewise.
13094         * config/i386/i386.c (ix86_data_alignment): Likewise.
13095         (ix86_local_alignment): Likewise.
13096         * dbxout.c (stabstr_O): Likewise.
13097         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13098         * expr.c (const_vector_from_tree): Likewise.
13099         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13100         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13101         (fold_negate_expr_1, int_const_binop_1, const_binop)
13102         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13103         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13104         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13105         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13106         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13107         (fold_not_const, round_up_loc): Likewise.
13108         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13109         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13110         (alloca_call_type): Likewise.
13111         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13112         * godump.c (go_output_typedef): Likewise.
13113         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13114         * internal-fn.c (get_min_precision): Likewise.
13115         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13116         * ipa-polymorphic-call.c
13117         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13118         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13119         (ipa_modify_call_arguments): Likewise.
13120         * match.pd: Likewise.
13121         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13122         * print-tree.c (print_node_brief, print_node): Likewise.
13123         * stmt.c (expand_case): Likewise.
13124         * stor-layout.c (layout_type): Likewise.
13125         * tree-affine.c (tree_to_aff_combination): Likewise.
13126         * tree-cfg.c (group_case_labels_stmt): Likewise.
13127         * tree-data-ref.c (dr_analyze_indices): Likewise.
13128         (prune_runtime_alias_test_list): Likewise.
13129         * tree-dump.c (dequeue_and_dump): Likewise.
13130         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13131         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13132         * tree-pretty-print.c (dump_generic_node): Likewise.
13133         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13134         (simple_iv_with_niters): Likewise.
13135         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13136         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13137         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13138         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13139         (refine_value_range_using_guard, number_of_iterations_ne_max)
13140         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13141         (get_cst_init_from_scev, record_nonwrapping_iv)
13142         (scev_var_range_cant_overflow): Likewise.
13143         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13144         * tree-ssa-pre.c (compute_avail): Likewise.
13145         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13146         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13147         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13148         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13149         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13150         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13151         (dump_case_nodes, try_switch_expansion): Likewise.
13152         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13153         (vect_do_peeling): Likewise.
13154         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13155         * tree-vect-stmts.c (vectorizable_load): Likewise.
13156         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13157         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13158         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13159         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13160         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13161         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13162         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13163         (evrp_dom_walker::before_dom_children): Likewise.
13164         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13165         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13166         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13167         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13168         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13169         (get_range_pos_neg): Likewise.
13170         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13171         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13172         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13173         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13174         * config/avr/avr.c (avr_fold_builtin): Likewise.
13175         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13176         * config/msp430/msp430.c (msp430_attr): Likewise.
13177         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13178         * config/powerpcspe/powerpcspe-c.c
13179         (altivec_resolve_overloaded_builtin): Likewise.
13180         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13181         (rs6000_expand_ternop_builtin): Likewise.
13182         * config/rs6000/rs6000-c.c
13183         (altivec_resolve_overloaded_builtin): Likewise.
13184         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13185         (rs6000_expand_ternop_builtin): Likewise.
13186         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13188 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13190         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13191         when copying loop nest with only one inner loop.
13193 2017-10-10  Richard Biener  <rguenther@suse.de>
13195         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13196         blocks if SCEV is active.
13197         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13198         dead code.
13199         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13200         (scev_initialize): Assert we are not yet initialized.
13202 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13204         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13205         inner loop's exit stmt by making it always exit the loop, otherwise
13206         we would generate an infinite empty loop.
13208 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13210         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13211         renaming variables in new preheader if it's deleted.
13213 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13215         * tree-loop-distribution.c (struct partition): Remove unused field
13216         loops of the structure.
13217         (partition_alloc, partition_free): Ditto.
13218         (build_rdg_partition_for_vertex): Ditto.
13220 2017-10-09  Jeff Law  <law@redhat.com>
13222         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13223         return type to match prototype and documentation.
13225 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13227         * config/rs6000/rs6000.c (processor_costs): Move to ...
13228         * config/rs6000/rs6000.h: ... here.
13229         (rs6000_cost): Declare.
13231 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13233         * except.c (setjmp_fn): New global variable.
13234         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13235         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13236         if DONT_USE_BUILTIN_SETJMP is defined.
13238 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13240         * target.def (insn_cost): New hook.
13241         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13242         * doc/tm.texi: Regenerate.
13243         * rtlanal.c (insn_cost): Use the new hook.
13245 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13247         * combine.c (combine_validate_cost): Compute the new insn_cost,
13248         not just pattern_cost.
13249         (try_combine): Adjust comment.
13251 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13253         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13254         insn_cost.
13255         * combine.c (uid_insn_cost): Adjust comment.
13256         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13257         of insn_rtx_cost
13258         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13259         * dse.c (find_shift_sequence): Ditto.
13260         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13261         (bb_valid_for_noce_process_p): Use pattern_cost.
13262         * rtl.h (insn_rtx_cost): Delete.
13263         (pattern_cost): New prototype.
13264         (insn_cost): New prototype.
13265         * rtlanal.c (insn_rtx_cost): Rename to...
13266         (pattern_cost): ... this.
13267         (insn_cost): New.
13269 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13271         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13272         (*jcc<mode>_0_r_i387): Ditto.
13273         (*jccxf_r_i387): Ditto.
13274         (*jcc<mode>_r_i387): Ditto.
13275         (*jccu<mode>_r_i387): Ditto.
13276         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13277         (*jcc): Rename from *jcc_1.
13279 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13281         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13282         deferred rescans after the lvx/stvx recombination pre-pass.
13284 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13286         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13287         memory operation instruction support.
13288         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13289         (rs6000-ibm-aix[789]*): Likewise.
13290         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13291         Document new functions.
13293 2017-10-09  Richard Biener  <rguenther@suse.de>
13295         PR tree-optimization/82397
13296         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13297         equality only for semantically equal trees.
13299 2017-10-09  Richard Biener  <rguenther@suse.de>
13301         PR tree-optimization/82449
13302         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13303         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13304         allow constant addresses.
13305         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13306         are linear.
13308 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13310         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13311         flags.
13313 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13315         PR target/82463
13316         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13317         definitions.
13319 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13321         PR target/82465
13322         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13324 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13326         PR target/82464
13327         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13328         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13330 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13332         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13333         (WI_BINARY_PREDICATE_RESULT): Likewise.
13334         (wi::binary_traits::operator_result): New type.
13335         (wi::binary_traits::predicate_result): Likewise.
13336         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13337         (generic_wide_int::operator==, generic_wide_int::operator!=)
13338         (generic_wide_int::operator&, generic_wide_int::and_not)
13339         (generic_wide_int::operator|, generic_wide_int::or_not)
13340         (generic_wide_int::operator^, generic_wide_int::operator+
13341         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13342         Delete.
13343         (operator~, unary operator-, operator==, operator!=, operator&)
13344         (operator|, operator^, operator+, binary operator-, operator*): New
13345         functions.
13346         * expr.c (get_inner_reference): Use wi::bit_and_not.
13347         * fold-const.c (fold_binary_loc): Likewise.
13348         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13349         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13350         (bit_value_binop): Likewise.
13351         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13352         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13353         (extract_range_from_binary_expr_1): Likewise.
13354         (masked_increment): Likewise.
13355         (simplify_bit_ops_using_ranges): Likewise.
13357 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13359         PR hsa/82416
13360         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13361         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13362         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13363         COMPLEX types.
13364         (hsa_fixup_mov_insn_type): New function.
13365         (hsa_op_with_type::get_in_type): Use it.
13366         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13367         immediates in an assert.
13368         (hsa_op_with_type::extend_int_to_32bit): New method.
13369         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13370         types.  Convert to dest type if necessary.
13371         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13372         (reg_for_gimple_ssa): Pass false as min32int to
13373         hsa_type_for_scalar_tree_type.
13374         (gen_hsa_addr): Fixup type when creating addresable temporary.
13375         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13376         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13377         necessary.  Call hsa_fixup_mov_insn_type.
13378         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13379         extend operands and convert to dest type if necessary.
13380         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13381         to dest type if necessary.
13382         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13383         if conversion nt necessary and size matches.
13384         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13385         to dest type if necessary.
13386         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13387         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13388         necessary.
13389         (gen_hsa_clrsb): Likewise.
13390         (gen_hsa_ffs): Likewise.
13391         (gen_hsa_divmod): Extend operands and convert to dest type if
13392         necessary.
13393         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13395 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13397         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13398         Remove empty default arguments.  Use a brace block as output
13399         statement.
13400         (conditional return): Ditto.
13401         (jump): Ditto.
13402         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13403         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13404         Ditto.
13405         (group_ending_nop): Ditto.
13406         (doloop_end): Ditto.
13407         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13408         (splitters for those): Ditto.
13410 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13412         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13413         a conditional jump (and the compare for it) so that pc_rtx is the
13414         last operand.
13415         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13416         for the deleted and renamed ctr<mode>_internal[234] patterns.
13417         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13418         Delete second conditional return pattern.
13419         (ctr<mode>_internal2): Delete this second bdnz pattern.
13420         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13421         (ctr<mode>_internal4): Delete this second bdz pattern.
13423 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13425         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13426         (always_initialized_rtx_for_ssa_name_p): New predicate.
13427         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13428         (finish_out_of_ssa): Free new field of SA.
13429         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13430         * tree-ssa-coalesce.c: Include tree-ssa.h.
13431         (get_parm_default_def_partitions): Remove extern keyword.
13432         (get_undefined_value_partitions): New function.
13433         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13434         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13435         uninitialized bits.
13436         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13438 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13440         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13442 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13444         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13445         for targets that preffer 128bit.
13447 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13449         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13451 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13453         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13454         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13455         prevent DSE.
13456         (thumb_set_return_address): Likewise.
13458 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13460         * common/config/arm/arm-common.c (arm_except_unwind_info):
13461         Handle DWARF2_UNWIND_INFO.
13463 2017-10-07  Michael Collison <michael.collison@arm.com>
13465         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13466         New pattern.
13468 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13470         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13471         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13472         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13473         defined, force the creation of a new block for a dispatch label.
13475 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13477         * invoke.texi (Wsuggest-attribute=cold): Document.
13478         * common.opt (Wsuggest-attribute=cold): New
13479         * ipa-pure-const.c (warn_function_cold): New function.
13480         * predict.c (compute_function_frequency): Use it.
13481         * predict.h (warn_function_cold): Declare.
13483 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13485         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13486         Update profile.
13488 2017-10-06  Martin Liska  <mliska@suse.cz>
13490         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13491         keyword for member functions.
13492         (struct sanopt_tree_couple): New struct.
13493         (struct sanopt_tree_couple_hash): New function.
13494         (struct sanopt_ctx): Add new hash_map.
13495         (has_dominating_ubsan_ptr_check): New function.
13496         (record_ubsan_ptr_check_stmt): Likewise.
13497         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13498         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13499         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13501 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13503         PR target/82440
13504         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13505         aarch64_simd_valid_immediate on CONST_VECTORs.
13506         (aarch64_reg_or_bic_imm): Likewise.
13508 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13510         PR rtl-optimization/82396
13511         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13513 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13515         * graphite-dependences.c (scop_get_reads): Move code to...
13516         (scop_get_must_writes): Move code to...
13517         (scop_get_may_writes): Move code to...
13518         (scop_get_reads_and_writes): ... here.
13519         (scop_get_dependences): Call scop_get_reads_and_writes.
13521 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13523         PR tree-optimization/82434
13524         * fold-const.h (can_native_encode_type_p,
13525         can_native_encode_string_p): Remove.
13526         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13527         don't encode anything, just return what would be otherwise returned.
13528         (native_encode_fixed, native_encode_complex, native_encode_vector):
13529         Likewise.
13530         (native_encode_string): Likewise.  Inline by hand
13531         can_native_encode_string_p.
13532         (can_native_encode_type_p): Remove.
13533         (can_native_encode_string_p): Remove.
13534         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13535         STRING_CSTs using can_native_encode_string_p, test all
13536         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13537         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13538         argument from native_encode_expr.
13539         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13540         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13541         but just 2.
13543 2017-10-06  Richard Biener  <rguenther@suse.de>
13545         PR tree-optimization/82397
13546         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13547         operand_equal_p but rely on data_ref_compare_tree for detecting
13548         equalities.
13549         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13550         to match up with dr_group_sort_cmp.
13552 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13554         PR target/82322
13555         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13556         builtin.
13557         * config/s390/s390-builtin-types.def: Regenerate.
13559 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13561         PR target/82317
13562         * config/s390/s390-builtin-types.def: Regenerate.
13563         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13564         Change flag from B_VXE to B_VX.
13565         (s390_vec_min_dbl): Remove B_VXE flag.
13567 2017-10-06  Richard Biener  <rguenther@suse.de>
13569         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13570         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13571         translate_isl_ast_to_gimple::is_valid_rename,
13572         translate_isl_ast_to_gimple::get_rename,
13573         translate_isl_ast_to_gimple::get_def_bb_for_const,
13574         translate_isl_ast_to_gimple::get_new_name,
13575         translate_isl_ast_to_gimple::collect_all_ssa_names,
13576         translate_isl_ast_to_gimple::copy_loop_phi_args,
13577         translate_isl_ast_to_gimple::collect_all_ssa_names,
13578         translate_isl_ast_to_gimple::copy_loop_phi_args,
13579         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13580         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13581         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13582         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13583         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13584         translate_isl_ast_to_gimple::copy_cond_phi_args,
13585         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13586         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13587         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13588         translate_isl_ast_to_gimple::rename_uses,
13589         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13590         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13591         (set_rename_for_each_def): Likewise.
13592         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13593         here.  Handle rewriting SCEV analyzable uses here.
13594         (copy_bb_and_scalar_dependences): Generate code for PHI
13595         copy-in/outs.
13596         (graphite_regenerate_ast_isl): Adjust.
13597         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13598         (add_write, add_read): New functions.
13599         (build_cross_bb_scalars_def): Use it and simplify.
13600         (build_cross_bb_scalars_use): Likewise.
13601         (graphite_find_cross_bb_scalar_vars): Inline into...
13602         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13603         simulating out-of-SSA.  Compute liveout and add dependencies.
13604         (build_scops): Force an empty entry block.
13605         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13606         members.
13607         (sese_build_liveouts): Declare.
13608         (sese_trivially_empty_bb_p): Likewise.
13609         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13610         compute liveout and debug_liveout.
13611         (sese_bad_liveouts_use): Remove.
13612         (sese_reset_debug_liveouts_bb): Likewise.
13613         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13614         (sese_build_liveouts): Build liveout and debug_liveout and store
13615         it in region.
13616         (new_sese_info): Adjust.
13617         (free_sese_info): Likewise.
13618         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13619         do not build liveout here.
13620         (move_sese_in_condition): Adjust region entry.
13621         (scev_analyzable_p): Match up with chrec_apply requirements.
13622         (sese_trivially_empty_bb_p): New.
13623         * tree-into-ssa.c (get_reaching_def): Properly support generating
13624         default-defs for incremental rewrite of anonymous names.
13626 2017-10-06  Richard Biener  <rguenther@suse.de>
13628         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13629         precision do not perform modulo reduction.
13631 2017-10-06  Richard Biener  <rguenther@suse.de>
13633         PR tree-optimization/82436
13634         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13635         conservatively choose the vectorization factor when checking
13636         whether we can perform the required load permutation.
13637         (vect_transform_slp_perm_load): Assert when we may not fail.
13639 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13641         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13642         message for incompatible -msdata=* and -mcall-* options.
13644 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13646         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13647         rate for post-reload scheduling.
13649 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13651         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13653 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13655         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13656         to improve monte carlo in scimark.
13658 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13660         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13661         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13662         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13663         pentium4_cost, nocona_cost): Set reassociation width to 1.
13664         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13665         width to 2 for fp operations and 1 otherwise.
13666         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13667         for int and fp.
13668         (atom_cost): Set reassociation width to 2.
13669         (slm_cost, generic_cost): Set fp reassociation width
13670         to 2 and 1 otherwise.
13671         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13672         (core_cost): Set fp reassociation width to 4 and vector to 2.
13673         (ix86_reassociation_width): Rewrite using cost table; special case
13674         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13675         and TARGET_AVX128_OPTIMAL.
13676         * config/i386/i386.h (processor_costs): Add
13677         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13678         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13679         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13680         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13681         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13682         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13684 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13686         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13688 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13690         * config/arm/arm.c (arm_test_fpu_data): New.
13691         (arm_run_selftests): Call arm_test_fpu_data.
13693 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13695         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13696         decl.
13697         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13699 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13701         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13702         check type for aarch64_simd_valid_immediate.
13703         (aarch64_output_simd_mov_immediate): Update prototype.
13704         (aarch64_simd_valid_immediate): Update prototype.
13705         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13706         support for ORR-immediate.
13707         (and<mode>3): modified pattern to add support for BIC-immediate.
13708         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13709         now checks for valid immediate for BIC and ORR based on new enum
13710         argument.
13711         (aarch64_output_simd_mov_immediate): Function now used to output
13712         BIC/ORR imm as well based on new enum argument.
13713         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13714         (Db) : Likewise.
13715         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13716         (aarch64_reg_or_bic_imm): Likewise.
13718 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13720         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13721         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13723 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13725         Revert r253399:
13727         PR rtl-optimization/82396
13728         * haifa-sched.c (autopref_multipass_init): Simplify
13729         initialization.
13730         (autopref_rank_data): Simplify sort order.
13731         * sched-int.h (autopref_multipass_data_): Remove
13732         multi_mem_insn_p, min_offset and max_offset.
13734 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13736         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13738 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13740         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13741         vect_doubleint_cvt.
13743 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13745         * doc/sourcebuild.texi: Document vect_long_mult.
13747 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13749         PR tree-optimization/82413
13750         * fold-const.c (build_range_check): Use widest_int when comparing
13751         the maximum ETYPE value with HIGH.
13753 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13755         PR rtl-optimization/82396
13756         * haifa-sched.c (autopref_multipass_init): Simplify
13757         initialization.
13758         (autopref_rank_data): Simplify sort order.
13759         * sched-int.h (autopref_multipass_data_): Remove
13760         multi_mem_insn_p, min_offset and max_offset.
13762 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13764         PR tree-optimization/82381
13765         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13766         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13767         is not.
13769         PR tree-optimization/82374
13770         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13771         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13772         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13773         current_function_decl to the new decl.
13775 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13777         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13778         helper macro for IEEE float128 hardware built-in functions.
13779         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13780         semantics.
13781         (TRUNCF128_ODD): Likewise.
13782         (ADDF128_ODD): Likewise.
13783         (SUBF128_ODD): Likewise.
13784         (MULF128_ODD): Likewise.
13785         (DIVF128_ODD): Likewise.
13786         (FMAF128_ODD): Likewise.
13787         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13788         UNSPEC_TRUNC_ROUND_TO_ODD.
13789         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13790         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13791         floating point round to odd instructions.
13792         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13793         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13794         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13795         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13796         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13797         (trunc<mode>sf2_hw): Change the truncate with round to odd
13798         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13799         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13800         to odd hardware instructions.
13801         (sub<mode>3_odd): Likewise.
13802         (mul<mode>3_odd): Likewise.
13803         (div<mode>3_odd): Likewise.
13804         (sqrt<mode>2_odd): Likewise.
13805         (fma<mode>4_odd): Likewise.
13806         (fms<mode>4_odd): Likewise.
13807         (nfma<mode>4_odd): Likewise.
13808         (nfms<mode>4_odd): Likewise.
13809         (trunc<mode>df2_odd): Change the truncate with round to odd
13810         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13811         function.
13812         * doc/extend.texi (PowerPC built-in functions): Update documentation
13813         for existing IEEE float128-bit built-in functions.  Add built-in
13814         functions that generate the IEEE 128-bit floating point round to
13815         odd instructions.
13817 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13819         PR rtl-optimization/77729
13820         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13821         to (X&(C1&~C2))|C2 transformations.
13823 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13825         PR tree-optimization/82363
13826         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13827         mismatch, mark lacc written regardless of racc.
13829 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13831         PR tree-optimization/82381
13832         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13833         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13834         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13835         fallthrough into reassoc_stmt_dominates_stmt_p.
13837         PR target/82386
13838         * combine.c (combine_instructions): Don't combine in unreachable
13839         basic blocks.
13841 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13843         PR target/80210
13844         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13845         function to not use the have_cpu variable.  Do not set cpu_index,
13846         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13847         or the default cpu.
13848         (rs6000_valid_attribute_p): Remove duplicate initializations of
13849         old_optimize and func_optimize.
13850         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13851         (rs6000_activate_target_options): Make global.
13852         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13853         prototype.
13855 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13857         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13858         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13859         Set *poffset to 0 and *psize and *pmax_size to -1 if
13860         *poffset + *psize overflows in HOST_WIDE_INT.
13862         PR tree-optimization/82387
13863         PR tree-optimization/82388
13864         PR tree-optimization/82389
13865         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13866         instead of live_bytes non-NULL.
13868 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13870         PR target/41076
13871         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13872         alternative.
13874 2017-10-02  Richard Biener  <rguenther@suse.de>
13876         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13877         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13878         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13880 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13882         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13883         requested precision matches the type's.
13884         * calls.c (alloc_max_size): Calculate the new candidate size as
13885         a widest_int and use wi::to_widest when comparing it with the
13886         current candidate size.
13887         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13888         zero rather than integer_zero_node.
13889         * match.pd: Check for a no-op conversion before using wi::add
13890         rather than after.  Use tree_to_uhwi when summing small shift
13891         counts into an unsigned int.
13893 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13894             Alan Hayward  <alan.hayward@arm.com>
13895             David Sherwood  <david.sherwood@arm.com>
13897         PR target/71307
13898         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13899         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13900         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13901         POINTER_AND_FP_REGS.
13903 2017-10-02  Richard Biener  <rguenther@suse.de>
13905         PR tree-optimization/82355
13906         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13907         a mapping for the enclosing loop but avoid generating one for
13908         the loop tree root.
13909         (copy_bb_and_scalar_dependences): Remove premature codegen
13910         error on PHIs in blocks duplicated into multiple places.
13911         * graphite-scop-detection.c
13912         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13913         in the region use it as loop and nest to analyze the DR in.
13914         (try_generate_gimple_bb): Likewise.
13915         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13916         (add_loop_constraints): For blocks in a loop not in the region
13917         create a dimension with a single iteration.
13918         * sese.h (gbb_loop_at_index): Remove assert.
13920 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13922         * omp-expand.c (adjust_context_scope): New function.
13923         (expand_parallel_call): Call adjust_context_scope.
13925 2017-10-01  Jeff Law  <law@redhat.com>
13927         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13928         dom_opt_dom_walker class with direct access to private members.
13929         Add comments.  Call test_for_singularity.
13930         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13931         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13932         m_dummy_cond anymore.
13933         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13934         class ctor.
13935         (pass_dominator:execute): Build the dummy_cond here and pass it
13936         to the dom_opt_dom_walker ctor.
13937         (test_for_singularity): New function.
13939 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
13940             Maya Rashish  <coypu@sdf.org>
13942         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
13943         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
13944         (alpha*-*-netbsd*) Use nbsd_tm_file.
13945         (arm*-*-netbsdelf*) Likewise.
13946         (i[34567]86-*-netbsdelf*) Likewise.
13947         (x86_64-*-netbsd*) Likewise.
13948         (mips*-*-netbsd*) Likewise.
13949         (powerpc-*-netbsd*) Likewise.
13950         (sh*-*-netbsd*) Likewise.
13951         (sparc-*-netbsdelf*) Likewise.
13952         (sparc64-*-netbsd*) Likewise.
13953         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
13954         to tm_defines.
13955         (vax-*-netbsdelf*) Likewise.
13956         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
13957         (UINT_FAST8_TYPE) Likewise.
13958         (INT_FAST16_TYPE) Check CHAR_FAST16.
13959         (UINT_FAST16_TYPE) Likewise.
13961 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
13963         PR target/82361
13964         * config/i386/i386.md
13965         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
13966         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
13967         *divmodsi4_zext_2): New define_insn_and_split.
13968         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
13969         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
13970         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
13971         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
13972         New define_insn_and_split.
13973         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
13974         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
13975         operands[1] having DImode when mode is SImode.
13977         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
13978         always SImode for DIV and MOD in REG_EQUAL notes.
13980 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
13982         PR middle-end/82319
13983         * match.pd: Fix handling of NaNs in pattern.
13985 2017-09-29  Jeff Law  <law@redhat.com>
13987         * sbitmap.c (bitmap_bit_in_range_p): New function.
13988         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
13989         * tree-ssa-dse.c (live_bytes_read): New function.
13990         (dse_classify_store): Ignore reads of dead bytes.
13992         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
13993         typos and whitespace errors.
13994         * config/i386/predicates.md (address_no_seg_operand): Likewise.
13995         * config/s390/s390.c (s390_emit_prologue): Likewise.
13997 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13999         PR target/81481
14000         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
14001         with a symbol for LRA.
14003 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14005         PR rtl-optimization/82338
14006         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
14008 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
14010         * genmodes.c (calc_wider_mode): Suppress qsort macro.
14011         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
14012         (qsort_chk): Declare.
14013         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
14014         (qsort_chk): New function.
14016 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14018         PR tree-optimization/82337
14019         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
14020         phi definition if the PHI result appears in an abnormal PHI.
14021         (find_basis_for_base_expr): Don't record a basis if the LHS of the
14022         basis appears in an abnormal PHI.
14024 2017-09-29  Richard Biener  <rguenther@suse.de>
14026         * graphite-isl-ast-to-gimple.c
14027         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14028         (binary_op_to_tree): Use it.
14029         (get_rename_from_scev): Likewise.
14030         (copy_loop_phi_nodes): Likewise.
14031         (copy_bb_and_scalar_dependences): Likewise.
14032         (translate_pending_phi_nodes): Likewise.
14034 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14036         PR target/82339
14037         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14038         for movabsq $(i32 << shift), r64.
14040 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14042         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14043         index when encoding %esp as %rsp to avoid 0x67 prefix.
14045 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14047         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14048         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14050 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14052         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14053         Extensions with more than 16 double VFP registers.
14054         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14055         to_clear_mask and all code related to it.  Replace the remaining
14056         entry by a sbitmap and adapt code accordingly.
14058 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14060         * brig-builtins.def: Change pure attributes to const.
14062 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14064         * config.gcc (default_gnu_indirect_function): Default to yes for
14065         sparc*-*-linux* with glibc.
14067 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14069         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14070         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14071         when creating .init_array and .fini_array sections with priority
14072         specified.
14074 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14076         PR target/71727
14077         * config/aarch64/aarch64.c
14078         (aarch64_builtin_support_vector_misalignment): Always return false
14079         when misalignment is unknown.
14081 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14083         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14084         this function to return false if the definition used by the swap
14085         instruction is artificial, or if the memory address from which the
14086         constant value is loaded is not represented by a base address held
14087         in a register or if the base address register is a frame or stack
14088         pointer.  Additionally, return false if the base address of the
14089         loaded constant is a SYMBOL_REF but is not considered to be a
14090         constant.
14091         (replace_swapped_load_constant): New function.
14092         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14093         loaded constant vector with a load of a swapped constant vector.
14095 2017-09-27  Carl Love  <cel@us.ibm.com>
14097         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14098         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14099         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14100         fctiw instruction.
14102 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14104         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14105         first, always call autopref_rank_data otherwise.
14107 2017-09-27  Richard Biener  <rguenther@suse.de>
14109         * graphite-scop-detection.c (find_scop_parameters): Move
14110         loop bound handling ...
14111         (gather_bbs::before_dom_children): ... here, avoiding the need
14112         to build scop_info->loop_nest.
14113         (record_loop_in_sese): Remove.
14114         * sese.h (sese_info_t::loop_nest): Remove.
14115         * sese.c (new_sese_info): Do not allocate loop_nest.
14116         (free_sese_info): Do not free loop_nest.
14118 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14120         PR c++/82159
14121         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14122         lhs from calls if the lhs has addressable type.
14124 2017-09-27  Richard Biener  <rguenther@suse.de>
14126         * graphite.h (scop::max_alias_set): New member.
14127         * graphite-scop-detection.c: Remove references to non-existing
14128         --param in comments.
14129         (build_alias_sets): Record the maximum alias set used for drs.
14130         (build_scops): Support zero as unlimited for
14131         --param graphite-max-arrays-per-scop.
14132         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14133         and inline into ...
14134         (build_poly_sr_1): ... here.  Compute alias set based on the
14135         maximum alias set used for drs rather than
14136         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14138 2017-09-27  Richard Biener  <rguenther@suse.de>
14140         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14141         --param loop-block-tile-size=0 to disable tiling.
14143 2017-09-27  Richard Biener  <rguenther@suse.de>
14145         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14146         (graphite-max-nb-scop-params): Document special value zero.
14147         * domwalk.h (dom_walker::STOP): New symbolical constant.
14148         (dom_walker::dom_walker): Add optional parameter for bb to
14149         RPO mapping.
14150         (dom_walker::~dom_walker): Declare.
14151         (dom_walker::before_dom_children): Document STOP return value.
14152         (dom_walker::m_user_bb_to_rpo): New member.
14153         (dom_walker::m_bb_to_rpo): Likewise.
14154         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14155         mapping here if not provided by the user.
14156         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14157         provided by the user.
14158         (dom_walker::STOP): Define.
14159         (dom_walker::walk): Do not compute bb to RPO mapping here.
14160         Support STOP return value from before_dom_children to stop
14161         walking.
14162         * graphite-optimize-isl.c (optimize_isl): If the schedule
14163         is the same still generate code if -fgraphite-identity
14164         or -floop-parallelize-all are given.
14165         * graphite-scop-detection.c: Include cfganal.h.
14166         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14167         mapping.
14168         (gather_bbs::before_dom_children): Return STOP for BBs
14169         not in the region.
14170         (build_scops): Compute bb to RPO mapping and pass it to
14171         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14172         as not limiting the number of params.
14173         * graphite.c (graphite_initialize): Remove limit on the
14174         number of basic-blocks in a function.
14175         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14176         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14177         default value of 10.
14179 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14181         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14182         Adjust code to eliminate needing to do the shift right 32-bits
14183         operation after XSCVDPSPN.
14185 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14187         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14188         ((X / Y) != 0 -> X >= Y): Likewise.
14190 2017-09-26  Carl Love  <cel@us.ibm.com>
14192         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14193         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14194         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14195         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14196         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14197         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14198         definitions and overloading.
14199         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14200         statement for P9V_BUILTIN_XST_LEN_R.
14201         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14202         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14203         define_expand and define_insn for the instructions and builtins.
14204         * doc/extend.texi: Update the built-in documentation file for the new
14205         built-in functions.
14206         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14207         define_insn for the instructions
14209 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14211         PR target/39570
14212         * gcc/config/netbsd-protos.h: New file.
14213         * gcc/config/netbsd.c: New file.
14214         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14215         * gcc/config/t-netbsd: New file.
14216         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14217         (tmake_file) Add t-netbsd.
14218         (extra_objs) Add netbsd.o.
14220 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14222         PR fortran/82143
14223         PR fortran/82324
14224         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14226 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14228         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14229         sign extension from a vector register to a GPR by doing a 32-bit
14230         direct move and then an EXTSW.
14231         (extendsi<mode>2 splitter): Likewise.
14232         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14233         right or vector extract after doing XSCVDPSPN.  Use
14234         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14235         the GPRs.
14236         (movdi_from_sf_zero_ext): Likewise.
14237         (reload_gpr_from_vsxsf): Likewise.
14238         (p8_mfvsrd_4_disf): Delete, no longer used.
14239         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14240         then needing to move the SFmode to a GPR to use the XSCVDPSP
14241         instruction instead of FRSP and XSCVDPSPN.
14242         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14243         it is adjacent to the other XSCVSPDP insns.
14244         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14245         SFmode to be in traditional Altivec registers.
14246         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14247         (vsx_xscvspdpn): Likewise.
14248         (vsx_xscvspdpn_scalar): Likewise.
14250 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14252         * tree-sra.c (compare_access_positions): Put integral types first,
14253         stabilize sorting of integral types, remove conditions putting
14254         non-full-precision integers last.
14255         (sort_and_splice_var_accesses): Disable scalarization if a
14256         non-integert would be represented by a non-full-precision integer.
14258 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14260         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14261         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14262         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14263         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14264         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14265         conditionals inside the function instead of around it.  Call
14266         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14267         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14269 2017-09-26  Richard Biener  <rguenther@suse.de>
14271         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14272         fold in ...
14273         (scop_detection::build_scop_breadth): ... this.  Removed.
14274         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14275         (scop_detection::harmful_stmt_in_bb): Likewise.
14276         (scop_detection::graphite_can_represent_stmt): Likewise.
14277         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14278         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14279         (scop_detection::can_represent_loop_1): ... this.  Removed.
14280         (scop_detection::harmful_loop_in_region): Simplify after inlining
14281         the above and remove more quadraticness.
14282         (build_scops): Adjust.
14283         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14284         quadraticness.
14286 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14288         PR target/82267
14289         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14290         REGNO (base) == SP_REG if base is a REG.
14292         PR middle-end/35691
14293         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14294         if it is different SSA_NAME.
14295         (optimize_range_tests_cmp_bitwise): New function.
14296         (optimize_range_tests): Call it.
14298 2017-09-26  Richard Biener  <rguenther@suse.de>
14300         PR tree-optimization/82321
14301         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14302         for the def being inside the loop.
14304 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14306         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14307         assembler output.
14308         * config/s390/s390-builtins.def: Fix constraint on op4.
14310 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14312         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14313         independent expanders.
14314         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14315         ("vec_ordered", "vec_unordered"): New expanders.
14317 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14319         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14320         for SFmode.
14322 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14324         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14325         vec_unpacks_lo_v16qi.
14326         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14328 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14330         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14331         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14332         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14334 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14336         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14337         predicate.
14338         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14339         and V16QI.
14340         ("*vec_slb<mode>"): New insn pattern.
14341         ("vec_shr_<mode>"): New expander.
14342         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14343         and force the shift count operand to V16QImode.
14344         ("vec_srb<mode>"): Set shift count mode to V16QI.
14346 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14348         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14349         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14350         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14352 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14354         PR target/82175
14355         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14357 2017-09-26  Richard Biener  <rguenther@suse.de>
14359         PR tree-optimization/82320
14360         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14361         isn't a change.
14363 2017-09-25  Jeff Law  <law@redhat.com>
14365         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14366         prototype for new argument.
14367         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14368         mostly extracted from rs6000_emit_allocate_stack.
14369         (rs6000_emit_probe_stack_range_stack_clash): New function.
14370         (rs6000_emit_allocate_stack): Call
14371         rs6000_emit_probe_stack_range_stack_clash as needed.
14372         (rs6000_emit_probe_stack_range): Add additional argument
14373         to call to gen_probe_stack_range{si,di}.
14374         (output_probe_stack_range): New.
14375         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14376         (output_probe_stack_range_stack_clash): New.
14377         (rs6000_emit_prologue): Emit notes into dump file as requested.
14378         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14379         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14380         Add additional operand and pass it to output_probe_stack_range.
14382 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14384         PR tree-optimization/82163
14385         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14386         (checking_verify_loop_closed_ssa): New parameter.
14387         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14388         (check_loop_closed_ssa_stmt): Delete.
14389         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14390         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14391         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14392         changed loops.
14394 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14396         * brig-builtins.def: Treat HSAIL barrier builtins as
14397         setjmp/longjump style functions.
14399 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14401         * target.def (constant_alignment): New hook.
14402         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14403         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14404         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14405         * doc/tm.texi: Regenerate.
14406         * targhooks.h (default_constant_alignment): Declare.
14407         (constant_alignment_word_strings): Likewise.
14408         * targhooks.c (default_constant_alignment): New function.
14409         (constant_alignment_word_strings): Likewise.
14410         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14411         instead of CONSTANT_ALIGNMENT.
14412         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14413         (force_const_mem): Likewise.
14414         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14415         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14416         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14417         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14418         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14419         definition.
14420         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14421         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14422         constant_alignment_word_strings.
14423         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14424         (CONSTANT_ALIGNMENT): Likewise.
14425         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14426         (arm_constant_alignment): New function.
14427         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14428         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14429         constant_alignment_word_strings.
14430         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14431         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14432         constant_alignment_word_strings.
14433         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14434         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14435         (cris_constant_alignment): New function.
14436         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14437         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14438         (epiphany_constant_alignment): New function.
14439         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14440         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14441         constant_alignment_word_strings.
14442         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14443         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14444         constant_alignment_word_strings.
14445         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14446         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14447         constant_alignment_word_strings.
14448         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14449         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14450         * config/i386/i386.c (ix86_constant_alignment): Make static.
14451         Use the same interface as the target hook.
14452         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14453         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14454         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14455         constant_alignment_word_strings.
14456         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14457         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14458         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14459         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14460         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14461         constant_alignment_word_strings.
14462         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14463         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14464         constant_alignment_word_strings.
14465         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14466         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14467         constant_alignment_word_strings.
14468         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14469         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14470         New function.
14471         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14472         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14473         * config/mips/mips.c (mips_constant_alignment): New function.
14474         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14475         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14476         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14477         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14478         (mmix_constant_alignment): Make static.  Use the same interface
14479         as the target hook.
14480         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14481         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14482         constant_alignment_word_strings.
14483         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14484         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14485         constant_alignment_word_strings.
14486         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14487         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14488         constant_alignment_word_strings.
14489         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14490         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14491         (rs6000_constant_alignment): New function.
14492         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14493         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14494         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14495         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14496         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14497         (rs6000_constant_alignment): New function.
14498         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14499         * config/s390/s390.c (s390_constant_alignment): New function.
14500         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14501         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14502         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14503         constant_alignment_word_strings.
14504         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14505         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14506         (sparc_constant_alignment): New function.
14507         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14508         * config/spu/spu.c (spu_constant_alignment): New function.
14509         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14510         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14511         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14512         constant_alignment_word_strings.
14513         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14514         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14515         constant_alignment_word_strings.
14516         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14517         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14518         constant_alignment_word_strings.
14519         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14520         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14521         (visium_constant_alignment): New function.
14522         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14523         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14524         (xtensa_constant_alignment): New function.
14525         * system.h (CONSTANT_ALIGNMENT): Poison.
14527 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14529         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14530         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14531         (rs6000_builtin_valid_without_lhs): New helper function.
14532         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14533         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14535 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14537         * target.h (vec_perm_indices): Use unsigned short rather than
14538         unsigned char.
14539         (auto_vec_perm_indices): Likewise.
14540         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14541         Use unsigned int rather than unsigned char.
14542         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14544 2017-09-25  Richard Biener  <rguenther@suse.de>
14546         * cfgloop.h (sort_sibling_loops): Declare.
14547         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14548         (sort_sibling_loops): New function sorting the sibling loop list
14549         in RPO order.
14550         * graphite.c (graphite_transform_loops): Sort sibling loops.
14552 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14554         * target.def (vec_perm_const_ok): Change sel parameter to
14555         vec_perm_indices.
14556         * optabs-query.c (can_vec_perm_p): Update accordingly.
14557         * doc/tm.texi: Regenerate.
14558         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14559         auto_vec_perm_indices and remove separate nelt field.
14560         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14561         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14562         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14563         (aarch64_expand_vec_perm_const): Update accordingly.
14564         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14565         to vec_perm_indices.
14566         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14567         auto_vec_perm_indices and remove separate nelt field.
14568         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14569         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14570         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14571         accordingly.
14572         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14573         to vec_perm_indices.
14574         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14575         sel to vec_perm_indices.
14576         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14577         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14578         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14579         Likewise.
14580         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14581         Likewise.
14583 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14585         PR debug/82155
14586         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14587         on the FUNCTION_DECL function context if it has a DIE that is a
14588         declaration.
14590 2017-09-25  Richard Biener  <rguenther@suse.de>
14592         PR tree-optimization/82285
14593         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14594         enumeral types.
14596 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14598         PR target/80035
14599         PR target/81069
14600         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14601         noreturn function.
14603 2017-09-25  Richard Biener  <rguenther@suse.de>
14605         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14606         ISL errors other than isl_error_quota happen.  Dump if the
14607         schedule is the same.
14608         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14609         errors instead of aborting inside ISL.
14611 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14613         PR target/80556
14614         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14615         of libgcc_eh for m64.
14616         * config/i386/darwin64.h: Likewise.
14618 2017-09-25  Richard Biener  <rguenther@suse.de>
14620         PR middle-end/82144
14621         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14622         attribute for incomplete types nor twice for complete ones.
14624 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14626         PR target/82267
14627         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14628         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14629         register.
14631 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14633         PR bootstrap/82306
14634         * config/i386/i386.opt (mprefer-avx256): Use
14635         ix86_target_flags variable.
14636         * config/i386/i386.c (ix86_target_string): Move
14637         -mprefer-avx256 to flag2_opts.
14639 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14641         PR middle-end/35691
14642         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14643         and x != -1 | y != -1 into (x & y) != -1.
14645 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14647         * config.gcc: Add new case statement to set
14648         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14649         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14650         s390x-*-linux* case statements.   Added aarch64 to the list of
14651         supported architectures.
14653 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14655         PR tree-optimization/82289
14656         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14657         STMT_VINFO_RELEVANT_P.
14659 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14660             Alan Hayward  <alan.hayward@arm.com>
14661             David Sherwood  <david.sherwood@arm.com>
14663         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14664         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14665         (extract_range_from_binary_expr_1): Don't call
14666         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14668 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14669             Alan Hayward  <alan.hayward@arm.com>
14670             David Sherwood  <david.sherwood@arm.com>
14672         * target.def (preferred_vector_alignment): New hook.
14673         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14674         hook.
14675         * doc/tm.texi: Regenerate.
14676         * targhooks.h (default_preferred_vector_alignment): Declare.
14677         * targhooks.c (default_preferred_vector_alignment): New function.
14678         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14679         Expand commentary.
14680         (DR_TARGET_ALIGNMENT): New macro.
14681         (aligned_access_p): Update commentary.
14682         (vect_known_alignment_in_bytes): New function.
14683         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14684         function.
14685         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14686         Calculate the misalignment based on the target alignment rather than
14687         the vector size.
14688         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14689         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14690         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14691         the target alignment, rather than masking the element misalignment
14692         with the number of elements in a vector.  Also use the target
14693         alignment when calculating the maximum number of peels.
14694         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14695         instead of TYPE_ALIGN_UNIT.
14696         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14697         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14698         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14699         (vect_create_data_ref_ptr): Likewise.
14700         (vect_setup_realignment): Realign by ANDing with
14701         -DR_TARGET_MISALIGNMENT.
14702         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14703         the number of peels based on DR_TARGET_ALIGNMENT.
14704         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14705         with the guaranteed alignment boundary when deciding whether
14706         overrun is OK.
14707         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14708         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14709         (ensure_base_align): Remove stmt_info parameter.  Get the
14710         target base alignment from DR_TARGET_ALIGNMENT.
14711         (vectorizable_store): Update call accordingly.   Interpret
14712         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14713         TYPE_ALIGN_UNIT.
14714         (vectorizable_load): Likewise.
14716 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14717             Alan Hayward  <alan.hayward@arm.com>
14718             David Sherwood  <david.sherwood@arm.com>
14720         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14721         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14722         (vect_enhance_data_refs_alignment): Likewise.
14724 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14726         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14727         error.  Only quit immediately if parsing is complete.
14728         (BEGIN): Initialize fatal_err and parse_done.
14729         (begin fpu, end fpu): Check number of arguments.
14730         (begin arch, end arch): Likewise.
14731         (begin cpu, end cpu): Likewise.
14732         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14733         (optalias): Likewise.
14735 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14737         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14738         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14739         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14740         * config/arm/arm.c (arm_option_override): Use lower case for feature
14741         bit names.
14742         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14743         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14744         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14745         (isa_pfx): Delete.
14746         (print_isa_bits_for): New function.
14747         (gen_isa): New function.
14748         (gen_comm_data): Use print_isa_bits_for.
14749         (define feature): New keyword.
14750         (define fgroup): New keyword.
14751         * config/arm/t-arm (TM_H): Remove.
14752         (GTM_H): Add arm-isa.h.
14753         (arm-isa.h): Add rule to generate file.
14754         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14755         case for feature bit names.
14757 2017-09-22  Richard Biener  <rguenther@suse.de>
14759         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14760         single caller.
14761         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14762         print of no dependency loops ...
14763         * graphite.c (graphite_transform_loops): ... here.
14764         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14765         loops.
14766         (same_close_phi_node, remove_duplicate_close_phi,
14767         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14768         (canonicalize_loop_closed_ssa): ... here and simplify.
14769         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14770         (optimize_isl): Use dump_printf_loc to tell when we stopped
14771         optimizing because of an ISL timeout.
14773 2017-09-22  Richard Biener  <rguenther@suse.de>
14775         PR tree-optimization/82291
14776         * tree-if-conv.c (predicate_mem_writes): Make sure to
14777         remove writes in blocks predicated with false.
14779 2017-09-22  Richard Biener  <rguenther@suse.de>
14781         * sese.c: Include cfganal.h.
14782         (if_region_set_false_region): Remove.
14783         (create_if_region_on_edge): Likewise.
14784         (move_sese_in_condition): Re-implement without destroying
14785         dominators.
14787 2017-09-22  Richard Biener  <rguenther@suse.de>
14789         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14790         Verify both BBs contain loop PHI nodes before dispatching to
14791         copy_loop_phi_args.
14792         (graphite_regenerate_ast_isl): Do not recompute dominators,
14793         do not verify three times.  Restructure for clarity.
14794         * graphite-scop-detection.c (same_close_phi_node,
14795         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14796         defined_in_loop_p, canonicalize_loop_closed_ssa,
14797         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14798         checking and SSA rewrite, move to ...
14799         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14800         (graphite_initialize): Do not pass in ctx, do not reset the
14801         SCEV cache, compute only dominators.
14802         (graphite_transform_loops): Allocate ISL ctx after
14803         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14804         Maintain post-dominators only around build_scops.
14805         * sese.c (if_region_set_false_region): Make static.  Free
14806         and recompute dominators.
14807         (move_sese_in_condition): Assert we don't get called with
14808         post-dominators computed.
14809         * sese.h (if_region_set_false_region): Remove.
14811 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14813         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14814         mode attribute for TARGET_AVX512VL.
14816 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14818         * config/i386/i386.opt (mprefer-avx256): New option.
14819         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14820         to flag_opts.
14821         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14822         for TARGET_PREFER_AVX256.
14823         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14825 2017-09-21  Jeff Law  <law@redhat.com>
14827         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14828         Fix dump output if the only stack space is for pushed registers.
14830 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14832         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14833         of insn_cost.
14835 2017-09-21  Martin Sebor  <msebor@redhat.com>
14837         PR c/81882
14838         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14839         code (in C++) or code that triggers warnings.
14841 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14843         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14845 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14847         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14848         * sched-rgn.c: Ditto.
14849         * sel-sched-ir.c: Ditto.
14851 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14853         * toplev.h (set_random_seed): Adjust return type.
14854         * toplev.c (init_local_tick): Move eager initialization of random_seed
14855         to get_random_seed.  Adjust comment.
14856         (init_random_seed): Inline to get_random_seed, delete.
14857         (get_random_seed): Initialize random_seed lazily.
14858         (set_random_seed): Do not return previous value.
14859         (print_switch_value): Do not call get_random_seed.
14861 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14863         * cgraph.c (delete_function_version): New, broken out from...
14864         (cgraph_node::delete_function_version): ...here.  Rename to
14865         cgraph_node::delete_function_version_by_decl.  Update all uses.
14866         (cgraph_node::remove): Call delete_function_version.
14868 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14870         PR sanitizer/81715
14871         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14872         VAR_DECLs to which addressable non-volatile parameters are mapped
14873         and for id->retvar after the return value assignment.  Clear
14874         id->retval and id->retbnd after inlining.
14876 2017-09-21  Richard Biener  <rguenther@suse.de>
14878         PR tree-optimization/82276
14879         PR tree-optimization/82244
14880         * tree-vrp.c (build_assert_expr_for): Set
14881         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14882         has it set.
14883         (remove_range_assertions): Revert earlier change.
14885 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14887         PR target/71951
14888         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14890 2017-09-21  Richard Biener  <rguenther@suse.de>
14892         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14893         Restore valid IL after code generation errors.
14894         * graphite.c (graphite_transform_loops): Diagnose code
14895         generation issues as MSG_MISSED_OPTIMIZATION and continue
14896         with processing SCOPs.
14898 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14899             Alan Hayward  <alan.hayward@arm.com>
14900             David Sherwood  <david.sherwood@arm.com>
14902         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14903         rather than choosing between plus_constant and gen_rtx_<CODE>.
14904         * expr.c (emit_push_insn): Likewise.
14905         (expand_expr_real_2): Likewise.
14907 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14908             Alan Hayward  <alan.hayward@arm.com>
14909             David Sherwood  <david.sherwood@arm.com>
14911         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14913 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14914             Alan Hayward  <alan.hayward@arm.com>
14915             David Sherwood  <david.sherwood@arm.com>
14917         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14918         calling tree_to_uhwi.
14920 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14921             Alan Hayward  <alan.hayward@arm.com>
14922             David Sherwood  <david.sherwood@arm.com>
14924         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14925         INTEGER_CST rather than a negative test for ADDR_EXPR.
14927 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14928             Alan Hayward  <alan.hayward@arm.com>
14929             David Sherwood  <david.sherwood@arm.com>
14931         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14932         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14934 2017-09-21  Richard Biener  <rguenther@suse.de>
14936         PR tree-optimization/71351
14937         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14938         graphite_create_new_loop_guard): Remove, fold remaining parts
14939         into caller ...
14940         (translate_isl_ast_node_for): ... here and simplify.
14942 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14944         PR target/82260
14945         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
14946         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
14947         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
14948         alternative always use QI mode, for -Os imov (=R,R) alternative
14949         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
14950         ignore -Os.
14952 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14953             Jeff Law  <law@redhat.com>
14955         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
14956         (allocate_stack_space): New function, partially extracted from
14957         s390_emit_prologue.
14958         (s390_emit_prologue): Track offset to most recent stack probe.
14959         Code to allocate space moved into allocate_stack_space.
14960         Dump actions when no stack is allocated.
14961         (s390_prologue_plus_offset): New function.
14962         (s390_emit_stack_probe): Likewise.
14964 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
14966         * common.opt (Wa, Wl, Wp, g, gz=): Add
14967         RejectNegative.
14968         (gno-column-info): Remove.
14969         (gcolumn-info): Drop RejectNegative.
14970         (gno-): New prefix.
14971         (gno-record-gcc-switches): Remove.
14972         (grecord-gcc-switches): Drop RejectNegative.
14973         (gno-split-dwarf): Remove.
14974         (gsplit-dwarf): Drop RejectNegative.
14975         (gno-strict-dwarf): Remove.
14976         (gstrict-dwarf): Drop RejectNegative.
14977         * config/darwin.opt (gfull, gused): Add RejectNegative.
14978         * dwarf2out.c (gen_producer_string): Drop
14979         gno-record-gcc-switches handler.
14980         * optc-gen.awk: Add g to prefixes with negative forms.
14981         * opts-common.c (remapping_prefix_p): New.
14982         (find_opt): Check it.
14983         (generate_canonical_option): Test g prefix.
14984         (option_map): Add -gno- mapping.
14985         (add_misspelling_candidates): Check remapping_prefix_p.
14987 2017-09-20  Jeff Law  <law@redhat.com>
14989         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
14990         thinko in stack clash protection support.
14992         * explow.c (compute_stack_clash_protection_loop_data): Use
14993         CONST_INT_P instead of explicit test.  Verify object is a
14994         CONST_INT_P before looking at INTVAL.
14995         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
14996         instead of explicit test.
14998 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
15000         PR target/77687
15001         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
15002         address instead of to r1 and r11.
15004 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
15006         * config.gcc: Support "knm".
15007         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
15008         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15009         PROCESSOR_KNM.
15010         * config/i386/i386.c (m_KNM): Define.
15011         (processor_target_table): Add "knm".
15012         (PTA_KNM): Define.
15013         (ix86_option_override_internal): Add "knm".
15014         (ix86_issue_rate): Add PROCESSOR_KNM.
15015         (ix86_adjust_cost): Ditto.
15016         (ia32_multipass_dfa_lookahead): Ditto.
15017         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
15018         (fold_builtin_cpu): Add M_INTEL_KNM.
15019         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
15020         (processor_type): Add PROCESSOR_KNM.
15021         * config/i386/x86-tune.def: Add m_KNM.
15022         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15024 2017-09-20  Richard Biener  <rguenther@suse.de>
15026         PR tree-optimization/80213
15027         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15028         are allowed in empty BBs as well.
15029         (canonicalize_loop_closed_ssa): Also look for other complex
15030         edges.
15031         (scop_detection::get_sese): Include the loop-closed PHI block
15032         in loop SESEs.
15033         (scop_detection::merge_sese): Remove code adding extra blocks.
15034         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15035         (build_scops): Assert the final returned scop is invalid.
15037 2017-09-20  Richard Biener  <rguenther@suse.de>
15039         PR tree-optimization/82264
15040         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15041         for GIMPLE_CONDs.
15042         (vn_phi_lookup): Likewise.
15043         (vn_phi_insert): Likewise.
15045 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15047         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15048         that fits into uhwi or shwi, add DW_AT_const_value regardless
15049         of early_dwarf without going through RTL, using add_AT_unsigned
15050         or add_AT_int.
15052         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15053         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15054         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15055         (DEBUG_MACRO_SECTION): Likewise.
15056         (DEBUG_LTO_MACRO_SECTION): Likewise.
15057         (DEBUG_STR_DWO_SECTION): Likewise.
15058         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15059         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15060         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15061         (DEBUG_LTO_LINE_STR_SECTION): Define.
15062         (init_sections_and_labels): Initialize debug_line_str_section
15063         variable.  Initialize debug_loc_section for -gdwarf-5 to
15064         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15066 2017-09-20  Richard Biener  <rguenther@suse.de>
15068         * graphite-sese-to-poly.c (extract_affine): Properly handle
15069         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15071 2017-09-20  Richard Biener  <rguenther@suse.de>
15073         PR tree-optimization/81373
15074         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15075         Force SESE live-out defs to be handled even if they are
15076         scev_analyzable_p.
15078 2017-09-19  Jeff Law  <law@redhat.com>
15080         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15081         nothing for stack adjustments with REG_STACK_CHECK.
15082         * sched-deps.c (parse_add_or_inc): Reject insns with
15083         REG_STACK_CHECK from dependency breaking.
15084         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15085         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15086         * reg-notes.def (STACK_CHECK): New note.
15088         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15089         (ix86_expand_prologue): Dump stack clash info as needed.
15090         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15092         * function.c (dump_stack_clash_frame_info): New function.
15093         * function.h (dump_stack_clash_frame_info): Prototype.
15094         (enum stack_clash_probes): New enum.
15096         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15097         flag_stack_clash_protection.
15098         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15099         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15100         (arm_frame_pointer_required): Likewise.
15101         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15102         (ia64_expand_prologue): Likewise.
15103         * config/mips/mips.c (mips_expand_prologue): Likewise.
15104         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15105         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15106         (sparc_flat_expand_prologue): Likewise.
15107         * config/spu/spu.c (spu_expand_prologue): Likewise.
15109         * explow.c: Include "params.h".
15110         (anti_adjust_stack_and_probe_stack_clash): New function.
15111         (get_stack_check_protect): Likewise.
15112         (compute_stack_clash_protection_loop_data): Likewise.
15113         (emit_stack_clash_protection_loop_start): Likewise.
15114         (emit_stack_clash_protection_loop_end): Likewise.
15115         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15116         Use anti_adjust_stack_and_probe_stack_clash.
15117         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15118         (emit_stack_clash_protection_loop_start): Likewise.
15119         (emit_stack_clash_protection_loop_end): Likewise.
15120         * rtl.h (get_stack_check_protect): Prototype.
15121         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15122         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15123         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15124         Prototype.
15125         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15126         Add @hook.
15127         * doc/tm.texi: Rebuilt.
15128         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15129         get_stack_check_protect.
15130         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15131         * config/arm/arm.c (arm_expand_prologue): Likewise.
15132         (arm_frame_pointer_required): Likewise.
15133         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15134         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15135         * config/mips/mips.c (mips_expand_prologue): Likewise.
15136         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15137         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15138         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15139         (sparc_flat_expand_prologue): Likewise.
15141         * common.opt (-fstack-clash-protection): New option.
15142         * flag-types.h (enum stack_check_type): Note difference between
15143         -fstack-check= and -fstack-clash-protection.
15144         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15145         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15146         * toplev.c (process_options): Issue warnings/errors for cases
15147         not handled with -fstack-clash-protection.
15148         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15149         (-fstack-check): Note additional problem with -fstack-check=generic.
15150         Note that -fstack-check is primarily for Ada and refer users
15151         to -fstack-clash-protection for stack-clash-protection.
15152         Document new params for stack clash protection.
15154 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15156         * config/i386/i386.c (ix86_split_long_move): Do not handle
15157         address used for LEA in a special way.
15159 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15161         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15163 2017-09-19  Martin Sebor  <msebor@redhat.com>
15165         PR c/81854
15166         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15167         of incompatible types.
15169 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15171         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15172         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15173         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15174         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15176 2017-09-19  Richard Biener  <rguenther@suse.de>
15178         PR tree-optimization/82244
15179         * tree-vrp.c (remove_range_assertions): Do not propagate
15180         a constant to abnormals but replace the assert with a copy.
15182 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15184         PR rtl-optimization/57878
15185         PR rtl-optimization/68988
15186         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15187         avoidance test involving non_reload_pseudos.  Move frequency test
15188         below the general fragmentation avoidance test.
15190 2017-09-19  Richard Biener  <rguenther@suse.de>
15192         PR tree-optimization/69728
15193         * graphite-sese-to-poly.c (schedule_error): New global.
15194         (add_loop_schedule): Handle empty domain by failing the
15195         schedule.
15196         (build_original_schedule): Handle schedule_error.
15198 2017-09-19  Richard Biener  <rguenther@suse.de>
15200         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15201         Do not iterate to sibling loops but only to siblings of inner
15202         loops.
15204 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15206         PR target/81613
15207         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15208         that the registers are different.
15210 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15212         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15213         to processor_model and "amdfam17h" to arch_names_table.
15214         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15216 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15218         PR c/82234
15219         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15221 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15222             Alan Hayward  <alan.hayward@arm.com>
15223             David Sherwood  <david.sherwood@arm.com>
15225         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15226         with a vec_info *.
15227         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15228         accordingly.
15229         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15230         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15231         vect_schedule_slp_instance.
15232         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15233         Update call to vect_slp_analyze_node_operations.  Simplify return
15234         value.
15235         (vect_slp_analyze_bb_1): Update call accordingly.
15236         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15237         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15238         (vect_schedule_slp): Update call accordingly.
15240 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15241             Alan Hayward  <alan.hayward@arm.com>
15242             David Sherwood  <david.sherwood@arm.com>
15244         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15245         with types that aren't in fact scalar.
15247 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15249         * tree-vect-slp.c (vect_record_max_nunits): New function,
15250         split out from...
15251         (vect_build_slp_tree_1): ...here.
15252         (vect_build_slp_tree_2): Call it for phis too.
15254 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15256         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15257         to vect_get_vec_def_for_operand when getting the mask operand.
15259 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15260             Alan Hayward  <alan.hayward@arm.com>
15261             David Sherwood  <david.sherwood@arm.com>
15263         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15264         bitstart.
15266 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15267             Alan Hayward  <alan.hayward@arm.com>
15268             David Sherwood  <david.sherwood@arm.com>
15270         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15271         calculation for vector booleans.
15273 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15274             Alan Hayward  <alan.hayward@arm.com>
15275             David Sherwood  <david.sherwood@arm.com>
15277         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15278         split out from...
15279         (vect_transform_stmt): ...here.
15280         (vect_analyze_stmt): Use it instead of calling
15281         vectorizable_live_operation directly.
15283 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15285         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15286         non-SIMT targets in acc vector loops.
15288 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15290         * configure.ac: Add arc and check if assembler supports gdwarf2.
15291         * configure: Regenerate.
15293 2017-09-18  Richard Biener  <rguenther@suse.de>
15295         PR tree-optimization/82220
15296         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15297         epilogue niters from the min_profitable_iters compute.
15299 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15301         PR target/82145
15302         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15303         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15304         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15306 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15308         PR target/81361
15309         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15310         switching to a new text section.
15312 2017-09-18  Richard Biener  <rguenther@suse.de>
15314         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15315         Simplify.
15316         (build_alias_set): Reject aliases with no access function.
15318 2017-09-18  Richard Biener  <rguenther@suse.de>
15320         PR tree-optimization/79622
15321         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15322         handle PHIs.
15323         (build_cross_bb_scalars_use): Likewise.
15325 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15327         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15329 2017-09-18  Alan Modra  <amodra@gmail.com>
15331         PR target/81996
15332         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15333         stack_pointer_rtx for count 0.  Update comments.  Break up
15334         large rtl expression.
15336 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15338         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15339         Increase to 20 bytes.
15340         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15341         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15342         or avx version of the stub.
15344 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15346         PR target/82166
15347         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15348         compute the minimum stack alignment.  Also update preferred stack
15349         boundary for leaf functions.
15351 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15353         PR tree-optimization/82228
15354         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15355         of ncopies.
15357 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15359         * common/config/nds32/nds32-common.c
15360         (nds32_option_optimization_table): Refine formatting.
15361         (nds32_option_optimization_table): Use -fsched-pressure and
15362         -fomit-frame-pointer for specific optimization level.
15364 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15366         * config/nds32/nds32.c: Refine formatting and comments.
15367         * config/nds32/nds32.h: Likewise.
15368         * config/nds32/nds32.md: Likewise.
15369         * config/nds32/nds32-cost.c: Likewise.
15370         * config/nds32/nds32-isr.c: Likewise.
15371         * config/nds32/nds32-md-auxiliary.c: Likewise.
15372         * config/nds32/nds32-multiple.md: Likewise.
15373         * config/nds32/nds32-predicates.c: Likewise.
15375 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15376             Jakub Jelinek  <jakub@redhat.com>
15378         Add support for -std=c++2a.
15379         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15380         or -std=gnu+2a.
15381         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15383 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15385         PR target/82066
15386         * doc/extend.texi (Common Function Attributes): Add 
15387         references to ARM, AArch64, and S/390 specific attributes.
15388         (Function Specific Option Pragmas): Add AArch64 and S/390
15389         to list of back ends that support the target pragma.
15391 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15393         * doc/standards.texi: Fix C++17 description.  Update URLs for
15394         C++11 & 14.
15396 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15398         * common.opt (Wcast-align=strict): New warning option.
15399         * doc/invoke.texi: Document -Wcast-align=strict. 
15401 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15403         * cgraph.h (cgraph_thunk_info): Add comments.
15404         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15405         assert for VIRTUAL_* arguments stricter.
15407 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15409         PR tree-optimization/71026
15410         * match.pd: Move RDIV patterns from fold-const.c
15411         * fold-const.c (distribute_real_division): Removed.
15412         (fold_binary_loc): Remove calls to distribute_real_divison.
15414 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15416         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15417         c++1z and gnu++1z as deprecated.  Change other references to
15418         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15419         Change -Wc++1z-compat to -Wc++17-compat.
15420         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15421         * dwarf2out.c (highest_c_language): Handle C++17.
15422         (gen_compile_unit_die): Likewise.
15424 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15426         PR rtl-optimization/82192
15427         * combine.c (make_extraction): Don't look through non-paradoxical
15428         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15429         inner's mode.
15431 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15432             Alan Hayward  <alan.hayward@arm.com>
15433             David Sherwood  <david.sherwood@arm.com>
15435         * target.def (function_arg_offset): New hook.
15436         * targhooks.h (default_function_arg_offset): Declare.
15437         * targhooks.c (default_function_arg_offset): New function.
15438         * function.c (locate_and_pad_parm): Use
15439         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15440         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15441         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15442         * doc/tm.texi: Regenerate.
15443         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15444         * config/spu/spu.c (spu_function_arg_offset): New function.
15445         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15446         * system.h (FUNCTION_ARG_OFFSET): Poison.
15448 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15449             Alan Hayard  <alan.hayward@arm.com>
15450             David Sherwood  <david.sherwood@arm.com>
15452         * target.def (truly_noop_truncation): New hook.
15453         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15454         than TRULY_NOOP_TRUNCATION.
15455         * hooks.h (hook_bool_uint_uint_true): Declare.
15456         * hooks.c (hook_bool_uint_uint_true): New function.
15457         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15458         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15459         * doc/tm.texi: Regenerate.
15460         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15461         rather than TRULY_NOOP_TRUNCATION in comments.
15462         (simplify_comparison): Likewise.
15463         (record_truncated_value): Likewise.
15464         * expmed.c (extract_bit_field_1): Likewise.
15465         (extract_split_bit_field): Likewise.
15466         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15467         instead of TRULY_NOOP_TRUNCATION.
15468         * function.c (assign_parm_setup_block): Likewise.
15469         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15470         * rtlhooks.c: Include target.h.
15471         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15472         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15473         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15474         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15475         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15476         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15477         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15478         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15479         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15480         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15481         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15482         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15483         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15484         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15485         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15486         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15487         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15488         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15489         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15490         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15491         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15492         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15493         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15494         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15495         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15496         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15497         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15498         rather than TRULY_NOOP_TRUNCATION in comments.
15499         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15500         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15501         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15502         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15503         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15504         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15505         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15506         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15507         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15508         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15509         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15510         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15511         rather than TRULY_NOOP_TRUNCATION in comments.
15512         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15513         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15514         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15515         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15516         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15517         TRULY_NOOP_TRUNCATION condition.
15518         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15519         (TRULY_NOOP_TRUNCATION): Delete.
15520         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15521         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15522         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15523         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15524         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15525         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15526         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15527         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15528         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15529         rather than TRULY_NOOP_TRUNCATION in comments.
15530         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15531         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15532         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15533         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15534         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15535         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15537 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15539         PR target/67591
15540         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15541         (*cmp_ior): Likewise.
15542         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15543         (*ior_scc_scc_cmp): Likewise.
15544         (*and_scc_scc): Likewise.
15545         (*and_scc_scc_cmp): Likewise.
15547 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15548             Alan Hayard  <alan.hayward@arm.com>
15549             David Sherwood  <david.sherwood@arm.com>
15551         * target.def (can_change_mode_class): New hook.
15552         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15553         (hard_regno_nregs): Likewise.
15554         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15555         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15556         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15557         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15558         (LOAD_EXTEND_OP): Update accordingly.
15559         * doc/tm.texi: Regenerate.
15560         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15561         CANNOT_CHANGE_MODE_CLASS.
15562         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15563         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15564         * combine.c (simplify_set): Update accordingly.
15565         * emit-rtl.c (validate_subreg): Likewise.
15566         * recog.c (general_operand): Likewise.
15567         * regcprop.c (mode_change_ok): Likewise.
15568         * reload1.c (choose_reload_regs): Likewise.
15569         (inherit_piecemeal_p): Likewise.
15570         * rtlanal.c (simplify_subreg_regno): Likewise.
15571         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15572         instead of CANNOT_CHANGE_MODE_CLASS.
15573         (reload_cse_simplify_operands): Likewise.
15574         * reload.c (push_reload): Use targetm.can_change_mode_class
15575         instead of CANNOT_CHANGE_MODE_CLASS.
15576         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15577         REG_CANNOT_CHANGE_MODE_P.
15578         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15579         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15580         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15581         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15582         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15583         (arm_can_change_mode_class): New function.
15584         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15585         than CANNOT_CHANGE_MODE_CLASS in comments.
15586         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15587         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15588         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15589         (ix86_can_change_mode_class): ...this new function, inverting the
15590         sense of the return value.
15591         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15592         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15593         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15594         (ia64_can_change_mode_class): New function.
15595         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15596         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15597         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15598         (m32c_can_change_mode_class): ...this new function, inverting the
15599         sense of the return value.
15600         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15601         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15602         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15603         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15604         (mips_can_change_mode_class): ...this new function, inverting the
15605         sense of the return value.
15606         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15607         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15608         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15609         (msp430_can_change_mode_class): New function.
15610         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15611         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15612         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15613         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15614         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15615         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15616         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15617         (pa_cannot_change_mode_class): Replace with...
15618         (pa_can_change_mode_class): ...this new function, inverting the
15619         sense of the return value.
15620         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15621         than CANNOT_CHANGE_MODE_CLASS in comments.
15622         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15623         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15624         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15625         (pdp11_cannot_change_mode_class): Replace with...
15626         (pdp11_can_change_mode_class): ...this new function, inverting the
15627         sense of the return value.
15628         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15629         * config/powerpcspe/powerpcspe-protos.h
15630         (rs6000_cannot_change_mode_class_ptr): Delete.
15631         * config/powerpcspe/powerpcspe.c
15632         (rs6000_cannot_change_mode_class_ptr): Delete.
15633         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15634         (rs6000_option_override_internal): Assign to
15635         targetm.can_change_mode_class instead of
15636         rs6000_cannot_change_mode_class_ptr.
15637         (rs6000_cannot_change_mode_class): Replace with...
15638         (rs6000_can_change_mode_class): ...this new function, inverting the
15639         sense of the return value.
15640         (rs6000_debug_cannot_change_mode_class): Replace with...
15641         (rs6000_debug_can_change_mode_class): ...this new function.
15642         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15643         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15644         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15645         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15646         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15647         Delete.
15648         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15649         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15650         (rs6000_option_override_internal): Assign to
15651         targetm.can_change_mode_class instead of
15652         rs6000_cannot_change_mode_class_ptr.
15653         (rs6000_cannot_change_mode_class): Replace with...
15654         (rs6000_can_change_mode_class): ...this new function, inverting the
15655         sense of the return value.
15656         (rs6000_debug_cannot_change_mode_class): Replace with...
15657         (rs6000_debug_can_change_mode_class): ...this new function.
15658         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15659         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15660         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15661         (s390_can_change_mode_class): ...this new function, inverting the
15662         sense of the return value.
15663         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15664         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15665         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15666         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15667         (sh_cannot_change_mode_class): Replace with...
15668         (sh_can_change_mode_class): ...this new function, inverting the
15669         sense of the return value.
15670         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15671         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15672         (sparc_can_change_mode_class): New function.
15673         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15674         * config/spu/spu.c (spu_can_change_mode_class): New function.
15675         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15676         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15677         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15678         (visium_can_change_mode_class): New function.
15679         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15681 2017-09-15  Richard Biener  <rguenther@suse.de>
15683         PR tree-optimization/82217
15684         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15685         but not undefined case.
15687 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15689         PR target/82145
15690         * postreload.c (reload_cse_simplify_operands): Skip
15691         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15693 2017-09-15  Richard Biener  <rguenther@suse.de>
15695         PR tree-optimization/68823
15696         * graphite-scop-detection.c (build_alias_set): If we have a
15697         possible dependence check whether we can handle them by just
15698         looking at the DRs DR_ACCESS_FNs.
15699         (build_scops): If build_alias_set fails, fail the SCOP.
15701 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15703         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15704         to support float128 built-in functions that require the ISA 3.0
15705         hardware.
15706         (BU_FLOAT128_3_HW): Likewise.
15707         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15708         built-in functions.
15709         (FMAF128): Likewise.
15710         (FMAQ): Likewise.
15711         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15712         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15713         floating point instructions.
15714         (rs6000_invalid_builtin): Likewise.
15715         (rs6000_builtin_mask_names): Likewise.
15716         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15717         (RS6000_BTM_FLOAT128_HW): Likewise.
15718         (RS6000_BTM_COMMON): Likewise.
15719         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15720         function.
15721         * doc/extend.texi (RS/6000 built-in functions): Document the
15722         IEEE 128-bit floating point square root and fused multiply-add
15723         built-in functions.
15725 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15727         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15728         reg (r2) isn't in the set of registers defined in the prologue.
15730 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15731             Alan Hayward  <alan.hayward@arm.com>
15732             David Sherwood  <david.sherwood@arm.com>
15734         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15735         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15736         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15737         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15738         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15739         accordingly.
15740         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15741         max_vectorization_factor.
15742         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15744 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15745             Alan Hayward  <alan.hayward@arm.com>
15746             David Sherwood  <david.sherwood@arm.com>
15748         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15749         (vect_worthwhile_without_simd_p): Declare.
15750         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15751         (vectorizable_reduction): Use it.
15752         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15753         (vectorizable_operation): Likewise.
15755 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15756             Alan Hayward  <alan.hayward@arm.com>
15757             David Sherwood  <david.sherwood@arm.com>
15759         * tree-vectorizer.h (vect_get_num_copies): New function.
15760         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15761         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15762         (vectorizable_induction): Likewise.
15763         (vectorizable_live_operation): Likewise.
15764         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15765         (vectorizable_bswap): Likewise.
15766         (vectorizable_call): Likewise.
15767         (vectorizable_conversion): Likewise.
15768         (vectorizable_assignment): Likewise.
15769         (vectorizable_shift): Likewise.
15770         (vectorizable_operation): Likewise.
15771         (vectorizable_store): Likewise.
15772         (vectorizable_load): Likewise.
15773         (vectorizable_condition): Likewise.
15774         (vectorizable_comparison): Likewise.
15775         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15777 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15778             Alan Hayward  <alan.hayward@arm.com>
15779             David Sherwood  <david.sherwood@arm.com>
15781         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15782         of vect_init_vector.
15784 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15785             Alan Hayward  <alan.hayward@arm.com>
15786             David Sherwood  <david.sherwood@arm.com>
15788         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15789         an inline wrapper that provides a location.
15790         (gimple_build_vector): Likewise.
15791         * gimple-fold.c (gimple_build_vector_from_val): New function.
15792         (gimple_build_vector): Likewise.
15793         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15794         functions to build the initial value.  Always return a gimple value.
15795         (get_initial_defs_for_reduction): Likewise.  Only compute
15796         neutral_vec once.
15797         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15798         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15799         (vectorizable_induction): Use gimple_build_vector rather than
15800         vect_init_vector.
15802 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15803             Alan Hayward  <alan.hayward@arm.com>
15804             David Sherwood  <david.sherwood@arm.com>
15806         * target.h (vec_perm_indices): New typedef.
15807         (auto_vec_perm_indices): Likewise.
15808         * optabs-query.h: Include target.h
15809         (can_vec_perm_p): Take a vec_perm_indices *.
15810         * optabs-query.c (can_vec_perm_p): Likewise.
15811         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15812         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15813         * tree-vect-generic.c (lower_vec_perm): Likewise.
15814         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15815         (vect_grouped_load_supported): Likewise.
15816         (vect_shift_permute_load_chain): Likewise.
15817         (vect_permute_store_chain): Use auto_vec_perm_indices.
15818         (vect_permute_load_chain): Likewise.
15819         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15820         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15821         Update uses of can_vec_perm_p.
15822         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15823         mode with a number of elements.  Take a vec_perm_indices *.
15824         (vect_create_epilog_for_reduction): Update accordingly.
15825         Use auto_vec_perm_indices.
15826         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15827         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15828         (vect_transform_slp_perm_load): Likewise.
15829         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15830         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15831         (vect_gen_perm_mask_checked): Likewise.
15832         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15833         (vect_gen_perm_mask_checked): Likewise.
15834         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15835         (vectorizable_store): Likewise.
15836         (vectorizable_load): Likewise.
15837         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15838         (vectorizable_bswap): Likewise.
15840 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15841             Alan Hayward  <alan.hayward@arm.com>
15842             David Sherwood  <david.sherwood@arm.com>
15844         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15845         * tree.c (build_vector): Likewise.
15846         (build_vector_from_ctor): Update accordingly.
15847         (build_vector_from_val): Likewise.
15848         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15849         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15850         * tree-vect-generic.c (add_rshift): Likewise.
15851         (expand_vector_divmod): Likewise.
15852         (optimize_vector_constructor): Likewise.
15853         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15854         (vect_transform_slp_perm_load): Likewise.
15855         (vect_schedule_slp_instance): Likewise.
15856         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15857         (vectorizable_call): Likewise.
15858         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15859         * expmed.c (make_tree): Likewise.
15860         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15861         a vector passed to build_vector.
15862         (fold_convert_const): Likewise.
15863         (exact_inverse): Likewise.
15864         (fold_ternary_loc): Likewise.
15865         (fold_relational_const): Likewise.
15866         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15867         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15868         (const_unop): Likewise.  Store the reduction accumulator in a
15869         variable rather than an array.
15870         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15871         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15872         the new vector, rather than constructing it after the input arrays.
15873         (native_interpret_vector): Use auto_vec<tree> when building
15874         a vector passed to build_vector.  Add elements in order.
15875         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15876         auto_vec<tree> when building a vector passed to build_vector.
15877         (vect_create_epilog_for_reduction): Likewise.
15878         (vectorizable_induction): Likewise.
15879         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15880         case statements.
15881         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15882         to a vec<tree> *.
15883         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15884         passed to build_vector.
15886 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15887             Alan Hayward  <alan.hayward@arm.com>
15888             David Sherwood  <david.sherwood@arm.com>
15890         * tree-core.h (tree_base::u): Add an "nelts" field.
15891         (tree_vector): Use VECTOR_CST_NELTS as the length.
15892         * tree.c (tree_size): Likewise.
15893         (make_vector): Initialize VECTOR_CST_NELTS.
15894         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15895         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15896         TYPE_VECTOR_SUBPARTS.
15897         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15898         as the number of units, setting it from VECTOR_CST_NELTS.
15899         (const_vector_from_tree): Likewise.
15900         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15901         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15902         (fold_negate_expr_1): Likewise.
15903         (fold_convert_const): Likewise.
15904         (const_binop): Likewise.  Differentiate the number of output and
15905         input elements.
15906         (const_unop): Likewise.
15907         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15908         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15909         in cases that did the opposite.
15911 2017-09-14  Richard Biener  <rguenther@suse.de>
15913         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15914         to VN_TOP.
15916 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15918         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15920 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15922         PR target/81325
15923         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15924         if and where to split a bb, except for splitting before debug insn
15925         sequences followed by non-label real insn.  Delete debug insns
15926         in between basic blocks.
15928         * combine.c (make_compound_operation_int): Formatting fixes.
15930         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15931         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15932         * config/netbsd.h (LINK_EH_SPEC): Likewise.
15933         * config/sol2.h (LINK_EH_SPEC): Likewise.
15934         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15935         * config/s390/linux.h (LINK_SPEC): Likewise.
15936         * config/freebsd.h (LINK_EH_SPEC): Likewise.
15937         * config/openbsd.h (LINK_EH_SPEC): Likewise.
15938         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15939         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15940         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
15941         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15942         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
15943         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
15945 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
15947         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
15948         support.
15949         (ENDFILE_LINUX_SPEC): Likewise.
15950         (LINK_EH_SPEC): Likewise.
15951         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
15952         (LINK_OS_LINUX_SPEC32): Likewise.
15953         (LINK_OS_LINUX_SPEC64): Likewise.
15954         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
15955         (LINK_OS_LINUX_SPEC): Likewise.
15957 2017-09-13  Martin Liska  <mliska@suse.cz>
15959         PR middle-end/82154
15960         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
15961         CASE_HIGH is NULL_TREE.
15963 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15964             Alan Hayward  <alan.hayward@arm.com>
15965             David Sherwood  <david.sherwood@arm.com>
15967         * target.def (secondary_memory_needed): New hook.
15968         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
15969         instead of SECONDARY_MEMORY_NEEDED.
15970         (secondary_memory_needed_mode): Likewise.
15971         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
15972         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
15973         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
15974         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
15975         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
15976         * doc/tm.texi: Regenerate.
15977         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
15978         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
15979         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15980         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
15981         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
15982         * config/i386/i386.c (inline_secondary_memory_needed): Put the
15983         mode argument first and change the reg_class arguments to reg_class_t.
15984         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
15985         Make static.  Update the call to inline_secondary_memory_needed.
15986         (ix86_register_move_cost): Update the call to
15987         inline_secondary_memory_needed.
15988         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15989         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
15990         definition.
15991         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
15992         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15993         in comment.
15994         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
15995         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
15996         * config/mips/mips.c (mips_secondary_memory_needed): Make static
15997         and match hook interface.  Add comment from mips.h.
15998         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15999         * config/mmix/mmix.md (truncdfsf2): Refer to
16000         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16001         in comment.
16002         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
16003         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
16004         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16005         (pa_secondary_memory_needed): New function.
16006         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
16007         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
16008         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16009         (pdp11_secondary_memory_needed): Make static and match hook interface.
16010         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
16011         * config/powerpcspe/powerpcspe-protos.h
16012         (rs6000_secondary_memory_needed_ptr): Delete.
16013         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
16014         Delete.
16015         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16016         (rs6000_option_override_internal): Assign to
16017         targetm.secondary_memory_needed rather than
16018         rs6000_secondary_memory_needed_ptr.
16019         (rs6000_secondary_memory_needed): Match hook interface.
16020         (rs6000_debug_secondary_memory_needed): Likewise.
16021         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
16022         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
16023         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16024         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16025         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16026         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16027         Delete.
16028         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
16029         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16030         (rs6000_option_override_internal): Assign to
16031         targetm.secondary_memory_needed rather than
16032         rs6000_secondary_memory_needed_ptr.
16033         (rs6000_secondary_memory_needed): Match hook interface.
16034         (rs6000_debug_secondary_memory_needed): Likewise.
16035         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16036         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16037         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16038         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16039         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16040         (sparc_secondary_memory_needed): New function.
16041         * lra-constraints.c (check_and_process_move): Refer to
16042         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16043         in comment.
16044         (curr_insn_transform): Likewise.
16045         (process_alt_operands): Use targetm.secondary_memory_needed
16046         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16047         (check_secondary_memory_needed_p): Likewise.
16048         (choose_split_class): Likewise.
16049         * reload.c: Unconditionally include code that was previously
16050         conditional on SECONDARY_MEMORY_NEEDED.
16051         (push_secondary_reload): Use targetm.secondary_memory_needed
16052         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16053         (push_reload): Likewise.
16054         * reload1.c: Unconditionally include code that was previously
16055         conditional on SECONDARY_MEMORY_NEEDED.
16056         (choose_reload_regs): Use targetm.secondary_memory_needed
16057         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16058         (gen_reload): Likewise.
16059         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16061 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16062             Alan Hayward  <alan.hayward@arm.com>
16063             David Sherwood  <david.sherwood@arm.com>
16065         * target.def (secondary_memory_needed_mode): New hook:
16066         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16067         * targhooks.h (default_secondary_memory_needed_mode): New function.
16068         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16069         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16070         * doc/tm.texi: Regenerate.
16071         * lra-constraints.c (check_and_process_move): Use
16072         targetm.secondary_memory_needed_mode instead of
16073         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16074         (curr_insn_transform): Likewise.
16075         * reload.c (get_secondary_mem): Likewise.
16076         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16077         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16078         function.
16079         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16080         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16081         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16082         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16083         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16084         Delete.
16085         * config/powerpcspe/powerpcspe-protos.h
16086         (rs6000_secondary_memory_needed_mode): Delete.
16087         * config/powerpcspe/powerpcspe.c
16088         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16089         (rs6000_secondary_memory_needed_mode): Make static.
16090         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16091         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16092         Delete.
16093         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16094         Redefine.
16095         (rs6000_secondary_memory_needed_mode): Make static.
16096         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16097         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16098         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16099         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16100         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16101         Redefine.
16102         (sparc_secondary_memory_needed_mode): New function.
16103         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16105 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16107         * config/aarch64/constraints.md (Umq): New constraint.
16108         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16109         Change to use Umq.
16110         (mov<mode>): Update condition.
16112 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16114         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16115         when bitposition is the same.
16117 2017-09-13  Richard Biener  <rguenther@suse.de>
16119         * dwarf2out.c (output_die_symbol): Remove.
16120         (output_die): Do not output a DIEs symbol.
16122 2017-09-13  Richard Biener  <rguenther@suse.de>
16124         PR middle-end/82128
16125         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16126         default-def to avoid breaking iterator update with the weird
16127         interaction with cgraph_update_edges_for_call_stmt_node.
16129 2017-09-13  Richard Biener  <rguenther@suse.de>
16131         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16132         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16133         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16134         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16136 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16138         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16139         Disable pc relative literal load irrespective of
16140         TARGET_FIX_ERR_A53_84341 for default.
16142 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16144         * config/sparc/sparc.c (output_return): Output the source location of
16145         the insn in the delay slot, if any.
16146         (output_sibcall): Likewise.
16148 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16150         PR driver/81498
16151         * common.opt (-static-pie): New alias.
16152         (shared): Negate static-pie.
16153         (-no-pie): Update help text.
16154         (-pie): Likewise.
16155         (static-pie): New option.
16156         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16157         -static-pie support.
16158         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16159         (LINK_EH_SPEC): Likewise.
16160         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16161         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16162         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16163         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16164         (init_gcc_specs): Likewise.
16165         (init_spec): Likewise.
16166         (display_help): Update help message for -pie.
16167         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16168         -static-pie.
16170 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16172         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16173         (movdi_aarch64): Likewise.
16174         (movti_aarch64): Likewise.
16176 2017-09-12 Simon Wright <simon@pushface.org>
16178         PR target/80204
16179         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16180         calculation of the minor version, always output as 0.
16182 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16184         PR target/82112
16185         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16186         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16187         on it early, rather than manual conversion late.  For
16188         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16189         instead of performing manual conversion.
16191 2017-09-12  Carl Love  <cel@us.ibm.com>
16193         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16194         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16195         vmulouw, vmulosw.
16196         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16197         VMULOSW): Add definitions.
16198         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16199         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16200         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16201         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16202         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16204 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16206         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16207         types correctly.
16208         (movti_aarch64): Likewise.
16209         (movdf_aarch64): Likewise.
16210         (movtf_aarch64): Likewise.
16211         (load_pairdi): Likewise.
16212         (store_pairdi): Likewise.
16213         (load_pairdf): Likewise.
16214         (store_pairdf): Likewise.
16215         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16216         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16217         (ldr_got_small_<mode>): Likewise.
16218         (ldr_got_small_28k_<mode>): Likewise.
16219         (ldr_got_tiny): Likewise.
16220         * config/aarch64/iterators.md (ldst_sz): New.
16221         (ldpstp_sz): Likewise.
16222         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16223         to store_16.
16224         (thunderx_load): Split load_8 to load_16.
16225         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16226         load_8 to load_16.
16227         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16228         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16229         (xgene1_store_pair): Split store_8 to store_16.
16230         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16231         (falkor_st_0_st_sd): Split store_8 to store_16.
16233 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16235         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16236         and store1/2/3/4 to store_4/8/12/16.
16237         * config/aarch64/aarch64.md: Update for rename.
16238         * config/arm/arm.md: Likewise.: Likewise.
16239         * config/arm/arm.c: Likewise.
16240         * config/arm/thumb1.md: Likewise.
16241         * config/arm/thumb2.md: Likewise.
16242         * config/arm/vfp.md: Likewise.
16243         * config/arm/arm-generic.md: Likewise.
16244         * config/arm/arm1020e.md: Likewise.
16245         * config/arm/arm1026ejs.md: Likewise.
16246         * config/arm/arm1136jfs.md: Likewise.
16247         * config/arm/arm926ejs.md: Likewise.
16248         * config/arm/cortex-a15.md: Likewise.
16249         * config/arm/cortex-a17.md: Likewise.
16250         * config/arm/cortex-a5.md: Likewise.
16251         * config/arm/cortex-a53.md: Likewise.
16252         * config/arm/cortex-a57.md: Likewise.
16253         * config/arm/cortex-a7.md: Likewise.
16254         * config/arm/cortex-a8.md: Likewise.
16255         * config/arm/cortex-a9.md: Likewise.
16256         * config/arm/cortex-m4.md: Likewise.
16257         * config/arm/cortex-m7.md: Likewise.
16258         * config/arm/cortex-r4.md: Likewise.
16259         * config/arm/exynos-m1.md: Likewise.
16260         * config/arm/fa526.md: Likewise.
16261         * config/arm/fa606te.md: Likewise.
16262         * config/arm/fa626te.md: Likewise.
16263         * config/arm/fa726te.md: Likewise.
16264         * config/arm/fmp626.md: Likewise.
16265         * config/arm/iwmmxt.md: Likewise.
16266         * config/arm/ldmstm.md: Likewise.
16267         * config/arm/marvell-pj4.md: Likewise.
16268         * config/arm/xgene1.md: Likewise.
16269         * config/aarch64/thunderx.md: Likewise.
16270         * config/aarch64/thunderx2t99.md: Likewise.
16271         * config/aarch64/falkor.md: Likewise.
16273 2017-09-12  Martin Liska  <mliska@suse.cz>
16275         * attribs.c (private_lookup_attribute): New function.
16276         * attribs.h (private_lookup_attribute): Declared here.
16277         (lookup_attribute): Called from this place.
16279 2017-09-12  Richard Biener  <rguenther@suse.de>
16281         PR tree-optimization/82157
16282         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16283         stmts with side-effects.
16285 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16286             Alan Hayward  <alan.hayward@arm.com>
16287             David Sherwood <david.sherwood@arm.com>
16289         * target.def (hard_regno_nregs): New hook.
16290         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16291         * targhooks.h (default_hard_regno_nregs): Declare.
16292         * targhooks.c (default_hard_regno_nregs): New function.
16293         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16294         (TARGET_HARD_REGNO_NREGS): ...this hook.
16295         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16296         (CLASS_MAX_NREGS): Likewise.
16297         * doc/tm.texi: Regenerate.
16298         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16299         instead of HARD_REGNO_NREGS.
16300         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16301         HARD_REGNO_NREGS in the comment.
16302         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16303         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16304         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16305         Return an unsigned int.
16306         (TARGET_HARD_REGNO_NREGS): Redefine.
16307         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16308         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16309         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16310         (arc_hard_regno_nregs): New function.
16311         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16312         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16313         (arm_hard_regno_nregs): New function.
16314         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16315         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16316         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16317         (TARGET_HARD_REGNO_NREGS): Redefine.
16318         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16319         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16320         (HARD_REGNO_NREGS): Delete.
16321         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16322         (cr16_hard_regno_nregs): New function.
16323         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16324         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16325         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16326         (cris_hard_regno_nregs): New function.
16327         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16328         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16329         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16330         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16331         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16332         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16333         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16334         (frv_hard_regno_nregs): Make static.  Take and return an
16335         unsigned int.
16336         (frv_class_max_nregs): Remove outdated copy of documentation.
16337         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16338         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16339         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16340         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16341         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16342         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16343         (TARGET_HARD_REGNO_NREGS): Redefine.
16344         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16345         (CLASS_MAX_NREGS): Update comment.
16346         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16347         (ia64_hard_regno_nregs): New function.
16348         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16349         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16350         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16351         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16352         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16353         an unsigned int.
16354         (m32c_hard_regno_nregs): Likewise.  Make static.
16355         (TARGET_HARD_REGNO_NREGS): Redefine.
16356         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16357         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16358         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16359         (m68k_hard_regno_nregs): New function.
16360         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16361         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16362         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16363         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16364         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16365         Take and return an unsigned int.
16366         (TARGET_HARD_REGNO_NREGS): Redefine.
16367         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16368         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16369         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16370         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16371         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16372         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16373         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16374         (msp430_hard_regno_nregs): Make static.  Take and return an
16375         unsigned int.
16376         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16377         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16378         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16379         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16380         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16381         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16382         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16383         (TARGET_HARD_REGNO_NREGS): Redefine.
16384         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16385         (PA_HARD_REGNO_NREGS): ...this.
16386         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16387         (PA_HARD_REGNO_NREGS): ...this.
16388         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16389         (pa_hard_regno_nregs): New function.
16390         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16391         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16392         (pdp11_hard_regno_nregs): New function.
16393         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16394         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16395         (rs6000_hard_regno_nregs_hook): New function.
16396         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16397         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16398         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16399         Take and return an unsigned int.  Move earlier in file.
16400         (TARGET_HARD_REGNO_NREGS): Redefine.
16401         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16402         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16403         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16404         (rl78_hard_regno_nregs): Make static.  Take and return an
16405         unsigned int.
16406         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16407         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16408         (rs6000_hard_regno_nregs_hook): New function.
16409         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16410         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16411         (TARGET_HARD_REGNO_NREGS): Redefine.
16412         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16413         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16414         instead of HARD_REGNO_NREGS.
16415         (s390_hard_regno_nregs): New function.
16416         (s390_hard_regno_mode_ok): Add comment from s390.h.
16417         (TARGET_HARD_REGNO_NREGS): Redefine.
16418         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16419         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16420         (sh_hard_regno_nregs): New function.
16421         (sh_pass_in_reg_p): Use it.
16422         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16423         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16424         (sparc_hard_regno_nregs): New function.
16425         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16426         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16427         (spu_function_arg_advance): Use it, supplying a valid register number.
16428         (TARGET_HARD_REGNO_NREGS): Redefine.
16429         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16430         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16431         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16432         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16433         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16434         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16435         (CLASS_MAX_NREGS): Remove copy of old documentation.
16436         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16437         (visium_hard_regno_nregs): New function.
16438         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16439         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16440         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16441         xtensa_hard_regno_nregs): New function.
16442         * system.h (HARD_REGNO_NREGS): Poison.
16444 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16446         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16447         hard_regno_nregs instead of HARD_REGNO_NREGS.
16448         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16449         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16450         (c6x_expand_epilogue): Likewise.
16451         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16452         (frv_read_iacc_argument): Likewise.
16453         * config/sh/sh.c: Include regs.h.
16454         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16455         (regs_used): Likewise.
16456         (output_stack_adjust): Likewise.
16457         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16458         * expmed.c: Include regs.h.
16459         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16460         * ree.c: Include regs.h.
16461         (combine_reaching_defs): Use hard_regno_nregs instead of
16462         HARD_REGNO_NREGS.
16463         (add_removable_extension): Likewise.
16465 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16467         * regs.h (hard_regno_nregs): Turn into a function.
16468         (end_hard_regno): Update accordingly.
16469         * caller-save.c (setup_save_areas): Likewise.
16470         (save_call_clobbered_regs): Likewise.
16471         (replace_reg_with_saved_mem): Likewise.
16472         (insert_restore): Likewise.
16473         (insert_save): Likewise.
16474         * combine.c (can_change_dest_mode): Likewise.
16475         (move_deaths): Likewise.
16476         (distribute_notes): Likewise.
16477         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16478         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16479         (rs6000_split_multireg_move): Likewise.
16480         (rs6000_register_move_cost): Likewise.
16481         (rs6000_memory_move_cost): Likewise.
16482         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16483         (rs6000_split_multireg_move): Likewise.
16484         (rs6000_register_move_cost): Likewise.
16485         (rs6000_memory_move_cost): Likewise.
16486         * cselib.c (cselib_reset_table): Likewise.
16487         (cselib_lookup_1): Likewise.
16488         * emit-rtl.c (set_mode_and_regno): Likewise.
16489         * function.c (aggregate_value_p): Likewise.
16490         * ira-color.c (setup_profitable_hard_regs): Likewise.
16491         (check_hard_reg_p): Likewise.
16492         (calculate_saved_nregs): Likewise.
16493         (assign_hard_reg): Likewise.
16494         (improve_allocation): Likewise.
16495         (calculate_spill_cost): Likewise.
16496         * ira-emit.c (modify_move_list): Likewise.
16497         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16498         (ira_hard_reg_in_set_p): Likewise.
16499         * ira.c (setup_reg_mode_hard_regset): Likewise.
16500         (clarify_prohibited_class_mode_regs): Likewise.
16501         (check_allocation): Likewise.
16502         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16503         (lra_setup_reg_renumber): Likewise.
16504         (setup_try_hard_regno_pseudos): Likewise.
16505         (spill_for): Likewise.
16506         (assign_hard_regno): Likewise.
16507         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16508         * lra-constraints.c (in_class_p): Likewise.
16509         (lra_constraint_offset): Likewise.
16510         (simplify_operand_subreg): Likewise.
16511         (lra_constraints): Likewise.
16512         (split_reg): Likewise.
16513         (split_if_necessary): Likewise.
16514         (invariant_p): Likewise.
16515         (inherit_in_ebb): Likewise.
16516         * lra-lives.c (process_bb_lives): Likewise.
16517         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16518         (get_hard_regs): Likewise.
16519         (do_remat): Likewise.
16520         * lra-spills.c (assign_spill_hard_regs): Likewise.
16521         * mode-switching.c (create_pre_exit): Likewise.
16522         * postreload.c (reload_combine_recognize_pattern): Likewise.
16523         * recog.c (peep2_find_free_register): Likewise.
16524         * regcprop.c (kill_value_regno): Likewise.
16525         (set_value_regno): Likewise.
16526         (copy_value): Likewise.
16527         (maybe_mode_change): Likewise.
16528         (find_oldest_value_reg): Likewise.
16529         (copyprop_hardreg_forward_1): Likewise.
16530         * regrename.c (check_new_reg_p): Likewise.
16531         (regrename_do_replace): Likewise.
16532         * reload.c (push_reload): Likewise.
16533         (combine_reloads): Likewise.
16534         (find_dummy_reload): Likewise.
16535         (operands_match_p): Likewise.
16536         (find_reloads): Likewise.
16537         (find_equiv_reg): Likewise.
16538         (reload_adjust_reg_for_mode): Likewise.
16539         * reload1.c (count_pseudo): Likewise.
16540         (count_spilled_pseudo): Likewise.
16541         (find_reg): Likewise.
16542         (clear_reload_reg_in_use): Likewise.
16543         (free_for_value_p): Likewise.
16544         (allocate_reload_reg): Likewise.
16545         (choose_reload_regs): Likewise.
16546         (reload_adjust_reg_for_temp): Likewise.
16547         (emit_reload_insns): Likewise.
16548         (delete_output_reload): Likewise.
16549         * rtlanal.c (subreg_get_info): Likewise.
16550         * sched-deps.c (sched_analyze_reg): Likewise.
16551         * sel-sched.c (init_regs_for_mode): Likewise.
16552         (mark_unavailable_hard_regs): Likewise.
16553         (choose_best_reg_1): Likewise.
16554         (verify_target_availability): Likewise.
16555         * valtrack.c (dead_debug_insert_temp): Likewise.
16556         * var-tracking.c (track_loc_p): Likewise.
16557         (emit_note_insn_var_location): Likewise.
16558         * varasm.c (make_decl_rtl): Likewise.
16559         * reginfo.c (choose_hard_reg_mode): Likewise.
16560         (init_reg_modes_target): Refer directly to
16561         this_target_regs->x_hard_regno_nregs.
16563 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16565         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16566         instead of hard_regno_nregs.
16568 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16570         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16571         end_hard_regno instead of hard_regno_nregs.
16572         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16573         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16574         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16575         * ira-color.c (improve_allocation): Likewise.
16576         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16577         * lra-lives.c (mark_regno_live): Likewise.
16578         (mark_regno_dead): Likewise.
16579         * lra-remat.c (operand_to_remat): Likewise.
16580         * lra.c (collect_non_operand_hard_regs): Likewise.
16581         * postreload.c (reload_combine_note_store): Likewise.
16582         (move2add_valid_value_p): Likewise.
16583         * reload.c (regno_clobbered_p): Likewise.
16585 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16587         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16588         hard_regno_nregs.
16589         * config/v850/v850.c (v850_reorg): Likewise.
16590         * reload.c (refers_to_regno_for_reload_p): Likewise.
16591         (find_equiv_reg): Likewise.
16592         * reload1.c (reload_reg_reaches_end_p): Likewise.
16594 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16596         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16597         hard_regno_nregs.
16598         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16599         * config/arm/arm.c (output_move_neon): Likewise.
16600         (arm_attr_length_move_neon): Likewise.
16601         (neon_split_vcombine): Likewise.
16602         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16603         (c6x_mark_reg_written): Likewise.
16604         (c6x_dwarf_register_span): Likewise.
16605         * config/i386/i386.c (ix86_save_reg): Likewise.
16606         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16607         (rws_access_reg): Likewise.
16608         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16609         * mode-switching.c (create_pre_exit): Likewise.
16610         * ree.c (combine_reaching_defs): Likewise.
16611         (add_removable_extension): Likewise.
16612         * regcprop.c (find_oldest_value_reg): Likewise.
16613         (copyprop_hardreg_forward_1): Likewise.
16614         * reload.c (reload_inner_reg_of_subreg): Likewise.
16615         (push_reload): Likewise.
16616         (combine_reloads): Likewise.
16617         (find_dummy_reload): Likewise.
16618         (reload_adjust_reg_for_mode): Likewise.
16619         * reload1.c (find_reload_regs): Likewise.
16620         (forget_old_reloads_1): Likewise.
16621         (reload_reg_free_for_value_p): Likewise.
16622         (reload_adjust_reg_for_temp): Likewise.
16623         (emit_reload_insns): Likewise.
16624         (delete_output_reload): Likewise.
16625         * sel-sched.c (choose_best_reg_1): Likewise.
16626         (choose_best_pseudo_reg): Likewise.
16628 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16629             Alan Hayward  <alan.hayward@arm.com>
16630             David Sherwood <david.sherwood@arm.com>
16632         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16633         * target.def (slow_unaligned_access): New hook.
16634         * targhooks.h (default_slow_unaligned_access): Declare.
16635         * targhooks.c (default_slow_unaligned_access): New function.
16636         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16637         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16638         * doc/tm.texi: Regenerate.
16639         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16640         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16641         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16642         definition.
16643         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16644         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16645         Redefine.
16646         (rs6000_slow_unaligned_access): New function.
16647         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16648         (expand_block_compare): Likewise.
16649         (expand_strn_compare): Likewise.
16650         (rs6000_rtx_costs): Likewise.
16651         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16652         (riscv_slow_unaligned_access): Likewise.
16653         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16654         (riscv_slow_unaligned_access_p): ...this and make static.
16655         (riscv_option_override): Update accordingly.
16656         (riscv_slow_unaligned_access): New function.
16657         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16658         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16659         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16660         (rs6000_slow_unaligned_access): New function.
16661         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16662         (rs6000_rtx_costs): Likewise.
16663         * config/rs6000/rs6000-string.c (expand_block_compare)
16664         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16665         of SLOW_UNALIGNED_ACCESS.
16666         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16667         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16668         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16669         of SLOW_UNALIGNED_ACCESS.
16670         * expmed.c (simple_mem_bitfield_p): Likewise.
16671         * expr.c (alignment_for_piecewise_move): Likewise.
16672         (emit_group_load_1): Likewise.
16673         (emit_group_store): Likewise.
16674         (copy_blkmode_from_reg): Likewise.
16675         (emit_push_insn): Likewise.
16676         (expand_assignment): Likewise.
16677         (store_field): Likewise.
16678         (expand_expr_real_1): Likewise.
16679         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16680         * lra-constraints.c (simplify_operand_subreg): Likewise.
16681         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16682         * gimple-ssa-store-merging.c: Likewise in block comment at start
16683         of file.
16684         * tree-ssa-strlen.c: Include target.h.
16685         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16686         of SLOW_UNALIGNED_ACCESS.
16687         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16689 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16691         PR rtl-optimization/82185
16692         * expmed.c (emit_store_flag_int): Only test tem if it has been
16693         initialized.
16695 2017-09-12  Richard Biener  <rguenther@suse.de>
16697         PR middle-end/82149
16698         * match.pd ((FTYPE) N CMP CST): Fix typo.
16700 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16702         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16703         attribute.
16704         (mips_near_type_p): Add 'short_call' attribute as a synonym
16705         for 'near'.
16706         * doc/extend.texi (short_call): Document new function attribute.
16708 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16710         PR target/82112
16711         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16712         assertion check that in the condition.
16713         (get_atomic_generic_size): Likewise.  Before testing if parameter
16714         has pointer type, if it has array type, call for C++
16715         default_conversion to perform array-to-pointer conversion.
16717 2017-09-12  Richard Biener  <rguenther@suse.de>
16719         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16720         for operations we cannot scalarize.
16722 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16724         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16725         vectors heap vectors.  Clean up comments.
16726         Make visited_bbs a reference.
16727         (profitable_jump_thread_path): Make GC
16728         vectors heap vectors.  Clean up comments.
16729         Misc cleanups.
16730         (convert_and_register_jump_thread_path): Make GC vectors heap
16731         vectors.
16732         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16733         Make visited_bbs a reference.
16734         (handle_phi): Abstract common code to to
16735         register_jump_thread_path_if_profitable.
16736         Rename VAR_BB to DEF_BB.
16737         Update comments.
16738         Make GC vectors heap vectors.
16739         Make visited_bbs a reference.
16740         (handle_assignment): Same.
16741         (register_jump_thread_path_if_profitable): New.
16742         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16743         DEF_BB.
16744         Make GC vectors heap vectors.  Clean up comments.
16745         Make visited_bbs a reference.
16746         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16747         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16748         comment.
16750 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16752         PR target/82181
16753         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16754         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16756 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16758         Revert r251800 and r251799.
16760 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16762         PR hsa/82119
16763         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16764         arguments in advance.
16765         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16766         use it to find predecessor edges.
16767         (naive_outof_ssa): Collect vector of predecessors.
16769 2017-09-08  Jason Merrill  <jason@redhat.com>
16771         PR c++/70029 - ICE with ref-qualifier and -flto
16772         * langhooks.h (struct lang_hooks_for_types): Add
16773         copy_lang_qualifiers.
16774         * attribs.c (build_type_attribute_qual_variant): Use it.
16775         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16776         NULL.
16777         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16778         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16780 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16782         PR target/81988
16783         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16784         (*mulsi3_sp64): New instruction.
16785         (mulsi3): New expander.
16787 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16789         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16791 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16793         * sancov.c: Include memmodel.h.
16795 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16797         PR target/80897
16798         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16799         large offsets.
16801 2017-09-07  Carl Love  <cel@us.ibm.com>
16803         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16804         the sldi instruction.
16806 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16808         * sancov.c: Include tm_p.h.
16810 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16812         PR target/81979
16813         * output.h (switch_to_other_text_partition): New declaration.
16814         * varasm.c (switch_to_other_text_partition): New function.
16815         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16816         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16817         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16818         to the other text partition before emitting LCL label and switch back
16819         after emitting the word after it.
16821 2017-09-07  Richard Biener  <rguenther@suse.de>
16823         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16824         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16825         critical edges here, after loop init.
16826         (pass_data_pre): Remove PROP_no_crit_edges flags.
16827         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16828         for valueization of call args to avoid leaking VN_TOP.
16829         (visit_use): Assert we do not visit default defs.
16830         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16831         Use error_mark_node to more easily detect leaking VN_TOP.
16832         All default-defs are varying, not VN_TOP.  Mark them visited.
16833         (run_scc_vn): Make code match comment.
16835 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16837         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16838         OPTION_MASK_FLOAT128_KEYWORD.
16839         (POWERPC_MASKS): Likewise.
16840         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16841         support for the -mfloat128-type option, and make -mfloat128
16842         default on PowerPC Linux systems.  Define or undefine
16843         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16844         Define __float128 to be __ieee128 if IEEE 128-bit support is
16845         enabled, or undefine it.
16846         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16847         Delete defining __FLOAT128_TYPE__.
16848         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16849         -mfloat128-type option and make -mfloat128 default on PowerPC
16850         Linux systems.
16851         (TARGET_FLOAT128_TYPE): Likewise.
16852         (-mfloat128-type): Likewise.
16853         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16854         Delete the -mfloat128-type option and make -mfloat128 default on
16855         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16856         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16857         128-bit floating point is enabled.  Change tests from using
16858         -mfloat128-type to -mfloat128.
16859         (rs6000_mangle_type): Use the correct mangling for the __float128
16860         type even if normal long double is restricted to 64-bits.
16861         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16862         systems.
16863         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16864         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16865         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16866         -mfloat128-type.
16867         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16868         documentation for -mfloat128.
16870 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16872         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16874 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16875             Jakub Jelinek  <jakub@redhat.com>
16877         * asan.c (initialize_sanitizer_builtins): Add
16878         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16879         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16880         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16881         BT_FN_VOID_UINT64_PTR variables.
16882         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16883         (BT_FN_VOID_UINT16_UINT16): Likewise.
16884         (BT_FN_VOID_UINT32_UINT32): Likewise.
16885         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16886         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16887         (BT_FN_VOID_UINT64_PTR): Likewise.
16888         * common.opt (flag_sanitize_coverage): New variable.
16889         (fsanitize-coverage=trace-pc): Remove.
16890         (fsanitize-coverage=): Add.
16891         * flag-types.h (enum sanitize_coverage_code): New enum.
16892         * fold-const.c (fold_range_test): Disable non-short-circuit
16893         optimization if flag_sanitize_coverage.
16894         (fold_truth_andor): Likewise.
16895         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16896         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16897         (coverage_sanitizer_opts): New array.
16898         (get_closest_sanitizer_option): Add OPTS argument, handle also
16899         OPT_fsanitize_coverage_.
16900         (parse_sanitizer_options): Adjusted to also handle
16901         OPT_fsanitize_coverage_.
16902         (common_handle_option): Add OPT_fsanitize_coverage_.
16903         * sancov.c (instrument_comparison, instrument_switch): New function.
16904         (sancov_pass): Add trace-cmp support.
16905         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16906         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16907         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16908         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16909         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16910         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16911         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16912         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16913         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16914         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16916 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16918         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16919         error.  Only quit immediately if parsing is complete.
16920         (BEGIN): Initialize fatal_err and parse_done.
16921         (begin fpu, end fpu): Check number of arguments.
16922         (begin arch, end arch): Likewise.
16923         (begin cpu, end cpu): Likewise.
16924         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16925         (optalias): Likewise.
16927 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16929         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16930         * config/arm/arm-isa.h: Delete.  Move definitions to ...
16931         * arm-cpus.in: ... here.  Use new feature and fgroup values.
16932         * config/arm/arm.c (arm_option_override): Use lower case for feature
16933         bit names.
16934         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16935         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16936         * config/arm/parsecpu.awk (END): Add new command 'isa'.
16937         (isa_pfx): Delete.
16938         (print_isa_bits_for): New function.
16939         (gen_isa): New function.
16940         (gen_comm_data): Use print_isa_bits_for.
16941         (define feature): New keyword.
16942         (define fgroup): New keyword.
16943         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
16944         (arm-isa.h): Add rule to generate file.
16945         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
16946         case for feature bit names.
16948 2017-09-06  Richard Biener  <rguenther@suse.de>
16950         * tree-ssa-pre.c (NECESSARY): Remove.
16951         (create_expression_by_pieces): Do not touch pass-local flags.
16952         (insert_into_preds_of_block): Likewise.
16953         (do_pre_regular_insertion): Likewise.
16954         (eliminate_insert): Likewise.
16955         (eliminate_dom_walker::before_dom_children): Likewise.
16956         (fini_eliminate): Do not look at inserted_exprs.
16957         (mark_operand_necessary): Remove.
16958         (remove_dead_inserted_code): Replace with simple work-list
16959         algorithm based on inserted_exprs and SSA uses.
16960         (pass_pre::execute): Re-order fini_eliminate and
16961         remove_dead_inserted_code.
16963 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16965         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
16966         for VxWorks 7.  Adjust surrounding comments.
16968 2017-09-06  Richard Biener  <rguenther@suse.de>
16970         * gimple-ssa-strength-reduction.c
16971         (find_candidates_dom_walker::before_dom_children): Also allow
16972         pointer types.
16974 2017-09-06  Richard Biener  <rguenther@suse.de>
16976         PR tree-optimization/82108
16977         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
16978         for gap in the non-permutation SLP case.
16980 2017-09-06  Martin Jambor  <mjambor@suse.cz>
16982         PR tree-optimization/82078
16983         * tree-sra.c (sort_and_splice_var_accesses): Move call to
16984         add_access_to_work_queue...
16985         (build_accesses_from_assign): ...here.
16986         (propagate_all_subaccesses): Make sure racc is the group
16987         representative, if there is one.
16989 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
16991         PR middle-end/82095
16992         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
16993         NULL DECL_INITIAL.
16995 2017-09-06  Richard Biener  <rguenther@suse.de>
16997         * gimple-ssa-strength-reduction.c
16998         (find_candidates_dom_walker::before_doom_children): Use a
16999         type and not a mode check.
17001 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17003         PR target/77308
17004         * config/arm/predicates.md (arm_general_adddi_operand): Create new
17005         non-vfp predicate.
17006         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
17008 2017-09-05  Jeff Law  <law@redhat.com>
17010         PR tree-optimization/64910
17011         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
17012         cases where we have 3 or more operands.
17014 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
17016         PR middle-end/81768
17017         * omp-low.c (lower_omp_for): Recompute tree invariant if
17018         gimple_omp_for_initial/final is ADDR_EXPR.
17020         PR middle-end/81768
17021         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
17022         into gimple val before gimplification fo the COND_EXPR.
17024 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17026         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17027         REGION_COPY argument.
17028         (thread_through_all_blocks): Remove unused argument to
17029         duplicate_thread_path.
17031 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17032             Alan Hayward  <alan.hayward@arm.com>
17033             David Sherwood  <david.sherwood@arm.com>
17035         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17036         Take a scalar_mode rather than a machine_mode.
17037         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17038         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17039         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17040         (aarch64_gen_adjusted_ldpstp): Likewise.
17041         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17043 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17044             Alan Hayward  <alan.hayward@arm.com>
17045             David Sherwood  <david.sherwood@arm.com>
17047         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17048         Take a scalar_int_mode instead of a machine_mode.
17049         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17050         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17051         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17052         (aarch64_simd_attr_length_rglist): Delete.
17053         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17054         a scalar_int_mode instead of a machine_mode.
17055         (aarch64_add_offset): Likewise.
17056         (aarch64_internal_mov_immediate): Likewise
17057         (aarch64_add_constant_internal): Likewise.
17058         (aarch64_add_constant): Likewise.
17059         (aarch64_movw_imm): Likewise.
17060         (aarch64_rtx_arith_op_extract_p): Likewise.
17061         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17062         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17063         Remove assert that the mode isn't a vector.
17064         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17065         (aarch64_expand_mov_immediate): Update calls after above changes.
17066         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17067         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17068         (aarch64_move_imm): Likewise.
17069         (aarch64_can_const_movi_rtx_p): Likewise.
17070         (aarch64_strip_extend): Likewise.
17071         (aarch64_extr_rtx_p): Likewise.
17072         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17073         a CONST_INT when the mode parameter is VOIDmode.
17074         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17076 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17078         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17079         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17080         (bitwise_type_for_mode): Update accordingly.
17082 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17084         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17085         * stor-layout.c (mode_for_size_tree): Likewise.
17086         (mode_for_array): Update accordingly.
17087         (layout_decl): Likewise.
17088         (compute_record_mode): Likewise.  Only set the mode once.
17090 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17092         * target.def (get_mask_mode): Change return type to opt_mode.
17093         Expand commentary.
17094         * doc/tm.texi: Regenerate.
17095         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17096         * targhooks.c (default_get_mask_mode): Likewise.
17097         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17098         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17099         targetm.get_mask_mode.
17100         * tree.c (build_truth_vector_type): Likewise.
17102 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17104         * machmode.h (mode_for_vector): Return an opt_mode.
17105         * stor-layout.c (mode_for_vector): Likewise.
17106         (mode_for_int_vector): Update accordingly.
17107         (layout_type): Likewise.
17108         * config/i386/i386.c (emit_memmov): Likewise.
17109         (ix86_expand_set_or_movmem): Likewise.
17110         (ix86_expand_vector_init): Likewise.
17111         (ix86_get_mask_mode): Likewise.
17112         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17113         Likewise.
17114         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17115         * expmed.c (extract_bit_field_1): Likewise.
17116         * expr.c (expand_expr_real_2): Likewise.
17117         * optabs-query.c (can_vec_perm_p): Likewise.
17118         (can_vec_mask_load_store_p): Likewise.
17119         * optabs.c (expand_vec_perm): Likewise.
17120         * targhooks.c (default_get_mask_mode): Likewise.
17121         * tree-vect-stmts.c (vectorizable_store): Likewise.
17122         (vectorizable_load): Likewise.
17123         (get_vectype_for_scalar_type_and_size): Likewise.
17125 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17127         * machmode.h (mode_for_int_vector): New function.
17128         * stor-layout.c (mode_for_int_vector): Likewise.
17129         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17130         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17131         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17132         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17133         (s390_expand_vcond): Likewise.
17135 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17137         * machmode.h (opt_machine_mode): New type.
17138         (opt_mode<T>): Allow construction from anything that can be
17139         converted to a T.
17140         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17141         (mode_for_size): Return an opt_machine_mode.
17142         * stor-layout.c (mode_for_size): Likewise.
17143         (mode_for_size_tree): Update call accordingly.
17144         (bitwise_mode_for_mode): Likewise.
17145         (make_fract_type): Likewise.
17146         (make_accum_type): Likewise.
17147         * caller-save.c (replace_reg_with_saved_mem): Update call
17148         accordingly.
17149         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17150         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17151         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17152         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17153         * expmed.c (extract_bit_field_1): Likewise.
17154         * reload.c (get_secondary_mem): Likewise.
17155         * varasm.c (assemble_integer): Likewise.
17156         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17157         early-out.
17159 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17161         * machmode.h (decimal_float_mode_for_size): New function.
17162         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17163         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17164         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17165         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17166         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17167         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17169 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17171         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17172         (get_builtin_sync_mode): Likewise.
17173         (expand_ifn_atomic_compare_exchange): Likewise.
17174         (expand_builtin_atomic_clear): Likewise.
17175         (expand_builtin_atomic_test_and_set): Likewise.
17176         (fold_builtin_atomic_always_lock_free): Likewise.
17177         * calls.c (compute_argument_addresses): Likewise.
17178         (emit_library_call_value_1): Likewise.
17179         (store_one_arg): Likewise.
17180         * combine.c (combine_instructions): Likewise.
17181         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17182         * config/arm/arm.c (arm_function_value): Likewise.
17183         (aapcs_allocate_return_reg): Likewise.
17184         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17185         * config/i386/i386.c (construct_container): Likewise.
17186         (ix86_gimplify_va_arg): Likewise.
17187         (ix86_expand_sse_cmp): Likewise.
17188         (emit_memmov): Likewise.
17189         (emit_memset): Likewise.
17190         (expand_small_movmem_or_setmem): Likewise.
17191         (ix86_expand_pextr): Likewise.
17192         (ix86_expand_pinsr): Likewise.
17193         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17194         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17195         Likewise.
17196         * config/mips/mips.c (mips_function_value_1) Likewise.
17197         (mips_block_move_straight): Likewise.
17198         (mips_expand_ins_as_unaligned_store): Likewise.
17199         * config/powerpcspe/powerpcspe.c
17200         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17201         (rs6000_darwin64_record_arg_flush): Likewise.
17202         * config/rs6000/rs6000.c
17203         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17204         (rs6000_darwin64_record_arg_flush): Likewise.
17205         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17206         (sparc_function_value_1): Likewise.
17207         * config/spu/spu.c (adjust_operand): Likewise.
17208         (spu_emit_branch_or_set): Likewise.
17209         (arith_immediate_p): Likewise.
17210         * emit-rtl.c (gen_lowpart_common): Likewise.
17211         * expr.c (expand_expr_real_1): Likewise.
17212         * function.c (assign_parm_setup_block): Likewise.
17213         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17214         * reload1.c (alter_reg): Likewise.
17215         * stor-layout.c (mode_for_vector): Likewise.
17216         (layout_type): Likewise.
17218 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17220         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17221         (spu_convert_move): Likewise.
17222         * lower-subreg.c (resolve_simple_move): Likewise.
17224 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17226         PR target/81833
17227         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17228         define_insn to a define_expand.
17229         (altivec_vsum2sws_direct): New define_insn.
17230         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17232 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17234         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17235         max_insns_skipped.
17237 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17239         PR target/59501
17240         PR target/81624
17241         PR target/81769
17242         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17243         realign stack if stack alignment needed is less than incoming
17244         stack boundary.
17246 2017-09-05  Marek Polacek  <polacek@redhat.com>
17248         PR sanitizer/82072
17249         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17250         check earlier.
17252 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17254         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17256 2017-09-05  Richard Biener  <rguenther@suse.de>
17258         PR tree-optimization/82084
17259         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17261 2017-09-05  Richard Biener  <rguenther@suse.de>
17263         PR tree-optimization/82102
17264         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17266 2017-09-05  Martin Liska  <mliska@suse.cz>
17268         PR tree-optimization/82032
17269         * tree-cfg.c (generate_range_test): New function.
17270         * tree-cfg.h (generate_range_test): Declared here.
17271         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17272         (cleanup_control_expr_graph): Use it.
17273         * tree-switch-conversion.c (try_switch_expansion): Remove
17274         assert.
17275         (emit_case_nodes): Use generate_range_test.
17277 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17279         PR target/82098
17280         * config/i386/i386.md (*<btsc><mode>_mask): Add
17281         TARGET_USE_BT to insn constraint.
17282         (*btr<mode>_mask): Ditto.
17284 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17286         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17287         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17289 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17291         PR target/77308
17292         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17293         TARGET_NEON and TARGET_IWMMXT.
17294         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17295         TARGET_NEON and TARGET_IWMMXT.
17296         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17298 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17300         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17301         (ix86_rewrite_tls_address): Ditto.
17302         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17303         (ix86_rewrite_tls_address_1): Ditto.
17304         (ix86_rewrite_tls_address): Ditto.
17305         * config/i386/predicates.md (tls_address_pattern): New predicate.
17306         * config/i386/i386.md (TLS address splitter): New splitter.
17308 2017-09-04  Richard Biener  <rguenther@suse.de>
17310         PR tree-optimization/82084
17311         * fold-const.h (can_native_encode_string_p): Declare.
17312         * fold-const.c (can_native_encode_string_p): Factor out from ...
17313         (native_encode_string): ... here.
17314         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17315         vectorizing stores from constants we later cannot handle.
17317 2017-09-04  Marek Polacek  <polacek@redhat.com>
17319         PR c/81783
17320         * doc/invoke.texi: Update -Wtautological-compare documentation.
17322 2017-09-04  Jeff Law  <law@redhat.com>
17324         PR tree-optimization/64910
17325         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17326         swap the first and last operand if the last is a constant.
17328 2017-09-04  Marek Polacek  <polacek@redhat.com>
17330         PR sanitizer/82072
17331         * convert.c (do_narrow): When sanitizing signed integer overflows,
17332         bail out for signed types.
17333         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17335 2017-09-04  Richard Biener  <rguenther@suse.de>
17337         PR tree-optimization/82060
17338         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17339         Move devirtualization after stmt folding and before EH/AB/noreturn
17340         cleanup to get the stmt refs canonicalized.  Use a bool instead
17341         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17342         NOPs generated by folding for removal.
17344 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17345             Alan Hayward  <alan.hayward@arm.com>
17346             David Sherwood  <david.sherwood@arm.com>
17348         * coretypes.h (pad_direction): New enum.
17349         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17350         (FUNCTION_ARG_PADDING): Likewise.
17351         * target.def (function_arg_padding): New hook.
17352         * targhooks.h (default_function_arg_padding): Declare.
17353         * targhooks.c (default_function_arg_padding): New function.
17354         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17355         (TARGET_FUNCTION_ARG_PADDING): ...this.
17356         * doc/tm.texi: Regenerate.
17357         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17358         instead of direction.
17359         (compute_argument_addresses): Likewise.
17360         (load_register_parameters): Likewise.
17361         (emit_library_call_value_1): Likewise.
17362         (store_one_arg): Use targetm.calls.function_arg_padding instead
17363         of FUNCTION_ARG_PADDING.
17364         (must_pass_in_stack_var_size_or_pad): Likewise.
17365         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17366         (emit_group_store): Likewise.
17367         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17368         instead of FUNCTION_ARG_PADDING.
17369         (emit_push_insn): Likewise, and propagate enum change throughout
17370         function.
17371         * function.h (direction): Delete.
17372         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17373         of direction.
17374         * function.c (assign_parm_find_stack_rtl): Likewise.
17375         (assign_parm_setup_block_p): Likewise.
17376         (assign_parm_setup_block): Likewise.
17377         (gimplify_parameters): Likewise.
17378         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17379         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17380         function.
17381         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17382         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17383         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17384         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17385         (aarch64_function_arg_padding): ...this new function.
17386         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17387         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17388         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17389         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17390         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17391         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17392         (arm_pad_arg_upward): Replace with...
17393         (arm_function_arg_padding): ...this new function.
17394         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17395         of direction.
17396         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17397         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17398         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17399         (ia64_hpux_function_arg_padding): Replace with...
17400         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17401         instead of direction.  Check for TARGET_HPUX.
17402         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17403         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17404         (iq2000_function_arg_padding): New function.
17405         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17406         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17407         (mips_function_arg_padding): ...this new function.
17408         (mips_pad_reg_upward): Update accordingly.
17409         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17410         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17411         targetm.calls.function_arg_padding.
17412         (FUNCTION_ARG_PADDING): Delete.
17413         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17414         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17415         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17416         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17417         (nios2_block_reg_padding): Return pad_direction instead of direction.
17418         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17419         instead of direction.
17420         (nios2_function_arg_padding): Likewise.  Make static.
17421         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17422         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17423         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17424         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17425         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17426         (pa_function_arg_padding): Make static.  Return pad_direction instead
17427         of direction.
17428         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17429         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17430         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17431         instead of direction.  Use targetm.calls.function_arg_padding.
17432         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17433         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17434         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17435         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17436         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17437         Redefine.
17438         (function_arg_padding): Rename to...
17439         (rs6000_function_arg_padding): ...this.  Make static.  Return
17440         pad_direction instead of direction.
17441         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17442         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17443         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17444         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17445         instead of direction.  Use targetm.calls.function_arg_padding.
17446         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17447         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17448         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17449         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17450         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17451         (function_arg_padding): Rename to...
17452         (rs6000_function_arg_padding): ...this.  Make static.  Return
17453         pad_direction instead of direction.
17454         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17455         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17456         * config/s390/s390.c (s390_function_arg_padding): New function.
17457         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17458         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17459         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17460         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17461         (function_arg_padding): Rename to...
17462         (sparc_function_arg_padding): ...this.  Make static.  Return
17463         pad_direction instead of direction.
17464         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17465         * config/spu/spu.c (spu_function_arg_padding): New function.
17466         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17467         * system.h (FUNCTION_ARG_PADDING): Poison.
17469 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17470             Alan Hayward  <alan.hayward@arm.com>
17471             David Sherwood  <david.sherwood@arm.com>
17473         * target.def (modes_tieable_p): New hook.
17474         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17475         (TARGET_MODES_TIEABLE_P): ...this.
17476         * doc/tm.texi.in: Regenerate.
17477         * hooks.h (hook_bool_mode_mode_true): Declare.
17478         * hooks.c (hook_bool_mode_mode_true): New function.
17479         * combine.c (subst): Use targetm.modes_tieable_p instead of
17480         MODES_TIEABLE_P.
17481         * dse.c (find_shift_sequence): Likewise.
17482         * expmed.c (extract_low_bits): Likewise.
17483         * lower-subreg.c: Include target.h.
17484         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17485         MODES_TIEABLE_P.
17486         * rtlanal.c (rtx_cost): Likewise.
17487         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17488         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17489         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17490         (TARGET_MODES_TIEABLE_P): Redefine.
17491         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17492         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17493         (TARGET_MODES_TIEABLE_P): Redefine.
17494         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17495         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17496         (arc_modes_tieable_p): New function.
17497         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17498         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17499         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17500         (arm_modes_tieable_p): Make static.
17501         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17502         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17503         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17504         (TARGET_MODES_TIEABLE_P): Redefine.
17505         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17506         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17507         (TARGET_MODES_TIEABLE_P): Redefine.
17508         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17509         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17510         (cr16_modes_tieable_p): New function.
17511         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17512         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17513         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17514         (TRULY_NOOP_TRUNCATION): Update comment.
17515         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17516         (TRULY_NOOP_TRUNCATION): Update comment.
17517         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17518         (frv_modes_tieable_p): New function.
17519         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17520         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17521         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17522         (TARGET_MODES_TIEABLE_P): Redefine.
17523         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17524         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17525         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17526         (TARGET_MODES_TIEABLE_P): Redefine.
17527         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17528         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17529         (ia64_modes_tieable_p): New function.
17530         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17531         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17532         (iq2000_modes_tieable_p): New function.
17533         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17534         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17535         (lm32_modes_tieable_p): New function.
17536         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17537         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17538         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17539         (TARGET_MODES_TIEABLE_P): Redefine.
17540         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17541         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17542         (m32r_modes_tieable_p): New function.
17543         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17544         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17545         (m68k_modes_tieable_p): New function.
17546         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17547         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17548         (mcore_modes_tieable_p): New function.
17549         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17550         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17551         function.
17552         (TARGET_MODES_TIEABLE_P): Redefine.
17553         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17554         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17555         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17556         (TARGET_MODES_TIEABLE_P): Redefine.
17557         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17558         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17559         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17560         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17561         (mn10300_modes_tieable_p): ...this and make static.
17562         (TARGET_MODES_TIEABLE_P): Redefine.
17563         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17564         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17565         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17566         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17567         (msp430_modes_tieable_p): Make static.
17568         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17569         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17570         (TARGET_MODES_TIEABLE_P): Redefine.
17571         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17572         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17573         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17574         (TARGET_MODES_TIEABLE_P): Redefine.
17575         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17576         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17577         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17578         (TARGET_MODES_TIEABLE_P): Redefine.
17579         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17580         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17581         (pdp11_modes_tieable_p): New function.
17582         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17583         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17584         (rs6000_modes_tieable_p): New function.
17585         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17586         * config/powerpcspe/powerpcspe.md: Update comment.
17587         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17588         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17589         (TARGET_MODES_TIEABLE_P): Redefine.
17590         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17591         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17592         (rl78_modes_tieable_p): New function.
17593         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17594         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17595         (rs6000_modes_tieable_p): New function.
17596         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17597         * config/rs6000/rs6000.md: Update comment.
17598         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17599         * config/rx/rx.c (rx_modes_tieable_p): New function.
17600         (TARGET_MODES_TIEABLE_P): Redefine.
17601         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17602         * config/s390/s390.c (s390_modes_tieable_p): New function.
17603         (TARGET_MODES_TIEABLE_P): Redefine.
17604         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17605         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17606         (sh_modes_tieable_p): New function.
17607         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17608         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17609         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17610         (sparc_modes_tieable_p): Make static.
17611         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17612         * config/spu/spu.c (spu_modes_tieable_p): New function.
17613         (TARGET_MODES_TIEABLE_P): Redefine.
17614         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17615         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17616         (TARGET_MODES_TIEABLE_P): Redefine.
17617         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17618         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17619         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17620         * config/v850/v850.c (v850_modes_tieable_p): New function.
17621         (TARGET_MODES_TIEABLE_P): Redefine.
17622         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17623         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17624         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17625         (visium_modes_tieable_p): New function.
17626         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17627         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17628         (xtensa_modes_tieable_p): New function.
17629         * system.h (MODES_TIEABLE_P): Poison.
17631 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17632             Alan Hayward  <alan.hayward@arm.com>
17633             David Sherwood  <david.sherwood@arm.com>
17635         * target.def (hard_regno_mode_ok): New hook.
17636         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17637         (TARGET_HARD_REGNO_MODE_OK): ...this.
17638         * doc/tm.texi.in: Regenerate.
17639         * hooks.h (hook_bool_uint_mode_true): Declare.
17640         * hooks.c (hook_bool_uint_mode_true): New function.
17641         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17642         HARD_REGNO_MODE_OK.
17643         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17644         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17645         instead of HARD_REGNO_MODE_OK.
17646         * caller-save.c: Include target.h.
17647         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17648         HARD_REGNO_MODE_OK.
17649         * combine.c (can_combine_p): Likewise.
17650         (combinable_i3pat): Likewise.
17651         (can_change_dest_mode): Likewise.
17652         * expr.c (init_expr_target): Likewise.
17653         (convert_move): Likewise.
17654         (convert_modes): Likewise.
17655         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17656         (setup_prohibited_mode_move_regs): Likewise.
17657         * ira.h (target_ira): Likewise.
17658         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17659         * lra-constraints.c (process_alt_operands): Likewise.
17660         (split_reg): Likewise.
17661         * recog.c (peep2_find_free_register): Likewise.
17662         * ree.c (combine_reaching_defs): Likewise.
17663         * regcprop.c (maybe_mode_change): Likewise.
17664         * reginfo.c (init_reg_sets_1): Likewise.
17665         (choose_hard_reg_mode): Likewise.
17666         (simplifiable_subregs): Likewise.
17667         * regrename.c (check_new_reg_p): Likewise.
17668         * reload.c (find_valid_class): Likewise.
17669         (find_valid_class_1): Likewise.
17670         (reload_inner_reg_of_subreg): Likewise.
17671         (push_reload): Likewise.
17672         (combine_reloads): Likewise.
17673         (find_dummy_reload): Likewise.
17674         (find_reloads): Likewise.
17675         * reload1.c (find_reg): Likewise.
17676         (set_reload_reg): Likewise.
17677         (allocate_reload_reg): Likewise.
17678         (choose_reload_regs): Likewise.
17679         (reload_adjust_reg_for_temp): Likewise.
17680         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17681         (simplify_subreg_regno): Likewise.
17682         * sel-sched.c (init_regs_for_mode): Likewise.
17683         * varasm.c (make_decl_rtl): Likewise.
17684         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17685         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17686         HARD_REGNO_MODE_OK.
17687         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17688         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17689         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17690         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17691         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17692         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17693         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17694         (arc_mode_class): Delete.
17695         (HARD_REGNO_MODE_OK): Delete.
17696         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17697         (arc_hard_regno_mode_ok): Rename old array to...
17698         (arc_hard_regno_mode_ok_modes): ...this.
17699         (arc_conditional_register_usage): Update accordingly.
17700         (arc_mode_class): Make static.
17701         (arc_hard_regno_mode_ok): New function.
17702         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17703         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17704         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17705         (arm_hard_regno_mode_ok): Make static.
17706         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17707         HARD_REGNO_MODE_OK.
17708         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17709         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17710         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17711         return a bool.
17712         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17713         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17714         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17715         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17716         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17717         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17718         * config/bfin/predicates.md (valid_reg_operand): Use
17719         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17720         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17721         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17722         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17723         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17724         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17725         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17726         (cr16_hard_regno_mode_ok): Make static and return a bool.
17727         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17728         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17729         (cris_hard_regno_mode_ok): New function.
17730         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17731         (epiphany_mode_class): Delete.
17732         (HARD_REGNO_MODE_OK): Delete.
17733         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17734         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17735         (hard_regno_mode_ok): Rename to...
17736         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17737         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17738         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17739         HARD_REGNO_MODE_OK.
17740         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17741         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17742         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17743         (frv_hard_regno_mode_ok): Make static and return a bool.
17744         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17745         HARD_REGNO_MODE_OK.
17746         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17747         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17748         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17749         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17750         and return a bool.
17751         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17752         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17753         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17754         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17755         return a bool.
17756         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17757         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17758         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17759         (ia64_hard_regno_mode_ok): New function.
17760         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17761         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17762         (iq2000_hard_regno_mode_ok): New function.
17763         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17764         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17765         (lm32_hard_regno_mode_ok): New function.
17766         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17767         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17768         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17769         instead of HARD_REGNO_MODE_OK.
17770         (m32c_hard_regno_ok): Rename to...
17771         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17772         (m32c_cannot_change_mode_class): Update accordingly.
17773         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17774         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17775         (m32r_mode_class): Delete.
17776         (HARD_REGNO_MODE_OK): Delete.
17777         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17778         (m32r_hard_regno_mode_ok): Rename to...
17779         (m32r_hard_regno_modes): ...this.
17780         (m32r_mode_class): Make static.
17781         (m32r_hard_regno_mode_ok): New function.
17782         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17783         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17784         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17785         (m68k_hard_regno_mode_ok): Make static.
17786         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17787         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17788         (mcore_hard_regno_mode_ok): New function.
17789         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17790         (HARD_REGNO_MODE_OK): Delete.
17791         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17792         Rename to...
17793         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17794         (microblaze_hard_regno_mode_ok): New function.
17795         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17796         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17797         (mips_hard_regno_mode_ok): Delete.
17798         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17799         (mips_hard_regno_mode_ok_p): ...this and make static.
17800         (mips_hard_regno_mode_ok_p): Rename to...
17801         (mips_hard_regno_mode_ok_uncached): ...this.
17802         (mips_hard_regno_mode_ok): New function.
17803         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17804         of HARD_REGNO_MODE_OK.
17805         (mips_option_override): Update after above name changes.
17806         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17807         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17808         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17809         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17810         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17811         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17812         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17813         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17814         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17815         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17816         (msp430_hard_regno_mode_ok): Make static and return a bool.
17817         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17818         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17819         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17820         and return a bool.
17821         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17822         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17823         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17824         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17825         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17826         (PA_HARD_REGNO_MODE_OK): ...this
17827         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17828         (PA_HARD_REGNO_MODE_OK): ...this.
17829         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17830         (pa_hard_regno_mode_ok): New function.
17831         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17832         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17833         (pdp11_hard_regno_mode_ok): New function.
17834         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17835         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17836         Delete.
17837         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17838         Make static.
17839         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17840         (rs6000_hard_regno_mode_ok): Rename to...
17841         (rs6000_hard_regno_mode_ok_uncached): ...this.
17842         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17843         (rs6000_hard_regno_mode_ok): New function.
17844         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17845         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17846         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17847         (riscv_hard_regno_mode_ok): ...this and make static.
17848         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17849         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17850         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17851         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17852         (rl78_hard_regno_mode_ok): Make static and return bool.
17853         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17854         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17855         Delete.
17856         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17857         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17858         (rs6000_hard_regno_mode_ok): Rename to...
17859         (rs6000_hard_regno_mode_ok_uncached): ...this.
17860         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17861         (rs6000_hard_regno_mode_ok): New function.
17862         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17863         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17864         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17865         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17866         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17867         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17868         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17869         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17870         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17871         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17872         (sh_hard_regno_mode_ok): Make static.
17873         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17874         instead of HARD_REGNO_MODE_OK.
17875         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17876         (sparc_mode_class): Delete.
17877         (HARD_REGNO_MODE_OK): Delete.
17878         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17879         (hard_regno_mode_classes): Make static.
17880         (sparc_mode_class): Likewise.
17881         (sparc_hard_regno_mode_ok): New function.
17882         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17883         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17884         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17885         function.
17886         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17887         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17888         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17889         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17890         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17891         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17892         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17893         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17894         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17895         (visium_hard_regno_mode_ok): New function.
17896         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17897         instead of HARD_REGNO_MODE_OK.
17898         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17899         (HARD_REGNO_MODE_OK): Delete.
17900         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17901         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17902         (xtensa_option_override): Update accordingly.
17903         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17904         (xtensa_hard_regno_mode_ok): New function.
17905         * system.h (HARD_REGNO_MODE_OK): Poison.
17907 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17908             Alan Hayward  <alan.hayward@arm.com>
17909             David Sherwood  <david.sherwood@arm.com>
17911         * target.def (hard_regno_call_part_clobbered): New hook.
17912         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17913         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17914         * doc/tm.texi: Regenerate.
17915         * hooks.h (hook_bool_uint_mode_false): Declare.
17916         * hooks.c (hook_bool_uint_mode_false): New function.
17917         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17918         * cselib.c (cselib_process_insn): Use
17919         targetm.hard_regno_call_part_clobbered instead of
17920         HARD_REGNO_CALL_PART_CLOBBERED.
17921         * ira-conflicts.c (ira_build_conflicts): Likewise.
17922         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17923         * lra-constraints.c (need_for_call_save_p): Likewise.
17924         * lra-lives.c: Include target.h.
17925         (check_pseudos_live_through_calls): Use
17926         targetm.hard_regno_call_part_clobbered instead of
17927         HARD_REGNO_CALL_PART_CLOBBERED.
17928         * regcprop.c: Include target.h.
17929         (copyprop_hardreg_forward_1): Use
17930         targetm.hard_regno_call_part_clobbered instead of
17931         HARD_REGNO_CALL_PART_CLOBBERED.
17932         * reginfo.c (choose_hard_reg_mode): Likewise.
17933         * regrename.c (check_new_reg_p): Likewise.
17934         * reload.c (find_equiv_reg): Likewise.
17935         * reload1.c (emit_reload_insns): Likewise.
17936         * sched-deps.c (deps_analyze_insn): Likewise.
17937         * sel-sched.c (init_regs_for_mode): Likewise.
17938         (mark_unavailable_hard_regs): Likewise.
17939         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17940         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17941         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
17942         New function.
17943         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17944         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17945         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
17946         Delete.
17947         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
17948         and return a bool.
17949         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17950         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17951         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
17952         function.
17953         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17954         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17955         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
17956         function.
17957         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17958         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
17959         Delete.
17960         * config/powerpcspe/powerpcspe.c
17961         (rs6000_hard_regno_call_part_clobbered): New function.
17962         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17963         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17964         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
17965         New function.
17966         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17967         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17968         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
17969         function.
17970         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17971         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17972         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
17974 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17975             Alan Hayward  <alan.hayward@arm.com>
17976             David Sherwood  <david.sherwood@arm.com>
17978         * rtl.h (subreg_memory_offset): Declare.
17979         * emit-rtl.c (subreg_memory_offset): New function.
17980         * expmed.c (store_bit_field_1): Use it.
17981         * expr.c (undefined_operand_subword_p): Likewise.
17982         * simplify-rtx.c (simplify_subreg): Likewise.
17984 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
17986         PR rtl-optimization/57448
17987         PR target/67458
17988         PR target/81316
17989         * optabs.c (expand_atomic_load): Place compiler memory barriers if
17990         using atomic_load pattern.
17991         (expand_atomic_store): Likewise.
17993 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
17995         PR sanitizer/81981
17996         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
17997         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
17998         handling.  Use replace_call_with_value with NULL instead of
17999         gsi_replace, unlink_stmt_vdef and release_defs.
18001         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
18002         instead of tab.
18004         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
18006 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18008         PR bootstrap/82045
18009         * rtl.h (emit_library_call_value_1): Declare.
18010         (emit_library_call): Replace declaration with a series of overloads.
18011         Remove the parameter count argument.
18012         (emit_library_call_value): Likewise.
18013         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
18014         with an "rtx_mode_t *".
18015         (emit_library_call_value): Delete.
18016         (emit_library_call): Likewise.
18017         * asan.c (asan_emit_stack_protection): Update calls accordingly.
18018         (asan_emit_allocas_unpoison): Likewise.
18019         * builtins.c (expand_builtin_powi): Likewise.
18020         (expand_asan_emit_allocas_unpoison): Likewise.
18021         * cfgexpand.c (expand_main_function): Likewise.
18022         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
18023         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18024         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18025         * config/arm/arm.c (arm_trampoline_init): Likewise.
18026         (arm_call_tls_get_addr): Likewise.
18027         (arm_expand_divmod_libfunc): Likewise.
18028         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18029         (smulsi3_highpart): Likewise.
18030         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18031         (c6x_expand_compare): Likewise.
18032         (c6x_expand_movmem): Likewise.
18033         * config/frv/frv.c (frv_trampoline_init): Likewise.
18034         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18035         (ix86_expand_divmod_libfunc): Likewise.
18036         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18037         (ia64_expand_compare): Likewise.
18038         (ia64_profile_hook): Likewise.
18039         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18040         (nonlocal_goto): Likewise.
18041         (restore_stack_nonlocal): Likewise.
18042         * config/m32r/m32r.c (block_move_call): Likewise.
18043         (m32r_trampoline_init): Likewise.
18044         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18045         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18046         (m68k_call_m68k_read_tp): Likewise.
18047         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18048         (microblaze_expand_divide): Likewise.
18049         * config/mips/mips.h (mips_args): Likewise.
18050         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18051         (MIPS_ICACHE_SYNC): Likewise.
18052         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18053         (nios2_trampoline_init): Likewise.
18054         * config/pa/pa.c (hppa_tls_call): Likewise.
18055         (pa_trampoline_init): Likewise.
18056         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18057         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18058         (expand_strn_compare): Likewise.
18059         (rs6000_generate_compare): Likewise.
18060         (rs6000_expand_float128_convert): Likewise.
18061         (output_profile_hook): Likewise.
18062         (rs6000_trampoline_init): Likewise.
18063         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18064         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18065         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18066         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18067         (rs6000_generate_compare): Likewise.
18068         (rs6000_expand_float128_convert): Likewise.
18069         (output_profile_hook): Likewise.
18070         (rs6000_trampoline_init): Likewise.
18071         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18072         * config/sh/sh.c (sh_trampoline_init): Likewise.
18073         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18074         (sparc_emit_float_lib_cmp): Likewise.
18075         (sparc32_initialize_trampoline): Likewise.
18076         (sparc64_initialize_trampoline): Likewise.
18077         (sparc_profile_hook): Likewise.
18078         * config/spu/spu.c (ea_load_store): Likewise.
18079         * config/spu/spu.md (floatunssidf2): Likewise.
18080         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18081         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18082         * config/visium/visium.c (expand_block_move_4): Likewise.
18083         (expand_block_move_2): Likewise.
18084         (expand_block_move_1): Likewise.
18085         (expand_block_set_4): Likewise.
18086         (expand_block_set_2): Likewise.
18087         (expand_block_set_1): Likewise.
18088         (visium_trampoline_init): Likewise.
18089         (visium_profile_hook): Likewise.
18090         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18091         (xtensa_setup_frame_addresses): Likewise.
18092         (xtensa_trampoline_init): Likewise.
18093         * except.c (sjlj_emit_function_enter): Likewise.
18094         (sjlj_emit_function_exit): Likewise.
18095         * explow.c (allocate_dynamic_stack_space): Likewise.
18096         (probe_stack_range): Likewise.
18097         * expr.c (convert_mode_scalar): Likewise.
18098         * optabs.c (expand_binop): Likewise.
18099         (expand_twoval_binop_libfunc): Likewise.
18100         (expand_unop): Likewise.
18101         (prepare_cmp_insn): Likewise.
18102         (prepare_float_lib_cmp): Likewise.
18103         (expand_float): Likewise.
18104         (expand_fix): Likewise.
18105         (expand_fixed_convert): Likewise.
18106         (maybe_emit_sync_lock_test_and_set): Likewise.
18107         (expand_atomic_compare_and_swap): Likewise.
18108         (expand_mem_thread_fence): Likewise.
18109         (expand_atomic_fetch_op): Likewise.
18111 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18113         * doc/generic.texi (OpenACC): Adjust URL.
18114         * doc/invoke.texi (C Dialect Options): Ditto.
18116 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18118         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18119         predicate for operand 1.  Add (m,<S>) constraint.
18120         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18121         Prevent memory operand 1 with register operand 2.
18123 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18125         PR rtl-optimization/82024
18126         * combine.c (try_combine): If the combination result is a PARALLEL,
18127         and we only need to retain the SET in there that would be placed
18128         at I2, check that we can place that at I3 instead, before doing so.
18130 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18132         PR target/81766
18133         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18134         instead of void.
18135         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18136         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18137         and label.
18139 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18140             Jeff Law  <law@redhat.com>
18142         * varasm.c (bss_initializer_p): Do not put constants into .bss
18143         (categorize_decl_for_section): Handle bss_initializer_p returning
18144         false when DECL_INITIAL is NULL.
18146 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18148         PR target/82012
18149         * config/s390/s390.c (s390_can_inline_p): New function.
18151 2017-09-01  Jeff Law  <law@redhat.com>
18153         PR tree-optimization/82052
18154         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18155         Always initialize the returned slot after a hash table miss
18156         when INSERT is true.
18158 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18160         * config/s390/s390.md (mem_signal_fence): Remove.
18161         * doc/md.texi (mem_signal_fence): Remove.
18162         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18163         Update comments.
18164         * target-insns.def (mem_signal_fence): Remove.
18166 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18168         PR sanitizer/81902
18169         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18171         PR sanitizer/81923
18172         * asan.c (create_odr_indicator): Strip name encoding from assembler
18173         name before appending it after __odr_asan_.
18175 2017-09-01  Martin Liska  <mliska@suse.cz>
18177         PR tree-optimization/82059
18178         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18179         frequency only when an edge is redirected.
18181 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18183         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18184         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18185         (arc_conditional_register_usage): Remove ARC600 lp_count
18186         exception.
18187         (arc_file_start): Emit Tag_ARC_CPU_variation.
18188         (arc_can_use_doloop_p): New conditions to use ZOLs.
18189         (hwloop_fail): New function.
18190         (hwloop_optimize): Likewise.
18191         (hwloop_pattern_reg): Likewise.
18192         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18193         (arc_reorg_loops): New function, calls reorg_loops.
18194         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18195         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18196         hwloop_optimize.
18197         (arc_loop_hazard): Remove function, functionality moved into
18198         hwloop_optimize.
18199         (arc_hazard): Remove arc_loop_hazard call.
18200         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18201         into hwloop_optimize.
18202         (arc_label_align): Remove ZOL handling.
18203         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18204         * config/arc/arc.md (doloop_begin): Remove pattern.
18205         (doloop_begin_i): Likewise.
18206         (doloop_end_i): Likewise.
18207         (doloop_fallback): Likewise.
18208         (doloop_fallback_m): Likewise.
18209         (doloop_end): Reimplement expand.
18210         (arc_lp): New pattern for LP instruction.
18211         (loop_end): New pattern.
18212         (loop_fail): Likewise.
18213         (decrement_and_branch_until_zero): Likewise.
18214         * config/arc/arc.opt (mlpc-width): New option.
18215         * doc/invoke.texi (mlpc-width): Document option.
18217 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18219         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18220         (arc_ccfsm_advance): Fix checking for delay slots.
18221         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18223 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18225         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18226         immediates.
18227         (movhi_insn): Update store instruction constraint which are saving
18228         6-bit short immediates.
18229         (movsi_insn): Consider also short scaled load operations.
18230         (zero_extendhisi2_i): Use Usd constraint instead of T.
18231         (extendhisi2_i): Add q constraint.
18232         (arc_clzsi2): Add type and length attributes.
18233         (arc_ctzsi2): Likewise.
18234         * config/arc/constraints.md (Usc): Update constraint, the
18235         assembler can parse two relocations for a single instruction.
18237 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18239         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18240         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18242 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18244         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18245         match as powerpc-wrs-vxworks*.
18247 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18249         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18250         register constraint for by-element operand.
18251         (aarch64_mls_elt_merge<mode>): Likewise.
18253 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18255         * config/arc/arc.c (arc_can_follow_jump): Check for short
18256         branches.
18258 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18260         * config.gcc: Use g.opt for arc.
18261         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18262         functionality moved to ...
18263         (legitimate_scaled_address_p): New function, ...here.
18264         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18265         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18266         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18267         condition.
18268         (arc_override_options): Handle G option.
18269         (arc_output_pic_addr_const): Correct function definition.
18270         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18271         (arc_decl_anon_ns_mem_p): Delete.
18272         (arc_in_small_data_p): Overhaul this function to take into
18273         consideration the value given via G option.
18274         (arc_rewrite_small_data_1): Renamed and corrected old
18275         arc_rewrite_small_data function.
18276         (arc_rewrite_small_data): New function.
18277         (small_data_pattern): Don't use pic_offset_table_rtx.
18278         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18279         * config/arc/simdext.md (movmisalignv2hi): Use
18280         prepare_move_operands function.
18281         (mov*): Likewise.
18282         (movmisalign*): Likewise.
18283         * doc/invoke.texi (ARC options): Document -G option.
18285 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18287         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18288         prototype.
18289         * config/arc/arc.c (arc_print_operand): Output scalled address for
18290         sdata whenever is possible.
18291         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18292         load/stores are available.
18293         (compact_sda_memory_operand): Check for the alignment required by
18294         code density instructions.
18295         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18296         constraint.
18297         * config/arc/constraints.md (Usd): Update constraint.
18298         (Us0): New constraint.
18299         (Usc): Update constraint.
18301 2017-08-31  Richard Biener  <rguenther@suse.de>
18303         PR middle-end/82054
18304         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18305         function only once.
18307 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18309         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18310         Resize type_signature.
18312 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18313             Alan Hayward  <alan.hayward@arm.com>
18314             David Sherwood  <david.sherwood@arm.com>
18316         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18317         subregs whose inner modes can be stored in GPRs.
18318         (aarch64_classify_index): Likewise.
18320 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18321             Alan Hayward  <alan.hayward@arm.com>
18322             David Sherwood  <david.sherwood@arm.com>
18324         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18325         (V_INT_EQUIV): ...this.
18326         (v_cmp_result): Rename to...
18327         (v_int_equiv): ...this.
18328         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18329         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18330         (copysign<mode>3): Likewise.
18331         (aarch64_simd_bsl<mode>_internal): Likewise.
18332         (aarch64_simd_bsl<mode>): Likewise.
18333         (vec_cmp<mode><mode>): Likewise.
18334         (vcond<mode><mode>): Likewise.
18335         (vcond<v_cmp_mixed><mode>): Likewise.
18336         (vcondu<mode><v_cmp_mixed>): Likewise.
18337         (aarch64_cm<optab><mode>): Likewise.
18338         (aarch64_cmtst<mode>): Likewise.
18339         (aarch64_fac<optab><mode>): Likewise.
18340         (vec_perm_const<mode>): Likewise.
18341         (vcond_mask_<mode><v_cmp_result>): Rename to...
18342         (vcond_mask_<mode><v_int_equiv>): ...this.
18343         (vec_cmp<mode><v_cmp_result>): Rename to...
18344         (vec_cmp<mode><v_int_equiv>): ...this.
18346 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18347             Alan Hayward  <alan.hayward@arm.com>
18348             David Sherwood  <david.sherwood@arm.com>
18350         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18351         vector modes.
18352         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18353         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18354         (UNSPEC_LD4_DREG): New unspecs.
18355         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18356         (aarch64_ld2<mode>_dreg_be): Replace with...
18357         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18358         unspec.
18359         (aarch64_ld3<mode>_dreg_le)
18360         (aarch64_ld3<mode>_dreg_be): Replace with...
18361         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18362         unspec.
18363         (aarch64_ld4<mode>_dreg_le)
18364         (aarch64_ld4<mode>_dreg_be): Replace with...
18365         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18366         unspec.
18368 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18370         PR tree-optimization/81987
18371         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18372         insert an initializer in a location not dominated by the stride
18373         definition.
18375 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18377         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18378         on the entire header of the finally block in the fallthru case.
18380 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18382         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18384 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18386         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18387         rs6000_emit_move_from_cr and call renamed function.
18388         (rs6000_emit_prologue): Call renamed functions.
18389         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18390         movesi_from_cr, remove volatile CRs.
18392 2017-08-30  Jon Beniston  <jon@beniston.com>
18393             Richard Biener  <rguenther@suse.de>
18395         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18396         of VECTOR_MODE_P check.
18397         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18398         element vector types.
18400 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18402         * df.h (df_read_modify_subreg_p): Remove in favor of...
18403         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18404         const_rtx instead of an rtx.
18405         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18406         * df-problems.c (df_word_lr_mark_ref): Likewise.
18407         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18408         (mark_pseudo_reg_dead): Likewise.
18409         (mark_ref_dead): Likewise.
18410         * reginfo.c (init_subregs_of_mode): Likewise.
18411         * sched-deps.c (sched_analyze_1): Likewise.
18412         * df-scan.c (df_def_record_1): Likewise.
18413         (df_uses_record): Likewise.
18414         (df_read_modify_subreg_p): Remove in favor of...
18415         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18416         const_rtx instead of an rtx.
18418 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18419             Alan Hayward  <alan.hayward@arm.com>
18420             David Sherwood  <david.sherwood@arm.com>
18422         * rtl.h (partial_subreg_p): New function.
18423         * caller-save.c (save_call_clobbered_regs): Use it.
18424         * calls.c (expand_call): Likewise.
18425         * combine.c (combinable_i3pat): Likewise.
18426         (simplify_set): Likewise.
18427         (make_extraction): Likewise.
18428         (make_compound_operation_int): Likewise.
18429         (gen_lowpart_or_truncate): Likewise.
18430         (force_to_mode): Likewise.
18431         (make_field_assignment): Likewise.
18432         (reg_truncated_to_mode): Likewise.
18433         (record_truncated_value): Likewise.
18434         (move_deaths): Likewise.
18435         * cse.c (record_jump_cond): Likewise.
18436         (cse_insn): Likewise.
18437         * cselib.c (cselib_lookup_1): Likewise.
18438         * expmed.c (extract_bit_field_using_extv): Likewise.
18439         * function.c (assign_parm_setup_reg): Likewise.
18440         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18441         * ira-build.c (create_insn_allocnos): Likewise.
18442         * lra-coalesce.c (merge_pseudos): Likewise.
18443         * lra-constraints.c (match_reload): Likewise.
18444         (simplify_operand_subreg): Likewise.
18445         (curr_insn_transform): Likewise.
18446         * lra-lives.c (process_bb_lives): Likewise.
18447         * lra.c (new_insn_reg): Likewise.
18448         (lra_substitute_pseudo): Likewise.
18449         * regcprop.c (mode_change_ok): Likewise.
18450         (maybe_mode_change): Likewise.
18451         (copyprop_hardreg_forward_1): Likewise.
18452         * reload.c (push_reload): Likewise.
18453         (find_reloads): Likewise.
18454         (find_reloads_subreg_address): Likewise.
18455         * reload1.c (alter_reg): Likewise.
18456         (eliminate_regs_1): Likewise.
18457         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18459 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18461         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18462         back to if statements, including unpack.
18464 2017-08-30  Martin Liska  <mliska@suse.cz>
18466         PR inline-asm/82001
18467         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18468         Rename to ...
18469         (func_checker::compare_asm_inputs_outputs): ... this function.
18470         (func_checker::compare_gimple_asm): Use the function to compare
18471         also ASM constrains.
18472         * ipa-icf-gimple.h: Rename the function.
18474 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18475             Alan Hayward  <alan.hayward@arm.com>
18476             David Sherwood  <david.sherwood@arm.com>
18478         * coretypes.h (complex_mode): New type.
18479         * gdbhooks.py (build_pretty_printer): Handle it.
18480         * machmode.h (complex_mode): New class.
18481         (complex_mode::includes_p): New function.
18482         (is_complex_int_mode): Likewise.
18483         (is_complex_float_mode): Likewise.
18484         * genmodes.c (get_mode_class): Handle complex mode classes.
18485         * function.c (expand_function_end): Use is_complex_int_mode.
18487 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18488             Alan Hayward  <alan.hayward@arm.com>
18489             David Sherwood  <david.sherwood@arm.com>
18491         * coretypes.h (scalar_mode_pod): New typedef.
18492         * gdbhooks.py (build_pretty_printer): Handle it.
18493         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18494         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18495         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18496         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18497         as_a <scalar_mode>.
18499 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18500             Alan Hayward  <alan.hayward@arm.com>
18501             David Sherwood  <david.sherwood@arm.com>
18503         * machmode.h (mode_for_vector): Take a scalar_mode instead
18504         of a machine_mode.
18505         * stor-layout.c (mode_for_vector): Likewise.
18506         * explow.c (promote_mode): Use as_a <scalar_mode>.
18507         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18509 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18510             Alan Hayward  <alan.hayward@arm.com>
18511             David Sherwood  <david.sherwood@arm.com>
18513         * target.def (preferred_simd_mode): Take a scalar_mode
18514         instead of a machine_mode.
18515         * targhooks.h (default_preferred_simd_mode): Likewise.
18516         * targhooks.c (default_preferred_simd_mode): Likewise.
18517         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18518         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18519         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18520         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18521         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18522         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18523         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18524         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18525         Likewise.
18526         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18527         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18528         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18529         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18530         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18531         * doc/tm.texi: Regenerate.
18532         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18533         non-scalar modes.
18535 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18536             Alan Hayward  <alan.hayward@arm.com>
18537             David Sherwood  <david.sherwood@arm.com>
18539         * target.def (scalar_mode_supported_p): Take a scalar_mode
18540         instead of a machine_mode.
18541         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18542         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18543         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18544         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18545         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18546         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18547         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18548         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18549         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18550         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18551         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18552         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18553         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18554         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18555         Likewise.
18556         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18557         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18558         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18559         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18560         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18561         Likewise.
18562         * doc/tm.texi: Regenerate.
18564 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18565             Alan Hayward  <alan.hayward@arm.com>
18566             David Sherwood  <david.sherwood@arm.com>
18568         * coretypes.h (opt_scalar_mode): New typedef.
18569         * gdbhooks.py (build_pretty_printers): Handle it.
18570         * machmode.h (mode_iterator::get_2xwider): Add overload for
18571         opt_mode<T>.
18572         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18573         over scalar modes.
18574         * expr.c (convert_mode_scalar): Likewise.
18575         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18576         * optabs.c (expand_float): Likewise.
18577         (expand_fix): Likewise.
18578         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18580 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18581             Alan Hayward  <alan.hayward@arm.com>
18582             David Sherwood  <david.sherwood@arm.com>
18584         * optabs.c (expand_float): Explicitly check for scalars before
18585         using a branching expansion.
18586         (expand_fix): Likewise.
18588 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18589             Alan Hayward  <alan.hayward@arm.com>
18590             David Sherwood  <david.sherwood@arm.com>
18592         * expr.c (convert_mode): Split scalar handling out into...
18593         (convert_mode_scalar): ...this new function.  Treat the modes
18594         as scalar_modes.
18596 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18597             Alan Hayward  <alan.hayward@arm.com>
18598             David Sherwood  <david.sherwood@arm.com>
18600         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18601         and scalar_mode.
18602         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18604 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18605             Alan Hayward  <alan.hayward@arm.com>
18606             David Sherwood  <david.sherwood@arm.com>
18608         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18609         rather than a machine_mode.
18610         (fixed_from_string): Likewise.
18611         (fixed_convert): Likewise.
18612         (fixed_convert_from_int): Likewise.
18613         (fixed_convert_from_real): Likewise.
18614         (real_convert_from_fixed): Likewise.
18615         * fixed-value.c (fixed_from_double_int): Likewise.
18616         (fixed_from_string): Likewise.
18617         (fixed_convert): Likewise.
18618         (fixed_convert_from_int): Likewise.
18619         (fixed_convert_from_real): Likewise.
18620         (real_convert_from_fixed): Likewise.
18621         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18623 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18624             Alan Hayward  <alan.hayward@arm.com>
18625             David Sherwood  <david.sherwood@arm.com>
18627         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18628         of separate mode class checks.  Do not allow vector modes here.
18629         (immed_wide_int_const): Use as_a <scalar_mode>.
18630         * explow.c (trunc_int_for_mode): Likewise.
18631         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18632         (wi::shwi): Likewise.
18633         (wi::min_value): Likewise.
18634         (wi::max_value): Likewise.
18635         * dwarf2out.c (loc_descriptor): Likewise.
18636         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18637         for CONST_WIDE_INT.
18639 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18640             Alan Hayward  <alan.hayward@arm.com>
18641             David Sherwood  <david.sherwood@arm.com>
18643         * tree.h (SCALAR_TYPE_MODE): New macro.
18644         * expr.c (expand_expr_addr_expr_1): Use it.
18645         (expand_expr_real_2): Likewise.
18646         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18647         (fold_convert_const_fixed_from_int): Likewise.
18648         (fold_convert_const_fixed_from_real): Likewise.
18649         (native_encode_fixed): Likewise
18650         (native_encode_complex): Likewise
18651         (native_encode_vector): Likewise.
18652         (native_interpret_fixed): Likewise.
18653         (native_interpret_real): Likewise.
18654         (native_interpret_complex): Likewise.
18655         (native_interpret_vector): Likewise.
18656         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18657         (simd_clone_adjust_argument_types): Likewise.
18658         (simd_clone_init_simd_arrays): Likewise.
18659         (simd_clone_adjust): Likewise.
18660         * stor-layout.c (layout_type): Likewise.
18661         * tree.c (build_minus_one_cst): Likewise.
18662         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18663         * tree-inline.c (estimate_move_cost): Likewise.
18664         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18665         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18666         (vectorizable_reduction): Likewise.
18667         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18668         (vect_recog_mixed_size_cond_pattern): Likewise.
18669         (check_bool_pattern): Likewise.
18670         (adjust_bool_pattern): Likewise.
18671         (search_type_for_mask_1): Likewise.
18672         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18673         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18674         (vectorizable_load): Likewise.
18675         (vectorizable_store): Likewise.
18676         * ubsan.c (ubsan_encode_value): Likewise.
18677         * varasm.c (output_constant): Likewise.
18679 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18680             Alan Hayward  <alan.hayward@arm.com>
18681             David Sherwood  <david.sherwood@arm.com>
18683         * coretypes.h (scalar_mode): New class.
18684         * machmode.h (scalar_mode): Likewise.
18685         (scalar_mode::includes_p): New function.
18686         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18687         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18688         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18689         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18690         * expmed.c (store_bit_field_1): Likewise.
18691         (extract_bit_field_1): Likewise.
18692         * expr.c (write_complex_part): Likewise.
18693         (read_complex_part): Likewise.
18694         (emit_move_complex_push): Likewise.
18695         (expand_expr_real_2): Likewise.
18696         * function.c (assign_parm_setup_reg): Likewise.
18697         (assign_parms_unsplit_complex): Likewise.
18698         * optabs.c (expand_binop): Likewise.
18699         * rtlanal.c (subreg_get_info): Likewise.
18700         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18701         * varasm.c (output_constant_pool_2): Likewise.
18703 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18704             Alan Hayward  <alan.hayward@arm.com>
18705             David Sherwood  <david.sherwood@arm.com>
18707         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18708         assertion.
18709         (expmed_mult_highpart_optab): Likewise.
18710         (expmed_mult_highpart): Likewise.
18712 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18713             Alan Hayward  <alan.hayward@arm.com>
18714             David Sherwood  <david.sherwood@arm.com>
18716         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18717         instead of a machine_mode.
18718         (builtin_memset_read_str): Likewise.
18719         * builtins.c (c_readstr): Likewise.
18720         (builtin_memcpy_read_str): Likewise.
18721         (builtin_strncpy_read_str): Likewise.
18722         (builtin_memset_read_str): Likewise.
18723         (builtin_memset_gen_str): Likewise.
18724         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18725         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18726         instead of a machine_mode.
18727         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18728         variables.
18729         (make_extraction): Likewise.
18730         (try_widen_shift_mode): Take and return scalar_int_modes instead
18731         of machine_modes.
18732         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18733         a scalar_int_mode instead of a machine_mode.
18734         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18735         (avr_addr_space_pointer_mode): Likewise.
18736         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18737         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18738         (msp430_unwind_word_mode): Likewise.
18739         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18740         (spu_addr_space_pointer_mode): Likewise.
18741         (spu_addr_space_address_mode): Likewise.
18742         (spu_libgcc_cmp_return_mode): Likewise.
18743         (spu_libgcc_shift_count_mode): Likewise.
18744         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18745         (rl78_addr_space_pointer_mode): Likewise.
18746         (fl78_unwind_word_mode): Likewise.
18747         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18748         machine_mode.
18749         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18750         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18751         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18752         (mips_valid_pointer_mode): Likewise.
18753         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18754         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18755         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18756         of a machine_mode.
18757         (ft32_addr_space_address_mode): Likewise.
18758         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18759         scalar_int_mode instead of a machine_mode.
18760         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18761         of a machine_mode.
18762         (m32c_addr_space_address_mode): Likewise.
18763         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18764         (rs6000_eh_return_filter_mode): Likewise.
18765         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18766         (rs6000_eh_return_filter_mode): Likewise.
18767         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18768         (s390_libgcc_shift_count_mode): Likewise.
18769         (s390_unwind_word_mode): Likewise.
18770         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18771         machine_mode.
18772         * target.def (mode_rep_extended): Likewise.
18773         (valid_pointer_mode): Likewise.
18774         (addr_space.valid_pointer_mode): Likewise.
18775         (eh_return_filter_mode): Return a scalar_int_mode rather than
18776         a machine_mode.
18777         (libgcc_cmp_return_mode): Likewise.
18778         (libgcc_shift_count_mode): Likewise.
18779         (unwind_word_mode): Likewise.
18780         (addr_space.pointer_mode): Likewise.
18781         (addr_space.address_mode): Likewise.
18782         * doc/tm.texi: Regenerate.
18783         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18784         a machine_mode.
18785         (do_jump): Use scalar_int_mode for local variables.
18786         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18787         rather than a machine_mode.
18788         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18789         (scompare_loc_descriptor_wide): Likewise.
18790         (scompare_loc_descriptor_narrow): Likewise.
18791         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18792         variables.
18793         * except.c (sjlj_emit_dispatch_table): Likewise.
18794         (expand_builtin_eh_copy_values): Likewise.
18795         * explow.c (convert_memory_address_addr_space_1): Likewise.
18796         Take a scalar_int_mode rather than a machine_mode.
18797         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18798         than a machine_mode.
18799         (memory_address_addr_space): Use scalar_int_mode for local variables.
18800         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18801         rather than a machine_mode.
18802         * expmed.c (mask_rtx): Likewise.
18803         (init_expmed_one_conv): Likewise.
18804         (expand_mult_highpart_adjust): Likewise.
18805         (extract_high_half): Likewise.
18806         (expmed_mult_highpart_optab): Likewise.
18807         (expmed_mult_highpart): Likewise.
18808         (expand_smod_pow2): Likewise.
18809         (expand_sdiv_pow2): Likewise.
18810         (emit_store_flag_int): Likewise.
18811         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18812         variables.
18813         (extract_low_bits): Likewise.
18814         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18815         a machine_mode.
18816         * expr.c (pieces_addr::adjust):  Likewise.
18817         (can_store_by_pieces): Likewise.
18818         (store_by_pieces): Likewise.
18819         (clear_by_pieces_1): Likewise.
18820         (expand_expr_addr_expr_1): Likewise.
18821         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18822         (expand_expr_real_1): Likewise.
18823         (try_casesi): Likewise.
18824         * final.c (shorten_branches): Likewise.
18825         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18826         type of "mode" to machine_mode.
18827         * internal-fn.c (expand_arith_overflow_result_store): Take a
18828         scalar_int_mode rather than a machine_mode.
18829         (expand_mul_overflow): Use scalar_int_mode for local variables.
18830         * loop-doloop.c (doloop_modify): Likewise.
18831         (doloop_optimize): Likewise.
18832         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18833         than a machine_mode.
18834         (expand_doubleword_shift_condmove): Likewise.
18835         (expand_doubleword_shift): Likewise.
18836         (expand_doubleword_clz): Likewise.
18837         (expand_doubleword_popcount): Likewise.
18838         (expand_doubleword_parity): Likewise.
18839         (expand_absneg_bit): Use scalar_int_mode for local variables.
18840         (prepare_float_lib_cmp): Likewise.
18841         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18842         rather than a machine_mode.
18843         (convert_memory_address_addr_space): Likewise.
18844         (get_mode_bounds): Likewise.
18845         (get_address_mode): Return a scalar_int_mode rather than a
18846         machine_mode.
18847         * rtlanal.c (get_address_mode): Likewise.
18848         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18849         than a machine_mode.
18850         * targhooks.c (default_mode_rep_extended): Likewise.
18851         (default_valid_pointer_mode): Likewise.
18852         (default_addr_space_valid_pointer_mode): Likewise.
18853         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18854         than a machine_mode.
18855         (default_libgcc_cmp_return_mode): Likewise.
18856         (default_libgcc_shift_count_mode): Likewise.
18857         (default_unwind_word_mode): Likewise.
18858         (default_addr_space_pointer_mode): Likewise.
18859         (default_addr_space_address_mode): Likewise.
18860         * targhooks.h (default_eh_return_filter_mode): Likewise.
18861         (default_libgcc_cmp_return_mode): Likewise.
18862         (default_libgcc_shift_count_mode): Likewise.
18863         (default_unwind_word_mode): Likewise.
18864         (default_addr_space_pointer_mode): Likewise.
18865         (default_addr_space_address_mode): Likewise.
18866         (default_mode_rep_extended): Take a scalar_int_mode rather than
18867         a machine_mode.
18868         (default_valid_pointer_mode): Likewise.
18869         (default_addr_space_valid_pointer_mode): Likewise.
18870         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18871         local variables.
18872         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18873         rather than a machine_mode.
18874         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18875         for local variables.
18876         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18877         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18878         than a machine_mode.
18880 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18881             Alan Hayward  <alan.hayward@arm.com>
18882             David Sherwood  <david.sherwood@arm.com>
18884         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18885         the mode argument to scalar_int_mode.
18886         (do_jump_by_parts_zero_rtx): Likewise.
18887         (do_jump_by_parts_equality_rtx): Likewise.
18888         (do_jump_by_parts_greater): Take a mode argument.
18889         (do_jump_by_parts_equality): Likewise.
18890         (do_jump_1): Update calls accordingly.
18892 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18893             Alan Hayward  <alan.hayward@arm.com>
18894             David Sherwood  <david.sherwood@arm.com>
18896         * is-a.h (safe_dyn_cast): New function.
18897         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18898         (jump_table_for_label): Likewise.
18899         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18900         instead of an rtx_insn *.
18901         (shorten_branches): Use dyn_cast instead of LABEL_P and
18902         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18903         rtx_jump_table_data::get_data_mode.
18904         (final_scan_insn): Likewise.
18906 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18907             Alan Hayward  <alan.hayward@arm.com>
18908             David Sherwood  <david.sherwood@arm.com>
18910         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18911         trying to combine a full-register integer set with a subreg
18912         integer set.
18914 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18915             Alan Hayward  <alan.hayward@arm.com>
18916             David Sherwood  <david.sherwood@arm.com>
18918         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18919         that is always either address_mode or pointer_mode.
18921 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18922             Alan Hayward  <alan.hayward@arm.com>
18923             David Sherwood  <david.sherwood@arm.com>
18925         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18926         when the two are known to be equal.
18928 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18929             Alan Hayward  <alan.hayward@arm.com>
18930             David Sherwood  <david.sherwood@arm.com>
18932         * simplify-rtx.c (simplify_const_unary_operation): Use
18933         is_a <scalar_int_mode> instead of checking for a nonzero
18934         precision.  Forcibly convert op_mode to a scalar_int_mode
18935         in that case.  More clearly differentiate the operand and
18936         result modes and use the former when deciding what the value
18937         of a count-bits operation should be.  Use is_int_mode instead
18938         of checking for a MODE_INT.  Remove redundant check for whether
18939         this mode has a zero precision.
18941 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18942             Alan Hayward  <alan.hayward@arm.com>
18943             David Sherwood  <david.sherwood@arm.com>
18945         * optabs.c (widen_leading): Change the type of the mode argument
18946         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
18947         (widen_bswap): Likewise.
18948         (expand_parity): Likewise.
18949         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
18950         (expand_ffs): Likewise.
18951         (epand_unop): Check for scalar integer modes before calling the
18952         above routines.
18954 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18955             Alan Hayward  <alan.hayward@arm.com>
18956             David Sherwood  <david.sherwood@arm.com>
18958         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
18959         Expand commentary.
18960         (expand_expr_real_1): Update call accordingly.
18962 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18963             Alan Hayward  <alan.hayward@arm.com>
18964             David Sherwood  <david.sherwood@arm.com>
18966         * expmed.c (store_bit_field_using_insv): Add op0_mode and
18967         value_mode arguments.  Use scalar_int_mode internally.
18968         (store_bit_field_1): Rename the new integer mode from imode
18969         to op0_mode and use it instead of GET_MODE (op0).  Update calls
18970         to store_split_bit_field, store_bit_field_using_insv and
18971         store_fixed_bit_field.
18972         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
18973         Use scalar_int_mode internally.  Use a bit count rather than a mode
18974         when calculating the largest bit size for get_best_mode.
18975         Update calls to store_split_bit_field and store_fixed_bit_field_1.
18976         (store_fixed_bit_field_1): Add mode and value_mode arguments.
18977         Remove assertion that OP0 has a scalar integer mode.
18978         (store_split_bit_field): Add op0_mode and value_mode arguments.
18979         Update calls to extract_fixed_bit_field.
18980         (extract_bit_field_using_extv): Add an op0_mode argument.
18981         Use scalar_int_mode internally.
18982         (extract_bit_field_1): Rename the new integer mode from imode to
18983         op0_mode and use it instead of GET_MODE (op0).  Update calls to
18984         extract_split_bit_field, extract_bit_field_using_extv and
18985         extract_fixed_bit_field.
18986         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
18987         to extract_split_bit_field and extract_fixed_bit_field_1.
18988         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
18989         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
18990         on the target mode.
18991         (extract_split_bit_field): Add an op0_mode argument.  Update call
18992         to extract_fixed_bit_field.
18994 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18995             Alan Hayward  <alan.hayward@arm.com>
18996             David Sherwood  <david.sherwood@arm.com>
18998         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
18999         * explow.c (hard_function_value): Likewise.
19000         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
19001         convert_to_mode call outside the loop.
19002         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
19003         for the mode iterator.  Require the mode specified by max_pieces
19004         to exist.
19005         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
19006         mode iterator.
19007         (copy_blkmode_to_reg): Likewise.
19008         (set_storage_via_setmem): Likewise.
19009         * optabs.c (prepare_cmp_insn): Likewise.
19010         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19011         * stor-layout.c (finish_bitfield_representative): Likewise.
19013 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19014             Alan Hayward  <alan.hayward@arm.com>
19015             David Sherwood  <david.sherwood@arm.com>
19017         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
19018         * expr.c (convert_move): Use them.
19019         (convert_modes): Likewise.
19020         (store_expr_with_bounds): Likewise.
19022 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19023             Alan Hayward  <alan.hayward@arm.com>
19024             David Sherwood  <david.sherwood@arm.com>
19026         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19027         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19028         and "known_ret" arguments.  Change the type of the mode argument
19029         to scalar_int_mode.
19030         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19031         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19032         (reg_num_sign_bit_copies_for_combine): Likewise.
19033         * rtlanal.c (nonzero_bits1): Likewise.
19034         (num_sign_bit_copies1): Likewise.
19035         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19036         (reg_num_sign_bit_copies_general): Likewise.
19037         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19038         (reg_nonzero_bits_general): Likewise.
19040 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19041             Alan Hayward  <alan.hayward@arm.com>
19042             David Sherwood  <david.sherwood@arm.com>
19044         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19045         than in subroutines.  Return 1 for non-integer modes.
19046         (cached_num_sign_bit_copies): Change the type of the mode parameter
19047         to scalar_int_mode.
19048         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19049         classes.  Handle CONST_INT_P first and then check whether X also
19050         has a scalar integer mode.  Check the same thing for inner registers
19051         of a SUBREG and for values that are being extended or truncated.
19053 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19054             Alan Hayward  <alan.hayward@arm.com>
19055             David Sherwood  <david.sherwood@arm.com>
19057         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19058         in subroutines.  Return the mode mask for non-integer modes.
19059         (cached_nonzero_bits): Change the type of the mode parameter
19060         to scalar_int_mode.
19061         (nonzero_bits1): Likewise.  Remove early exit for other mode
19062         classes.  Handle CONST_INT_P first and then check whether X
19063         also has a scalar integer mode.
19065 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19066             Alan Hayward  <alan.hayward@arm.com>
19067             David Sherwood  <david.sherwood@arm.com>
19069         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19070         Return a scalar_int_mode and assert that the size is greater than
19071         one byte.
19072         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19073         (op_by_pieces_d::op_by_pieces_d): Likewise.
19074         (op_by_pieces_d::run): Likewise.
19075         (can_store_by_pieces): Likewise.
19077 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19078             Alan Hayward  <alan.hayward@arm.com>
19079             David Sherwood  <david.sherwood@arm.com>
19081         * combine.c (extract_left_shift): Add a mode argument and update
19082         recursive calls.
19083         (make_compound_operation_int): Change the type of the mode parameter
19084         to scalar_int_mode and update the call to extract_left_shift.
19086 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19087             Alan Hayward  <alan.hayward@arm.com>
19088             David Sherwood  <david.sherwood@arm.com>
19090         * combine.c (simplify_and_const_int): Change the type of the mode
19091         parameter to scalar_int_mode.
19092         (simplify_and_const_int_1): Likewise.  Update recursive call.
19094 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19095             Alan Hayward  <alan.hayward@arm.com>
19096             David Sherwood  <david.sherwood@arm.com>
19098         * combine.c (simplify_compare_const): Check that the mode is a
19099         scalar_int_mode (rather than VOIDmode) before testing its
19100         precision.
19101         (simplify_comparison): Move COMPARISON_P handling out of the
19102         loop and restrict the latter part of the loop to scalar_int_modes.
19103         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19104         and when considering SUBREG_REGs.  Use is_int_mode instead of
19105         checking GET_MODE_CLASS against MODE_INT.
19107 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19108             Alan Hayward  <alan.hayward@arm.com>
19109             David Sherwood  <david.sherwood@arm.com>
19111         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19112         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19113         shift_unit_mode and for modes involved in scalar shifts.
19115 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19116             Alan Hayward  <alan.hayward@arm.com>
19117             David Sherwood  <david.sherwood@arm.com>
19119         * combine.c (force_int_to_mode): New function, split out from...
19120         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19121         instead of GET_MODE (x).
19123 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19124             Alan Hayward  <alan.hayward@arm.com>
19125             David Sherwood  <david.sherwood@arm.com>
19127         * optabs-query.h (extraction_insn::struct_mode): Change type to
19128         opt_scalar_int_mode and update comment.
19129         (extraction_insn::field_mode): Change type to scalar_int_mode.
19130         (extraction_insn::pos_mode): Likewise.
19131         * combine.c (make_extraction): Update accordingly.
19132         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19133         (get_optab_extraction_insn): Likewise.
19134         * recog.c (simplify_while_replacing): Likewise.
19135         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19136         parameter to opt_scalar_int_mode.
19138 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19139             Alan Hayward  <alan.hayward@arm.com>
19140             David Sherwood  <david.sherwood@arm.com>
19142         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19143         to a scalar_int_mode instead of a machine_mode.
19144         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19145         (get_best_mode): Return a boolean and use a pointer argument to store
19146         the selected mode.  Replace the limit mode parameter with a bit limit.
19147         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19148         for the values returned by bit_field_mode_iterator::next_mode.
19149         (store_bit_field): Update call to get_best_mode.
19150         (store_fixed_bit_field): Likewise.
19151         (extract_fixed_bit_field): Likewise.
19152         * expr.c (optimize_bitfield_assignment_op): Likewise.
19153         * fold-const.c (optimize_bit_field_compare): Likewise.
19154         (fold_truth_andor_1): Likewise.
19155         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19156         Update for new type of m_mode.
19157         (get_best_mode): As above.
19159 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19160             Alan Hayward  <alan.hayward@arm.com>
19161             David Sherwood  <david.sherwood@arm.com>
19163         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19164         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19165         (store_bit_field): Check is_a <scalar_int_mode> before calling
19166         strict_volatile_bitfield_p.
19167         (extract_bit_field): Likewise.
19169 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19170             Alan Hayward  <alan.hayward@arm.com>
19171             David Sherwood  <david.sherwood@arm.com>
19173         * target.def (cstore_mode): Return a scalar_int_mode.
19174         * doc/tm.texi: Regenerate.
19175         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19176         * targhooks.h (default_cstore_mode): Likewise.
19177         * targhooks.c (default_cstore_mode): Likewise, using a forced
19178         conversion.
19179         * expmed.c (emit_cstore): Expect the target of the cstore to be
19180         a scalar_int_mode.
19182 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19183             Alan Hayward  <alan.hayward@arm.com>
19184             David Sherwood  <david.sherwood@arm.com>
19186         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19187         scalar_int_mode.
19188         (niter_desc): Likewise mode.
19189         (iv_analyze): Add a mode parameter.
19190         (biv_p): Likewise.
19191         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19192         and change its type to scalar_int_mode.
19193         * loop-iv.c: Update commentary at head of file.
19194         (iv_constant): Pass the mode paraeter before the rtx it describes
19195         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19196         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19197         (iv_extend): Likewise.
19198         (shorten_into_mode): Likewise.
19199         (iv_add): Use scalar_int_mode.
19200         (iv_mult): Likewise.
19201         (iv_shift): Likewise.
19202         (canonicalize_iv_subregs): Likewise.
19203         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19204         it describes and change its mode to scalar_int_mode.   Also change
19205         the type of the returned inner_mode to scalar_int_mode.
19206         (get_biv_step): Likewise, turning outer_mode from a pointer
19207         into a direct parameter.  Update call to get_biv_step_1.
19208         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19209         iv_constant and get_biv_step.
19210         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19211         and change its type to scalar_int_mode.  Don't initialise iv->mode
19212         to VOIDmode and remove later checks for its still being VOIDmode.
19213         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19214         is_a <scalar_int_mode> when changing the mode under consideration.
19215         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19216         Update call to iv_analyze_expr.
19217         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19218         inner register is not also a scalar_int_mode.  Update call to
19219         iv_analyze_biv.
19220         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19221         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19222         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19223         separate mode class checks.  Update calls to iv_analyze.  Remove
19224         fix-up of VOIDmodes after iv_analyze_biv.
19225         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19226         don't have a scalar_int_mode.  Update call to biv_p.
19228 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19229             Alan Hayward  <alan.hayward@arm.com>
19230             David Sherwood  <david.sherwood@arm.com>
19232         * cfgexpand.c (convert_debug_memory_address): Use
19233         as_a <scalar_int_mode>.
19234         * combine.c (expand_compound_operation): Likewise.
19235         (make_extraction): Likewise.
19236         (change_zero_ext): Likewise.
19237         (simplify_comparison): Likewise.
19238         * cse.c (cse_insn): Likewise.
19239         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19240         (mem_loc_descriptor): Likewise.
19241         (loc_descriptor): Likewise.
19242         * expmed.c (init_expmed_one_mode): Likewise.
19243         (synth_mult): Likewise.
19244         (emit_store_flag_1): Likewise.
19245         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19246         of a comparison with size.
19247         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19248         (reduce_to_bit_field_precision): Likewise.
19249         * function.c (expand_function_end): Likewise.
19250         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19251         * loop-doloop.c (doloop_modify): Likewise.
19252         * optabs.c (expand_binop): Likewise.
19253         (expand_unop): Likewise.
19254         (expand_copysign_absneg): Likewise.
19255         (prepare_cmp_insn): Likewise.
19256         (maybe_legitimize_operand): Likewise.
19257         * recog.c (const_scalar_int_operand): Likewise.
19258         * rtlanal.c (get_address_mode): Likewise.
19259         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19260         (simplify_cond_clz_ctz): Likewise.
19261         * tree-nested.c (get_nl_goto_field): Likewise.
19262         * tree.c (build_vector_type_for_mode): Likewise.
19263         * var-tracking.c (use_narrower_mode): Likewise.
19265 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19266             Alan Hayward  <alan.hayward@arm.com>
19267             David Sherwood  <david.sherwood@arm.com>
19269         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19270         * builtins.c (expand_builtin_signbit): Use it.
19271         * cfgexpand.c (expand_debug_expr): Likewise.
19272         * dojump.c (do_jump): Likewise.
19273         (do_compare_and_jump): Likewise.
19274         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19275         * expmed.c (make_tree): Likewise.
19276         * expr.c (expand_expr_real_2): Likewise.
19277         (expand_expr_real_1): Likewise.
19278         (try_casesi): Likewise.
19279         * fold-const-call.c (fold_const_call_ss): Likewise.
19280         * fold-const.c (unextend): Likewise.
19281         (extract_muldiv_1): Likewise.
19282         (fold_single_bit_test): Likewise.
19283         (native_encode_int): Likewise.
19284         (native_encode_string): Likewise.
19285         (native_interpret_int): Likewise.
19286         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19287         * internal-fn.c (expand_addsub_overflow): Likewise.
19288         (expand_neg_overflow): Likewise.
19289         (expand_mul_overflow): Likewise.
19290         (expand_arith_overflow): Likewise.
19291         * match.pd: Likewise.
19292         * stor-layout.c (layout_type): Likewise.
19293         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19294         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19295         * tree-ssanames.c (get_range_info): Likewise.
19296         * tree-switch-conversion.c (array_value_type) Likewise.
19297         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19298         (vect_recog_divmod_pattern): Likewise.
19299         (vect_recog_mixed_size_cond_pattern): Likewise.
19300         * tree-vrp.c (extract_range_basic): Likewise.
19301         (simplify_float_conversion_using_ranges): Likewise.
19302         * tree.c (int_fits_type_p): Likewise.
19303         * ubsan.c (instrument_bool_enum_load): Likewise.
19304         * varasm.c (mergeable_string_section): Likewise.
19305         (narrowing_initializer_constant_valid_p): Likewise.
19306         (output_constant): Likewise.
19308 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19309             Alan Hayward  <alan.hayward@arm.com>
19310             David Sherwood  <david.sherwood@arm.com>
19312         * machmode.h (NARROWEST_INT_MODE): New macro.
19313         * expr.c (alignment_for_piecewise_move): Use it instead of
19314         GET_CLASS_NARROWEST_MODE (MODE_INT).
19315         (push_block): Likewise.
19316         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19317         Likewise.
19318         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19320 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19321             Alan Hayward  <alan.hayward@arm.com>
19322             David Sherwood  <david.sherwood@arm.com>
19324         * postreload.c (move2add_valid_value_p): Change the type of the
19325         mode parameter to scalar_int_mode.
19326         (move2add_use_add2_insn): Add a mode parameter and use it instead
19327         of GET_MODE (reg).
19328         (move2add_use_add3_insn): Likewise.
19329         (reload_cse_move2add): Update accordingly.
19331 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19332             Alan Hayward  <alan.hayward@arm.com>
19333             David Sherwood  <david.sherwood@arm.com>
19335         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19336         double-word mode.
19337         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19338         * optabs.c (expand_unop): Likewise.
19340 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19341             Alan Hayward  <alan.hayward@arm.com>
19342             David Sherwood  <david.sherwood@arm.com>
19344         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19345         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19346         (popcount_loc_descriptor): Likewise.
19347         (bswap_loc_descriptor): Likewise.
19348         (rotate_loc_descriptor): Likewise.
19349         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19350         calling the functions above.
19352 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19353             Alan Hayward  <alan.hayward@arm.com>
19354             David Sherwood  <david.sherwood@arm.com>
19356         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19357         checks.
19358         (try_combine): Likewise.
19359         (simplify_if_then_else): Likewise.
19360         * cse.c (cse_insn): Likewise.
19361         * dwarf2out.c (mem_loc_descriptor): Likewise.
19362         * emit-rtl.c (gen_lowpart_common): Likewise.
19363         * simplify-rtx.c (simplify_truncation): Likewise.
19364         (simplify_binary_operation_1): Likewise.
19365         (simplify_const_relational_operation): Likewise.
19366         (simplify_ternary_operation): Likewise.
19367         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19369 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19370             Alan Hayward  <alan.hayward@arm.com>
19371             David Sherwood  <david.sherwood@arm.com>
19373         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19374         * reload.c (push_reload): Likewise.
19375         (find_reloads): Likewise.
19377 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19378             Alan Hayward  <alan.hayward@arm.com>
19379             David Sherwood  <david.sherwood@arm.com>
19381         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19382         (make_compound_operation_int): Likewise.
19383         (change_zero_ext): Likewise.
19384         * expr.c (convert_move): Likewise.
19385         (convert_modes): Likewise.
19386         * fwprop.c (forward_propagate_subreg): Likewise.
19387         * loop-iv.c (get_biv_step_1): Likewise.
19388         * optabs.c (widen_operand): Likewise.
19389         * postreload.c (move2add_valid_value_p): Likewise.
19390         * recog.c (simplify_while_replacing): Likewise.
19391         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19392         (simplify_binary_operation_1): Likewise.  Remove redundant
19393         mode equality check.
19395 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19396             Alan Hayward  <alan.hayward@arm.com>
19397             David Sherwood  <david.sherwood@arm.com>
19399         * combine.c (combine_simplify_rtx): Add checks for
19400         is_a <scalar_int_mode>.
19401         (simplify_if_then_else): Likewise.
19402         (make_field_assignment): Likewise.
19403         (simplify_comparison): Likewise.
19404         * ifcvt.c (noce_try_bitop): Likewise.
19405         * loop-invariant.c (canonicalize_address_mult): Likewise.
19406         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19408 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19409             Alan Hayward  <alan.hayward@arm.com>
19410             David Sherwood  <david.sherwood@arm.com>
19412         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19413         is_a <scalar_int_mode> instead of != BLKmode.
19415 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19416             Alan Hayward  <alan.hayward@arm.com>
19417             David Sherwood  <david.sherwood@arm.com>
19419         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19420         instead of != VOIDmode.
19421         * combine.c (if_then_else_cond): Likewise.
19422         (change_zero_ext): Likewise.
19423         * dwarf2out.c (mem_loc_descriptor): Likewise.
19424         (loc_descriptor): Likewise.
19425         * rtlanal.c (canonicalize_condition): Likewise.
19426         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19428 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19429             Alan Hayward  <alan.hayward@arm.com>
19430             David Sherwood  <david.sherwood@arm.com>
19432         * simplify-rtx.c (simplify_binary_operation_1): Use
19433         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19435 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19436             Alan Hayward  <alan.hayward@arm.com>
19437             David Sherwood  <david.sherwood@arm.com>
19439         * wide-int.h (int_traits<unsigned char>) New class.
19440         (int_traits<unsigned short>) Likewise.
19441         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19442         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19443         SCALAR_INT_MODE_P.
19444         * combine.c (set_nonzero_bits_and_sign_copies): Use
19445         is_a <scalar_int_mode>.
19446         (find_split_point): Likewise.
19447         (combine_simplify_rtx): Likewise.
19448         (simplify_logical): Likewise.
19449         (expand_compound_operation): Likewise.
19450         (expand_field_assignment): Likewise.
19451         (make_compound_operation): Likewise.
19452         (extended_count): Likewise.
19453         (change_zero_ext): Likewise.
19454         (simplify_comparison): Likewise.
19455         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19456         (ucompare_loc_descriptor): Likewise.
19457         (minmax_loc_descriptor): Likewise.
19458         (mem_loc_descriptor): Likewise.
19459         (loc_descriptor): Likewise.
19460         * expmed.c (init_expmed_one_mode): Likewise.
19461         * lra-constraints.c (lra_constraint_offset): Likewise.
19462         * optabs.c (prepare_libcall_arg): Likewise.
19463         * postreload.c (move2add_note_store): Likewise.
19464         * reload.c (operands_match_p): Likewise.
19465         * rtl.h (load_extend_op): Likewise.
19466         * rtlhooks.c (gen_lowpart_general): Likewise.
19467         * simplify-rtx.c (simplify_truncation): Likewise.
19468         (simplify_unary_operation_1): Likewise.
19469         (simplify_binary_operation_1): Likewise.
19470         (simplify_const_binary_operation): Likewise.
19471         (simplify_const_relational_operation): Likewise.
19472         (simplify_subreg): Likewise.
19473         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19474         * var-tracking.c (adjust_mems): Likewise.
19475         (prepare_call_arguments): Likewise.
19477 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19478             Alan Hayward  <alan.hayward@arm.com>
19479             David Sherwood  <david.sherwood@arm.com>
19481         * machmode.h (is_int_mode): New fuction.
19482         * combine.c (find_split_point): Use it.
19483         (combine_simplify_rtx): Likewise.
19484         (simplify_if_then_else): Likewise.
19485         (simplify_set): Likewise.
19486         (simplify_shift_const_1): Likewise.
19487         (simplify_comparison): Likewise.
19488         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19489         * cse.c (notreg_cost): Likewise.
19490         (cse_insn): Likewise.
19491         * cselib.c (cselib_lookup_1): Likewise.
19492         * dojump.c (do_jump_1): Likewise.
19493         (do_compare_rtx_and_jump): Likewise.
19494         * dse.c (get_call_args): Likewise.
19495         * dwarf2out.c (rtl_for_decl_init): Likewise.
19496         (native_encode_initializer): Likewise.
19497         * expmed.c (emit_store_flag_1): Likewise.
19498         (emit_store_flag): Likewise.
19499         * expr.c (convert_modes): Likewise.
19500         (store_field): Likewise.
19501         (expand_expr_real_1): Likewise.
19502         * fold-const.c (fold_read_from_constant_string): Likewise.
19503         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19504         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19505         * optabs.c (expand_binop): Likewise.
19506         (expand_unop): Likewise.
19507         (expand_abs_nojump): Likewise.
19508         (expand_one_cmpl_abs_nojump): Likewise.
19509         * simplify-rtx.c (mode_signbit_p): Likewise.
19510         (val_signbit_p): Likewise.
19511         (val_signbit_known_set_p): Likewise.
19512         (val_signbit_known_clear_p): Likewise.
19513         (simplify_relational_operation_1): Likewise.
19514         * tree.c (vector_type_mode): Likewise.
19516 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19517             Alan Hayward  <alan.hayward@arm.com>
19518             David Sherwood  <david.sherwood@arm.com>
19520         * machmode.h (smallest_mode_for_size): Fix formatting.
19521         (smallest_int_mode_for_size): New function.
19522         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19523         instead of smallest_mode_for_size.
19524         * combine.c (make_extraction): Likewise.
19525         * config/arc/arc.c (arc_expand_movmem): Likewise.
19526         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19527         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19528         * config/s390/s390.c (s390_expand_insv): Likewise.
19529         * config/sparc/sparc.c (assign_int_registers): Likewise.
19530         * config/spu/spu.c (spu_function_value): Likewise.
19531         (spu_function_arg): Likewise.
19532         * coverage.c (get_gcov_type): Likewise.
19533         (get_gcov_unsigned_t): Likewise.
19534         * dse.c (find_shift_sequence): Likewise.
19535         * expmed.c (store_bit_field_1): Likewise.
19536         * expr.c (convert_move): Likewise.
19537         (store_field): Likewise.
19538         * internal-fn.c (expand_arith_overflow): Likewise.
19539         * optabs-query.c (get_best_extraction_insn): Likewise.
19540         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19541         * stor-layout.c (layout_type): Likewise.
19542         (initialize_sizetypes): Likewise.
19543         * targhooks.c (default_get_mask_mode): Likewise.
19544         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19546 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19547             Alan Hayward  <alan.hayward@arm.com>
19548             David Sherwood  <david.sherwood@arm.com>
19550         * machmode.h (opt_mode::else_blk): New function.
19551         (int_mode_for_mode): Declare.
19552         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19553         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19554         return type.
19555         * cfgexpand.c (expand_debug_expr): Likewise.
19556         * combine.c (gen_lowpart_or_truncate): Likewise.
19557         (gen_lowpart_for_combine): Likewise.
19558         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19559         * config/avr/avr.c (avr_to_int_mode): Likewise.
19560         (avr_out_plus_1): Likewise.
19561         (avr_out_plus): Likewise.
19562         (avr_out_round): Likewise.
19563         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19564         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19565         (s390_expand_vcond): Likewise.
19566         * config/spu/spu.c (spu_split_immediate): Likewise.
19567         (spu_expand_mov): Likewise.
19568         * dse.c (get_stored_val): Likewise.
19569         * expmed.c (store_bit_field_1): Likewise.
19570         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19571         int_mode_for_size.
19572         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19573         (extract_low_bits): Likewise.
19574         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19575         handling rather than repeating the check.
19576         (emit_group_store): Likewise.
19577         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19578         * optabs.c (expand_absneg_bit): Likewise.
19579         (expand_copysign_absneg): Likewise.
19580         (expand_copysign_bit): Likewise.
19581         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19582         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19583         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19584         * var-tracking.c (prepare_call_arguments):  Likewise.
19585         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19586         int_mode_for_mode instead of mode_for_size.
19587         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19589 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19590             Alan Hayward  <alan.hayward@arm.com>
19591             David Sherwood  <david.sherwood@arm.com>
19593         * machmode.h (int_mode_for_size): New function.
19594         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19595         instead of mode_for_size.
19596         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19597         explicit.
19598         * combine.c (expand_field_assignment): Use int_mode_for_size
19599         instead of mode_for_size.
19600         (make_extraction): Likewise.
19601         (simplify_shift_const_1): Likewise.
19602         (simplify_comparison): Likewise.
19603         * dojump.c (do_jump): Likewise.
19604         * dwarf2out.c (mem_loc_descriptor): Likewise.
19605         * emit-rtl.c (init_derived_machine_modes): Likewise.
19606         * expmed.c (flip_storage_order): Likewise.
19607         (convert_extracted_bit_field): Likewise.
19608         * expr.c (copy_blkmode_from_reg): Likewise.
19609         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19610         * internal-fn.c (expand_mul_overflow): Likewise.
19611         * lower-subreg.c (simple_move): Likewise.
19612         * optabs-libfuncs.c (init_optabs): Likewise.
19613         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19614         * tree.c (vector_type_mode): Likewise.
19615         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19616         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19617         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19618         * tree-vect-stmts.c (vectorizable_load): Likewise.
19619         (vectorizable_store): Likewise.
19621 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19622             Alan Hayward  <alan.hayward@arm.com>
19623             David Sherwood  <david.sherwood@arm.com>
19625         * coretypes.h (pod_mode): New type.
19626         (scalar_int_mode_pod): New typedef.
19627         * machmode.h (pod_mode): New class.
19628         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19629         * genmodes.c (emit_mode_int_n): Update accordingly.
19630         * lower-subreg.h (target_lower_subreg): Change type to
19631         scalar_int_mode_pod.
19632         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19633         scalar_int_mode_pod.
19635 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19636             Alan Hayward  <alan.hayward@arm.com>
19637             David Sherwood  <david.sherwood@arm.com>
19639         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19640         machine_mode to scalar_int_mode.
19641         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19642         (rs6000_option_override_internal): Remove cast to int.
19643         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19644         machine_mode to scalar_int_mode.
19645         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19646         (rs6000_option_override_internal): Remove cast to int.
19647         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19648         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19649         to machine_mode.
19650         * config/s390/s390.c (s390_expand_builtin): Likewise.
19651         * coretypes.h (scalar_int_mode): New type.
19652         (opt_scalar_int_mode): New typedef.
19653         * machmode.h (scalar_int_mode): New class.
19654         (scalar_int_mode::includes_p): New function.
19655         (byte_mode): Change type to scalar_int_mode.
19656         (word_mode): Likewise.
19657         (ptr_mode): Likewise.
19658         * emit-rtl.c (byte_mode): Likewise.
19659         (word_mode): Likewise.
19660         (ptr_mode): Likewise.
19661         (init_derived_machine_modes): Update accordingly.
19662         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19663         and MODE_PARTIAL_INT.
19664         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19665         opt_scalar_int_mode.
19667 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19668             Alan Hayward  <alan.hayward@arm.com>
19669             David Sherwood  <david.sherwood@arm.com>
19671         * target.def (libgcc_floating_mode_supported_p): Take a
19672         scalar_float_mode.
19673         * doc/tm.texi: Regenerate.
19674         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19675         scalar_float_mode.
19676         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19677         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19678         Likewise.
19680 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19681             Alan Hayward  <alan.hayward@arm.com>
19682             David Sherwood  <david.sherwood@arm.com>
19684         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19685         * doc/tm.texi: Regenerate.
19686         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19687         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19688         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19689         * targhooks.h (default_floatn_mode): Likewise.
19690         * targhooks.c (default_floatn_mode): Likewise.
19691         * tree.c (build_common_tree_nodes): Update accordingly.
19693 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19694             Alan Hayward  <alan.hayward@arm.com>
19695             David Sherwood  <david.sherwood@arm.com>
19697         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19698         (mode_iterator::iterate_p): Likewise.
19699         (mode_iterator::get_wider): Likewise.
19700         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19702 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19703             Alan Hayward  <alan.hayward@arm.com>
19704             David Sherwood  <david.sherwood@arm.com>
19706         * coretypes.h (opt_scalar_float_mode): New typedef.
19707         * machmode.h (float_mode_for_size): New function.
19708         * emit-rtl.c (double_mode): Delete.
19709         (init_emit_once): Use float_mode_for_size.
19710         * stor-layout.c (layout_type): Likewise.
19711         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19713 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19714             Alan Hayward  <alan.hayward@arm.com>
19715             David Sherwood  <david.sherwood@arm.com>
19717         * output.h (assemble_real): Take a scalar_float_mode.
19718         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19719         * config/arm/arm.md (consttable_4): Likewise.
19720         (consttable_8): Likewise.
19721         (consttable_16): Likewise.
19722         * config/mips/mips.md (consttable_float): Likewise.
19723         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19724         * varasm.c (assemble_real): Take a scalar_float_mode.
19725         (output_constant_pool_2): Update accordingly.
19726         (output_constant): Likewise.
19728 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19729             Alan Hayward  <alan.hayward@arm.com>
19730             David Sherwood  <david.sherwood@arm.com>
19732         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19733         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19734         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19735         (native_encode_real): Likewise.
19736         (native_interpret_real): Likewise.
19737         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19738         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19740 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19741             Alan Hayward  <alan.hayward@arm.com>
19742             David Sherwood  <david.sherwood@arm.com>
19744         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19745         <scalar_float_mode>.  Simplify.
19746         (gen_extend_conv_libfunc): Likewise.
19748 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19749             Alan Hayward  <alan.hayward@arm.com>
19750             David Sherwood  <david.sherwood@arm.com>
19752         * coretypes.h (scalar_float_mode): New type.
19753         * machmode.h (mode_traits::from_int): Use machine_mode if
19754         USE_ENUM_MODES is defined.
19755         (is_a): New function.
19756         (as_a): Likewise.
19757         (dyn_cast): Likewise.
19758         (scalar_float_mode): New class.
19759         (scalar_float_mode::includes_p): New function.
19760         (is_float_mode): Likewise.
19761         * gdbhooks.py (MachineModePrinter): New class.
19762         (build_pretty_printer): Use it for scalar_float_mode.
19763         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19764         (format_helper::format_helper): Turn into a template.
19765         * genmodes.c (get_mode_class): New function.
19766         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19767         or machine_mode if none.
19768         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19769         as_a <scalar_float_mode>.
19770         * dwarf2out.c (mem_loc_descriptor): Likewise.
19771         (insert_float): Likewise.
19772         (add_const_value_attribute): Likewise.
19773         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19774         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19775         (expand_unop): Update accordingly.
19776         (expand_abs_nojump): Likewise.
19777         (expand_copysign_absneg): Take a scalar_float_mode.
19778         (expand_copysign_bit): Likewise.
19779         (expand_copysign): Update accordingly.
19781 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19782             Alan Hayward  <alan.hayward@arm.com>
19783             David Sherwood  <david.sherwood@arm.com>
19785         * coretypes.h (opt_mode): New class.
19786         * machmode.h (opt_mode): Likewise.
19787         (opt_mode::else_void): New function.
19788         (opt_mode::require): Likewise.
19789         (opt_mode::exists): Likewise.
19790         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19791         (GET_MODE_2XWIDER_MODE): Likewise.
19792         (mode_iterator::get_wider): Update accordingly.
19793         (mode_iterator::get_2xwider): Likewise.
19794         (mode_iterator::get_known_wider): Likewise, turning into a template.
19795         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19796         forcing a wider mode to exist.
19797         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19798         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19799         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19800         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19801         * lower-subreg.c (init_lower_subreg): Likewise.
19802         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19803         on the final iteration.
19804         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19805         a wider mode exists before asking for a move pattern.
19806         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19807         forcing a wider mode to exist.
19808         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19809         returning false if no such mode exists.
19810         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19811         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19812         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19813         Avoid checking for a MODE_INT if we already know the mode is not a
19814         SCALAR_INT_MODE_P.
19815         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19816         forcing a wider mode to exist.
19817         (expmed_mult_highpart_optab): Likewise.
19818         (expmed_mult_highpart): Likewise.
19819         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19820         using else_void.
19821         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19822         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19823         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19824         * internal-fn.c (expand_mul_overflow): Update use of
19825         GET_MODE_2XWIDER_MODE.
19826         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19827         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19828         GET_MODE_WIDER_MODE.
19829         (convert_plusminus_to_widen): Likewise.
19830         * tree-switch-conversion.c (array_value_type): Likewise.
19831         * var-tracking.c (emit_note_insn_var_location): Likewise.
19832         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19833         Return false inside rather than outside the loop if no wider mode
19834         exists
19835         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19836         and GET_MODE_2XWIDER_MODE
19837         (can_compare_p): Use else_void.
19838         * gdbhooks.py (OptMachineModePrinter): New class.
19839         (build_pretty_printer): Use it for opt_mode.
19841 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19842             Alan Hayward  <alan.hayward@arm.com>
19843             David Sherwood  <david.sherwood@arm.com>
19845         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19846         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19848 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19849             Alan Hayward  <alan.hayward@arm.com>
19850             David Sherwood  <david.sherwood@arm.com>
19852         * machmode.h (mode_traits): New structure.
19853         (get_narrowest_mode): New function.
19854         (mode_iterator::start): Likewise.
19855         (mode_iterator::iterate_p): Likewise.
19856         (mode_iterator::get_wider): Likewise.
19857         (mode_iterator::get_known_wider): Likewise.
19858         (mode_iterator::get_2xwider): Likewise.
19859         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19860         (FOR_EACH_MODE): Likewise.
19861         (FOR_EACH_MODE_FROM): Likewise.
19862         (FOR_EACH_MODE_UNTIL): Likewise.
19863         (FOR_EACH_WIDER_MODE): Likewise.
19864         (FOR_EACH_2XWIDER_MODE): Likewise.
19865         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19866         * combine.c (simplify_comparison): Likewise
19867         * config/i386/i386.c (type_natural_mode): Likewise.
19868         * cse.c (cse_insn): Likewise.
19869         * dse.c (find_shift_sequence): Likewise.
19870         * emit-rtl.c (init_derived_machine_modes): Likewise.
19871         (init_emit_once): Likewise.
19872         * explow.c (hard_function_value): Likewise.
19873         * expmed.c (extract_fixed_bit_field_1): Likewise.
19874         (extract_bit_field_1): Likewise.
19875         (expand_divmod): Likewise.
19876         (emit_store_flag_1): Likewise.
19877         * expr.c (init_expr_target): Likewise.
19878         (convert_move): Likewise.
19879         (alignment_for_piecewise_move): Likewise.
19880         (widest_int_mode_for_size): Likewise.
19881         (emit_block_move_via_movmem): Likewise.
19882         (copy_blkmode_to_reg): Likewise.
19883         (set_storage_via_setmem): Likewise.
19884         (compress_float_constant): Likewise.
19885         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19886         * optabs-query.c (get_best_extraction_insn): Likewise.
19887         * optabs.c (expand_binop): Likewise.
19888         (expand_twoval_unop): Likewise.
19889         (expand_twoval_binop): Likewise.
19890         (widen_leading): Likewise.
19891         (widen_bswap): Likewise.
19892         (expand_parity): Likewise.
19893         (expand_unop): Likewise.
19894         (prepare_cmp_insn): Likewise.
19895         (prepare_float_lib_cmp): Likewise.
19896         (expand_float): Likewise.
19897         (expand_fix): Likewise.
19898         (expand_sfix_optab): Likewise.
19899         * postreload.c (move2add_use_add2_insn): Likewise.
19900         * reg-stack.c (reg_to_stack): Likewise.
19901         * reginfo.c (choose_hard_reg_mode): Likewise.
19902         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19903         * stor-layout.c (mode_for_size): Likewise.
19904         (smallest_mode_for_size): Likewise.
19905         (mode_for_vector): Likewise.
19906         (finish_bitfield_representative): Likewise.
19907         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19908         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19909         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19910         * var-tracking.c (prepare_call_arguments): Likewise.
19912 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19913             Alan Hayward  <alan.hayward@arm.com>
19914             David Sherwood  <david.sherwood@arm.com>
19916         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19917         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19918         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19919         * machmode.h (mode_size): Move earlier in file.
19920         (mode_precision): Likewise.
19921         (mode_inner): Likewise.
19922         (mode_nunits): Likewise.
19923         (mode_unit_size): Likewise.
19924         (unit_unit_precision): Likewise.
19925         (mode_wider): Likewise.
19926         (mode_2xwider): Likewise.
19927         (machine_mode): New class.
19928         (mode_to_bytes): New function.
19929         (mode_to_bits): Likewise.
19930         (mode_to_precision): Likewise.
19931         (mode_to_inner): Likewise.
19932         (mode_to_unit_size): Likewise.
19933         (mode_to_unit_precision): Likewise.
19934         (mode_to_nunits): Likewise.
19935         (GET_MODE_SIZE): Use mode_to_bytes.
19936         (GET_MODE_BITSIZE): Use mode_to_bits.
19937         (GET_MODE_PRECISION): Use mode_to_precision.
19938         (GET_MODE_INNER): Use mode_to_inner.
19939         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19940         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
19941         (GET_MODE_NUNITS): Use mode_to_nunits.
19942         * system.h (ALWAYS_INLINE): New macro.
19943         * config/powerpcspe/powerpcspe-c.c
19944         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
19945         int for arg1_mode and arg2_mode.
19947 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19948             Alan Hayward  <alan.hayward@arm.com>
19949             David Sherwood  <david.sherwood@arm.com>
19951         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
19952         Prefix mode names with E_ in case statements.
19953         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19954         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
19955         (aarch64_split_simd_move): Likewise.
19956         (aarch64_gen_storewb_pair): Likewise.
19957         (aarch64_gen_loadwb_pair): Likewise.
19958         (aarch64_gen_store_pair): Likewise.
19959         (aarch64_gen_load_pair): Likewise.
19960         (aarch64_get_condition_code_1): Likewise.
19961         (aarch64_constant_pool_reload_icode): Likewise.
19962         (get_rsqrte_type): Likewise.
19963         (get_rsqrts_type): Likewise.
19964         (get_recpe_type): Likewise.
19965         (get_recps_type): Likewise.
19966         (aarch64_gimplify_va_arg_expr): Likewise.
19967         (aarch64_simd_container_mode): Likewise.
19968         (aarch64_emit_load_exclusive): Likewise.
19969         (aarch64_emit_store_exclusive): Likewise.
19970         (aarch64_expand_compare_and_swap): Likewise.
19971         (aarch64_gen_atomic_cas): Likewise.
19972         (aarch64_emit_bic): Likewise.
19973         (aarch64_emit_atomic_swap): Likewise.
19974         (aarch64_emit_atomic_load_op): Likewise.
19975         (aarch64_evpc_trn): Likewise.
19976         (aarch64_evpc_uzp): Likewise.
19977         (aarch64_evpc_zip): Likewise.
19978         (aarch64_evpc_ext): Likewise.
19979         (aarch64_evpc_rev): Likewise.
19980         (aarch64_evpc_dup): Likewise.
19981         (aarch64_gen_ccmp_first): Likewise.
19982         (aarch64_gen_ccmp_next): Likewise.
19983         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
19984         (alpha_emit_xfloating_libcall): Likewise.
19985         (emit_insxl): Likewise.
19986         (alpha_arg_type): Likewise.
19987         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
19988         (arc_preferred_simd_mode): Likewise.
19989         (arc_secondary_reload): Likewise.
19990         (get_arc_condition_code): Likewise.
19991         (arc_print_operand): Likewise.
19992         (arc_legitimate_constant_p): Likewise.
19993         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19994         * config/arc/arc.md (casesi_load): Likewise.
19995         (casesi_compact_jump): Likewise.
19996         * config/arc/predicates.md (proper_comparison_operator): Likewise.
19997         (cc_use_register): Likewise.
19998         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19999         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
20000         (arm_init_iwmmxt_builtins): Likewise.
20001         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
20002         (neon_expand_vector_init): Likewise.
20003         (arm_attr_length_move_neon): Likewise.
20004         (maybe_get_arm_condition_code): Likewise.
20005         (arm_emit_vector_const): Likewise.
20006         (arm_preferred_simd_mode): Likewise.
20007         (arm_output_iwmmxt_tinsr): Likewise.
20008         (thumb1_output_casesi): Likewise.
20009         (thumb2_output_casesi): Likewise.
20010         (arm_emit_load_exclusive): Likewise.
20011         (arm_emit_store_exclusive): Likewise.
20012         (arm_expand_compare_and_swap): Likewise.
20013         (arm_evpc_neon_vuzp): Likewise.
20014         (arm_evpc_neon_vzip): Likewise.
20015         (arm_evpc_neon_vrev): Likewise.
20016         (arm_evpc_neon_vtrn): Likewise.
20017         (arm_evpc_neon_vext): Likewise.
20018         (arm_validize_comparison): Likewise.
20019         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
20020         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
20021         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
20022         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
20023         (c6x_preferred_simd_mode): Likewise.
20024         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20025         (epiphany_rtx_costs): Likewise.
20026         * config/epiphany/predicates.md (proper_comparison_operator):
20027         Likewise.
20028         * config/frv/frv.c (condexec_memory_operand): Likewise.
20029         (frv_emit_move): Likewise.
20030         (output_move_single): Likewise.
20031         (output_condmove_single): Likewise.
20032         (frv_hard_regno_mode_ok): Likewise.
20033         (frv_matching_accg_mode): Likewise.
20034         * config/h8300/h8300.c (split_adds_subs): Likewise.
20035         (h8300_rtx_costs): Likewise.
20036         (h8300_print_operand): Likewise.
20037         (compute_mov_length): Likewise.
20038         (output_logical_op): Likewise.
20039         (compute_logical_op_length): Likewise.
20040         (compute_logical_op_cc): Likewise.
20041         (h8300_shift_needs_scratch_p): Likewise.
20042         (output_a_shift): Likewise.
20043         (compute_a_shift_length): Likewise.
20044         (compute_a_shift_cc): Likewise.
20045         (expand_a_rotate): Likewise.
20046         (output_a_rotate): Likewise.
20047         * config/i386/i386.c (classify_argument): Likewise.
20048         (function_arg_advance_32): Likewise.
20049         (function_arg_32): Likewise.
20050         (function_arg_64): Likewise.
20051         (function_value_64): Likewise.
20052         (ix86_gimplify_va_arg): Likewise.
20053         (ix86_legitimate_constant_p): Likewise.
20054         (put_condition_code): Likewise.
20055         (split_double_mode): Likewise.
20056         (ix86_avx256_split_vector_move_misalign): Likewise.
20057         (ix86_expand_vector_logical_operator): Likewise.
20058         (ix86_split_idivmod): Likewise.
20059         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20060         (ix86_build_const_vector): Likewise.
20061         (ix86_build_signbit_mask): Likewise.
20062         (ix86_match_ccmode): Likewise.
20063         (ix86_cc_modes_compatible): Likewise.
20064         (ix86_expand_branch): Likewise.
20065         (ix86_expand_sse_cmp): Likewise.
20066         (ix86_expand_sse_movcc): Likewise.
20067         (ix86_expand_int_sse_cmp): Likewise.
20068         (ix86_expand_vec_perm_vpermi2): Likewise.
20069         (ix86_expand_vec_perm): Likewise.
20070         (ix86_expand_sse_unpack): Likewise.
20071         (ix86_expand_int_addcc): Likewise.
20072         (ix86_split_to_parts): Likewise.
20073         (ix86_vectorize_builtin_gather): Likewise.
20074         (ix86_vectorize_builtin_scatter): Likewise.
20075         (avx_vpermilp_parallel): Likewise.
20076         (inline_memory_move_cost): Likewise.
20077         (ix86_tieable_integer_mode_p): Likewise.
20078         (x86_maybe_negate_const_int): Likewise.
20079         (ix86_expand_vector_init_duplicate): Likewise.
20080         (ix86_expand_vector_init_one_nonzero): Likewise.
20081         (ix86_expand_vector_init_one_var): Likewise.
20082         (ix86_expand_vector_init_concat): Likewise.
20083         (ix86_expand_vector_init_interleave): Likewise.
20084         (ix86_expand_vector_init_general): Likewise.
20085         (ix86_expand_vector_set): Likewise.
20086         (ix86_expand_vector_extract): Likewise.
20087         (emit_reduc_half): Likewise.
20088         (ix86_emit_i387_round): Likewise.
20089         (ix86_mangle_type): Likewise.
20090         (ix86_expand_round_sse4): Likewise.
20091         (expand_vec_perm_blend): Likewise.
20092         (canonicalize_vector_int_perm): Likewise.
20093         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20094         (expand_vec_perm_1): Likewise.
20095         (expand_vec_perm_interleave3): Likewise.
20096         (expand_vec_perm_even_odd_pack): Likewise.
20097         (expand_vec_perm_even_odd_1): Likewise.
20098         (expand_vec_perm_broadcast_1): Likewise.
20099         (ix86_vectorize_vec_perm_const_ok): Likewise.
20100         (ix86_expand_vecop_qihi): Likewise.
20101         (ix86_expand_mul_widen_hilo): Likewise.
20102         (ix86_expand_sse2_abs): Likewise.
20103         (ix86_expand_pextr): Likewise.
20104         (ix86_expand_pinsr): Likewise.
20105         (ix86_preferred_simd_mode): Likewise.
20106         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20107         * config/i386/sse.md (*andnot<mode>3): Likewise.
20108         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20109         (*<code><mode>3): Likewise.
20110         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20111         (ia64_expand_atomic_op): Likewise.
20112         (ia64_arg_type): Likewise.
20113         (ia64_mode_to_int): Likewise.
20114         (ia64_scalar_mode_supported_p): Likewise.
20115         (ia64_vector_mode_supported_p): Likewise.
20116         (expand_vec_perm_broadcast): Likewise.
20117         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20118         (iq2000_function_arg_advance): Likewise.
20119         (iq2000_function_arg): Likewise.
20120         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20121         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20122         (m68k_libcall_value): Likewise.
20123         (m68k_function_value): Likewise.
20124         (sched_attr_op_type): Likewise.
20125         * config/mcore/mcore.c (mcore_output_move): Likewise.
20126         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20127         Likewise.
20128         (microblaze_function_arg): Likewise.
20129         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20130         (mips_print_operand): Likewise.
20131         (mips_mode_ok_for_mov_fmt_p): Likewise.
20132         (mips_vector_mode_supported_p): Likewise.
20133         (mips_preferred_simd_mode): Likewise.
20134         (mips_expand_vpc_loongson_even_odd): Likewise.
20135         (mips_expand_vec_unpack): Likewise.
20136         (mips_expand_vi_broadcast): Likewise.
20137         (mips_expand_vector_init): Likewise.
20138         (mips_expand_vec_reduc): Likewise.
20139         (mips_expand_msa_cmp): Likewise.
20140         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20141         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20142         (cc_flags_for_mode): Likewise.
20143         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20144         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20145         (nds32_output_casesi_pc_relative): Likewise.
20146         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20147         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20148         (nvptx_gen_unpack): Likewise.
20149         (nvptx_gen_pack): Likewise.
20150         (nvptx_gen_shuffle): Likewise.
20151         (nvptx_gen_wcast): Likewise.
20152         (nvptx_preferred_simd_mode): Likewise.
20153         * config/pa/pa.c (pa_secondary_reload): Likewise.
20154         * config/pa/predicates.md (base14_operand): Likewise.
20155         * config/powerpcspe/powerpcspe-c.c
20156         (altivec_resolve_overloaded_builtin): Likewise.
20157         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20158         Likewise.
20159         (rs6000_preferred_simd_mode): Likewise.
20160         (output_vec_const_move): Likewise.
20161         (rs6000_expand_vector_extract): Likewise.
20162         (rs6000_split_vec_extract_var): Likewise.
20163         (reg_offset_addressing_ok_p): Likewise.
20164         (rs6000_legitimate_offset_address_p): Likewise.
20165         (rs6000_legitimize_address): Likewise.
20166         (rs6000_emit_set_const): Likewise.
20167         (rs6000_const_vec): Likewise.
20168         (rs6000_emit_move): Likewise.
20169         (spe_build_register_parallel): Likewise.
20170         (rs6000_darwin64_record_arg_recurse): Likewise.
20171         (swap_selector_for_mode): Likewise.
20172         (spe_init_builtins): Likewise.
20173         (paired_init_builtins): Likewise.
20174         (altivec_init_builtins): Likewise.
20175         (do_load_for_compare): Likewise.
20176         (rs6000_generate_compare): Likewise.
20177         (rs6000_expand_float128_convert): Likewise.
20178         (emit_load_locked): Likewise.
20179         (emit_store_conditional): Likewise.
20180         (rs6000_output_function_epilogue): Likewise.
20181         (rs6000_handle_altivec_attribute): Likewise.
20182         (rs6000_function_value): Likewise.
20183         (emit_fusion_gpr_load): Likewise.
20184         (emit_fusion_p9_load): Likewise.
20185         (emit_fusion_p9_store): Likewise.
20186         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20187         (fusion_gpr_mem_load): Likewise.
20188         (fusion_addis_mem_combo_load): Likewise.
20189         (fusion_addis_mem_combo_store): Likewise.
20190         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20191         (fusion_gpr_mem_load): Likewise.
20192         (fusion_addis_mem_combo_load): Likewise.
20193         (fusion_addis_mem_combo_store): Likewise.
20194         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20195         Likewise.
20196         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20197         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20198         (rs6000_preferred_simd_mode): Likewise.
20199         (output_vec_const_move): Likewise.
20200         (rs6000_expand_vector_extract): Likewise.
20201         (rs6000_split_vec_extract_var): Likewise.
20202         (reg_offset_addressing_ok_p): Likewise.
20203         (rs6000_legitimate_offset_address_p): Likewise.
20204         (rs6000_legitimize_address): Likewise.
20205         (rs6000_emit_set_const): Likewise.
20206         (rs6000_const_vec): Likewise.
20207         (rs6000_emit_move): Likewise.
20208         (rs6000_darwin64_record_arg_recurse): Likewise.
20209         (swap_selector_for_mode): Likewise.
20210         (paired_init_builtins): Likewise.
20211         (altivec_init_builtins): Likewise.
20212         (rs6000_expand_float128_convert): Likewise.
20213         (emit_load_locked): Likewise.
20214         (emit_store_conditional): Likewise.
20215         (rs6000_output_function_epilogue): Likewise.
20216         (rs6000_handle_altivec_attribute): Likewise.
20217         (rs6000_function_value): Likewise.
20218         (emit_fusion_gpr_load): Likewise.
20219         (emit_fusion_p9_load): Likewise.
20220         (emit_fusion_p9_store): Likewise.
20221         * config/rx/rx.c (rx_gen_move_template): Likewise.
20222         (flags_from_mode): Likewise.
20223         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20224         (s390_slb_comparison): Likewise.
20225         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20226         (s390_vector_mode_supported_p): Likewise.
20227         (s390_cc_modes_compatible): Likewise.
20228         (s390_match_ccmode_set): Likewise.
20229         (s390_canonicalize_comparison): Likewise.
20230         (s390_emit_compare_and_swap): Likewise.
20231         (s390_branch_condition_mask): Likewise.
20232         (s390_rtx_costs): Likewise.
20233         (s390_secondary_reload): Likewise.
20234         (__SECONDARY_RELOAD_CASE): Likewise.
20235         (s390_expand_cs): Likewise.
20236         (s390_preferred_simd_mode): Likewise.
20237         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20238         * config/sh/sh.c (sh_print_operand): Likewise.
20239         (dump_table): Likewise.
20240         (sh_secondary_reload): Likewise.
20241         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20242         * config/sh/sh.md (casesi_worker_1): Likewise.
20243         (casesi_worker_2): Likewise.
20244         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20245         (fcc_comparison_operator): Likewise.
20246         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20247         (emit_soft_tfmode_cvt): Likewise.
20248         (sparc_preferred_simd_mode): Likewise.
20249         (output_cbranch): Likewise.
20250         (sparc_print_operand): Likewise.
20251         (sparc_expand_vec_perm_bmask): Likewise.
20252         (vector_init_bshuffle): Likewise.
20253         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20254         (spu_vector_mode_supported_p): Likewise.
20255         (spu_expand_insv): Likewise.
20256         (spu_emit_branch_or_set): Likewise.
20257         (spu_handle_vector_attribute): Likewise.
20258         (spu_builtin_splats): Likewise.
20259         (spu_builtin_extract): Likewise.
20260         (spu_builtin_promote): Likewise.
20261         (spu_expand_sign_extend): Likewise.
20262         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20263         (tilegx_simd_int): Likewise.
20264         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20265         (tilepro_simd_int): Likewise.
20266         * config/v850/v850.c (const_double_split): Likewise.
20267         (v850_print_operand): Likewise.
20268         (ep_memory_offset): Likewise.
20269         * config/vax/vax.c (vax_rtx_costs): Likewise.
20270         (vax_output_int_move): Likewise.
20271         (vax_output_int_add): Likewise.
20272         (vax_output_int_subtract): Likewise.
20273         * config/visium/predicates.md (visium_branch_operator): Likewise.
20274         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20275         (visium_print_operand_address): Likewise.
20276         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20277         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20278         (xtensa_expand_conditional_branch): Likewise.
20279         (xtensa_copy_incoming_a7): Likewise.
20280         (xtensa_output_literal): Likewise.
20281         * dfp.c (decimal_real_maxval): Likewise.
20282         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20284 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20285             Alan Hayward  <alan.hayward@arm.com>
20286             David Sherwood  <david.sherwood@arm.com>
20288         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20289         (mode_nunits_inline): Likewise.
20290         (mode_inner_inline): Likewise.
20291         (mode_unit_size_inline): Likewise.
20292         (mode_unit_precision_inline): Likewise.
20293         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20294         unprefixed name.
20295         (emit_mode_wider): Add an E_ prefix to mode names.
20296         (emit_mode_complex): Likewise.
20297         (emit_mode_inner): Likewise.
20298         (emit_mode_adjustments): Likewise.
20299         (emit_mode_int_n): Likewise.
20300         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20301         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20302         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20303         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20304         (CRC32_BUILTIN, ENTRY): Likewise.
20305         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20306         (aarch64_pop_regs): Likewise.
20307         (aarch64_process_components): Likewise.
20308         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20309         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20310         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20311         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20312         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20313         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20314         Likewise.
20315         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20316         (pcmpistr): Likewise.
20317         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20318         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20319         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20320         Likewise.
20321         * config/rl78/rl78.c (mduc_regs): Likewise.
20322         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20323         (htm_expand_builtin): Likewise.
20324         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20325         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20326         * config/xtensa/xtensa.c (print_operand): Likewise.
20327         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20328         (NUM_MODE_VECTOR_INT): Likewise.
20329         * genoutput.c (null_operand): Likewise.
20330         (output_operand_data): Likewise.
20331         * genrecog.c (print_parameter_value): Likewise.
20332         * lra.c (debug_operand_data): Likewise.
20334 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20336         * dwarf2out.c (scompare_loc_descriptor_wide)
20337         (scompare_loc_descriptor_narrow): New functions, split out from...
20338         (scompare_loc_descriptor): ...here.
20339         * expmed.c (emit_store_flag_int): New function, split out from...
20340         (emit_store_flag): ...here.
20342 2017-08-30  Richard Biener  <rguenther@suse.de>
20344         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20345         (dwarf2out_early_finish): Move setting of AT_pubnames from
20346         early debug output to early finish.
20348 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20350         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20351         and -mdata-region to the assembler.
20353 2017-08-30  Richard Biener  <rguenther@suse.de>
20355         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20356         attributes.
20357         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20358         (dwarf2out_early_global_decl): Only generate a DIE for the
20359         abstract origin if it doesn't already exist or is a declaration DIE.
20360         (resolve_addr): Do not add the linkage name twice when
20361         generating a stub DIE for the DW_TAG_GNU_call_site target.
20363 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20365         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20366         Use machine_mode rather than int for arg1_mode.
20368 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20370         PR target/82015
20371         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20372         that the second argument of the built-in functions to unpack
20373         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20374         switch statement instead a lot of if statements.
20375         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20376         Allow 64-bit values to be in Altivec registers as well as
20377         traditional floating point registers.
20378         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20380 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20382         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20383         MAX_REGS_PER_ADDRESS == 1.
20385 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20387         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20388         * config/i386/i386.c (ix86_option_override_internal): Simplify
20389         setting of opts->x_flag_entry.
20391 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20392             Jakub Jelinek  <jakub@redhat.com>
20393             Richard Biener  <rguenther@suse.de>
20395         PR tree-optimization/81503
20396         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20397         folded constant fits in the target type; reorder tests for clarity.
20399 2017-08-29  Martin Liska  <mliska@suse.cz>
20401         * passes.def: Include pass_lower_switch.
20402         * stmt.c (dump_case_nodes): Remove and move to
20403         tree-switch-conversion.
20404         (case_values_threshold): Likewise.
20405         (expand_switch_as_decision_tree_p): Likewise.
20406         (emit_case_decision_tree): Likewise.
20407         (expand_case): Likewise.
20408         (balance_case_nodes): Likewise.
20409         (node_has_low_bound): Likewise.
20410         (node_has_high_bound): Likewise.
20411         (node_is_bounded): Likewise.
20412         (emit_case_nodes): Likewise.
20413         (struct simple_case_node): New struct.
20414         (add_case_node): Remove.
20415         (emit_case_dispatch_table): Use vector instead of case_list.
20416         (reset_out_edges_aux): Remove.
20417         (compute_cases_per_edge): Likewise.
20418         (expand_case): Build list of simple_case_node.
20419         (expand_sjlj_dispatch_table): Use it.
20420         * tree-switch-conversion.c (struct case_node): Moved from
20421         stmt.c and adjusted.
20422         (emit_case_nodes): Likewise.
20423         (node_has_low_bound): Likewise.
20424         (node_has_high_bound): Likewise.
20425         (node_is_bounded): Likewise.
20426         (case_values_threshold): Likewise.
20427         (reset_out_edges_aux): Likewise.
20428         (compute_cases_per_edge): Likewise.
20429         (add_case_node): Likewise.
20430         (dump_case_nodes): Likewise.
20431         (balance_case_nodes): Likewise.
20432         (expand_switch_as_decision_tree_p): Likewise.
20433         (emit_jump): Likewise.
20434         (emit_case_decision_tree): Likewise.
20435         (try_switch_expansion): Likewise.
20436         (do_jump_if_equal): Likewise.
20437         (emit_cmp_and_jump_insns): Likewise.
20438         (fix_phi_operands_for_edge): New function.
20439         (record_phi_operand_mapping): Likewise.
20440         (class pass_lower_switch): New pass.
20441         (pass_lower_switch::execute): New function.
20442         (make_pass_lower_switch): Likewise.
20443         (conditional_probability):
20444         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20445         * tree-pass.h: Add make_pass_lower_switch.
20447 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20449         PR target/80993
20450         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20451         handlers as used.
20453 2017-08-29  Richard Biener  <rguenther@suse.de>
20455         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20456         we do not add a DW_AT_inline attribute twice.
20457         (gen_subprogram_die): Remove code setting DW_AT_inline on
20458         DECL_ABSTRACT_P nodes.
20460 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20462         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20463         calls to internal functions.
20464         (gimplify_modify_expr): Likewise.
20465         * tree-call-cdce.c (use_internal_fn): Likewise.
20466         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20467         (convert_to_divmod): Set the nothrow flag.
20468         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20469         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20470         (vectorizable_call): Likewise.
20471         (vectorizable_store): Likewise.
20472         (vectorizable_load): Likewise.
20473         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20474         (vect_recog_mask_conversion_pattern): Likewise.
20476 2017-08-29  Martin Liska  <mliska@suse.cz>
20478         PR other/39851
20479         * gcc.c (driver_handle_option): Add new argument.
20480         * opts-common.c (handle_option): Pass
20481         target_option_override_hook.
20482         * opts-global.c (lang_handle_option): Add new option.
20483         (set_default_handlers):  Add new argument.
20484         (decode_options): Likewise.
20485         * opts.c (target_handle_option): Likewise.
20486         (common_handle_option): Call target_option_override_hook.
20487         * opts.h (struct cl_option_handler_func): Add hook for
20488         target option override.
20489         (struct cl_option_handlers): Likewise.
20490         (set_default_handlers): Add new argument.
20491         (decode_options): Likewise.
20492         (common_handle_option): Likewise.
20493         (target_handle_option): Likewise.
20494         * toplev.c (toplev::main): Pass targetm.target_option.override
20495         hook.
20497 2017-08-29  Richard Biener  <rguenther@suse.de>
20498         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20500         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20501         life to the active subtree.
20503 2017-08-28  Jeff Law  <law@redhat.com>
20505         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20506         derive_equivalences.
20507         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20508         Code moved into....
20509         (edge_info::derive_equivalences): New private member function
20511         * tree-ssa-dom.c (class edge_info): Changed from a struct
20512         to a class.  Add ctor/dtor, methods and data members.
20513         (edge_info::edge_info): Renamed from allocate_edge_info.
20514         Initialize additional members.
20515         (edge_info::~edge_info): New.
20516         (free_dom_edge_info): Delete the edge info.
20517         (record_edge_info): Use new class & associated member functions.
20518         Tighten forms for testing for edge equivalences.
20519         (record_temporary_equivalences): Iterate over the simple
20520         equivalences rather than assuming there's only one per edge.
20521         (cprop_into_successor_phis): Iterate over the simple
20522         equivalences rather than assuming there's only one per edge.
20523         (optimize_stmt): Use operand_equal_p rather than pointer
20524         equality for mini-DSE code.
20526 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20528         * gcc.c (execute): Fold SIGPIPE handling into switch
20529         statement.  Adjust internal error message.
20531 2017-08-28  Richard Biener  <rguenther@suse.de>
20533         PR debug/81993
20534         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20535         Do nothing for removed DIEs.
20537 2017-08-28  Richard Biener  <rguenther@suse.de>
20539         PR tree-optimization/81977
20540         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20541         memcpy.
20543 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20545         PR target/80640
20546         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20547         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20548         using targetm.gen_mem_thread_fence.
20550 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20552         PR target/81995
20553         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20554         predicate to register_operand.  Reorder operands.
20555         (*btr<mode>): Ditto.
20556         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20557         (*btr<mode>_mask): Ditto.
20559 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20561         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20562         * config/rs6000/xmmintrin.h: New file.
20563         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20565 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20567         PR target/81504
20568         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20569         parameter and_insn and return it.
20570         (recombine_lvx_pattern): Insert a copy to ensure availability of
20571         the base register of the copied masking operation at the point of
20572         the instruction replacement.
20573         (recombine_stvx_pattern): Likewise.
20575 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20577         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20578         undocumented switches.
20579         (-mpower9-dform-vector): Likewise.
20580         (-mpower9-dform): Likewise.
20581         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20582         comments to delete references to -mpower9-dform* switches.
20583         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20584         Delete reference to -mpower9-dform* switches, test for
20585         -mpower9-vector instead.
20586         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20587         (OTHER_P9_VECTOR_MASKS): Likewise.
20588         (POWERPC_MASKS): Likewise.
20589         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20590         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20591         that checked for -mpower9-dform* consistancy with other options.
20592         Add test for -mpower9-misc to enable other power9 switches.
20593         (rs6000_init_hard_regno_mode_ok): Likewise.
20594         (rs6000_option_override_internal): Likewise.
20595         (rs6000_emit_prologue): Likewise.
20596         (rs6000_emit_epilogue): Likewise.
20597         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20598         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20599         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20600         -mpower9-vector.
20601         (emit_fusion_p9_store): Likewise.
20602         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20603         resetting these macros if the assembler does not support ISA 3.0
20604         instructions.
20605         (TARGET_P9_DFORM_VECTOR): Likewise.
20606         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20607         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20609 2017-08-25  Alan Modra  <amodra@gmail.com>
20611         PR rtl-optimization/81747
20612         * cse.c (cse_extended_basic_block): Don't attempt to record
20613         equivalences for degenerate conditional branches that branch
20614         to their fall-through.
20616 2017-08-24  Martin Sebor  <msebor@redhat.com>
20618         PR middle-end/81908
20619         * gimple-fold.c (size_must_be_zero_p): New function.
20620         (gimple_fold_builtin_memory_op): Call it.
20622 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20624         * config/rs6000/mm_malloc.h: New file.
20626 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20628         PR tree-optimization/81913
20629         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20630         analysis when either IVs in condition can wrap.
20632 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20634         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20635         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20637 2017-08-24  Richard Biener  <rguenther@suse.de>
20639         PR target/81921
20640         * targhooks.c (default_target_can_inline_p): Properly
20641         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20642         is present and always compare.
20643         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20644         infer -mfpmath=sse from TARGET_SSE_P.
20645         (ix86_can_inline_p): Properly use target_option_default_node when
20646         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20648 2017-08-24  Richard Biener  <rguenther@suse.de>
20650         PR debug/81936
20651         * dwarf2out.c (output_die): Handle flag_generate_offload like
20652         flag_generate_lto.
20653         (output_comp_unit): Likewise.
20654         (gen_array_type_die): Likewise.
20655         (dwarf2out_early_finish): Likewise.
20656         (note_variable_value_in_expr): Likewise.
20657         (dwarf2out_finish): Likewise.  Adjust assert.
20658         * cgraphunit.c (symbol_table::compile): Move setting of
20659         flag_generate_offload earlier ...
20660         (symbol_table::finalize_compilation_unit): ... here, before
20661         early debug finalization.
20663 2017-08-24  Richard Biener  <rguenther@suse.de>
20665         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20666         and ipa-fnsummary.h.
20667         (ix86_can_inline_p): When ix86_fpmath flags do not match
20668         check whether the callee uses FP math at all.
20670 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20672         PR middle-end/81931
20673         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20674         instead of TYPE_PRECISION.
20676 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20677             Alan Hayward  <alan.hayward@arm.com>
20678             David Sherwood  <david.sherwood@arm.com>
20680         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20682 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20684         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20685         Remove field.
20686         (ix86_frame::stack_realign_allocate): New field.
20687         (struct machine_frame_state): Modify comments.
20688         (machine_frame_state::sp_realigned_fp_end): New field.
20689         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20690         layout calculation.
20691         (sp_valid_at): Add assertion to assure no attempt to access invalid
20692         offset of a realigned stack.
20693         (fp_valid_at): Likewise.
20694         (choose_baseaddr): Modify comments.
20695         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20696         ix86_expand_prologue.
20697         (ix86_expand_prologue): Modify stack realignment and allocation.
20698         (ix86_expand_epilogue): Modify comments.
20699         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20700         avx2_runtime, avx512f, and avx512f_runtime.
20702 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20704         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20705         (mstackrealign): Do not init to -1.
20706         * config/i386/i386.c (ix86_option_override_internal):
20707         Check opts_set, not opts when setting default value of
20708         opts->x_ix86_force_align_arg_pointer.
20710 2017-08-23  Richard Biener  <rguenther@suse.de>
20712         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20713         lang_hooks.decl_printable_name.
20714         * print-rtl-function.c (print_rtx_function): Likewise.
20715         * tree-pretty-print.c (dump_function_header): Likewise.
20717 2017-08-23  Richard Biener  <rguenther@suse.de>
20719         PR lto/81940
20720         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20721         -g0 at compile-time.
20723 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20725         PR middle-end/19706
20726         * doc/sourcebuild.texi (Other hardware attributes):
20727         Document xorsign.
20729 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20731         PR middle-end/19706
20732         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20733         Fix single-use check.
20735 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20737         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20739 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20741         * doc/install.texi: Modify to add more details on running selected
20742         tests.
20744 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20746         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20747         is combined with -mabi=ms.
20748         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20749         ms_abi.
20751 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20753         PR tree-optimization/81488
20754         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20755         and cached_basis fields.
20756         (MAX_SPREAD): New constant.
20757         (alloc_cand_and_find_basis): Initialize new fields.
20758         (clear_visited): New function.
20759         (create_phi_basis_1): Rename from create_phi_basis, set visited
20760         and cached_basis fields.
20761         (create_phi_basis): New wrapper function.
20762         (phi_add_costs_1): Rename from phi_add_costs, add spread
20763         parameter, set visited field, short-circuit when limits reached.
20764         (phi_add_costs): New wrapper function.
20765         (record_phi_increments_1): Rename from record_phi_increments, set
20766         visited field.
20767         (record_phi_increments): New wrapper function.
20768         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20769         (phi_incr_cost): New wrapper function.
20770         (all_phi_incrs_profitable_1): Rename from
20771         all_phi_incrs_profitable, set visited field.
20772         (all_phi_incrs_profitable): New wrapper function.
20774 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20775             Alan Hayward  <alan.hayward@arm.com>
20776             David Sherwood  <david.sherwood@arm.com>
20778         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20779         that takes the outer and inner modes.
20780         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20781         comparison as the canonical test for a paradoxical subreg.
20782         * combine.c (simplify_set): Use paradoxical_subreg_p.
20783         (make_extraction): Likewise.
20784         (force_to_mode): Likewise.
20785         (rtx_equal_for_field_assignment_p): Likewise.
20786         (gen_lowpart_for_combine): Likewise.
20787         (simplify_comparison): Likewise.
20788         * cse.c (equiv_constant): Likewise.
20789         * expmed.c (store_bit_field_1): Likewise.
20790         * final.c (alter_subreg): Likewise.
20791         * fwprop.c (propagate_rtx): Likewise.
20792         (forward_propagate_subreg): Likewise.
20793         * ira-conflicts.c (ira_build_conflicts): Likewise.
20794         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20795         * lra-constraints.c (curr_insn_transform): Likewise.
20796         (split_reg): Likewise.
20797         * lra-eliminations.c (move_plus_up): Likewise.
20798         (lra_eliminate_regs_1): Likewise.
20799         * recog.c (general_operand): Likewise.
20800         * ree.c (combine_reaching_defs): Likewise.
20801         * reload.c (push_reload): Likewise.
20802         (find_reloads): Likewise.
20803         * reload1.c (elimination_effects): Likewise.
20804         (compute_reload_subreg_offset): Likewise.
20805         (choose_reload_regs): Likewise.
20806         * rtlanal.c (subreg_lsb_1): Likewise.
20807         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20808         (simplify_subreg): Likewise.
20809         * var-tracking.c (track_loc_p): Likewise.
20810         * emit-rtl.c (byte_lowpart_offset): Likewise.
20811         (paradoxical_subreg_p): Delete out-of-line definition.
20813 2017-08-22  Jeff Law  <law@redhat.com>
20815         PR tree-optimization/81741
20816         PR tree-optimization/71947
20817         * tree-ssa-dom.c: Include tree-inline.h.
20818         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20819         equivalences if one is more expensive to compute than the other.
20820         * tree-ssa-scopedtables.h (class const_or_copies): Make
20821         record_const_or_copy_raw method private.
20822         (class avail_exprs_stack): New method simplify_binary_operation.
20823         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20824         avail_exprs_stack::simplify_binary_operation as needed.
20825         (avail_exprs_stack::simplify_binary_operation): New function.
20827 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20829         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20830         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20831         (DOT_SYMBOLS): Likewise.
20832         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20833         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20834         (RS6000_ABI_NAME): Likewise.
20835         (TARGET_CMODEL): Likewise.
20836         (TOC_SECTION_ASM_OP): Likewise.
20837         (SET_CMODEL): New macro.
20838         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20840 2017-08-22  Richard Biener  <rguenther@suse.de>
20842         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20843         to allow for free-lang-data replacements similar to verify_type_variant.
20845 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20846             Alan Hayward  <alan.hayward@arm.com>
20847             David Sherwood  <david.sherwood@arm.com>
20849         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20850         VOIDmode for the LABEL_REF.
20852 2017-08-22  Richard Biener  <rguenther@suse.de>
20854         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20856 2017-08-22  Richard Biener  <rguenther@suse.de>
20858         * common.opt (feliminate-dwarf2-dups): Ignore.
20859         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20860         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20861         same_die_p_wrap, compute_section_prefix,
20862         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20863         (comdat_symbol_id, comdat_symbol_number): Likewise.
20864         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20865         Likewise.
20866         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20867         (output_die): Mark unreachable path unreachable.
20868         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20869         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20870         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20871         (dwarf2out_early_finish): Likewise.
20873 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20875         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20877 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20879         PR target/81910
20880         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20881         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20882         (avr_attribute_table) <io, io_low, address>: Initialize
20883         .decl_required with true.
20885 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20887         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20888         undocumented debugging options.
20889         (-mvsx-scalar-double): Likewise.
20890         (-mallow-df-permute): Likewise.
20891         (-mvectorize-builtins): Likewise.
20892         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20893         (rs6000_builtin_vectorized_function): Likewise.
20894         (rs6000_builtin_md_vectorized_function): Likewise.
20895         (rs6000_opt_vars): Likewise.
20897 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20899         PR target/46091
20900         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20901         (*btrq_imm): Rename from *btrq.
20902         (*btcq_imm): Rename from *btcq.
20903         (btsc): New code attribute.
20904         (*<btsc><mode>): New insn pattern.
20905         (*btr<mode>): Ditto.
20906         (*<btsc><mode>_mask): New insn_and_split pattern.
20907         (*btr<mode>_mask): Ditto.
20909 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20910             Alan Hayward  <alan.hayward@arm.com>
20911             David Sherwood  <david.sherwood@arm.com>
20913         * function.c (pad_below): Simplify padding calculation.
20915 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20916             Alan Hayward  <alan.hayward@arm.com>
20917             David Sherwood  <david.sherwood@arm.com>
20919         * target.def (function_prologue): Remove frame size argument.
20920         (function_epilogue): Likewise.
20921         * doc/tm.texi: Regenerate.
20922         * final.c (final_start_function): Update call to function_prologue.
20923         (final_end_function): Update call to function_epilogue.
20924         (default_function_pro_epilogue): Remove frame size argument.
20925         * output.h (default_function_pro_epilogue): Likewise.
20926         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20927         (arm_output_function_prologue): Likewise.
20928         * config/frv/frv.c (frv_function_prologue): Likewise.
20929         (frv_function_epilogue): Likewise.
20930         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20931         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20932         (ia64_output_function_epilogue): Likewise.
20933         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20934         (m32r_output_function_epilogue): Likewise.
20935         * config/microblaze/microblaze.c (microblaze_function_prologue)
20936         (microblaze_function_epilogue): Likewise.
20937         * config/mips/mips.c (mips_output_function_prologue): Likewise.
20938         (mips_output_function_epilogue): Likewise.
20939         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20940         (mmix_target_asm_function_epilogue): Likewise.
20941         * config/msp430/msp430.c (msp430_start_function): Likewise.
20942         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
20943         (nds32_asm_function_epilogue): Likewise.
20944         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
20945         * config/pa/pa.c (pa_output_function_prologue): Likewise.
20946         (pa_output_function_epilogue): Likewise.
20947         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
20948         (rs6000_output_function_epilogue): Likewise.
20949         * config/rl78/rl78.c (rl78_start_function): Likewise.
20950         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
20951         (rs6000_output_function_epilogue): Likewise.
20952         * config/rx/rx.c (rx_output_function_prologue): Likewise.
20953         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
20954         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
20955         (sparc_asm_function_epilogue): Likewise.
20957 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20959         * tree.h (type_has_mode_precision_p): New function.
20960         * convert.c (convert_to_integer_1): Use it.
20961         * expr.c (expand_expr_real_2): Likewise.
20962         (expand_expr_real_1): Likewise.
20963         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
20964         * match.pd: Likewise.
20965         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
20966         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
20967         * tree-tailcall.c (process_assignment): Likewise.
20968         * tree-vect-loop.c (vectorizable_reduction): Likewise.
20969         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
20970         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
20971         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20972         (vectorizable_assignment): Likewise.
20973         (vectorizable_shift): Likewise.
20974         (vectorizable_operation): Likewise.
20975         * tree-vrp.c (register_edge_assert_for_2): Likewise.
20977 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
20979         * match.pd: Add pow (C, x) simplification.
20981 2017-08-21  Richard Biener  <rguenther@suse.de>
20983         PR tree-optimization/81900
20984         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
20985         for blocks with abnormal predecessors.
20986         (compute_antic): Do not set visited flag prematurely.
20988 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
20990         PR target/79883
20991         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
20993 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20995         * stor-layout.h (vector_type_mode): Move to...
20996         * tree.h (vector_type_mode): ...here.
20997         * stor-layout.c (vector_type_mode): Move to...
20998         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
21000 2017-08-21  Richard Biener  <rguenther@suse.de>
21002         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
21003         register_external_die hooks.
21004         (debug_false_tree_charstarstar_uhwistar): Declare.
21005         (debug_nothing_tree_charstar_uhwi): Likewise.
21006         * debug.c (do_nothing_debug_hooks): Adjust.
21007         (debug_false_tree_charstarstar_uhwistar): New do nothing.
21008         (debug_nothing_tree_charstar_uhwi): Likewise.
21009         * dbxout.c (dbx_debug_hooks): Adjust.
21010         (xcoff_debug_hooks): Likewise.
21011         * sdbout.c (sdb_debug_hooks): Likewise.
21012         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21013         * dwarf2out.c (macinfo_label_base): New global.
21014         (dwarf2out_register_external_die): New function for the
21015         register_external_die hook.
21016         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
21017         (dwarf2_debug_hooks): Use them.
21018         (dwarf2_lineno_debug_hooks): Adjust.
21019         (struct die_struct): Add with_offset flag.
21020         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
21021         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
21022         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
21023         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21024         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21025         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21026         defining section names for the early LTO debug variants.
21027         (reset_indirect_string): New helper.
21028         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21029         (print_dw_val): Add support for offsetted symbol references.
21030         (get_ultimate_context): Split out from is_cxx.
21031         (is_cxx): Use get_ultimate_context.
21032         (is_fortran): Add decl overload.
21033         (compute_comp_unit_symbol): Split out worker from
21034         compute_section_prefix.
21035         (compute_section_prefix): Call compute_comp_unit_symbol and
21036         set comdat_type_p here.
21037         (output_die): Skip DIE symbol output for the LTO added one.
21038         Handle DIE symbol references with offset.
21039         (output_comp_unit): Guard section name mangling properly.
21040         For LTO debug sections emit a symbol at the section beginning
21041         which we use to refer to its DIEs.
21042         (add_abstract_origin_attribute): For DIEs registered via
21043         dwarf2out_register_external_die directly refer to the early
21044         DIE rather than indirectly through the shadow one we created.
21045         Remove obsolete call to dwarf2out_abstract_function for
21046         non-function/block origins.
21047         (gen_array_type_die): When generating early LTO debug do
21048         not emit DW_AT_string_length.
21049         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21050         late when in LTO.  As suggested place a gcc_unreachable for
21051         the DECL_ABSTRACT_P case.
21052         (gen_subprogram_die): Avoid another specification DIE
21053         for early built declarations/definitions for the late LTO case.
21054         (gen_variable_die): Add type references for late duplicated VLA dies
21055         when in late LTO.
21056         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21057         we have the abstract instance already.
21058         (process_scope_var): Adjust decl DIE contexts in LTO which
21059         first puts them in limbo.
21060         (gen_decl_die): Do not generate type DIEs late apart from
21061         types for VLAs or for decls we do not yet have a DIE.  Do not
21062         call dwarf2out_abstract_function late.
21063         (dwarf2out_early_global_decl): Make sure to create DIEs
21064         for abstract instances of a decl first.
21065         (dwarf2out_late_global_decl): Adjust comment.
21066         (output_macinfo_op): With multiple macro sections use
21067         macinfo_label_base to distinguish labels.
21068         (output_macinfo): Likewise.  Update macinfo_label_base.
21069         Pass in the line info label.
21070         (note_variable_value_in_expr): When generating LTO resolve
21071         all variable values here by generating DIEs as needed.
21072         (init_sections_and_labels): Add early LTO debug flag parameter
21073         and generate different sections and names if set.  Add generation
21074         counter for the labels so we can have multiple of them.
21075         (reset_dies): Helper to allow DIEs to be output multiple times.
21076         (dwarf2out_finish): When outputting DIEs to the fat part of an
21077         LTO object first reset DIEs.
21078         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21079         (modified_type_die): Check for decl_ultimate_origin being self
21080         before recursing.
21081         (gen_type_die_with_usage): Likewise.
21082         (gen_typedef_die): Allow decl_ultimate_origin being self.
21083         (set_decl_abstract_flags): Remove.
21084         (set_block_abstract_flags): Likewise.
21085         (dwarf2out_abstract_function): Treat the early generated DIEs
21086         as the abstract copy and only add DW_AT_inline and
21087         DW_AT_artificial here and call set_decl_origin_self.
21088         If the DIE has an abstract origin don't do anything.
21089         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21090         if we have none yet (Go fails to build one, PR78628).
21091         (variably_modified_type_p): Prevent endless recursion for Ada
21092         cyclic pointer types.
21093         * lto-streamer-in.c: Include debug.h.
21094         (dref_queue): New global.
21095         (lto_read_tree_1): Stream in DIE references.
21096         (lto_input_tree): Register DIE references.
21097         (input_function): Stream DECL_DEBUG_ARGS.
21098         * lto-streamer-out.c: Include debug.h.
21099         (lto_write_tree_1): Output DIE references.
21100         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21101         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21102         (output_function): Stream DECL_DEBUG_ARGS.
21103         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21104         Stream DECL_ABSTRACT_ORIGIN.
21105         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21106         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21107         DECL_CONTEXT for file-scope decls.
21108         * lto-streamer.h (struct dref_entry): Declare.
21109         (dref_queue): Likewise.
21110         * cfgexpand.c (pass_expand::execute): Do not call the
21111         outlining_inline_function hook here.
21112         * lto-wrapper.c (debug_obj): New global.
21113         (tool_cleanup): Unlink it if required.
21114         (debug_objcopy): New function.
21115         (run_gcc): Handle early debug sections in the IL files by
21116         extracting them to separate files, partially linkin them and
21117         feeding the result back as result to the linker.
21118         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21119         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21120         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21121         sections into a separate segment.
21122         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21123         segments.
21124         (darwin_asm_dwarf_section): Likewise.
21125         (darwin_asm_output_dwarf_offset): Likewise.
21126         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21128 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21129             Alan Hayward  <alan.hayward@arm.com>
21130             David Sherwood  <david.sherwood@arm.com>
21132         * read-md.h (md_reader::record_potential_iterator_use): Replace
21133         pointer argument with an rtx and an index.
21134         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21135         (apply_mode_iterator): Likewise.
21136         (apply_code_iterator): Likewise.
21137         (apply_int_iterator): Likewise.
21138         (apply_subst_iterator): Likewise.
21139         (record_iterator_use): Likewise.
21140         (record_attribute_use): Likewise.
21141         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21142         to record_iterator_use and apply_iterator.
21143         (iterator_use): Replace ptr with x and index.
21144         (attribute_use): Likewise.
21145         (apply_attribute_uses): Update calls to apply_iterator.
21146         (apply_iterators): Likewise.  Update initialization of iterator_use.
21147         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21148         and record_potential_iterator_use.
21149         (rtx_reader::read_rtx_operand): Likewise.
21151 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21152             Alan Hayward  <alan.hayward@arm.com>
21153             David Sherwood  <david.sherwood@arm.com>
21155         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21156         CONST_WIDE_INT.
21158 2017-08-21  Richard Biener  <rguenther@suse.de>
21160         PR middle-end/81884
21161         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21162         at struct end conservatively when comparing common bases.
21164 2017-08-21  Richard Biener  <rguenther@suse.de>
21166         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21167         (mem_ref_in_stmt): Remove.
21168         (determine_max_movement): Use ref index to get at the reference.
21169         (invariantness_dom_walker::before_dom_children): Deal with
21170         lim data already initialized.
21171         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21173 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21175         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21176         (TARGET_ISA_ROUND): Ditto.
21177         (TARGET_ROUND): Ditto.
21178         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21179         * config/i386/i386.md: Ditto.
21180         * config/i386/sse.md: Ditto.
21181         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21182         with OPTION_MASK_ISA_SSE4_1.
21184 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21186         PR target/81894
21187         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21188         __builtin_ia32_lzcnt_u16.
21190 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21192         PR target/80210
21193         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21194         (rs6000_set_current_function): Rewrite function to use it.
21196 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21198         PR c/53037
21199         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21200         and TYPE_WARN_IF_NOT_ALIGN.
21201         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21202         (handle_warn_if_not_align): New.
21203         (place_union_field): Call handle_warn_if_not_align.
21204         (place_field): Call handle_warn_if_not_align.
21205         Copy TYPE_WARN_IF_NOT_ALIGN.
21206         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21207         (layout_type): Likewise.
21208         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21209         spare to 18.
21210         (tree_decl_common): Add warn_if_not_align.
21211         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21212         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21213         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21214         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21215         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21216         * doc/extend.texi: Document warn_if_not_aligned attribute.
21217         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21219 2017-08-17  Martin Liska  <mliska@suse.cz>
21221         PR bootstrap/81864
21222         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21223         (get_data_dependence): Use it as pointer type.
21224         (distribute_loop): Likewise.
21226 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21228         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21229         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21230         (p8_vmrgow_<mode>_direct): New define_insn.
21231         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21232         handle endianness for vmrgew and vmrgow permute patterns.
21234 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21236         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21237         * config/rs6000/rs6000-cpus.def: Remove comment.
21238         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21239         (POWERPC_MASKS): Likewise.
21240         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21241         use of TARGET_VSX_TIMODE.
21242         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21243         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21244         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21245         (rs6000_option_override_internal): Remove dead code.
21246         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21247         (rs6000_legitimize_reload_address): Likewise.
21248         (rs6000_legitimate_address_p): Likewise.
21249         (rs6000_opt_masks): Delete "vsx-timode".
21250         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21251         from function comment.
21252         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21253         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21254         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21255         condition.
21256         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21257         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21258         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21259         (VSX_TI): Likewise.
21260         (VSX_M): Likewise.
21261         (define_peephole2): Likewise.
21263 2017-08-17  Martin Sebor  <msebor@redhat.com>
21265         PR c/81859
21266         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21267         past the end of an array.
21268         (test_pp_format): Add test cases.
21270 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21272         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21273         missing ECF_NOTHROW flags.
21275 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21277         PR target/72804
21278         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21279         operands residing in integer registers.
21280         (*vsx_le_perm_load_<mode>): Likewise.
21281         (*vsx_le_perm_store_<mode>): Likewise.
21282         (define_peephole2): Add peepholes to optimize the above.
21284 2017-08-17  Marek Polacek  <polacek@redhat.com>
21286         PR middle-end/81814
21287         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21288         to mimic what shorten_compare did.  Change the return type to bool.
21289         (fold_cond_expr_with_comparison): Update call to
21290         operand_equal_for_comparison_p.
21291         (fold_ternary_loc): Likewise.
21293 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21295         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21296         register.
21297         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21299 2017-08-17  Richard Biener  <rguenther@suse.de>
21301         * tree-ssa-structalias.c (solve_graph): When propagating
21302         to successors update the graphs succ edges and avoid duplicate work.
21304 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21306         PR target/81861
21307         * config/i386/i386.c (ix86_option_override_internal): Save target
21308         specific options after ix86_stack_protector_guard_reg was changed.
21310 2017-08-17  Richard Biener  <rguenther@suse.de>
21312         PR tree-optimization/81827
21313         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21314         (new_var_info): Initialize it conservatively.
21315         (get_call_vi): Mark register vars.
21316         (new_scalar_tmp_constraint_exp): Likewise.
21317         (handle_rhs_call): Likewise.
21318         (handle_const_call): Likewise.
21319         (create_function_info_for): Likewise.
21320         (solve_constraints): Sort varinfos to separate register from
21321         non-register vars to pack points-to solution bitmaps during
21322         iteration.
21324 2017-08-17  Marek Polacek  <polacek@redhat.com>
21326         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21328 2017-08-17  Richard Biener  <rguenther@suse.de>
21330         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21331         to true when overflow is undefined and we saturated the result.
21333 2017-08-17  Alan Modra  <amodra@gmail.com>
21335         PR target/80938
21336         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21337         Don't use store multiple if only one reg needs saving.
21338         (interesting_frame_related_regno): New function.
21339         (rs6000_frame_related): Don't emit frame info for regs that
21340         don't need saving.
21341         (rs6000_emit_epilogue): Likewise.
21343 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21345         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21346         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21347         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21348         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21349         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21350         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21351         (verify_type): Adjust for TYPE_BINFO move.
21352         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21353         process TYPE_BINFO directly.
21354         (hash_tree): Likewise.
21355         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21356         Likewise.
21357         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21358         Likewise.
21360 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21362         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21364 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21366         PR target/46091
21367         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21368         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21369         Add ix86_binary_operator_ok to insn constraint.
21370         (*iordi_1_bts): Ditto.
21371         (*xordi_1_btc): Ditto.
21372         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21373         Update corresponding peephole2 pattern.
21374         (*btrq): Ditto.
21375         (*btcq): Ditto.
21377 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21379         PR tree-optimization/81832
21380         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21381         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21383 2017-08-16  Marek Polacek  <polacek@redhat.com>
21385         PR middle/81695
21386         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21387         regarding size_zero_node.
21389 2017-08-16  Martin Liska  <mliska@suse.cz>
21391         PR target/81753
21392         * config.gcc: Respect previously set extra_objs in case
21393         of darwin target.
21395 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21397         PR tree-optimization/81835
21398         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21399         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21400         not depend on the phi.
21402 2017-08-16  Alan Modra  <amodra@gmail.com>
21404         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21405         dead code.
21407 2017-08-16  Alan Modra  <amodra@gmail.com>
21409         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21410         (save_reg_p): ..into this.  Update all callers.
21411         (first_reg_to_save): Simplify.
21413 2017-08-16  Alan Modra  <amodra@gmail.com>
21415         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21416         fixed regs.
21418 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21420         PR target/78460
21421         PR target/67712
21422         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21423         constant count if that count is less than 32.
21425 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21427         * gcc.c (execute): Emit friendlier message if inferior is killed
21428         by an external cause.
21430 2017-08-15  Richard Biener  <rguenther@suse.de>
21432         PR tree-optimization/81790
21433         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21434         CONSTRUCTORs from simplifying and VN.
21436 2017-08-14  Martin Sebor  <msebor@redhat.com>
21438         * builtin-attrs.def: Add comments.
21440 2017-08-14  Martin Sebor  <msebor@redhat.com>
21442         PR c/81117
21443         * doc/extend.texi (attribute nonstring): Document new attribute.
21445 2017-08-14  Martin Sebor  <msebor@redhat.com>
21447         PR c/81117
21448         * tree-diagnostic.c (default_tree_printer): Handle %G.
21449         * gimple-pretty-print.h (percent_G_format): Declare new function.
21450         * gimple-pretty-print.c (percent_G_format): Define.
21451         * tree-pretty-print.c (percent_K_format): Add argument.
21453 2017-08-14  Martin Sebor  <msebor@redhat.com>
21455         PR translation/79998
21456         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21457         Remove a stray space.
21459 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21461         PR target/46091
21462         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21463         (*iordi_1_bts): Ditto.
21464         (*xordi_1_btc): Ditto.
21466 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21468         PR target/79845
21469         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21470         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21471         Likewise.
21472         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21473         quoted strings, and make more translator-friendly.
21474         (darwin_rs6000_override_options): Likewise.
21475         (rs6000_option_override_internal): Likewise.
21476         (rs6000_return_in_memory): Fix overlong line.
21477         (init_cmulative_args): Use quoted strings, and make more
21478         translator-friendly.
21479         (rs6000_pass_by_reference): Fix overlong line.
21480         (def_builtin): Use quoted strings.
21481         (altivec_expand_predicate_builtin): Use quoted strings, and make
21482         more translator-friendly.
21483         (htm_expand_builtin): Use quoted strings.
21484         (cpu_expand_builtin): Use quoted strings, and make more
21485         translator-friendly.
21486         (altivec_expand_builtin): Likewise.
21487         (paired_expand_predicate_builtin): Likewise.
21488         (rs6000_invalid_builtin): Likewise.
21489         (builtin_function_type): Use quoted strings.
21490         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21491         more translator-friendly.
21492         (rs6000_trampoline_init): Likewise.
21493         (rs6000_handle_altivec_attribute): Likewise.
21494         (rs6000_inner_target_options): Use quoted strings.
21495         (rs6000_disable_incompatible_switches): Likewise.
21496         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21497         strings, and make more translator-friendly.
21498         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21500 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21502         PR tree-optimization/81799
21503         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21504         cond_expr to simple gimple operand.
21506 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21508         PR middle-end/46932
21509         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21511 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21513         PR target/81754
21514         PR target/81268
21515         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21516         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21517         TARGET_GASISR_PROLOGUES.
21518         * config/avr/avr.c (avr_option_override): Same.
21519         (avr_pass_pre_proep::execute): Same.
21521 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21523         PR target/81820
21524         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21525         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21527 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21529         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21530         define_insn_and_split.  Split to a memory load from 0 in
21531         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21532         using PTR mode iterator.
21533         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21534         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21535         (*add_tp_<mode>): Redefine as define_insn_and_split.
21536         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21537         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21538         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21539         Split to an add with a  memory load from 0 in
21540         DEFAULT_TLS_SEG_REG address space.
21542 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21544         * config/aarch64/aarch64-option-extensions.def (rdma):
21545         Fix feature string to what Linux prints out in /proc/cpuinfo.
21547 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21549         PR ada/79542
21550         * dwarf2out.c (modified_type_die): For C typedef types that have
21551         an ultimate origin, process the ultimate origin instead of the
21552         input type.
21553         (gen_typedef_die): Assert that input DECLs have no ultimate
21554         origin.
21555         (gen_type_die_with_usage): For typedef variants that have an
21556         ultimate origin, just call gen_decl_die on the original DECL.
21557         (process_scope_var): Avoid creating DIEs for local typedefs and
21558         concrete static variables.
21560 2017-08-12  Alan Modra  <amodra@gmail.com>
21562         PR target/81170
21563         PR target/81295
21564         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21565         match gnu-user.h startfile.
21566         (ENDFILE_LINUX_SPEC): Similarly.
21568 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21570         PR lto/81430
21571         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21572         Remove function.
21573         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21575 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21576         * config/aarch64/aarch64.md (mov<mode>): Change.
21577         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21578         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21579         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21581 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21583         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21584         for storage order barriers.
21586 2017-08-11  Martin Liska  <mliska@suse.cz>
21588         PR tree-opt/79987
21589         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21590         variables of void type.
21592 2017-08-11  Martin Liska  <mliska@suse.cz>
21594         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21595         TARGET_SUPPORTS_ALIASES.
21596         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21597         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21598         (optimize_weakref): Likewise.
21599         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21600         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21601         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21603 2017-08-11  Martin Liska  <mliska@suse.cz>
21605         PR ipa/81213
21606         * config/i386/i386.c (make_resolver_func): Do complete
21607         refactoring of the function.
21609 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21611         PR target/81708
21612         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21613         * config/i386/i386.c (ix86_stack_protect_guard): Use
21614         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21615         * doc/invoke.texi (x86 Options): Document
21616         -mstack-protector-guard-symbol= option.
21618 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21620         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21621         * config/i386/i386.c (ix86_split_stack_guard): New function.
21622         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21623         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21624         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21625         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21626         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21627         (split_stack_space_check): Call ix86_split_stack_guard.
21629 2017-08-10  Martin Sebor  <msebor@redhat.com>
21631         * print-tree.c (print_node): Print location using the established
21632         format %s:%i%i.
21633         Replace spaces with colons.
21634         (debug_raw, debug): Ditto.
21636 2017-08-10  Martin Sebor  <msebor@redhat.com>
21638         PR c++/81586
21639         * pretty-print.c (pp_format): Correct the handling of %s precision.
21641 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21643         PR target/81736
21644         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21645         to ...
21646         (ix86_finalize_stack_frame_flags): This.  Also clear
21647         frame_pointer_needed if -fno-omit-frame-pointer is used without
21648         stack access.
21649         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21650         with ix86_finalize_stack_frame_flags.
21651         (ix86_expand_epilogue): Likewise.
21652         (ix86_expand_split_stack_prologue): Likewise.
21653         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21655 2017-08-10  Martin Liska  <mliska@suse.cz>
21657         PR c++/81355
21658         * c-attribs.c (handle_target_attribute):
21659         Report warning for an empty string argument of target attribute.
21661 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21663         PR c/81687
21664         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21665         LABEL_DECLs.
21666         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21667         or DECL_NONLOCAL labels.
21668         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21669         or DECL_NONLOCAL labels here.
21671 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21673         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21674         to indicate when early gimple folding has been disabled.
21675         (rs6000_gimple_fold_builtin): Add debug content.
21676         (rs6000_invalid_builtin): Fix whitespace.
21677         (rs6000_expand_builtin): Fix whitespace.
21678         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21680 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21682         PR target/80938
21683         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21684         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21686 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21688         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21689         (qdf24xx): Likewise.
21690         * config/aarch64/aarch64.md: Include falkor.md.
21691         * config/aarch64/falkor.md: New.
21693 2017-08-09  Marek Polacek  <polacek@redhat.com>
21695         PR c/81233
21696         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21697         * diagnostic.c (emit_diagnostic): Add a comment.
21698         (emit_diagnostic_valist): New function.
21700 2017-08-09  Marek Polacek  <polacek@redhat.com>
21702         PR c/81417
21703         * input.c (make_location): New overload.
21704         * input.h (make_location): Declare.
21706 2017-08-08  Alan Modra  <amodra@gmail.com>
21707             H.J. Lu  <hongjiu.lu@intel.com>
21709         PR driver/81523
21710         * gcc.c (NO_PIE_SPEC): Delete.
21711         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21712         exclusion..
21713         (LINK_PIE_SPEC): ..to here.
21714         (LINK_COMMAND_SPEC): Support -no-pie.
21715         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21716         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21717         format.
21718         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21719         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21720         (ENDFILE_CRTEND_SPEC): Similarly.
21722 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21724         PR target/81708
21725         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21726         (mstack-protector-guard-offset=): Ditto.
21727         * config/i386/i386.c (ix86_option_override): Handle
21728         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21729         options.
21730         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21731         ix86_stack_protect_guard_offset variables.
21732         (TARGET_STACK_PROTECT_GUARD): Always define.
21733         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21734         and -mstack-protector-guard-offset= options.
21736 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21738         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21739         boundary case for the last candidate.
21741 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21743         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21744         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21746 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21748         PR middle-end/19706
21749         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21750         * config/aarch64/aarch64-builtins.c
21751         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21752         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21753         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21755 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21756             Andrew Pinski <pinskia@gmail.com>
21758         PR middle-end/19706
21759         * internal-fn.def (XORSIGN): New.
21760         * optabs.def (xorsign_optab): New.
21761         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21762         (convert_expand_mult_copysign): New.
21763         (pass_optimize_widening_mul::execute): Call
21764         convert_expand_mult_copysign.
21766 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21768         PR tree-optimization/81354
21769         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21770         Insert on edges rather than explicitly creating landing pads.
21771         (analyze_candidates_and_replace): Commit edge inserts.
21773 2017-08-08  Richard Biener  <rguenther@suse.de>
21775         PR middle-end/81719
21776         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21777         (expand_simple_operations): Also look through ADDR_EXPRs with
21778         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21780 2017-08-08  Richard Biener  <rguenther@suse.de>
21782         PR tree-optimization/81723
21783         * tree-vect-slp.c (struct bst_traits): New hash traits.
21784         (bst_fail): New global.
21785         (vect_build_slp_tree_2): New worker, split out from ...
21786         (vect_build_slp_tree): ... this now wrapping it with using
21787         bst_fail set to cache SLP tree build fails.  Properly handle
21788         max_tree_size.
21789         (vect_analyze_slp_instance): Allocate and free bst_fail.
21791 2017-08-08  Martin Liska  <mliska@suse.cz>
21793         PR tree-opt/81696
21794         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21795         LABEL_DECLs that can be from a different function.
21797 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21799         PR tree-optimization/81744
21800         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21801         loop's number of iterations.
21803 2017-08-08  Martin Liska  <mliska@suse.cz>
21805         * asan.c: Include header files.
21806         * attribs.c (build_decl_attribute_variant): New function moved
21807         from tree.[ch].
21808         (build_type_attribute_qual_variant): Likewise.
21809         (cmp_attrib_identifiers): Likewise.
21810         (simple_cst_list_equal): Likewise.
21811         (omp_declare_simd_clauses_equal): Likewise.
21812         (attribute_value_equal): Likewise.
21813         (comp_type_attributes): Likewise.
21814         (build_type_attribute_variant): Likewise.
21815         (lookup_ident_attribute): Likewise.
21816         (remove_attribute): Likewise.
21817         (merge_attributes): Likewise.
21818         (merge_type_attributes): Likewise.
21819         (merge_decl_attributes): Likewise.
21820         (merge_dllimport_decl_attributes): Likewise.
21821         (handle_dll_attribute): Likewise.
21822         (attribute_list_equal): Likewise.
21823         (attribute_list_contained): Likewise.
21824         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21825         (lookup_attribute_by_prefix): Likewise.
21826         * bb-reorder.c: Include header files.
21827         * builtins.c: Likewise.
21828         * calls.c: Likewise.
21829         * cfgexpand.c: Likewise.
21830         * cgraph.c: Likewise.
21831         * cgraphunit.c: Likewise.
21832         * convert.c: Likewise.
21833         * dwarf2out.c: Likewise.
21834         * final.c: Likewise.
21835         * fold-const.c: Likewise.
21836         * function.c: Likewise.
21837         * gimple-expr.c: Likewise.
21838         * gimple-fold.c: Likewise.
21839         * gimple-pretty-print.c: Likewise.
21840         * gimple.c: Likewise.
21841         * gimplify.c: Likewise.
21842         * hsa-common.c: Likewise.
21843         * hsa-gen.c: Likewise.
21844         * internal-fn.c: Likewise.
21845         * ipa-chkp.c: Likewise.
21846         * ipa-cp.c: Likewise.
21847         * ipa-devirt.c: Likewise.
21848         * ipa-fnsummary.c: Likewise.
21849         * ipa-inline.c: Likewise.
21850         * ipa-visibility.c: Likewise.
21851         * ipa.c: Likewise.
21852         * lto-cgraph.c: Likewise.
21853         * omp-expand.c: Likewise.
21854         * omp-general.c: Likewise.
21855         * omp-low.c: Likewise.
21856         * omp-offload.c: Likewise.
21857         * omp-simd-clone.c: Likewise.
21858         * opts-global.c: Likewise.
21859         * passes.c: Likewise.
21860         * predict.c: Likewise.
21861         * sancov.c: Likewise.
21862         * sanopt.c: Likewise.
21863         * symtab.c: Likewise.
21864         * toplev.c: Likewise.
21865         * trans-mem.c: Likewise.
21866         * tree-chkp.c: Likewise.
21867         * tree-eh.c: Likewise.
21868         * tree-into-ssa.c: Likewise.
21869         * tree-object-size.c: Likewise.
21870         * tree-parloops.c: Likewise.
21871         * tree-profile.c: Likewise.
21872         * tree-ssa-ccp.c: Likewise.
21873         * tree-ssa-live.c: Likewise.
21874         * tree-ssa-loop.c: Likewise.
21875         * tree-ssa-sccvn.c: Likewise.
21876         * tree-ssa-structalias.c: Likewise.
21877         * tree-ssa.c: Likewise.
21878         * tree-streamer-in.c: Likewise.
21879         * tree-vectorizer.c: Likewise.
21880         * tree-vrp.c: Likewise.
21881         * tsan.c: Likewise.
21882         * ubsan.c: Likewise.
21883         * varasm.c: Likewise.
21884         * varpool.c: Likewise.
21885         * tree.c: Remove functions moved to attribs.[ch].
21886         * tree.h: Likewise.
21887         * config/aarch64/aarch64.c: Add attrs.h header file.
21888         * config/alpha/alpha.c: Likewise.
21889         * config/arc/arc.c: Likewise.
21890         * config/arm/arm.c: Likewise.
21891         * config/avr/avr.c: Likewise.
21892         * config/bfin/bfin.c: Likewise.
21893         * config/c6x/c6x.c: Likewise.
21894         * config/cr16/cr16.c: Likewise.
21895         * config/cris/cris.c: Likewise.
21896         * config/darwin.c: Likewise.
21897         * config/epiphany/epiphany.c: Likewise.
21898         * config/fr30/fr30.c: Likewise.
21899         * config/frv/frv.c: Likewise.
21900         * config/ft32/ft32.c: Likewise.
21901         * config/h8300/h8300.c: Likewise.
21902         * config/i386/winnt.c: Likewise.
21903         * config/ia64/ia64.c: Likewise.
21904         * config/iq2000/iq2000.c: Likewise.
21905         * config/lm32/lm32.c: Likewise.
21906         * config/m32c/m32c.c: Likewise.
21907         * config/m32r/m32r.c: Likewise.
21908         * config/m68k/m68k.c: Likewise.
21909         * config/mcore/mcore.c: Likewise.
21910         * config/microblaze/microblaze.c: Likewise.
21911         * config/mips/mips.c: Likewise.
21912         * config/mmix/mmix.c: Likewise.
21913         * config/mn10300/mn10300.c: Likewise.
21914         * config/moxie/moxie.c: Likewise.
21915         * config/msp430/msp430.c: Likewise.
21916         * config/nds32/nds32-isr.c: Likewise.
21917         * config/nds32/nds32.c: Likewise.
21918         * config/nios2/nios2.c: Likewise.
21919         * config/nvptx/nvptx.c: Likewise.
21920         * config/pa/pa.c: Likewise.
21921         * config/pdp11/pdp11.c: Likewise.
21922         * config/powerpcspe/powerpcspe.c: Likewise.
21923         * config/riscv/riscv.c: Likewise.
21924         * config/rl78/rl78.c: Likewise.
21925         * config/rx/rx.c: Likewise.
21926         * config/s390/s390.c: Likewise.
21927         * config/sh/sh.c: Likewise.
21928         * config/sol2.c: Likewise.
21929         * config/sparc/sparc.c: Likewise.
21930         * config/spu/spu.c: Likewise.
21931         * config/stormy16/stormy16.c: Likewise.
21932         * config/tilegx/tilegx.c: Likewise.
21933         * config/tilepro/tilepro.c: Likewise.
21934         * config/v850/v850.c: Likewise.
21935         * config/vax/vax.c: Likewise.
21936         * config/visium/visium.c: Likewise.
21937         * config/xtensa/xtensa.c: Likewise.
21939 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21941         PR target/81593
21942         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
21943         constraints since the -mupper-regs-* switches have been
21944         eliminated.
21945         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
21946         into a vector from a double word element that was extracted from
21947         another vector, and eliminate extra XXPERMDI instructions.
21948         (vsx_concat_<mode>_2): Likewise.
21949         (vsx_concat_<mode>_3): Likewise.
21950         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
21951         concat to allow optimizing inserts from previous extracts.
21953 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
21955         * config/i386/i386.c (ix86_stack_protect_guard): Generate
21956         memory reference to a SSP offset in TLS address space.
21957         (ix86_print_operand) <case '@'>: Remove.
21958         (ix86_print_operand_punct_valid_p): Remove '@' code.
21959         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
21960         UNSPEC_SP_TLS_TEST.
21961         (stack_tls_protect_set_<mode>): Remove.
21962         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
21963         (stack_tls_protect_test_<mode>): Remove.
21964         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
21966 2017-08-07  Olivier Hainque  <hainque@adacore.com>
21968         PR target/81755
21969         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
21971 2017-08-07  Douglas Rupp  <rupp@adacore.com>
21973         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
21974         variable was referenced as multidir in command.
21976 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21978         PR c/69389
21979         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
21980         BIT_FIELD_REF.
21982 2017-08-07  Martin Liska  <mliska@suse.cz>
21984         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
21985         * config/rl78/rl78.c: Add include of attribs.h.
21986         * config/sh/sh.c: Likewise.
21987         * config/v850/v850.c: Likewise.
21989 2017-08-07  Tom de Vries  <tom@codesourcery.com>
21991         PR middle-end/78266
21992         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
21994 2017-08-07  Martin Liska  <mliska@suse.cz>
21996         * config/mips/mips.c: Include attribs.h.
21998 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22000         PR fortran/68829
22001         * doc/invoke.texi: Document change in behvaior for -Ofast for
22002         Fortran.
22004 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
22006         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
22007         Use gen_frame_mem.
22008         (aarch64_pop_regs): Likewise.
22009         (aarch64_gen_load_pair): Likewise.
22010         (aarch64_save_callee_saves): Likewise.
22011         (aarch64_restore_callee_saves): Likewise.
22013 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22015         * config/i386/i386.c: Revert the last change.
22017 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22019         PR target/81736
22020         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
22021         to ...
22022         (ix86_finalize_stack_frame_flags): This.  Also clear
22023         frame_pointer_needed if -fno-omit-frame-pointer is used without
22024         stack access.
22025         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22026         with ix86_finalize_stack_frame_flags.
22027         (ix86_expand_epilogue): Likewise.
22028         (ix86_expand_split_stack_prologue): Likewise.
22030 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22032         PR target/81743
22033         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22034         to P_AES for Westmere.
22036 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22038         * config/i386/mingw.opt (fset-stack-executable): Removed.
22039         * config/i386/cygming.opt (fset-stack-executable): Moved
22040         from mingw.opt.
22041         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22043 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22045         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22047 2017-08-07  Marek Polacek  <polacek@redhat.com>
22049         PR middle-end/81737
22050         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22052 2017-08-07  Martin Liska  <mliska@suse.cz>
22054         * attribs.h (canonicalize_attr_name): New function.
22055         (cmp_attribs): Move from c-format.c and adjusted.
22056         (is_attribute_p): Moved from tree.h.
22057         * tree-inline.c: Add new includes.
22058         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22059         (private_is_attribute_p): Remove.
22060         (private_lookup_attribute): Likewise.
22061         (private_lookup_attribute_by_prefix): Simplify.
22062         (remove_attribute): Use is_attribute_p.
22063         * tree.h: Remove removed declarations.
22065 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22067         PR middle-end/81698
22068         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22069         instead of computing it in the function.  Formatting fix.
22070         (expand_case): Don't rely on default_edge being the first edge,
22071         clear it if removing it, pass default_edge to
22072         emit_case_dispatch_table.
22073         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22074         fix.
22076 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22078         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22079         insn in the function, emit NOP after the insn.
22081 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22083         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22084         and element loops.
22086 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22088         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22089         loop.
22091 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22093         PR tree-optimization/57371
22094         * match.pd: New pattern.
22096 2017-08-04  Marek Polacek  <polacek@redhat.com>
22098         PR middle-end/81695
22099         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22100         perform the computation in offset_int.
22102 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22104         PR tree-optimization/81136
22105         * tree-vectorizer.h: Include tree-hash-traits.h.
22106         (vec_base_alignments): New typedef.
22107         (vec_info): Add a base_alignments field.
22108         (vect_record_base_alignments): Declare.
22109         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22110         field.
22111         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22112         (create_data_ref): Add an is_conditional_in_stmt argument.
22113         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22114         the is_conditional_in_stmt field.
22115         (data_ref_loc): Add an is_conditional_in_stmt field.
22116         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22117         (find_data_references_in_stmt): Update call to create_data_ref.
22118         (graphite_find_data_references_in_stmt): Likewise.
22119         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22120         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22121         (vect_record_base_alignment): New function.
22122         (vect_record_base_alignments): Likewise.
22123         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22124         for nested statements even if we fail to compute a misalignment.
22125         Use pooled base alignments for unconditional references.
22126         (vect_find_same_alignment_drs): Compare base addresses instead
22127         of base objects.
22128         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22129         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22131 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22133         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22134         Add an explicit name for the enum.  Use auto_vec for slp_instances
22135         and grouped_stores.
22136         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22137         for all vectors.
22138         (_bb_vec_info): Add a constructor and destructor.
22139         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22140         (destroy_loop_vec_info): Delete.
22141         (vect_destroy_datarefs): Likewise.
22142         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22143         (vec_info::vec_info): New function.
22144         (vec_info::~vec_info): Likewise.
22145         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22146         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22147         destroy_loop_vec_info.
22148         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22149         (_loop_vec_info::_loop_vec_info): ...this.
22150         (destroy_loop_vec_info): Replace with...
22151         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22152         the stmt_vec_infos.  Leave handling of vec_info information to its
22153         destructor.  Remove explicit vector releases.
22154         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22155         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22156         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22157         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22158         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22159         (destroy_bb_vec_info): Replace with...
22160         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22161         information to its destructor.
22162         (vect_slp_analyze_bb_1): Use new and delete instead of
22163         new_bb_vec_info and destroy_bb_vec_info.
22164         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22165         single delete.
22167 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22169         * tree-data-ref.h (subscript): Add access_fn field.
22170         (data_dependence_relation): Add could_be_independent_p.
22171         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22172         (same_access_functions): Move to tree-data-ref.c.
22173         * tree-data-ref.c (ref_contains_union_access_p): New function.
22174         (access_fn_component_p): Likewise.
22175         (access_fn_components_comparable_p): Likewise.
22176         (dr_analyze_indices): Add a reference to access_fn_component_p.
22177         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22178         DR_ACCESS_FN.
22179         (constant_access_functions): Likewise.
22180         (add_other_self_distances): Likewise.
22181         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22182         (initialize_data_dependence_relation): Use XCNEW and remove
22183         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22184         of access functions that have the same type.  Allow the
22185         subsequence to end with different bases in some circumstances.
22186         Record the chosen access functions in SUB_ACCESS_FN.
22187         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22188         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22189         (subscript_dependence_tester_1): Likewise dra and drb.
22190         (build_classic_dist_vector): Update calls accordingly.
22191         (subscript_dependence_tester): Likewise.
22192         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22193         DDR_COULD_BE_INDEPENDENT_P.
22194         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22195         comp_alias_ddrs instead of may_alias_ddrs.
22196         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22197         New function.
22198         (vect_analyze_data_ref_dependence): Use it if
22199         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22200         distance vectors if that fails.
22201         (dependence_distance_ge_vf): New function.
22202         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22203         LOOP_VINFO_MAY_ALIAS_DDRS.
22205 2017-08-04  Richard Biener  <rguenther@suse.de>
22207         PR middle-end/81705
22208         * fold-const.c (fold_binary_loc): Properly restrict
22209         minus_var0 && minus_var1 case when associating undefined overflow
22210         entities.
22212 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22214         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22216 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22218         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22219         Don't start diagnostic messages with a capital letter.
22220         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22221         Likewise.
22222         (rs6000_invalid_builtin): Likewise.
22223         (rs6000_trampoline_init): Likewise.
22225 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22227         PR target/81621
22228         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22229         after setting changeable df flags.
22231 2017-08-03  Richard Biener  <rguenther@suse.de>
22233         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22234         up if the use is in USE - X.
22236 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22238         * toplev.c (dumpfile.h): New include.
22239         (internal_error_reentered): New static function.  Use it...
22240         (internal_error_function): ...here to handle reentered internal_error.
22242 2017-08-03  Richard Biener  <rguenther@suse.de>
22244         PR middle-end/81148
22245         * fold-const.c (split_tree): Add minus_var and minus_con
22246         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22247         here but always use minus_*.
22248         (associate_trees): Assert we never associate with MINUS_EXPR
22249         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22250         when associating as MINUS_EXPR either.
22251         (fold_binary_loc): Track minus_var and minus_con.
22253 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22255         PR lto/81430
22256         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22257         ACCEL_COMPILER, apply finish_options on
22258         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22260 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22262         PR target/81662
22263         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22264         function_entry_patch_area_size > 0.
22266 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22268         PR driver/81650
22269         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22270         instead of 10??LU, perform unit multiplication in wide_int,
22271         don't change alloc_object_size_limit if the limit is larger
22272         than SSIZE_MAX.
22274         PR tree-optimization/81655
22275         PR tree-optimization/81588
22276         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22277         the case when ranges[i].low and high are 1 for unsigned type with
22278         precision 1.
22280         PR middle-end/81052
22281         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22282         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22284 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22286         * tree-vrp.h: Add include guard.
22288 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22290         PR target/81644
22291         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22292         (ud2): New insn pattern.
22293         * config/i386/i386.c (ix86_expand_epilogue):
22294         For naked functions, generate ud2 instead of trap insn.
22296 2017-08-02  Marek Polacek  <polacek@redhat.com>
22298         PR other/81667
22299         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22301 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22302             Cesar Philippidis  <cesar@codesourcery.com>
22304         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22305         Add missing edge probabilities.
22307 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22309         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22310         Correct endianness.
22312 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22314         PR middle-end/79499
22315         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22316         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22317         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22318         prologue_seq sequences - if any.
22320 2017-08-02  Richard Biener  <rguenther@suse.de>
22322         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22323         via vectors if supported, integer extracts via punning if supported
22324         or otherwise vector extracts.
22326 2017-08-02  Richard Biener  <rguenther@suse.de>
22328         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22329         into ...
22330         (bitmap_insert_into_set): ... this.
22332 2017-08-02  Richard Biener  <rguenther@suse.de>
22334         PR tree-optimization/81633
22335         Revert
22336         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22338         PR tree-optimization/71752
22339         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22341 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22343         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22344         (machine_function::call_ms2sysv_pad_out): Remove field.
22345         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22346         (ix86_compute_frame_layout): Likewise.
22348 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22350         PR target/81654
22351         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22352         attribute with interrupt attribute.
22354 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22356         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22357         BIT_INSERT_EXPR's operand 1
22358         to see if the types precision matches.
22360 2017-08-01  Martin Liska  <mliska@suse.cz>
22362         PR middle-end/70140
22363         * builtins.c (expand_builtin_memcpy_args): Remove.
22364         (expand_builtin_memcpy): Call newly added function
22365         expand_builtin_memory_copy_args.
22366         (expand_builtin_memcpy_with_bounds): Likewise.
22367         (expand_builtin_mempcpy): Remove last argument.
22368         (expand_builtin_mempcpy_with_bounds): Likewise.
22369         (expand_builtin_memory_copy_args): New function created from
22370         expand_builtin_mempcpy_args with small modifications.
22371         (expand_builtin_mempcpy_args): Remove.
22372         (expand_builtin_stpcpy): Remove unused argument.
22373         (expand_builtin): Likewise.
22374         (expand_builtin_with_bounds): Likewise.
22376 2017-08-01  Martin Liska  <mliska@suse.cz>
22378         Revert r250771
22379         Make mempcpy more optimal (PR middle-end/70140).
22381 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22383         PR target/81622
22384         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22385         __builtin_vec_cmpne verify both arguments are compatible vectors
22386         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22387         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22388         move computation of aligned to after checking the argument types.
22389         Formatting fixes.
22391         PR target/80846
22392         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22393         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22394         calls.
22396 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22397             Doug Rupp  <rupp@adacore.com>
22398             Olivier Hainque  <hainque@adacore.com>
22400         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22401         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22402         arm8 (arch v4).
22403         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22404         for TARGET_OS_CPP_BUILTIN.
22405         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22406         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22407         arm_arch7.
22408         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22409         passing required abi options to the assembler for EABI configurations.
22410         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22411         of .text.hot and .text.unlikely sections for kernel modules when
22412         using ARM style exceptions.
22413         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22414         options. Add EXTRA_CC1_SPEC.
22415         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22416         toolchain options.
22417         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22418         transition.
22419         (ARM_TARGET2_DWARF_FORMAT): Define.
22420         * config/arm/t-vxworks: Adjust multilib control to removal of the
22421         Diab command line options.
22423 2017-08-01  Martin Liska  <mliska@suse.cz>
22425         PR gcov-profile/81561
22426         * gcov.c (unblock): Make unblocking safe as we need to preserve
22427         index correspondence of blocks and block_lists.
22429 2017-08-01  Richard Biener  <rguenther@suse.de>
22431         PR tree-optimization/81181
22432         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22433         (compute_antic): ... end of iteration here.
22435 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22437         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22438         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22439         (ftree-slp-vectorize): Likewise.
22440         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22441         can no longer be set independent of flag_tree_loop_vectorize.
22442         * omp-general.c (emp_max_vf): Likewise.
22443         * opts.c (enable_fdo_optimizations): Remove references to
22444         flag_tree_vectorize, these are now implicit.
22445         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22446         and leave it for the options machinery.
22448 2017-08-01  Martin Liska  <mliska@suse.cz>
22450         PR middle-end/70140
22451         * builtins.c (expand_builtin_memcpy_args): Remove.
22452         (expand_builtin_memcpy): Call newly added function
22453         expand_builtin_memory_copy_args.
22454         (expand_builtin_memcpy_with_bounds): Likewise.
22455         (expand_builtin_mempcpy): Remove last argument.
22456         (expand_builtin_mempcpy_with_bounds): Likewise.
22457         (expand_builtin_memory_copy_args): New function created from
22458         expand_builtin_mempcpy_args with small modifications.
22459         (expand_builtin_mempcpy_args): Remove.
22460         (expand_builtin_stpcpy): Remove unused argument.
22461         (expand_builtin): Likewise.
22462         (expand_builtin_with_bounds): Likewise.
22464 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22466         PR target/81641
22467         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22468         print "ds:" only for immediates in generic address space.
22470 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22472         PR target/81639
22473         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22474         (ix86_function_ok_for_sibcall): Return false for naked functions.
22476 2017-08-01  Richard Biener  <rguenther@suse.de>
22478         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22479         (compute_antic): Seed worklist with exit block predecessors.
22480         * cfganal.c (dfs_find_deadend): For a cycle return the source
22481         of the edge closing it.
22483 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22485         * config/aarch64/aarch64.c
22486         (aarch64_can_const_movi_rtx_p): Move 0 check.
22488 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22490         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22491         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22492         above macro.
22493         * match.pd: Ditto in address comparison pattern.
22495 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22497         PR tree-optimization/81627
22498         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22499         closed ssa form for store-store chain.
22501 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22503         PR tree-optimization/81620
22504         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22505         for store-store chain.
22507 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22509         PR tree-optimization/81588
22510         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22511         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22512         swap rhs1 and rhs2 and comparison code unconditionally,
22513         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22514         ranges[i].in_p, instead invert comparison code ccode if
22515         opcode or oe->rank is BIT_IOR_EXPR.
22517         PR target/80846
22518         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22519         a direct optab to conversion optab.
22520         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22521         with GET_MODE_INNER as last argument instead of optab_handler.
22522         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22523         vector extraction if possible and optab is available.
22524         * expr.c (store_constructor): Use convert_optab_handler instead
22525         of optab_handler.  Use vector initialization from smaller
22526         vectors if possible and optab is available.
22527         * tree-vect-stmts.c (vectorizable_load): Likewise.
22528         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22529         now have two modes.
22530         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22531         of vec_init from half-sized vectors with the same element mode.
22532         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22533         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22534         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22535         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22536         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22537         after mode in gen_vec_extract* calls.
22538         (vec_extract<mode>): Renamed to ...
22539         (vec_extract<mode><ssescalarmodelower>): ... this.
22540         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22541         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22542         element mode after mode in gen_vec_init* calls.
22543         (VEC_INIT_HALF_MODE): New mode iterator.
22544         (vec_init<mode>): Renamed to ...
22545         (vec_init<mode><ssescalarmodelower>): ... this.
22546         (vec_init<mode><ssehalfvecmodelower>): New expander.
22547         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22548         (vec_extractv2sfsf): ... this.
22549         (vec_initv2sf): Renamed to ...
22550         (vec_initv2sfsf): ... this.
22551         (vec_extractv2si): Renamed to ...
22552         (vec_extractv2sisi): ... this.
22553         (vec_initv2si): Renamed to ...
22554         (vec_initv2sisi): ... this.
22555         (vec_extractv4hi): Renamed to ...
22556         (vec_extractv4hihi): ... this.
22557         (vec_initv4hi): Renamed to ...
22558         (vec_initv4hihi): ... this.
22559         (vec_extractv8qi): Renamed to ...
22560         (vec_extractv8qiqi): ... this.
22561         (vec_initv8qi): Renamed to ...
22562         (vec_initv8qiqi): ... this.
22563         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22564         (vec_init<mode>): Renamed to ...
22565         (vec_init<mode><VEC_base_l>): ... this.
22566         (vec_extract<mode>): Renamed to ...
22567         (vec_extract<mode><VEC_base_l>): ... this.
22568         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22569         (vec_initv2sfsf): ... this.
22570         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22571         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22572         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22573         element mode after mode in gen_vec_init* calls.
22574         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22575         (vec_init<mode><Vel>): ... this.
22576         (vec_extract<mode>): Renamed to ...
22577         (vec_extract<mode><Vel>): ... this.
22578         * config/aarch64/iterators.md (Vel): New mode attribute.
22579         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22580         Add element mode after mode in gen_vec_extract* calls.
22581         * config/s390/vector.md (non_vec_l): New mode attribute.
22582         (vec_extract<mode>): Renamed to ...
22583         (vec_extract<mode><non_vec_l>): ... this.
22584         (vec_init<mode>): Renamed to ...
22585         (vec_init<mode><non_vec_l>): ... this.
22586         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22587         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22588         vec_extract mode.
22589         * config/arm/iterators.md (V_elem_l): New mode attribute.
22590         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22591         (vec_extract<mode><V_elem_l>): ... this.
22592         (vec_extractv2di): Renamed to ...
22593         (vec_extractv2didi): ... this.
22594         (vec_init<mode>): Renamed to ...
22595         (vec_init<mode><V_elem_l>): ... this.
22596         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22597         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22598         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22599         Add element mode after gen_vec_extract* calls.
22600         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22601         (vec_init<mode><unitmode>): ... this.
22602         (vec_extract<mode>): Renamed to ...
22603         (vec_extract<mode><unitmode>): ... this.
22604         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22605         (vec_init<mode><unitmode>): ... this.
22606         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22607         (vec_initv2sfsf): ... this.
22608         (vec_extractv2sf): Renamed to ...
22609         (vec_extractv2sfsf): ... this.
22610         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22611         Add element mode after gen_vec_extract* calls.
22612         * config/mips/mips.md (unitmode): New mode iterator.
22613         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22614         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22615         * config/spu/spu.md (inner_l): New mode attribute.
22616         (vec_init<mode>): Renamed to ...
22617         (vec_init<mode><inner_l>): ... this.
22618         (vec_extract<mode>): Renamed to ...
22619         (vec_extract<mode><inner_l>): ... this.
22620         * config/sparc/sparc.md (veltmode): New mode iterator.
22621         (vec_init<VMALL:mode>): Renamed to ...
22622         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22623         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22624         (vec_initv2sisi): ... this.
22625         (vec_initv2sf): Renamed to ...
22626         (vec_initv2sfsf): ... this.
22627         (vec_extractv2sf): Renamed to ...
22628         (vec_extractv2sfsf): ... this.
22629         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22630         (vec_init<mode>): Renamed to ...
22631         (vec_init<mode><VEC_base_l>): ... this.
22632         (vec_extract<mode>): Renamed to ...
22633         (vec_extract<mode><VEC_base_l>): ... this.
22634         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22635         (vec_initv2sfsf): ... this.
22636         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22637         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22638         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22639         gen_vec_init* calls.
22641 2017-08-01  Richard Biener  <rguenther@suse.de>
22643         PR tree-optimization/81297
22644         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22645         TREE_OVERFLOW from INTEGER_CSTs.
22647 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22649         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22651 2017-07-31  Carl Love  <cel@us.ibm.com>
22653         * config/rs6000/rs6000-c: Add support for built-in functions
22654         vector signed char vec_xl_be (signed long long, signed char *);
22655         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22656         vector signed int vec_xl_be (signed long long, signed int *);
22657         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22658         vector signed long long vec_xl_be (signed long long, signed long long *);
22659         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22660         vector signed short vec_xl_be (signed long long, signed short *);
22661         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22662         vector double vec_xl_be (signed long long, double *);
22663         vector float vec_xl_be (signed long long, float *);
22664         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22665         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22666         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22667         for the builtins.
22668         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22669         (altivec_expand_builtin): Add switch statement to call
22670         altivec_expand_xl_be for each builtin.
22671         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22672         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22673         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22674         __builtin_vsx_le_be_v16qi.
22675         * doc/extend.texi: Update the built-in documentation file for the
22676         new built-in functions.
22678 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22680         PR target/25967
22681         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22682         New function.
22683         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22685 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22687         * config.gcc: Add z14.
22688         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22689         CPU model numbers for z13s and z14.
22690         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22691         arch12 with z14.
22692         * config/s390/s390-opts.h (enum processor_type): Rename
22693         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22694         * config/s390/s390.c (processor_table): Add field for CPU name to
22695         be passed to Binutils.
22696         (s390_asm_output_machine_for_arch): Use the new field in
22697         processor_table for Binutils.
22698         (s390_expand_builtin): Replace arch12 with z14.
22699         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22700         (s390_get_sched_attrmask): Likewise.
22701         (s390_get_unit_mask): Likewise.
22702         * config/s390/s390.opt: Add z14 to processor_type enum.
22704 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22706         PR hsa/81477
22707         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22708         regardless of optimization level.
22710 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22711             Martin Liska  <mliska@suse.cz>
22713         * predict.def: Remove old comment and adjust probability.
22714         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22715         PREDICT statements.
22717 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22719         PR target/25967
22720         * config/i386/i386.c (ix86_function_naked): New function.
22721         (ix86_can_use_return_insn_p): Return false for naked functions.
22722         (ix86_expand_prologue): Skip prologue for naked functions.
22723         (ix86_expand_epilogue): Skip epilogue for naked functions
22724         and emit trap instruction.
22725         (ix86_warn_func_return): New function.
22726         (ix86_attribute_table): Add "naked" attribute specification.
22727         (TARGET_WARN_FUNC_RETURN): Define.
22728         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22730 2017-07-31  Martin Liska  <mliska@suse.cz>
22732         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22733         (dump_gimple_bb_header): Always dump BB info.
22734         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22736 2017-07-31  Martin Liska  <mliska@suse.cz>
22738         PR sanitize/81530
22739         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22740         also with current_function_decl non-null equality.
22742 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22744         PR sanitizer/81604
22745         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22746         change type to the element type, instead add eltype variable and
22747         use it where we are interested in the element type.
22749         PR tree-optimization/81603
22750         * ipa-polymorphic-call.c
22751         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22752         offset arithmetic in offset_int, bail out if the resulting bit offset
22753         doesn't fit into shwi.
22755 2017-07-31  Martin Liska  <mliska@suse.cz>
22757         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22758         (gimplify_save_expr): Fix comment.
22760 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22762         PR target/79793
22763         * config/i386/i386.c (ix86_function_arg): Update arguments for
22764         exception handler.
22765         (ix86_compute_frame_layout): Set the initial stack offset to
22766         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22767         INCOMING_FRAME_SP_OFFSET.
22768         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22769         stack before exception handler returns.
22770         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22771         the 'ERROR_CODE' for exception handler.
22773 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22775         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22776         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22777         (ASM_OUTPUT_REG_POP): Ditto.
22778         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22779         instead of asm_fprintf to output pure string.
22781 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22783         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22784         to imported_module_or_decl hook.
22785         (debug_nothing_tree_tree_tree_bool): Remove.
22786         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22787         * debug.c (do_nothing_debug_hooks): Use
22788         debug_nothing_tree_tree_tree_bool_bool instead of
22789         debug_nothing_tree_tree_tree_bool.
22790         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22791         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22792         * sdbout.c (sdb_debug_hooks): Likewise.
22793         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22794         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22795         langhook wants it.
22796         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22797         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22798         attribute, don't add anything.
22800 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22802         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22803         (fold_build2_stat_loc): Likewise.
22804         (fold_build3_stat_loc): Likewise.
22805         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22806         (fold_build1_loc): Remove macro.
22807         (fold_build2_loc): Likewise.
22808         (fold_build3_loc): Likewise.
22810 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22812         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22813         (gimple_build_debug_bind_source_stat): Likewise.
22814         * gimple.h (gimple_build_debug_bind): Remove macro.
22815         (gimple_build_debug_bind_source): Likewise.
22817 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22819         * bitmap.c (bitmap_alloc): Adjust.
22820         (bitmap_gc_alloc): Likewise.
22821         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22823 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22825         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22826         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22827         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22828         (bitmap_gc_alloc_stat): Likewise.
22829         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22831 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22833         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22834         * rtl.h (shallow_copy_rtx): Remove macro.
22836 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22838         * emit-rtl.c (gen_raw_REG): Adjust.
22839         * gengenrtl.c (gendef): Likewise.
22840         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22841         * rtl.h (rtx_alloc): Remove macro.
22843 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22845         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22846         (build_tree_list_stat): Likewise.
22847         * tree.h (build_tree_list): Remove macro.
22848         (build_tree_list_vec): Likewise.
22850 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22852         * tree.c (make_vector_stat): Remove _stat from name.
22853         (build_vector_stat): Likewise.
22854         * tree.h (make_vector_stat): Remove macro.
22855         (build_vector_stat): Likewise.
22857 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22859         * tree.h (build_var_debug_value): Remove prototype.
22861 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22863         * tree.c (tree_cons_stat): Remove _stat from name.
22864         * tree.h (tree_cons): Remove macro.
22866 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22868         * tree.c (build_vl_exp_stat): Remove _stat from name.
22869         * tree.h (build_vl_exp): Remove macro.
22871 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22873         * tree.c (build_decl_stat): Remove _stat from name.
22874         * tree.h (build_decl): Remove macro.
22876 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22878         * gimple.c (gimple_build_with_ops_stat): Adjust.
22879         (gimple_alloc_stat): Remove _stat from name.
22880         * gimple.h (gimple_alloc): Remove macro.
22882 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22884         * tree.c (make_tree_vec_stat): Remove _stat from name.
22885         (grow_tree_vec_stat): Likewise.
22886         * tree.h (make_tree_vec_stat): Adjust prototype.
22887         (grow_tree_vec_stat): Likewise.
22888         (make_tree_vec): Remove macro.
22889         (grow_tree_vec): Likewise.
22891 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22893         * fold-const.c (fold_build1_stat_loc): Adjust.
22894         (fold_build2_stat_loc): Likewise.
22895         (fold_build3_stat_loc): Likewise.
22896         * tree.c (build0_stat): Remove _stat from name.
22897         (build1_stat): Likewise.
22898         (build2_stat): Likewise.
22899         (build3_stat): Likewise.
22900         (build4_stat): Likewise.
22901         (build5_stat): Likewise.
22902         * tree.h (build1_loc): Remove macro, and rename _stat function
22903         to this.
22904         (build2_loc): Likewise.
22905         (build3_loc): Likewise.
22906         (build4_loc): Likewise.
22907         (build5_loc): Likewise.
22909 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22911         * tree.c (make_int_cst_stat): Remove _stat from name.
22912         * tree.h (make_int_cst_stat): Adjust prototype.
22913         (make_int_cst): Remove macro.
22915 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22917         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22918         * tree.h (make_tree_binfo_stat): Adjust prototype.
22919         (make_tree_binfo): Remove.
22921 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22923         * tree.c (copy_node_stat): Rename to copy_node.
22924         (build_distinct_type_copy): Adjust.
22925         * tree.h (copy_node_stat): Adjust prototype.
22926         (copy_node): Remove macro.
22928 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22930         * tree.c (make_node_stat): rename to make_node.
22931         (build_tree_list_stat): Adjust.
22932         (build0_stat): Likewise.
22933         (build2_stat): Likewise.
22934         (build3_stat): Likewise.
22935         (build4_stat): Likewise.
22936         (build5_stat): Likewise.
22937         (build_decl_stat): Likewise.
22938         * tree.h (make_node_stat): Adjust prototype.
22939         (make_node): remove macro.
22941 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
22943         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
22944         (PPC_FEATURE2_SCV): Likewise.
22945         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
22947 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22949         * config/aarch64/aarch64.c
22950         (aarch64_internal_mov_immediate): Add new special pattern.
22951         * config/aarch64/aarch64.md (*movdi_aarch64):
22952         Add reg/32bit const mov case.
22954 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22955             Richard Sandiford <richard.sandiford@linaro.org>
22957         * config/aarch64/aarch64.md (mov<mode>): Generalize.
22958         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
22959         Add integer and movi cases.
22960         (movi-split-hf-df-sf split, fp16): New.
22961         (enabled): Added TARGET_FP_F16INST.
22962         * config/aarch64/iterators.md (GPF_HF): New.
22963         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
22965 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22967         * config/aarch64/aarch64.c
22968         (aarch64_simd_container_mode): Add prototype.
22969         (aarch64_expand_mov_immediate): Add HI support.
22970         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
22971         (aarch64_can_const_movi_rtx_p): New.
22972         (aarch64_preferred_reload_class):
22973         Remove restrictions of using FP registers for certain SIMD operations.
22974         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
22975         (aarch64_valid_floating_const): Add integer move validation.
22976         (aarch64_simd_imm_scalar_p): Remove.
22977         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
22978         (aarch64_legitimate_constant_p): Expand list of supported cases.
22979         * config/aarch64/aarch64-protos.h
22980         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
22981         (aarch64_reinterpret_float_as_int): New.
22982         (aarch64_simd_imm_scalar_p): Remove.
22983         * config/aarch64/constraints.md (Uvi): New.
22984         (Dd): Split into Ds and new Dd.
22985         * config/aarch64/aarch64.md (*movsi_aarch64):
22986         Add SIMD mov case.
22987         (*movdi_aarch64): Add SIMD mov case.
22989 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22991         * tree-predcom.c: (struct chain): Handle store-store chain in which
22992         stores for elimination only store loop invariant values.
22993         (execute_pred_commoning_chain): Ditto.
22994         (prepare_initializers_chain_store_elim): Ditto.
22995         (prepare_finalizers): Ditto.
22996         (is_inv_store_elimination_chain): New function.
22997         (initialize_root_vars_store_elim_1): New function.
22999 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23001         * tree-predcom.c: Revise general description of the pass.
23002         (enum chain_type): New enum type for store elimination.
23003         (struct chain): New field supporting store elimination.
23004         (struct component): Ditto.
23005         (dump_chain): Dump store-stores chain.
23006         (release_chain): Release resources.
23007         (split_data_refs_to_components): Compute and create component
23008         contains only stores for elimination.
23009         (get_chain_last_ref_at): New function.
23010         (make_invariant_chain): Initialization.
23011         (make_rooted_chain): Specify chain type in parameter and record it.
23012         (add_looparound_copies): Skip for store-stores chain.
23013         (determine_roots_comp): Compute type of chain and pass it to
23014         make_rooted_chain.
23015         (initialize_root_vars_store_elim_2): New function.
23016         (finalize_eliminated_stores): New function.
23017         (remove_stmt): Handle store for elimination.
23018         (execute_pred_commoning_chain): Execute predictive commoning on
23019         store-store chains.
23020         (determine_unroll_factor): Skip unroll for store-stores chain.
23021         (prepare_initializers_chain_store_elim): New function.
23022         (prepare_initializers_chain): Hanlde store-store chain.
23023         (prepare_finalizers_chain, prepare_finalizers): New function.
23024         (tree_predictive_commoning_loop): Return integer value indicating
23025         if loop is unrolled or lcssa form is corrupted.
23026         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23028 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23030         * tree-predcom.c (initialize_root): Delete.
23031         (execute_pred_commoning_chain): Initialize root vars and replace
23032         reference of non-combined chain directly, rather than call above
23033         function.
23035 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23037         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23038         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23040 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23042         * tree-predcom.c (struct chain): New field init_seq.
23043         (release_chain): Release init_seq.
23044         (prepare_initializers_chain): Record intialization stmts in above
23045         field.
23046         (insert_init_seqs): New function.
23047         (tree_predictive_commoning_loop): Call insert_init_seqs.
23049 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23051         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23053 2017-07-28  Richard Biener  <rguenther@suse.de>
23055         * match.pd: Remove superfluous :c.
23056         * genmatch.c (simplify::id): Add member.
23057         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23058         Copy id.
23059         (current_id): New global.
23060         (dt_node::parent): Move from ...
23061         (dt_operand::parent): ... here.  Add for_id member.
23062         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23063         (decision_tree::find_node): Relax order requirement when
23064         merging DT_TRUE nodes to ones inbetween the current simplify
23065         and the one we try to merge with.  Add diagnostic whenever
23066         we need to enforce pattern order by not merging.
23067         (decision_tree::insert): Set current_id.
23068         (decision_tree::print_node): Dump parent node and for_id.
23069         (parser::last_id): Add member.
23070         (parser::push_simplify): Assign unique id.
23071         (parser::parser): Initialize last_id.
23073 2017-07-28  Martin Liska  <mliska@suse.cz>
23075         PR sanitizer/81340
23076         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23077         gimple_build_debug_bind.
23079 2017-07-28  Richard Biener  <rguenther@suse.de>
23081         PR tree-optimization/81502
23082         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23083         BIT_FIELD_REF.
23084         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23085         size/pos operands.
23086         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23087         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23088         for BIT_FIELD_REF args.
23089         * fold-const.c (make_bit_field_ref): Likewise.
23090         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23092 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23094         PR sanitizer/80998
23095         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23096         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23097         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23098         Or it into SANITIZER_UNDEFINED.
23099         * ubsan.c: Include gimple-fold.h and varasm.h.
23100         (ubsan_expand_ptr_ifn): New function.
23101         (instrument_pointer_overflow): New function.
23102         (maybe_instrument_pointer_overflow): New function.
23103         (instrument_object_size): Formatting fix.
23104         (pass_ubsan::execute): Call instrument_pointer_overflow
23105         and maybe_instrument_pointer_overflow.
23106         * internal-fn.c (expand_UBSAN_PTR): New function.
23107         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23108         * sanitizer.def (__ubsan_handle_pointer_overflow,
23109         __ubsan_handle_pointer_overflow_abort): New builtins.
23110         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23111         * internal-fn.def (UBSAN_PTR): New internal function.
23112         * opts.c (sanitizer_opts): Add pointer-overflow.
23113         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23114         * fold-const.c (build_range_check): Compute pointer range check in
23115         integral type if pointer arithmetics would be needed.  Formatting
23116         fixes.
23118 2017-07-28  Martin Liska  <mliska@suse.cz>
23120         PR sanitizer/81460
23121         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23122         parameters that are of a variable-length.
23124 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23126         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23127         rs6000/biarch64.h.
23128         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23129         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23130         (CRT_CALL_STATIC_FUNCTION): Likewise.
23131         (ASM_DEFAULT_SPEC): New define.
23132         (ASM_SPEC32): Likewise.
23133         (ASM_SPEC64): Likewise.
23134         (ASM_SPEC_COMMON): Likewise.
23135         (ASM_SPEC): Likewise.
23136         (INVALID_64BIT): Likewise.
23137         (LINK_OS_DEFAULT_SPEC): Likewise.
23138         (LINK_OS_SPEC32): Likewise.
23139         (LINK_OS_SPEC64): Likewise.
23140         (POWERPC_LINUX): Likewise.
23141         (PTRDIFF_TYPE): Likewise.
23142         (RESTORE_FP_PREFIX): Likewise.
23143         (RESTORE_FP_SUFFIX): Likewise.
23144         (SAVE_FP_PREFIX): Likewise.
23145         (SAVE_FP_SUFFIX): Likewise.
23146         (SIZE_TYPE): Likewise.
23147         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23148         (TARGET_64BIT): Likewise.
23149         (TARGET_64BIT): Likewise.
23150         (TARGET_AIX): Likewise.
23151         (WCHAR_TYPE_SIZE): Likewise.
23152         (WCHAR_TYPE): Undefine.
23153         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23154         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23155         (CPP_OS_RTEMS_SPEC): Delete.
23156         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23157         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23158         link_os_spec64.
23159         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23161 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23163         PR tree-optimization/81578
23164         * tree-parloops.c (build_new_reduction): Bail out if
23165         reduction_code isn't one of the standard OpenMP reductions.
23166         Move the details printing after that decision.
23168 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23170         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23171         related to reload_in_progress.
23172         (splat_input_operand): Likewise.
23173         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23174         Delete prototype.
23175         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23176         field.
23177         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23178         (TARGET_INSTANTIATE_DECLS): Likewise.
23179         (legitimate_indexed_address_p): Delete reload_in_progress code.
23180         (rs6000_debug_legitimate_address_p): Likewise.
23181         (rs6000_eliminate_indexed_memrefs): Likewise.
23182         (rs6000_emit_le_vsx_store): Likewise.
23183         (rs6000_emit_move_si_sf_subreg): Likewise.
23184         (rs6000_emit_move): Likewise.
23185         (register_to_reg_type): Likewise.
23186         (rs6000_pre_atomic_barrier): Likewise.
23187         (rs6000_machopic_legitimize_pic_address): Likewise.
23188         (rs6000_allocate_stack_temp): Likewise.
23189         (rs6000_address_for_fpconvert): Likewise.
23190         (rs6000_address_for_altivec): Likewise.
23191         (rs6000_secondary_memory_needed_rtx): Delete function.
23192         (rs6000_check_sdmode): Likewise.
23193         (rs6000_alloc_sdmode_stack_slot): Likewise.
23194         (rs6000_instantiate_decls): Likewise.
23195         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23196         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23197         Delete reload_in_progress.
23198         (*vec_reload_and_plus_<mptrsize>): Likewise.
23199         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23200         (vsx_div_v2di): Likewise.
23201         (vsx_udiv_v2di): Likewise.
23203 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23205         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23206         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23207         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23208         (rs6000_debug_reg_global): Delete print of LRA status.
23209         (rs6000_option_override_internal): Delete dead LRA related code.
23210         (rs6000_lra_p): Delete function.
23211         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23213 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23215         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23216         * config/riscv/rtems.h: New file.
23218 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23219             Sudakshina Das  <sudi.das@arm.com>
23221         * config/aarch64/aarch64.md
23222         (define_split for and<mode>3nr_compare): Move
23223         non aarch64_logical_operand to a register.
23224         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23225         register immediate operand to a register.
23226         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23228 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23230         PR middle-end/81564
23231         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23233 2017-07-27  Richard Biener  <rguenther@suse.de>
23235         PR tree-optimization/81573
23236         PR tree-optimization/81494
23237         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23238         multi defuse cycle case.
23240 2017-07-27  Richard Biener  <rguenther@suse.de>
23242         PR tree-optimization/81571
23243         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23244         PHIs.
23246 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23248         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23249         earlier and only if MASK_FPU is set.  Adjust formatting.
23251 2017-07-27  Martin Liska  <mliska@suse.cz>
23253         * opt-functions.awk: Add validation of value of Init.
23254         * optc-gen.awk: Pass new argument.
23256 2017-07-27  Martin Liska  <mliska@suse.cz>
23258         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23259         Fix wrong condition.
23261 2017-07-27  Martin Liska  <mliska@suse.cz>
23263         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23264         BBs and edges seen by autoFDO.
23266 2017-07-27  Richard Biener  <rguenther@suse.de>
23268         PR tree-optimization/81502
23269         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23270         with incompatible but same sized type.
23271         (execute_update_addresses_taken): Likewise.
23273 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23275         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23276         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23278 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23280         PR target/81534
23281         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23282         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23283         Change s_operand to memory_operand.
23285 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23287         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23288         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23289         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23290         Emit instructions rather than returning an expression.  Handle TFmode
23291         and KFmode by casting to TImode.
23292         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23293         (rs6000_emit_le_vsx_store): Likewise.
23294         * config/rs6000/vsx.md (VSX_TI): New iterator.
23295         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23296         (*vsx_le_undo_permute_<mode>): Likewise.
23297         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23298         emit the split sequence.
23299         (*vsx_le_perm_store_<mode>): Likewise.
23301 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23303         PR tree-optimization/81555
23304         PR tree-optimization/81556
23305         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23306         if true, force CHANGED for the recursive invocation.
23307         (reassociate_bb): Remember original length of ops array, pass
23308         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23310         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23311         attributes for noipa attribute.  For naked attribute use
23312         lookup_attribute first before lookup_attribute_spec.
23313         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23314         noipa attribute.
23315         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23316         for functions with noipa attribute.
23317         (cgraph_externally_visible_p): Return true for functions with noipa
23318         attribute.
23319         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23320         for functions with noipa attribute.
23321         * doc/extend.texi: Document noipa function attribute.
23322         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23323         also for functions with noipa attribute.
23324         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23326 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23328         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23329         vec_unalign_load_cost and vec_unalign_store_cost.
23331 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23333         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23334         -mvsx-small-integer option.
23335         (ISA_3_0_MASKS_IEEE): Likewise.
23336         (OTHER_VSX_VECTOR_MASKS): Likewise.
23337         (POWERPC_MASKS): Likewise.
23338         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23339         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23340         code, only testing for DImode being allowed in non-VSX floating
23341         point registers.
23342         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23343         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23344         another VSX test.
23345         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23346         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23347         TARGET_P8_VECTOR test.
23348         (rs6000_secondary_reload_simple_move): Likewise.
23349         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23350         since TARGET_P9_VECTOR was already tested.
23351         (rs6000_opt_masks): Remove -mvsx-small-integer.
23352         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23353         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23354         used.
23355         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23356         test for TARGET_VEXTRACTUB was used, and that uses
23357         TARGET_P9_VECTOR.
23358         (p9 extract splitter): Likewise.
23359         (vsx_extract_<mode>_di_p9): Likewise.
23360         (vsx_extract_<mode>_store_p9): Likewise.
23361         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23362         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23363         the elimination of TARGET_VSX_SMALL_INTEGER.
23364         (vsx_extract_<mode>_p8): Likewise.
23365         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23366         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23367         (vsx_set_<mode>_p9): Likewise.
23368         (vsx_set_v4sf_p9): Likewise.
23369         (vsx_set_v4sf_p9_zero): Likewise.
23370         (vsx_insert_extract_v4sf_p9): Likewise.
23371         (vsx_insert_extract_v4sf_p9_2): Likewise.
23372         * config/rs6000/rs6000.md (sign extend splitter): Change
23373         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23374         (floatsi<mode>2_lfiwax_mem): Likewise.
23375         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23376         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23377         since a test for TARGET_P9_VECTOR was used.
23378         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23379         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23380         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23381         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23382         TARGET_P8_VECTOR test.
23383         (fix_trunc<mode>si2_stfiwx): Likewise.
23384         (fix_trunc<mode>si2_internal): Likewise.
23385         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23386         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23387         used.
23388         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23389         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23390         TARGET_P8_VECTOR test.
23391         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23392         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23393         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23394         used.
23395         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23396         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23397         since a test for TARGET_P9_VECTOR was used.
23398         (splitter for loading small constants): Likewise.
23400 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23402         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23403         vec_fp_stmt_cost.
23405 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23407         PR target/81563
23408         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23409         (fp_valid_at): Likewise.
23411 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23413         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23414         (qdf24xx_addrcost_table): Likewise.
23415         (cortexa57_tunings): Update to use generic_branch_cost.
23416         (cortexa72_tunings): Likewise.
23417         (cortexa73_tunings): Likewise.
23418         (qdf24xx_tunings): Likewise.
23420 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23422         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23423         (thunderx2t99_branch_cost): Likewise.
23424         (cortexa35_tunings): Update to use generic_branch_cost.
23425         (cortexa53_tunings): Likewise.
23426         (cortexa57_tunings): Likewise.
23427         (cortexa72_tunings): Likewise.
23428         (cortexa73_tunings): Likewise.
23429         (thunderx2t99_tunings): Likewise.
23431 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23433         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23434         (sparc_option_override): Honour MASK_FSMULD.
23435         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23436         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23437         * config/sparc/sparc.opt (mfsmuld): New option.
23438         * doc/invoke.texi (mfsmuld): Document option.
23440 2017-07-26  Marek Polacek  <polacek@redhat.com>
23442         PR middle-end/70992
23443         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23445 2017-07-26  Richard Biener  <rguenther@suse.de>
23447         * gimple-match-head.c (do_valueize): Return OP if valueize
23448         returns NULL_TREE.
23449         (get_def): New helper to get at the def stmt of a SSA name
23450         if valueize allows.
23451         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23452         do_valueize to get at the def stmt.
23453         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23455 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23457         PR middle-end/46932
23458         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23460 2017-07-26  Martin Liska  <mliska@suse.cz>
23462         PR sanitize/81186
23463         * function.c (expand_function_start): Make expansion of
23464         nonlocal_goto_save_area after parm_birth_insn.
23466 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23468         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23469         from all CPU target flags enable members.
23471 2017-07-26  Richard Biener  <rguenther@suse.de>
23473         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23474         (decision_tree::gen): Make 'type' const.
23475         (write_predicate): Likewise.
23477 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23479         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23480         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23481         (rs6000_option_override_internal): Likewise.
23482         (rs6000_expand_vector_set): Likewise.
23483         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23484         (TARGET_UPPER_REGS_SF): Likewise.
23485         (TARGET_UPPER_REGS_DI): Likewise.
23486         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23487         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23488         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23489         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23490         (Splitters for DI constants in Altivec registers): Likewise.
23491         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23492         (vsx_set_v4sf_p9): Likewise.
23493         (vsx_set_v4sf_p9_zero): Likewise.
23494         (vsx_insert_extract_v4sf_p9): Likewise.
23495         (vsx_insert_extract_v4sf_p9_2): Likewise.
23497 2017-07-25  Carl Love  <cel@us.ibm.com>
23499         * doc/extend.texi: Update the built-in documentation file for the
23500         existing built-in functions
23501         vector signed char vec_cnttz (vector signed char);
23502         vector unsigned char vec_cnttz (vector unsigned char);
23503         vector signed short vec_cnttz (vector signed short);
23504         vector unsigned short vec_cnttz (vector unsigned short);
23505         vector signed int vec_cnttz (vector signed int);
23506         vector unsigned int vec_cnttz (vector unsigned int);
23507         vector signed long long vec_cnttz (vector signed long long);
23508         vector unsigned long long vec_cnttz (vector unsigned long long);
23510 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23512         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23513         accesses where the use is for the first operand of a BIT_INSERT.
23515 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23517         PR bootstrap/81521
23518         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23519         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23521 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23523         * config/i386/gstabs.h: Delete.
23524         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23526 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23528         * config/i386/i386.c (ix86_decompose_address): Do not check for
23529         register RTX when looking at index_reg or base_reg.
23530         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23532 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23534         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23535         to update EH info here.
23537 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23539         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23541 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23543         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23545 2017-07-25  Torsten Duwe  <duwe@suse.de>
23547         * common.opt: Introduce -fpatchable-function-entry
23548         command line option, and its variables function_entry_patch_area_size
23549         and function_entry_patch_area_start.
23550         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23551         including a two-value parser.
23552         * target.def (print_patchable_function_entry): New target hook.
23553         * targhooks.h (default_print_patchable_function_entry): New function.
23554         * targhooks.c (default_print_patchable_function_entry): Likewise.
23555         * toplev.c (process_options): Switch off IPA-RA if
23556         patchable function entries are being generated.
23557         * varasm.c (assemble_start_function): Look at the
23558         patchable-function-entry command line switch and current
23559         function attributes and maybe generate NOP instructions by
23560         calling the print_patchable_function_entry hook.
23561         * doc/extend.texi: Document patchable_function_entry attribute.
23562         * doc/invoke.texi: Document -fpatchable_function_entry
23563         command line option.
23564         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23565         New target hook.
23566         * doc/tm.texi: Re-generate.
23568 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23570         PR target/81532
23571         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23572         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23574 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23576         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23578 2017-07-25  Richard Biener  <rguenther@suse.de>
23580         PR tree-optimization/81455
23581         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23582         not walk in cycles when looking for guards.
23584 2017-07-25  Richard Biener  <rguenther@suse.de>
23586         PR tree-optimization/81529
23587         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23588         when optimizing backedge uses.
23590 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23592         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23593         character for AIX.
23594         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23595         to dl_section_ref.  On AIX, append an expression to subtract
23596         the size of the section length to dl_section_ref.
23598 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23600         * configure.ac: If any of the config.* scripts fail, exit 1.
23601         * configure: Regenerate.
23603 2017-07-25  Richard Biener  <rguenther@suse.de>
23605         PR middle-end/81546
23606         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23607         of immediate uses, be more verbose on errors.
23609 2017-07-25  Richard Biener  <rguenther@suse.de>
23611         PR tree-optimization/81510
23612         * tree-vect-loop.c (vect_is_simple_reduction): When the
23613         reduction stmt is not inside the loop bail out.
23615 2017-07-25  Richard Biener  <rguenther@suse.de>
23617         PR tree-optimization/81303
23618         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23619         profitability check against LOOP_VINFO_NITERSM1.
23621 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23623         * domwalk.c (cmp_bb_postorder): Simplify.
23624         (sort_bbs_postorder): New function.  Use it...
23625         (dom_walker::walk): ...here to optimize common cases.
23627 2017-07-25  Martin Liska  <mliska@suse.cz>
23629         PR ipa/81520
23630         * ipa-visibility.c (function_and_variable_visibility): Make the
23631         redirection just on target that supports aliasing.
23632         Fix GNU coding style.
23634 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23636         PR libgcc/61152
23637         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23638         Format changes.
23639         * config/arm/rtems.h: Likewise.
23640         * config/bfin/rtems.h: Likewise.
23641         * config/i386/rtemself.h: Likewise.
23642         * config/lm32/rtems.h: Likewise.
23643         * config/m32c/rtems.h: Likewise.
23644         * config/m68k/rtemself.h: Likewise.
23645         * config/microblaze/rtems.h: Likewise.
23646         * config/mips/rtems.h: Likewise.
23647         * config/moxie/rtems.h: Likewise.
23648         * config/nios2/rtems.h: Likewise.
23649         * config/powerpcspe/rtems.h: Likewise.
23650         * config/rs6000/rtems.h: Likewise.
23651         * config/rtems.h: Likewise.
23652         * config/sh/rtems.h: Likewise.
23653         * config/sh/rtemself.h: Likewise.
23654         * config/sparc/rtemself.h: Likewise.
23656 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23658         PR 81487
23659         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23660         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23661         * tree-ssa-structalias.c (alias_get_name): Same.
23663 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23665         PR target/81414
23666         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23667         instructions if no du chain is found.
23669 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23671         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23673 2017-07-25  Richard Biener  <rguenther@suse.de>
23675         PR middle-end/81505
23676         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23677         sticky.
23679 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23681         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23682         upper-regs options.
23683         (ISA_2_7_MASKS_SERVER): Likewise.
23684         (ISA_3_0_MASKS_IEEE): Likewise.
23685         (OTHER_P8_VECTOR_MASKS): Likewise.
23686         (OTHER_VSX_VECTOR_MASKS): Likewise.
23687         (POWERPC_MASKS): Likewise.
23688         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23689         duplicate list of options.
23690         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23691         explicit -mupper-regs options.
23692         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23693         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23694         alias for -mupper-regs-df.
23695         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23696         (rs6000_init_hard_regno_mode_ok): Likewise.
23697         (rs6000_option_override_internal): Likewise.
23698         (rs6000_opt_masks): Likewise.
23699         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23700         options in terms of whether -mvsx or -mpower8-vector was used.
23701         (TARGET_UPPER_REGS_DI): Likewise.
23702         (TARGET_UPPER_REGS_SF): Likewise.
23703         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23704         -mupper-regs-* options.
23706 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23708         * passes.c (emergency_dump_function): Print some empty lines and a
23709         header before the RTL dump.
23711 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23713         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23715 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23717         PR target/79041
23718         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23719         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23721 2017-07-24  Carl Love  <cel@us.ibm.com>
23723         * config/rs6000/rs6000-c.c: Add support for built-in functions
23724         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23725         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23726         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23727         vec_extract_fp_from_shortl): Add defines for the two builtins.
23728         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23729         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23730         new builtins.
23731         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23732         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23733         * doc/extend.texi: Update the built-in documentation file for the
23734         new built-in function.
23736 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23738         PR bootstrap/81521
23739         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23740         documentation.
23741         * doc/generic.texi: Likewise.
23742         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23743         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23745 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23747         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23748         (aarch64_mls_elt_merge<mode>): Likewise.
23750 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23752         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23753         having __cxa_atexit.
23755 2017-07-23  Michael Collison  <michael.collison@arm.com>
23757         * config/arm/arm.c (arm_option_override): Deprecate
23758         use of -mstructure-size-boundary.
23759         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23760         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23762 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23764         PR target/80695
23765         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23766         Reduce cost estimate for direct moves.
23768 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23770         PR target/80569
23771         * config/i386/i386.c (ix86_option_override_internal): Disable
23772         BMI, BMI2 and TBM instructions for -m16.
23774 2017-07-21  Carl Love  <cel@us.ibm.com>
23776         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23777         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23778         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23779         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23780         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23781         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23782         VMULOSW): New enum "unspec" values.
23783         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23784         altivec_vmulosw): New patterns.
23785         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23786         VMULOSW): Add definitions.
23788 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23790         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23791         (qdf24xx): Likewise.
23792         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23793         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23794         (AARCH64_FL_V8_1): Renumber.
23795         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23796         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23797         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23798         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23799         rdma to feature modifiers list.
23801 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23803         PR middle-end/56727
23804         * ipa-visibility (function_and_variable_visibility): Convert
23805         recursive PLT call to direct call if appropriate.
23807 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23809         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23810         operand 1 to see if the types precision matches.
23811         * fold-const.c (operand_equal_p): Likewise.
23813 2017-07-21  Richard Biener  <rguenther@suse.de>
23815         PR tree-optimization/81303
23816         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23817         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23818         (vect_peeling_hash_get_lowest_cost): Adjust.
23819         (vect_enhance_data_refs_alignment): Likewise.  Use
23820         vect_get_peeling_costs_all_drs to compute the penalty for no
23821         peeling to match up costs.
23823 2017-07-21  Richard Biener  <rguenther@suse.de>
23825         PR tree-optimization/81500
23826         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23827         we didn't identify a reduction path.
23829 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23830             Cesar Philippidis  <cesar@codesourcery.com>
23832         PR gcov-profile/81442
23833         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23834         probabilities.
23836 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23838         PR lto/81430
23839         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23840         function.
23841         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23842         nvptx_override_options_after_change.
23844 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23846         * dwarf2out.c (output_file_names): Avoid double testing for
23847         dwarf_version >= 5.
23849 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23851         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23853 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23855         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23856         hot/cold regions.
23857         (try_crossjump_to_edge): Do not punt on partitioned functions.
23859 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23861         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23862         Put all BBs reachable only via paths crossing cold region to cold
23863         region.
23864         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23866 2016-07-21  Richard Biener  <rguenther@suse.de>
23868         PR tree-optimization/81303
23869         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23870         into account prologue and epilogue iterations when raising
23871         min_profitable_iters to sth at least covering one vector iteration.
23873 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23875         * config/arm/arm.c (arm_test_cpu_arch_dat):
23876         Check for overlap.
23878 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23880         Remove TYPE_METHODS.
23881         * tree.h (TYPE_METHODS): Delete.
23882         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23883         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23884         (dbxout_type_methods): Scan TYPE_FIELDS.
23885         (dbxout_type): Don't check TYPE_METHODS here.
23886         * function.c (use_register_for_decl): Always ignore register for
23887         class types when not optimizing.
23888         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23889         * tree.c (free_lang_data_in_type): Stitch out member functions and
23890         templates from TYPE_FIELDS.
23891         (build_distinct_type_copy, verify_type_variant,
23892         verify_type): Member fns are on TYPE_FIELDS.
23893         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23894         * tree-pretty-print.c (dump_generic_node): Likewise.
23896 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23898         PR target/80846
23899         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23900         V2TImode and V4TImode.
23901         (ix86_expand_vector_extract): Likewise.
23902         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23903         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23904         (ssescalarmode): Handle V4TImode and V2TImode.
23905         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23906         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23907         (VEXTRACTI128_MODE): New mode iterator.
23908         (splitter for *vec_extractv?ti first element): New.
23909         (VEC_INIT_MODE): New mode iterator.
23910         (vec_init<mode>): Consolidate 3 expanders into one using
23911         VEC_INIT_MODE mode iterator.
23913 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23915         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23916         non_spilled_static_chain_regno_p.
23918 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23920         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23922 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23924         * bb-reorder.c (connect_traces): Allow copying of blocks within
23925         single partition.
23927 2017-07-20  Richard Biener  <rguenther@suse.de>
23929         * gimple.h (gimple_phi_result): Add gphi * overload.
23930         (gimple_phi_result_ptr): Likewise.
23931         (gimple_phi_arg): Likewise.  Adjust index assert to only
23932         allow actual argument accesses rather than all slots available
23933         by capacity.
23934         (gimple_phi_arg_def): Add gphi * overload.
23935         * tree-phinodes.c (make_phi_node): Initialize only actual
23936         arguments.
23937         (resize_phi_node): Clear memory not covered by old node,
23938         do not initialize excess argument slots.
23939         (reserve_phi_args_for_new_edge): Initialize new argument slot
23940         completely.
23942 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
23944         PR tree-optimization/81388
23945         Revert r238585:
23946         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
23948         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
23949         by removing computation of may_be_zero.
23951 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23952             Tom de Vries  <tom@codesourcery.com>
23954         PR middle-end/81030
23955         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
23956         when gimple level profile disagrees with what RTL expander did.
23958 2017-07-20  Richard Biener  <rguenther@suse.de>
23960         PR tree-optimization/61171
23961         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
23962         (vect_analyze_stmt): Add slp instance parameter.
23963         (vectorizable_reduction): Likewise.
23964         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
23965         (vect_is_simple_reduction): Deal with chains not detected
23966         as SLP reduction chain, specifically not properly associated
23967         chains containing a mix of plus/minus.
23968         (get_reduction_op): Remove.
23969         (get_initial_defs_for_reduction): Simplify, pass in whether
23970         this is a reduction chain, pass in the SLP node for the PHIs.
23971         (vect_create_epilog_for_reduction): Get the SLP instance as
23972         arg and adjust.
23973         (vectorizable_reduction): Get the SLP instance as arg.
23974         During analysis remember the SLP node with the PHIs in the
23975         instance.  Simplify getting at the vectorized reduction PHIs.
23976         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
23977         through SLP instance.
23978         (vect_slp_analyze_operations): Likewise.
23979         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
23980         (vect_transform_stmt): Likewise.
23982 2017-07-20  Tom de Vries  <tom@codesourcery.com>
23984         PR tree-optimization/81489
23985         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
23986         read of phi arg location to before loop that modifies phi.
23988 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
23990         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
23991         New pattern.
23993 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23995         PR middle-end/81331
23996         * except.c (execute): Fix ordering issue.
23998 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24000         PR rtl-optimization/81423
24001         * combine.c (make_compound_operation_int): Don't try to optimize
24002         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
24004 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24006         PR rtl-optimization/81423
24007         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
24008         with a constant that is -1 in the truncated to mode.
24010 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24012         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
24013         (determine_unlikely_bbs): ... here.
24014         * predict.h (propagate_unlikely_bbs_forward): Declare.
24015         * cfgexpand.c (pass_expand::execute): Use it.
24016         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
24017         unlikely edges.
24018         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
24019         propagate_unlikely_bbs_forward.
24021 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24023         PR middle-end/81331
24024         * except.c (maybe_add_nop_after_section_switch): New function.
24025         (execute): Use it.
24027 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24029         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24031 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24033         * gimple.h (gimple_phi_arg): Make assert more strict.
24035 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24037         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24038         * config/rs6000/mmintrin.h: New file.
24039         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24041 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24043         PR tree-optimization/81346
24044         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24046 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24048         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24050 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24052         * config/nvptx/nvptx-modes.def: Add V2DImode.
24053         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24054         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24055         (nvptx_output_mov_insn): Handle lack of mov.b128.
24056         (nvptx_print_operand): Handle 'H' and 'L' codes.
24057         (nvptx_vector_mode_supported): Allow V2DImode.
24058         (nvptx_preferred_simd_mode): New function.
24059         (nvptx_data_alignment): New function.
24060         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24061         nvptx_preferred_simd_mode.
24062         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24063         64 to 128 bits.
24064         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24066 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24068         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24069         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24070         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24071         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24072         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24073         (mov<VECIM>_insn): New define_insn.
24074         (define_expand "mov<VECIM>): New define_expand.
24076 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24078         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24080 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24082         PR tree-optimization/81346
24083         * fold-const.h (fold_div_compare, range_check_type): Declare.
24084         * fold-const.c (range_check_type): New function.
24085         (build_range_check): Use range_check_type.
24086         (fold_div_compare): No longer static, rewritten into
24087         a match.pd helper function.
24088         (fold_comparison): Don't call fold_div_compare here.
24089         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24090         as helper function.
24092 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24094         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24095         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24096         * tree.c (find_decls_types_r, verify_type): Use
24097         TYPE_{MIN,MAX}_VALUE_RAW.
24098         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24099         (hash_tree): Likewise.
24100         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24101         Likewise.
24102         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24103         Likewise.
24105 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24107         PR middle-end/81464
24108         * omp-expand.c (expand_omp_for_static_chunk): Handle
24109         equal-argument loop exit phi.
24111 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24113         PR target/81471
24114         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24115         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24116         operand 2 predicate.
24117         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24118         operand 2 predicate.
24119         (ror,rol -> rorx splitters): Use const_int_operand as
24120         operand 2 predicate.
24122 2017-06-18  Richard Biener  <rguenther@suse.de>
24124         PR tree-optimization/81410
24125         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24126         the gap in the ! slp_perm SLP case after each group.
24128 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24130         PR middle-end/81463
24131         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24132         again.
24134 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24136         PR middle-end/81462
24137         * predict.c (set_even_probabilities): Cleanup; do not affect
24138         probabilities that are already known.
24139         (combine_predictions_for_bb): Call even when count is set.
24141 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24143         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24144         TYPE_MAX_VALUE.
24146 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24148         PR target/81408
24149         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24150         optimization for loop niter analysis.
24152 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24154         PR target/81473
24155         * config/avr/avr.c (avr_optimize_casesi): Don't use
24156         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24158 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24160         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24161         body_cost_vec from _vect_peel_extended_info.
24162         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24163         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24164         npeel.
24166 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24168         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24170 2017-07-18  Richard Biener  <rguenther@suse.de>
24172         PR tree-optimization/80620
24173         PR tree-optimization/81403
24174         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24175         info when re-using a VN table entry.
24177 2017-07-18  Richard Biener  <rguenther@suse.de>
24179         PR tree-optimization/81418
24180         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24181         vectype_in.  Verify that with lane-reducing reduction operations
24182         we have a single def-use cycle.
24184 2017-07-17  Carl Love  <cel@us.ibm.com>
24186         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24188         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24189         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24190         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24191         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24192         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24193         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24194         VMULOSW): New enum "unspec" values.
24195         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24196         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24197         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24198         altivec_vmulosw): New patterns.
24199         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24200         VMULOSW): Add definitions.
24202 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24204         * config/alpha/alpha.c: Include predict.h.
24206 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24208         * tree-vrp.c (compare_assert_loc): Fix comparison function
24209         to return predictable results.
24211 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24213         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24214         option.
24215         (subdi3): Likewise.
24216         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24217         * doc/invoke.texi (mexpand-adddi): Update text.
24219 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24221         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24222         that also clobbers the CC register. The old expand code is moved
24223         to ...
24224         (*arc_clzsi2): ... here.
24225         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24226         the CC register. The old expand code is moved to ...
24227         (arc_ctzsi2): ... here.
24229 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24231         * config/arc/arc.opt (mindexed-loads): Use initial value
24232         TARGET_INDEXED_LOADS_DEFAULT.
24233         (mauto-modify-reg): Use initial value
24234         TARGET_AUTO_MODIFY_REG_DEFAULT.
24235         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24236         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24237         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24238         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24240 2017-07-17  Martin Liska  <mliska@suse.cz>
24242         PR sanitizer/81302
24243         * opts.c (finish_options): Do not allow -fgnu-tm
24244         w/ -fsanitize={kernel-,}address.  Say sorry.
24246 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24248         PR target/81369
24249         * tree-loop-distribution.c (classify_partition): Only assert on
24250         numer of iterations.
24251         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24252         (distribute_loop): Remove code handling loop with unknown niters.
24253         (pass_loop_distribution::execute): Skip loop with unknown niters.
24255 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24257         PR target/81369
24258         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24259         function sort_partitions_by_post_order.
24261 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24263         PR tree-optimization/81374
24264         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24265         the max index of basic blocks, rather than number of basic blocks.
24267 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24269         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24270         proto.
24271         (arc_legitimate_pic_operand_p): Likewise.
24272         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24273         function.
24274         (arc_needs_pcl_p): Likewise.
24275         (arc_legitimate_pc_offset_p): Likewise.
24276         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24277         function is also used in constrains.md.
24278         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24279         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24280         PLUS.  Only return true/false in known cases, otherwise assert.
24281         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24282         is already called in arc_legitimate_constant_p.
24283         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24284         pic addresses.
24285         (LEGITIMATE_PIC_OPERAND_P): Use
24286         arc_raw_symbolic_reference_mentioned_p function.
24287         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24288         function.
24289         (Cal): Likewise.
24290         (C32): Likewise.
24292 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24293         Andrew Burgess  <andrew.burgess@embecosm.com>
24295         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24296         (arc_return_address_register): New function.
24297         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24298         (arc_handle_fndecl_attribute): Add naked attribute.
24299         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24300         (TARGET_WARN_FUNC_RETURN): Likewise.
24301         (arc_allocate_stack_slots_for_args): New function.
24302         (arc_warn_func_return): Likewise.
24303         (machine_function): Change type fn_type.
24304         (arc_compute_function_type): Consider new naked function type,
24305         change function return type.
24306         (arc_must_save_register): Adapt to handle new
24307         arc_compute_function_type's return type.
24308         (arc_expand_prologue): Likewise.
24309         (arc_expand_epilogue): Likewise.
24310         (arc_return_address_regs): Delete.
24311         (arc_return_address_register): New function.
24312         (arc_epilogue_uses): Use above function.
24313         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24314         (arc_function_type): Change encoding, add naked type.
24315         (ARC_INTERRUPT_P): Change to handle the new encoding.
24316         (ARC_FAST_INTERRUPT_P): Likewise.
24317         (ARC_NORMAL_P): Define.
24318         (ARC_NAKED_P): Likewise.
24319         (arc_compute_function_type): Delete prototype.
24320         * config/arc/arc.md (in_ret_delay_slot): Use
24321         arc_return_address_register function.
24322         (simple_return): Likewise.
24323         (p_return_i): Likewise.
24325 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24327         PR tree-optimization/81428
24328         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24329         can't be built for those types.
24331 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24333         Remove stuff dead since r239246.
24335         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24336         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24337         (avr_inform_devices): Remove dead stuff.
24339 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24341         * config/arm/arm_neon.h: Fix softp typo.
24343 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24345         PR tree-optimization/81365
24346         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24347         aggregate moves onto bb predecessor edges, make sure there are no
24348         loads that could alias the lhs in between the start of bb and the
24349         loads from *phi.
24351 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24353         PR 80929
24354         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24355         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24356         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24358 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24360         PR tree-optimization/81396
24361         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24362         (init_symbolic_number): Initialize it to 1.
24363         (perform_symbolic_merge): Add n_ops from both operands into the new
24364         n_ops.
24365         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24366         without base_addr as useless if they need more than one operation.
24367         (bswap_replace): Handle !bswap case for NULL base_addr.
24369 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24371         PR target/81069
24372         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24373         as possible.
24375 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24377         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24378         conditional builtin define __FIX_LEON3FT_B2BST.
24380 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24382         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24383         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24384         with -mfix-ut700.
24386 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24388         PR rtl-optimization/81424
24389         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24390         to remove potential trapping from operands if -fnon-call-exceptions.
24392 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24394         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24395         profile_proability for scalling.
24396         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24398 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24400         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24402 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24404         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24405         fixpoint arithmetics.
24407 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24409         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24410         fixpoint arithmetics.
24412 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24414         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24415         fixpoint arithmetics.
24417 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24419         * profile-count.h (profile_probability::from_reg_br_prob_note,
24420         profile_probability::to_reg_br_prob_note): New functions.
24421         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24422         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24423         * predict.c (probability_reliable_p): Update.
24424         (edge_probability_reliable_p): Update.
24425         (br_prob_note_reliable_p): Update.
24426         (invert_br_probabilities): Update.
24427         (add_reg_br_prob_note): New function.
24428         (combine_predictions_for_insn): Update.
24429         * asan.c (asan_clear_shadow): Update.
24430         * cfgbuild.c (compute_outgoing_frequencies): Update.
24431         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24432         (update_br_prob_note): Update.
24433         (rtl_verify_edges): Update.
24434         (purge_dead_edges): Update.
24435         (fixup_reorder_chain): Update.
24436         * emit-rtl.c (try_split): Update.
24437         * ifcvt.c (cond_exec_process_insns): Update.
24438         (cond_exec_process_if_block): Update.
24439         (dead_or_predicable): Update.
24440         * internal-fn.c (expand_addsub_overflow): Update.
24441         (expand_neg_overflow): Update.
24442         (expand_mul_overflow): Update.
24443         * loop-doloop.c (doloop_modify): Update.
24444         * loop-unroll.c (compare_and_jump_seq): Update.
24445         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24446         * predict.h: Update.
24447         * reorg.c (mostly_true_jump): Update.
24448         * rtl.h: Update.
24449         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24450         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24451         * config/arc/arc.c: (emit_unlikely_jump): Update.
24452         * config/arm/arm.c: (emit_unlikely_jump): Update.
24453         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24454         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24455         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24456         (ix86_print_operand): Update.
24457         (ix86_split_fp_branch): Update.
24458         (predict_jump): Update.
24459         * config/ia64/ia64.c (ia64_print_operand): Update.
24460         * config/mmix/mmix.c (mmix_print_operand): Update.
24461         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24462         (rs6000_expand_split_stack_prologue): Update.
24463         * config/rs6000/rs6000.c: Update.
24464         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24465         (s390_expand_vec_movstr): Update.
24466         (s390_expand_cs_tdsi): Update.
24467         (s390_expand_split_stack_prologue): Update.
24468         * config/sh/sh.c (sh_print_operand): Update.
24469         (expand_cbranchsi4): Update.
24470         (expand_cbranchdi4): Update.
24471         * config/sparc/sparc.c (output_v9branch): Update.
24472         * config/spu/spu.c (get_branch_target): Update.
24473         (ea_load_store_inline): Update.
24474         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24475         * config/tilepro/tilepro.c: Update.
24477 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24479         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24480         (gimplify_save_expr): Likewise.
24482 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24484         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24486 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24488         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24489         TV_IPA_FNSUMMARY.
24490         * timevar.def (TV_IPA_FNSUMMARY): Define.
24492 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24494         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24495         to back store errata sensitive sequence from being generated.
24496         (sqrtdf2_fix): Likewise.
24498 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24500         * tree-ssa-threadupdate.c (compute_path_counts,
24501         update_joiner_offpath_counts): Use profile_probability.
24503 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24505         Revert:
24506         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24508         * config/arm/arm-c.c (arm_cpu_builtins): Define
24509         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24511 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24513         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24514         array entries to represent __ieee128 versions of the
24515         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24516         scalar_extract_sig, and scalar_insert_exp built-in functions.
24517         (altivec_resolve_overloaded_builtin): Add special case handling
24518         for the __builtin_scalar_insert_exp function, as represented by
24519         the P9V_BUILTIN_VEC_VSIEDP constant.
24520         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24521         exponent support for __ieee128 argument.
24522         (VSESQP): Add scalar extract signature support for __ieee128
24523         argument.
24524         (VSTDCNQP): Add scalar test negative support for __ieee128
24525         argument.
24526         (VSIEQP): Add scalar insert exponent support for __int128 argument
24527         with __ieee128 result.
24528         (VSIEQPF): Add scalar insert exponent support for __ieee128
24529         argument with __ieee128 result.
24530         (VSTDCQP): Add scalar test data class support for __ieee128
24531         argument.
24532         (VSTDCNQP): Add overload support for scalar test negative with
24533         __ieee128 argument.
24534         (VSTDCQP): Add overload support for scalar test data class
24535         __ieee128 argument.
24536         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24537         UNSPEC_VSX_SXSIGDP.
24538         (UNSPEC_VSX_SIEXPQP): New constant.
24539         (xsxexpqp): New insn for VSX scalar extract exponent quad
24540         precision.
24541         (xsxsigqp): New insn for VSX scalar extract significand quad
24542         precision.
24543         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24544         precision with floating point argument.
24545         (xststdcqp): New expand for VSX scalar test data class quad
24546         precision.
24547         (xststdcnegqp): New expand for VSX scalar test negative quad
24548         precision.
24549         (xststdcqp): New insn to match expansions for VSX scalar test data
24550         class quad precision and VSX scalar test negative quad precision.
24551         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24552         special case operand checking to enforce that second operand of
24553         VSX scalar test data class with quad precision argument is a 7-bit
24554         unsigned literal.
24555         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24556         prototypes and descriptions of __ieee128 versions of
24557         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24558         scalar_test_data_class, and scalar_test_neg built-in functions.
24560 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24562         PR tree-optimization/81162
24563         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24564         replace a negate with an add.
24566 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24568         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24570 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24572         * config/arm/arm-c.c (arm_cpu_builtins): Define
24573         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24575 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24577         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24578         (armv8-r): Set ARM Cortex-R52 as default CPU.
24579         * config/arm/arm-tables.opt: Regenerate.
24580         * config/arm/arm-tune.md: Regenerate.
24581         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24582         Cortex-R52.
24583         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24584         extension for -mcpu=cortex-r52.
24586 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24588         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24589         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24590         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24591         (fp-armv8): Define it as FP_ARMv8 only.
24592         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24593         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24594         TARGET_FPU_ARMV8.
24595         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24596         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24597         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24598         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24599         than TARGET_FPU_ARMV8.
24600         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24601         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24602         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24603         TARGET_FPU_ARMV8.
24604         * config/arm/neon.md (neon_vrint): Likewise.
24605         (neon_vcvt): Likewise.
24606         (neon_<fmaxmin_op><mode>): Likewise.
24607         (<fmaxmin><mode>3): Likewise.
24608         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24609         * config/arm/predicates.md (arm_cond_move_operator): Check against
24610         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24612 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24614         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24615         to top of function.
24617 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24619         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24620         loop in comment with memset.
24622 2017-07-14  Martin Liska  <mliska@suse.cz>
24624         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24625         * dwarf2out.c (is_java): Remove the function.
24626         (output_pubname): Remove usage of the function.
24627         (lower_bound_default): Remove usage of DW_LANG_Java.
24628         (gen_compile_unit_die): Likewise.
24629         * gcc.c: Remove compiler defaults for .java and .zip files.
24630         * gimple-expr.c (remove_suffix): Change as there's no longer
24631         extension than 4-letter one.
24632         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24633         (gimplify_save_expr): Likewise.
24634         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24635         as it's possible even for other languages than Java.
24636         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24637         * lto-opts.c (lto_write_options): Remove reference to Java.
24638         * opts.c (strip_off_ending): Update file extension handling.
24639         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24640         * tree-eh.c (lower_resx): Likewise.
24641         * tree.c (free_lang_data_in_type): Remove dead code.
24642         (find_decls_types_r): Likewise.
24643         (build_common_builtin_nodes): Remove Java from a comment.
24644         (verify_type): Remove dead code.
24645         * varasm.c (assemble_external): Remove Java from a comment.
24647 2017-07-14  Martin Liska  <mliska@suse.cz>
24649         * opts.c (finish_options): Add quotes.
24650         (common_handle_option): Likewise.
24652 2017-07-14  Martin Liska  <mliska@suse.cz>
24654         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24655         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24656         Remove N_SO_PASCAL.
24657         * dwarf2out.c (lower_bound_default): Do not handle
24658         DW_LANG_Pascal83.
24659         (gen_compile_unit_die): Likewise.
24660         * gcc.c: Remove default extension binding for GNU Pascal.
24661         * stmt.c: Remove Pascal language from a comment.
24662         * xcoffout.c: Likewise.
24664 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24666         PR c/81405
24667         * diagnostic-show-locus.c (fixit_cmp): New function.
24668         (layout::layout): Sort m_fixit_hints.
24669         (column_range::column_range): Assert that the values are valid.
24670         (struct char_span): New struct.
24671         (correction::overwrite): New method.
24672         (struct source_line): New struct.
24673         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24674         calls in terms of classes source_line and char_span, and
24675         correction::overwrite.
24676         (selftest::test_overlapped_fixit_printing_2): New function.
24677         (selftest::diagnostic_show_locus_c_tests): Call it.
24679 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24681         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24682         early if there is no lhs.
24684 2017-07-13  Martin Liska  <mliska@suse.cz>
24686         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24687         (gen_reference_type_die): Likewise.
24688         * stor-layout.c: Remove Pascal-related comment.
24690 2017-07-13  Martin Liska  <mliska@suse.cz>
24692         * opts.c (finish_options): Add quotes to error messages.
24693         (parse_sanitizer_options): Likewise.
24695 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24697         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24699 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24701         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24703 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24705         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24706         during expansion.
24707         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24709 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24711         PR target/81193
24712         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24713         provides the hardware capability bits, define the macro
24714         __BUILTIN_CPU_SUPPORTS__.
24715         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24716         if GLIBC does not provide the hardware capability bits.  Add a
24717         gcc_unreachable call if the built-in cpu function is neither
24718         __builtin_cpu_is nor __builtin_cpu_supports.
24719         (rs6000_get_function_versions_dispatcher): Change the warning
24720         that an old GLIBC is used which does not export the capability
24721         bits to be an error.
24722         * doc/extend.texi (target_clones attribute): Document the
24723         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24724         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24725         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24726         the macros defined by GCC if the newer GLIBC is available.
24728 2017-07-12  Jeff Law  <law@redhat.com>
24730         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24731         remaining includes slightly.
24732         * config/riscv/riscv-builtins.c: Include profile-count.h.
24734 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24736         PR target/79883
24737         * config/avr/avr.c (avr_set_current_function): In diagnostic
24738         messages: Quote keywords and (parts of) identifiers.
24739         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24740         "INTERUPT".
24742 2017-07-12  Carl Love  <cel@us.ibm.com>
24744         * config/rs6000/rs6000-c.c: Add support for built-in functions
24745         vector bool char vec_revb (vector bool char);
24746         vector bool short vec_revb (vector short char);
24747         vector bool int vec_revb (vector bool int);
24748         vector bool long long vec_revb (vector bool long long);
24749         * doc/extend.texi: Update the built-in documentation file for the
24750         new built-in functions.
24752 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24754         * config/s390/s390.md: Remove movcc splitter.
24756 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24758         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24759         load/store on condition.
24761 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24763         PR target/81407
24764         * config/avr/avr.c (avr_encode_section_info)
24765         [progmem && !TREE_READONLY]: Error if progmem object needs
24766         constructing.
24768 2017-07-11  Michael Collison  <michael.collison@arm.com>
24770         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24771         New pattern.
24773 2017-07-11  Carl Love  <cel@us.ibm.com>
24775         * config/rs6000/rs6000-c.c: Add support for builtins
24776         vector unsigned int vec_parity_lsbb (vector signed int);
24777         vector unsigned int vec_parity_lsbb (vector unsigned int);
24778         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24779         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24780         vector unsigned long long vec_parity_lsbb (vector signed long long);
24781         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24782         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24783         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24784         * doc/extend.texi: Update the built-in documentation file for the
24785         new built-in functions.
24787 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24789         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24790         (layout::m_primary_loc): New field.
24791         (layout::layout): Initialize new field.  Move location filtering
24792         logic from here to...
24793         (layout::maybe_add_location_range): ...this new method.  Add
24794         support for filtering to just the lines already specified by other
24795         locations.
24796         (layout::will_show_line_p): New method.
24797         (gcc_rich_location::add_location_if_nearby): New method.
24798         (selftest::test_add_location_if_nearby): New test function.
24799         (selftest::diagnostic_show_locus_c_tests): Call it.
24800         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24801         New method.
24803 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24805         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24806         (bb_first_real_insn): New function.
24807         (nvptx_single): Add extra initialization of broadcasted condition
24808         variables.
24810 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24812         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24814 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24816         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24817         Binutils doc as TEXI will mess them up.
24818         * doc/invoke.texi (AVR Options): Same here.
24820 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24822         * config/sparc/sparc.opt (mfix-ut700): New option.
24823         (mfix-gr712rc): Likewise.
24824         (sparc_fix_b2bst): New variable.
24825         * doc/invoke.texi (SPARC options): Document them.
24826         (ARM options): Fix warnings.
24827         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24828         instructions to prevent sequences that can trigger the store-store
24829         errata for certain LEON3FT processors.
24830         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24831         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24832         * config/sparc/sparc.md (fix_b2bst): New attribute.
24833         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24835 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24837         PR target/81375
24838         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24839         (rcpps): Ditto.
24840         (*rsqrtsf2_sse): Ditto.
24841         (rsqrtsf2): Ditto.
24842         (div<mode>3): Macroize insn from divdf3 and divsf3
24843         using MODEF mode iterator.
24845 2017-07-10  Martin Sebor  <msebor@redhat.com>
24847         PR tree-optimization/80397
24848         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24849         instead of testing for equality to INTEGER_TYPE.
24851 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24853         * config.gcc: Remove uclibc from arc target spec.
24855 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24857         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24859 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24861         PR lto/80838
24862         * lto-wrapper.c (remove_option): New function.
24863         (merge_and_complain): Merge PIC/PIE options more realistically.
24865 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24867         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24869         PR target/20296
24870         PR target/81268
24871         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24872         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24873         * config.in: Regenerate.
24874         * configure: Regenerate.
24875         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24876         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24877         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24878         (TARGET_GASISR_PROLOGUES): ...target mask.
24879         * common/config/avr/avr-common.c
24880         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24881         Set -mgas-isr-prologues.
24882         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24883         INSERT_PASS_BEFORE for it.
24884         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24885         * config/avr/avr.c (avr_option_override)
24886         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24887         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24888         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24889         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24890         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24891         and rtl_opt_pass.
24892         (make_avr_pass_pre_proep): New function.
24893         (emit_push_sfr) <treg>: Add argument to function and use it
24894         instead of TMP_REG.
24895         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24896         and set machine->gasisr.yes.
24897         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24898         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24899         __gcc_isr.n_pushed to .L__stack_usage.
24900         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24901         (avr_asm_final_postscan_insn): ...this new static function.
24902         * config/avr/avr.h (machine_function)
24903         <is_no_gccisr, use_L__stack_usage>: New fields.
24904         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24905         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24906         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24907         (gasisr, *gasisr): New expander and insn.
24908         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24909         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24910         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24912 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24914         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24915         in quoted strings.
24917 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24919         Move jump-tables out of .text again.
24921         PR target/81075
24922         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24923         (ASM_OUTPUT_ADDR_VEC): New function.
24924         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24925         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24926         INSN_ADDRESSes as asm comment.
24927         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24928         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24929         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24930         * config/avr/avr.md (*tablejump): Adjust comment.
24931         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24932         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24933         New detail.
24934         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24935         (avr_output_addr_vec): New proto.
24936         (avr_log_t) <insn_addresses>: New field.
24938 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
24940         PR target/81313
24941         * config/i386/i386.c (ix86_function_arg_advance): Set
24942         outgoing_args_on_stack to true if there are outgoing arguments
24943         on stack.
24944         (ix86_function_arg): Likewise.
24945         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
24946         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
24947         * config/i386/i386.h (machine_function): Add
24948         outgoing_args_on_stack.
24950 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
24952         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
24953         supporting pthreds.
24954         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
24956 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
24958         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
24959         (REAL_H): Remove $(MACHMODE_H).
24960         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
24961         double-int.h.
24962         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
24963         $(MACHMODE_H) and double-int.h.
24964         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
24965         $(MACHMODE_H).
24966         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
24967         double-int.h.
24969 2017-07-07  Andrew Pinski  <apinski@cavium.com>
24971         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
24972         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
24974 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24976         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
24977         Add warning if GCC was not configured to link against a GLIBC that
24978         exports the hardware capability bits.
24979         (make_resolver_func): Make resolver function private and not a
24980         COMDAT function.  Create the name with clone_function_name instead
24981         of make_unique_name.
24983         PR target/81348
24984         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
24985         correct operand in doing the split.
24987 2017-07-07 Carl Love  <cel@us.ibm.com>
24989         * config/rs6000/rs6000-c: Add support for built-in function
24990         vector unsigned short vec_pack_to_short_fp32 (vector float,
24991                                                       vector float).
24992         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
24993         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
24994         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
24995         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
24996         (convert_4f32_8i16): Add define_expand.
24997         * doc/extend.texi: Update the built-in documentation file for the
24998         new built-in function.
25000 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25002         * config/sparc/m8.md: New file.
25003         * config/sparc/sparc.md: Include m8.md.
25005 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25007         * config/sparc/sparc.opt: New option -mvis4b.
25008         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
25009         (sparc_option_override): Handle VIS4B.
25010         (enum sparc_builtins): Define
25011         SPARC_BUILTIN_DICTUNPACK{8,16,32},
25012         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
25013         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
25014         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
25015         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
25016         (check_constant_argument): New function.
25017         (sparc_vis_init_builtins): Define builtins
25018         __builtin_vis_dictunpack{8,16,32},
25019         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
25020         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
25021         __builtin_vis_fpcmpde{8,16,32}shl and
25022         __builtin_vis_fpcmpur{8,16,32}shl.
25023         (sparc_expand_builtin): Check that the constant operands to
25024         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25025         constant and in range.
25026         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25027         TARGET_VIS4B.
25028         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25029         (SPARC_IMM5_P): Likewise.
25030         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25031         (enabled): Handle vis4b.
25032         (UNSPEC_DICTUNPACK): New unspec.
25033         (UNSPEC_FPCMPSHL): Likewise.
25034         (UNSPEC_FPUCMPSHL): Likewise.
25035         (UNSPEC_FPCMPDESHL): Likewise.
25036         (UNSPEC_FPCMPURSHL): Likewise.
25037         (cpu_feature): New CPU feature `vis4b'.
25038         (dictunpack{8,16,32}): New insns.
25039         (FPCSMODE): New mode iterator.
25040         (fpcscond): New code iterator.
25041         (fpcsucond): Likewise.
25042         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25043         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25044         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25045         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25046         * config/sparc/constraints.md: Define constraints `q' for unsigned
25047         2-bit integer constants and `t' for unsigned 5-bit integer
25048         constants.
25049         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25050         predicate.
25051         (imm5_operand_dictunpack16): Likewise.
25052         (imm5_operand_dictunpack32): Likewise.
25053         (imm2_operand): Likewise.
25054         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25055         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25056         ditunpack* and fpcmp*shl builtins.
25058 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25060         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25061         * config.in: Add HAVE_AS_SPARC6 define.
25062         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25063         M8.
25064         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25065         TARGET_CPU_m8.
25066         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25067         (CPP_CPU_SPEC): Handle m8.
25068         (ASM_CPU_SPEC): Likewise.
25069         * config/sparc/sparc-opts.h (enum processor_type): Add
25070         PROCESSOR_M8.
25071         * config/sparc/sparc.c (m8_costs): New struct.
25072         (sparc_option_override): Handle TARGET_CPU_m8.
25073         (sparc32_initialize_trampoline): Likewise.
25074         (sparc64_initialize_trampoline): Likewise.
25075         (sparc_issue_rate): Likewise.
25076         (sparc_register_move_cost): Likewise.
25077         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25078         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25079         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25080         (CPP_CPU_SPEC): Handle M8.
25081         (ASM_CPU_SPEC): Likewise.
25082         (AS_M8_FLAG): Define.
25083         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25084         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25085         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25086         M8 instructions.
25087         * configure: Regenerate.
25088         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25089         -mtune=m8.
25091 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25093         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25094         subtypes.
25095         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25096         ("*movdi_insn_sp32"): Do not set v3pipe.
25097         ("*movsi_insn"): Likewise.
25098         ("*movdi_insn_sp64"): Likewise.
25099         ("*movsf_insn"): Likewise.
25100         ("*movdf_insn_sp32"): Likewise.
25101         ("*movdf_insn_sp64"): Likewise.
25102         ("*zero_extendsidi2_insn_sp64"): Likewise.
25103         ("*sign_extendsidi2_insn"): Likewise.
25104         ("*mov<VM32:mode>_insn"): Likewise.
25105         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25106         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25107         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25108         ("<vlop:code><VL:mode>3"): Likewise.
25109         ("*not_<vlop:code><VL:mode>3"): Likewise.
25110         ("*nand<VL:mode>_vis"): Likewise.
25111         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25112         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25113         ("one_cmpl<VL:mode>2"): Likewise.
25114         ("faligndata<VM64:mode>_vis"): Likewise.
25115         ("alignaddrsi_vis"): Likewise.
25116         ("alignaddrdi_vis"): Likweise.
25117         ("alignaddrlsi_vis"): Likewise.
25118         ("alignaddrldi_vis"): Likewise.
25119         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25120         ("bmaskdi_vis"): Likewise.
25121         ("bmasksi_vis"): Likewise.
25122         ("bshuffle<VM64:mode>_vis"): Likewise.
25123         ("cmask8<P:mode>_vis"): Likewise.
25124         ("cmask16<P:mode>_vis"): Likewise.
25125         ("cmask32<P:mode>_vis"): Likewise.
25126         ("pdistn<P:mode>_vis"): Likewise.
25127         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25129 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25131         * config/sparc/sparc.md ("subtype"): New insn attribute.
25132         ("*wrgsr_sp64"): Set insn subtype.
25133         ("*rdgsr_sp64"): Likewise.
25134         ("alignaddrsi_vis"): Likewise.
25135         ("alignaddrdi_vis"): Likewise.
25136         ("alignaddrlsi_vis"): Likewise.
25137         ("alignaddrldi_vis"): Likewise.
25138         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25139         ("fexpand_vis"): Likewise.
25140         ("fpmerge_vis"): Likewise.
25141         ("faligndata<VM64:mode>_vis"): Likewise.
25142         ("bshuffle<VM64:mode>_vis"): Likewise.
25143         ("cmask8<P:mode>_vis"): Likewise.
25144         ("cmask16<P:mode>_vis"): Likewise.
25145         ("cmask32<P:mode>_vis"): Likewise.
25146         ("fchksm16_vis"): Likewise.
25147         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25148         ("fmean16_vis"): Likewise.
25149         ("fp<plusminus_insn>64_vis"): Likewise.
25150         ("<plusminus_insn>v8qi3"): Likewise.
25151         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25152         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25153         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25154         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25155         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25156         ("*movqi_insn"): Likewise.
25157         ("*movhi_insn"): Likewise.
25158         ("*movsi_insn"): Likewise.
25159         ("movsi_pic_gotdata_op"): Likewise.
25160         ("*movdi_insn_sp32"): Likewise.
25161         ("*movdi_insn_sp64"): Likewise.
25162         ("movdi_pic_gotdata_op"): Likewise.
25163         ("*movsf_insn"): Likewise.
25164         ("*movdf_insn_sp32"): Likewise.
25165         ("*movdf_insn_sp64"): Likewise.
25166         ("*zero_extendhisi2_insn"): Likewise.
25167         ("*zero_extendqihi2_insn"): Likewise.
25168         ("*zero_extendqisi2_insn"): Likewise.
25169         ("*zero_extendqidi2_insn"): Likewise.
25170         ("*zero_extendhidi2_insn"): Likewise.
25171         ("*zero_extendsidi2_insn_sp64"): Likewise.
25172         ("ldfsr"): Likewise.
25173         ("prefetch_64"): Likewise.
25174         ("prefetch_32"): Likewise.
25175         ("tie_ld32"): Likewise.
25176         ("tie_ld64"): Likewise.
25177         ("*tldo_ldub_sp32"): Likewise.
25178         ("*tldo_ldub1_sp32"): Likewise.
25179         ("*tldo_ldub2_sp32"): Likewise.
25180         ("*tldo_ldub_sp64"): Likewise.
25181         ("*tldo_ldub1_sp64"): Likewise.
25182         ("*tldo_ldub2_sp64"): Likewise.
25183         ("*tldo_ldub3_sp64"): Likewise.
25184         ("*tldo_lduh_sp32"): Likewise.
25185         ("*tldo_lduh1_sp32"): Likewise.
25186         ("*tldo_lduh_sp64"): Likewise.
25187         ("*tldo_lduh1_sp64"): Likewise.
25188         ("*tldo_lduh2_sp64"): Likewise.
25189         ("*tldo_lduw_sp32"): Likewise.
25190         ("*tldo_lduw_sp64"): Likewise.
25191         ("*tldo_lduw1_sp64"): Likewise.
25192         ("*tldo_ldx_sp64"): Likewise.
25193         ("*mov<VM32:mode>_insn"): Likewise.
25194         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25195         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25197 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25199         * config/sparc/sparc.md ("type"): New insn type viscmp.
25200         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25201         viscmp.
25202         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25203         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25204         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25205         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25206         viscmp.
25207         ("n7_vis_logical_11cycle"): Likewise.
25208         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25209         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25210         * config/sparc/niagara.md ("niag_vis"): Likewise.
25211         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25212         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25214 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25216         * config/sparc/sparc.md: New instruction type `bmask'.
25217         (bmaskdi_vis): Use the `bmask' type.
25218         (bmasksi_vis): Likewise.
25219         * config/sparc/ultra3.md (us3_array): Likewise.
25220         * config/sparc/niagara7.md (n7_array): Likewise.
25221         * config/sparc/niagara4.md (n4_array): Likewise.
25222         * config/sparc/niagara2.md (niag2_vis): Likewise.
25223         (niag3_vis): Likewise.
25224         * config/sparc/niagara.md (niag_vis): Likewise.
25226 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25228         * ipa-comdats.c: Remove optimize check from gate.
25229         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25230         for functions not optimized.
25231         (ipa_fn_summary_read): Skip optimize check.
25232         (ipa_fn_summary_write): Likewise.
25233         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25234         is optimized.
25235         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25236         uninlinable.
25237         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25238         (check_callers): Give up on caller which is not optimized.
25239         (inline_small_functions): Likewise.
25240         (ipa_inline): Do not give up when not optimizing.
25241         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25242         away unoptimizes cdtors.
25243         (whole_program_function_and_variable_visibility): Do
25244         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25245         * ipa.c (process_references): Do not check optimize.
25246         (symbol_table::remove_unreachable_nodes): Update optimize check.
25247         (set_writeonly_bit): Update optimize check.
25248         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25249         (pass_ipa_single_use::gate): Remove.
25251 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25253         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25254         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25255         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25256         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25257         permute_load, permute_store, adjust_extract, adjust_splat,
25258         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25259         replace_swap_with_copy, dump_swap_insn_table,
25260         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25261         recombine_lvx_pattern, recombine_stvx_pattern,
25262         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25263         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25264         to file rs6000-p8swap.c.
25265         * config/rs6000/rs6000-p8swap.c: New file.
25266         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25267         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25268         and rs6000*-*-* targets.
25270 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25272         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25274 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25276         * lto-wrapper.c (merge_and_complain): Do not merge
25277         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25278         fsigned_zeros, ftrapping_math, fwrapv.
25279         (append_compiler_options): Do not track these options.
25280         (append_linker_options): Likewie
25282 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25284         * cgraphunit.c (cgraph_node::finalize_function): When
25285         !flag_toplevel_reorde set no_reorder flag.
25286         (varpool_node::finalize_decl): Likewise.
25287         (symbol_table::compile): Drop no toplevel reorder path.
25289 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25291         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25292         edges; zero probability is not better than uninitialized.
25294 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25296         * asan.h (asan_sanitize_allocas_p): Declare.
25297         * asan.c (asan_sanitize_allocas_p): New function.
25298         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25299         (handle_builtin_alloca): Likewise.
25300         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25301         if !asan_sanitize_allocas_p.
25302         * params.def (asan-instrument-allocas): Add new option.
25303         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25304         * opts.c (common_handle_option): Disable allocas sanitization for
25305         KASan by default.
25307 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25309         * asan.c: Include gimple-fold.h.
25310         (get_last_alloca_addr): New function.
25311         (handle_builtin_stackrestore): Likewise.
25312         (handle_builtin_alloca): Likewise.
25313         (asan_emit_allocas_unpoison): Likewise.
25314         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25315         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25316         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25317         (instrument_builtin_call): Pass gimple iterator to
25318         get_mem_refs_of_builtin_call.
25319         (last_alloca_addr): New global.
25320         * asan.h (asan_emit_allocas_unpoison): Declare.
25321         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25322         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25323         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25324         if function calls alloca.
25325         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25326         * gimple-fold.h (replace_call_with_value): Declare.
25327         * internal-fn.c: Include asan.h.
25328         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25329         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25331 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25333         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25334         (C_SELFTEST_FLAGS): New.
25335         (CPP_SELFTEST_FLAGS): New.
25336         (SELFTEST_DEPS): New, from deps of s-selftest.
25337         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25338         (CPP_SELFTEST_DEPS): New.
25339         (selftest): Add dependency on s-selftest-c++.
25340         (s-selftest): Rename to...
25341         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25342         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25343         than SELFTEST_FLAGS.
25344         (selftest-gdb): Rename to...
25345         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25346         C_SELFTEST_FLAGS.
25347         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25348         (selftest-valgrind): Rename to...
25349         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25350         C_SELFTEST_FLAGS.
25351         (selftest-valgrind): Reintroduce as an alias for
25352         selftest-c-valgrind.
25353         (s-selftest-c++): New.
25354         (selftest-c++-gdb): New.
25355         (selftest-c++-valgrind): New.
25357 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25359         * gcc.c (process_command): When deciding if undefined variables
25360         should be ignored when processing specs, accept "gcc -v" as well.
25362 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25364         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25365         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25367 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25369         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25370         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25371         * config/arm/arm-tables.opt: Regenerate.
25372         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25373         enumerator.
25374         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25376 2017-07-06  Carl Love  <cel@us.ibm.com>
25378         * ChangeLog: Clean up from mid air collision
25380 2017-07-06  Carl Love  <cel@us.ibm.com>
25382         * config/rs6000/rs6000-c.c: Add support for built-in functions
25383         vector signed int vec_subc (vector signed int, vector signed int);
25384         vector signed __int128 vec_subc (vector signed __int128,
25385                                          vector signed __int128);
25386         vector unsigned __int128 vec_subc (vector unsigned __int128,
25387                                            vector unsigned __int128);
25388         vector signed int vec_sube (vector signed int, vector signed int,
25389                                     vector signed int);
25390         vector unsigned int vec_sube (vector unsigned int,
25391                                       vector unsigned int,
25392                                       vector unsigned int);
25393         vector signed __int128 vec_sube (vector signed __int128,
25394                                          vector signed __int128,
25395                                          vector signed__int128);
25396         vector unsigned __int128 vec_sube (vector unsigned __int128,
25397                                            vector unsigned __int128,
25398                                            vector unsigned __int128);
25399         vector signed int vec_subec (vector signed int, vector signed int,
25400                                      vector signed int);
25401         vector unsigned int vec_subec (vector unsigned int,
25402                                        vector unsigned int,
25403                                        vector unsigned int);
25404         vector signed __int128 vec_subec (vector signed __int128,
25405                                           vector signed __int128,
25406                                           vector signed__int128);
25407         vector unsigned __int128 vec_subec (vector unsigned __int128,
25408                                             vector unsigned __int128,
25409                                             vector unsigned __int128);
25410         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25411         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25412         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25413         BU_ALTIVEC_OVERLOAD_X definitions.
25414         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25415         * doc/extend.texi: Update the built-in documentation file for the new
25416         built-in functions.
25418 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25420         PR c++/79300
25421         * diagnostic-show-locus.c (layout::layout): Use start and finish
25422         spelling location for the start and finish of each range.
25423         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25424         Add unused aspect param.
25425         * input.c (expand_location_1): Add "aspect" param, and use it
25426         to access the correct part of the location.
25427         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25428         expand_location_1.
25429         (expand_location_to_spelling_point): Likewise.
25430         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25431         param, and pass it to expand_location_1.
25433 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25435         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25436         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25437         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25438         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25439         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25440         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25441         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25442         _mm_maskz_getmant_ss): New intrinsics.
25443         (__builtin_ia32_getexpss128_mask): Changed to ...
25444         __builtin_ia32_getexpss128_round ... this.
25445         (__builtin_ia32_getexpsd128_mask): Changed to ...
25446         __builtin_ia32_getexpsd128_round ... this.
25447         * config/i386/i386-builtin-types.def
25448         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25449         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25450         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25451         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25452         __builtin_ia32_getmantss_mask_round): New builtins.
25453         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25454         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25455         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25456         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25457         * config/i386/sse.md
25458         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25459         avx512f_sgetexp<mode><mask_scalar_name>
25460         <round_saeonly_scalar_name> ... this.
25461         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25462         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25463         vgetexp<ssescalarmodesuffix>
25464         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25465         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25466         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25467         avx512f_vgetmant<mode><mask_scalar_name>
25468         <round_saeonly_scalar_name> ... this.
25469         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25470         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25471         vgetmant<ssescalarmodesuffix>
25472         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25473         %0<mask_scalar_operand4>, %1, %2
25474         <round_saeonly_scalar_mask_op4>, %3} ... this.
25475         * config/i386/subst.md (mask_scalar_operand4,
25476         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25477         round_saeonly_scalar_nimm_predicate): New subst attributes.
25479 2017-07-06  Julia Koval  <julia.koval@intel.com>
25481         * config/i386/i386.c (ix86_erase_embedded_rounding):
25482         Remove code for old rounding pattern.
25484 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25486         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25488 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25490         * doc/sourcebuild.texi (Test Directives, Variants of
25491         dg-require-support): Add documentation for dg-require-stack-check.
25493 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25495         * config/i386/subst.md (mask_scalar, round_scalar,
25496         round_saeonly_scalar): New meta-templates.
25497         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25498         round_scalar_mask_operand3, round_scalar_mask_op3,
25499         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25500         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25501         round_saeonly_scalar_constraint,
25502         round_saeonly_scalar_prefix): New subst attribute.
25503         * config/i386/sse.md
25504         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25505         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25506         <round_scalar_name> ... this.
25507         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25508         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25509         <round_scalar_name> ... this.
25510         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25511         <sse>_vm<code><mode>3<mask_scalar_name>
25512         <round_saeonly_scalar_name> ... this.
25513         (v<plusminus_mnemonic><ssescalarmodesuffix>
25514         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25515         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25516         v<plusminus_mnemonic><ssescalarmodesuffix>
25517         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25518         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25519         (v<multdiv_mnemonic><ssescalarmodesuffix>
25520         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25521         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25522         v<multdiv_mnemonic><ssescalarmodesuffix>
25523         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25524         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25525         (v<maxmin_float><ssescalarmodesuffix>
25526         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25527         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25528         v<maxmin_float><ssescalarmodesuffix>
25529         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25530         %0<mask_scalar_operand3>, %1, %<iptr>2
25531         <round_saeonly_scalar_mask_op3>} ... this.
25533 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25535         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25536         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25538 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25539             Alan Hayward  <alan.hayward@arm.com>
25540             David Sherwood  <david.sherwood@arm.com>
25542         * combine.c (simplify_if_then_else): Remove "enum" before
25543         "machine_mode".
25544         * compare-elim.c (can_eliminate_compare): Likewise.
25545         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25546         Likewise.
25547         (aarch64_lookup_simd_builtin_type): Likewise.
25548         (aarch64_simd_builtin_type): Likewise.
25549         (aarch64_init_simd_builtin_types): Likewise.
25550         (aarch64_simd_expand_args): Likewise.
25551         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25552         Likewise.
25553         (aarch64_reverse_mask): Likewise.
25554         (aarch64_simd_emit_reg_reg_move): Likewise.
25555         (aarch64_gen_adjusted_ldpstp): Likewise.
25556         (aarch64_ccmp_mode_to_code): Likewise.
25557         (aarch64_operands_ok_for_ldpstp): Likewise.
25558         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25559         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25560         Likewise.
25561         (aarch64_min_divisions_for_recip_mul): Likewise.
25562         (aarch64_reassociation_width): Likewise.
25563         (aarch64_get_condition_code_1): Likewise.
25564         (aarch64_simd_emit_reg_reg_move): Likewise.
25565         (aarch64_simd_attr_length_rglist): Likewise.
25566         (aarch64_reverse_mask): Likewise.
25567         (aarch64_operands_ok_for_ldpstp): Likewise.
25568         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25569         (aarch64_gen_adjusted_ldpstp): Likewise.
25570         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25571         Likewise.
25572         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25573         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25574         (arm_lookup_simd_builtin_type): Likewise.
25575         (arm_simd_builtin_type): Likewise.
25576         (arm_init_simd_builtin_types): Likewise.
25577         (arm_expand_builtin_args): Likewise.
25578         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25579         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25580         (ft32_setup_incoming_varargs): Likewise.
25581         (ft32_function_arg): Likewise.
25582         (ft32_function_arg_advance): Likewise.
25583         (ft32_pass_by_reference): Likewise.
25584         (ft32_arg_partial_bytes): Likewise.
25585         (ft32_valid_pointer_mode): Likewise.
25586         (ft32_addr_space_pointer_mode): Likewise.
25587         (ft32_addr_space_legitimate_address_p): Likewise.
25588         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25589         Likewise.
25590         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25591         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25592         (iamcu_alignment): Likewise.
25593         (canonicalize_vector_int_perm): Likewise.
25594         (ix86_noce_conversion_profitable_p): Likewise.
25595         (ix86_mpx_bound_mode): Likewise.
25596         (ix86_operands_ok_for_move_multiple): Likewise.
25597         * config/microblaze/microblaze-protos.h
25598         (microblaze_expand_conditional_branch_reg): Likewise.
25599         * config/microblaze/microblaze.c
25600         (microblaze_expand_conditional_branch_reg): Likewise.
25601         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25602         Likewise.
25603         (rs6000_reassociation_width): Likewise.
25604         (rs6000_invalid_binary_op): Likewise.
25605         (fusion_p9_p): Likewise.
25606         (emit_fusion_p9_load): Likewise.
25607         (emit_fusion_p9_store): Likewise.
25608         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25609         Likewise.
25610         (riscv_hard_regno_mode_ok_p): Likewise.
25611         (riscv_address_insns): Likewise.
25612         (riscv_split_symbol): Likewise.
25613         (riscv_legitimize_move): Likewise.
25614         (riscv_function_value): Likewise.
25615         (riscv_hard_regno_nregs): Likewise.
25616         (riscv_expand_builtin): Likewise.
25617         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25618         (riscv_build_integer): Likewise.
25619         (riscv_split_integer): Likewise.
25620         (riscv_legitimate_constant_p): Likewise.
25621         (riscv_cannot_force_const_mem): Likewise.
25622         (riscv_regno_mode_ok_for_base_p): Likewise.
25623         (riscv_valid_base_register_p): Likewise.
25624         (riscv_valid_offset_p): Likewise.
25625         (riscv_valid_lo_sum_p): Likewise.
25626         (riscv_classify_address): Likewise.
25627         (riscv_legitimate_address_p): Likewise.
25628         (riscv_address_insns): Likewise.
25629         (riscv_load_store_insns): Likewise.
25630         (riscv_force_binary): Likewise.
25631         (riscv_split_symbol): Likewise.
25632         (riscv_force_address): Likewise.
25633         (riscv_legitimize_address): Likewise.
25634         (riscv_move_integer): Likewise.
25635         (riscv_legitimize_const_move): Likewise.
25636         (riscv_legitimize_move): Likewise.
25637         (riscv_address_cost): Likewise.
25638         (riscv_subword): Likewise.
25639         (riscv_output_move): Likewise.
25640         (riscv_canonicalize_int_order_test): Likewise.
25641         (riscv_emit_int_order_test): Likewise.
25642         (riscv_function_arg_boundary): Likewise.
25643         (riscv_pass_mode_in_fpr_p): Likewise.
25644         (riscv_pass_fpr_single): Likewise.
25645         (riscv_pass_fpr_pair): Likewise.
25646         (riscv_get_arg_info): Likewise.
25647         (riscv_function_arg): Likewise.
25648         (riscv_function_arg_advance): Likewise.
25649         (riscv_arg_partial_bytes): Likewise.
25650         (riscv_function_value): Likewise.
25651         (riscv_pass_by_reference): Likewise.
25652         (riscv_setup_incoming_varargs): Likewise.
25653         (riscv_print_operand): Likewise.
25654         (riscv_elf_select_rtx_section): Likewise.
25655         (riscv_save_restore_reg): Likewise.
25656         (riscv_for_each_saved_reg): Likewise.
25657         (riscv_register_move_cost): Likewise.
25658         (riscv_hard_regno_mode_ok_p): Likewise.
25659         (riscv_hard_regno_nregs): Likewise.
25660         (riscv_class_max_nregs): Likewise.
25661         (riscv_memory_move_cost): Likewise.
25662         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25663         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25664         (rl78_addr_space_address_mode): Likewise.
25665         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25666         Likewise.
25667         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25668         (rs6000_reassociation_width): Likewise.
25669         (rs6000_invalid_binary_op): Likewise.
25670         (fusion_p9_p): Likewise.
25671         (emit_fusion_p9_load): Likewise.
25672         (emit_fusion_p9_store): Likewise.
25673         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25674         (ok_for_simple_move_operands): Likewise.
25675         (ok_for_simple_move_strict_operands): Likewise.
25676         (ok_for_simple_arith_logic_operands): Likewise.
25677         (visium_legitimize_reload_address): Likewise.
25678         (visium_select_cc_mode): Likewise.
25679         (output_cbranch): Likewise.
25680         (visium_split_double_move): Likewise.
25681         (visium_expand_copysign): Likewise.
25682         (visium_expand_int_cstore): Likewise.
25683         (visium_expand_fp_cstore): Likewise.
25684         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25685         (visium_function_arg): Likewise.
25686         (visium_function_arg_advance): Likewise.
25687         (visium_libcall_value): Likewise.
25688         (visium_setup_incoming_varargs): Likewise.
25689         (visium_legitimate_constant_p): Likewise.
25690         (visium_legitimate_address_p): Likewise.
25691         (visium_legitimize_address): Likewise.
25692         (visium_secondary_reload): Likewise.
25693         (visium_register_move_cost): Likewise.
25694         (visium_memory_move_cost): Likewise.
25695         (prepare_move_operands): Likewise.
25696         (ok_for_simple_move_operands): Likewise.
25697         (ok_for_simple_move_strict_operands): Likewise.
25698         (ok_for_simple_arith_logic_operands): Likewise.
25699         (visium_function_value_1): Likewise.
25700         (rtx_ok_for_offset_p): Likewise.
25701         (visium_legitimize_reload_address): Likewise.
25702         (visium_split_double_move): Likewise.
25703         (visium_expand_copysign): Likewise.
25704         (visium_expand_int_cstore): Likewise.
25705         (visium_expand_fp_cstore): Likewise.
25706         (visium_split_cstore): Likewise.
25707         (visium_select_cc_mode): Likewise.
25708         (visium_split_cbranch): Likewise.
25709         (output_cbranch): Likewise.
25710         (visium_print_operand_address): Likewise.
25711         * expmed.c (flip_storage_order): Likewise.
25712         * expmed.h (emit_cstore): Likewise.
25713         (flip_storage_order): Likewise.
25714         * genrecog.c (validate_pattern): Likewise.
25715         * hsa-gen.c (gen_hsa_addr): Likewise.
25716         * internal-fn.c (expand_arith_overflow): Likewise.
25717         * ira-color.c (allocno_copy_cost_saving): Likewise.
25718         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25719         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25720         (process_invariant_for_inheritance): Likewise.
25721         * lra-eliminations.c (move_plus_up): Likewise.
25722         * omp-low.c (lower_oacc_reductions): Likewise.
25723         * simplify-rtx.c (simplify_subreg): Likewise.
25724         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25725         (TARGET_CHKP_BOUND_MODE): Likewise..
25726         * targhooks.c (default_chkp_bound_mode): Likewise.
25727         (default_setup_incoming_vararg_bounds): Likewise.
25728         * targhooks.h (default_chkp_bound_mode): Likewise.
25729         (default_setup_incoming_vararg_bounds): Likewise.
25730         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25731         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25732         (have_whole_vector_shift): Likewise.
25733         * tree-vect-stmts.c (vectorizable_load): Likewise.
25734         * doc/tm.texi: Regenerate.
25736 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25738         Graceful degrade if Binutils PR21472 is not available.
25740         PR target/81072
25741         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25742         .rodata in flash test fails.
25743         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25744         * confgure: Regenerate.
25745         * config.in: Regenerate.
25746         * config/avr/avr.c (avr_asm_named_section)
25747         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25748         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25749         (avr_asm_init_sections): Same.
25751 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25753         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25754         (fma<VH:mode>4_intrinsic): Likewise.
25755         (*fmsub<VCVTF:mode>4): Likewise.
25756         (*fmsub<VH:mode>4_intrinsic): Likewise.
25758 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25760         PR target/81305
25761         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25762         Don't depend on "optimize > 0".
25763         (out_movhi_r_mr, out_movqi_mr_r): Same.
25764         (out_movhi_mr_r, out_movqi_r_mr): Same.
25765         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25766         io_address_operand on "optimize > 0".
25768 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25770         * tree-loop-distribution.c: Add general explanantion on the pass.
25771         (generate_loops_for_partition): Mark distributed loop.
25772         (pg_add_dependence_edges): New parameter.  Handle alias data
25773         dependence specially and record it in the parameter if asked.
25774         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25775         (init_partition_graph_vertices, add_partition_graph_edge): New.
25776         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25777         (free_partition_graph_vdata, build_partition_graph): New.
25778         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25779         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25780         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25781         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25782         (version_for_distribution_p, finalize_partitions): New.
25783         (distribute_loop): Handle alias data dependence specially.  Factor
25784         out loop fusion code as functions and call these functions.
25786 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25788         * tree-loop-distribution.c (classify_partition): New parameter and
25789         better handle reduction statement.
25790         (rdg_build_partitions): Revise comment.
25791         (distribute_loop): Compute statements in all partitions and pass it
25792         to classify_partition.
25794 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25796         * tree-loop-distribution.c (enum partition_type): New.
25797         (struct partition): New field type.
25798         (partition_merge_into): Add parameter.  Update partition type.
25799         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25800         (build_rdg_partition_for_vertex): Compute partition type.
25801         (rdg_build_partitions): Dump partition type.
25802         (distribute_loop): Update calls to partition_merge_into.
25804 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25806         * tree-loop-distribution.c (struct ddr_hasher): New.
25807         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25808         (ddrs_table): New.
25809         (classify_partition): Call get_data_dependence.
25810         (pg_add_dependence_edges): Ditto.
25811         (distribute_loop): Release data dependence hash table.
25813 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25815         * tree-loop-distribution.c (ref_base_address): Delete.
25816         (similar_memory_accesses): Rename ...
25817         (share_memory_accesses): ... to this.  Check if partitions access
25818         the same memory reference.
25819         (distribute_loop): Call share_memory_accesses.
25821 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25823         * tree-loop-distribution.c (struct partition): New field recording
25824         its data reference.
25825         (partition_alloc, partition_free): Init and release data refs.
25826         (partition_merge_into): Merge data refs.
25827         (build_rdg_partition_for_vertex): Collect data refs for partition.
25828         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25829         Update uses.
25830         (distribute_loop): Remve data refs from vertice data of partition
25831         graph.
25833 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25835         * tree-loop-distribution.c (params.h): Include header file.
25836         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25837         (datarefs_vec): New global var.
25838         (create_rdg_vertices): Use datarefs_vec directly.
25839         (free_rdg): Don't free data references.
25840         (build_rdg): Update use.  Don't free data references.
25841         (distribute_loop): Compute global variable for data references.
25842         Bail out if there are too many data references.
25844 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25846         * tree-loop-distribution.c (loop_nest): New global var.
25847         (build_rdg): Use loop directly, rather than loop nest.
25848         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25849         variable directly.
25850         (distribute_loop): Compute global variable loop nest.  Update use.
25852 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25854         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25855         (partition_merge_into): New parameter.  Dump reason for fusion.
25856         (distribute_loop): Update use of partition_merge_into.
25858 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25860         * tree-loop-distribution.c (bb_top_order_index): New.
25861         (bb_top_order_index_size, bb_top_order_cmp): New.
25862         (stmts_from_loop): Use topological order.
25863         (pass_loop_distribution::execute): Compute and release topological
25864         order for basic blocks.
25866 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25868         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25869         if no loops.
25871 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25873         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25874         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25875         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25876         * internal-fn.def (LOOP_DIST_ALIAS): New.
25877         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25878         (fold_loop_internal_call): ... this.
25879         (vect_loop_dist_alias_call): New function.
25880         (set_uid_loop_bbs): Call fold_loop_internal_call.
25881         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25882         internal calls.
25884 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25886         PR target/81300
25887         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25888         Require dead FLAGS_REG at the beginning of a peephole.
25890 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25892         PR target/81294
25893         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25894         arguments in the call to __builtin_ia32_sbb_u32.
25895         (_subborrow_u64): Swap _X and _Y arguments in the call to
25896         __builtin_ia32_sbb_u64.
25898 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25900         PR debug/81278
25901         * tree-vrp.c (compare_assert_loc): Turn into a function template
25902         with stable template parameter.  Only test if a->e is NULL,
25903         !a->e == !b->e has been verified already.  Use e == NULL or
25904         e != NULL instead of e or ! e tests.  If stable is true, don't use
25905         iterative_hash_expr, on the other side allow a or b or both NULL
25906         and sort the NULLs last.
25907         (process_assert_insertions): Sort using compare_assert_loc<false>
25908         instead of compare_assert_loc, later sort using
25909         compare_assert_loc<true> before calling process_assert_insertions_for
25910         in a loop.  Use break instead of continue once seen NULL pointer.
25912 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25914         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25915         Cortex-R7 and Cortex-R8 processors.
25917 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25919         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25920         uninitialized while src is not.
25922 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25924         * common/config/arm/arm-common.c: Adjust include path for
25925         arm-cpu-cdata.h
25926         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25927         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
25928         (arm-cpu-data.h): Likewise.
25929         (arm-cpu-cdata.h): Likewise.
25930         * config/arm/arm-cpu.h: Delete.
25931         * config/arm/arm-cpu-cdata.h: Delete.
25932         * config/arm/arm-cpu-data.h: Delete.
25934 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
25936         * config/arm/arm-cpus.in (cortex-a55): New.
25937         (cortex-a75): Likewise.
25938         (cortex-a75.cortex-a55): Likewise.
25939         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25940         cortex-a75.
25941         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
25942         * config/arm/arm-cpu-cdata.h: Regenerate.
25943         * config/arm/arm-cpu-data.h: Regenerate.
25944         * config/arm/arm-cpu.h: Regenerate.
25945         * config/arm/arm-tables.opt: Regenerate.
25946         * config/arm/arm-tune.md: Regenerate.
25948 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25950         * haifa-sched.c (sched_create_recovery_edges): Update profile.
25952 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25954         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
25955         probability.
25957 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
25959         PR tree-optimization/81292
25960         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
25961         full_string_p, also call adjust_related_strinfos if the adjustment
25962         is simple, otherwise invalidate related strinfos.
25964 2017-07-04  Martin Liska  <mliska@suse.cz>
25966         PR sanitizer/81040
25967         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
25968         newly created variable as DECL_IGNORED_P.
25970 2017-07-04  Martin Liska  <mliska@suse.cz>
25972         PR ipa/81293
25973         * ipa-inline.c (inline_small_functions):
25974         Use xstrdup_for_dump.
25976 2017-07-04  Tom de Vries  <tom@codesourcery.com>
25978         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
25980 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25982         PR target/81033
25983         * config/darwin.c (darwin_function_switched_text_sections):
25984         Fix spaces.
25986 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
25988         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
25990 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
25992         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
25994 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25996         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
25997         min_profitable_iters, and th as inclusive lower bounds.
25998         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
25999         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
26000         for min_profitable_iters and min_profitable_estimate.
26001         (vect_transform_loop): Treat th as an inclusive lower bound.
26002         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
26004 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26006         PR target/81033
26007         * config/darwin.c (darwin_function_switched_text_sections):
26008         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
26009         in two pieces, and suppress the use of buf.
26011 2017-07-03  Nathan Sidwell  <nathan@acm.org>
26013         * hash-table.h (hash_table_mod1): Fix indentation.
26015 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26017         PR middle-end/81290
26018         * predict.c (force_edge_cold): Be more careful about propagation
26019         backward.
26020         * profile-count.h (profile_probability::guessed,
26021         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
26022         New.
26023         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26025 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26027         * doc/invoke.texi (rcpc architecture extension): Document it.
26029 2017-07-03  Richard Biener  <rguenther@suse.de>
26031         PR tree-optimization/60510
26032         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26033         the scalar reduction PHI and use it.
26034         (vectorizable_reduction): Properly guard the single_defuse_cycle
26035         path for non-SLP reduction chains where we cannot use it.
26036         Rework reduc_def/index and vector type deduction.  Rework
26037         vector operand gathering during reduction op code-gen.
26038         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26039         chains dissolve the chain and leave it to non-SLP reduction
26040         handling.
26042 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26044         * tree-data-ref.h (dr_alignment): Declare.
26045         * tree-data-ref.c (dr_alignment): New function.
26046         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26047         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26048         set it.
26049         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26051 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26053         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26054         and base_misalignment fields.
26055         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26056         * tree-data-ref.c: Include builtins.h.
26057         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26058         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26059         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26060         * tree-vect-data-refs.c: Include tree-cfg.h.
26061         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26062         fields instead of calculating an alignment here.
26063         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26064         innermost_loop_behavior fields.
26066 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26068         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26069         field.
26070         (DR_STEP_ALIGNMENT): New macro.
26071         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26072         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26073         (create_data_ref): Print it.
26074         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26075         to tell whether the step preserves vector (mis)alignment.
26076         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26077         Move the check for an integer step and generalise to all INTEGER_CST.
26078         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26079         Print the outer step alignment.
26081 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26083         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26084         with offset_alignment.
26085         (DR_ALIGNED_TO): Delete.
26086         (DR_OFFSET_ALIGNMENT): New macro.
26087         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26088         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26089         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26090         (create_data_ref): Likewise.
26091         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26092         (vect_analyze_data_refs): Likewise.
26093         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26094         creating dummy innermost behavior.
26096 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26098         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26099         with a "innermost_loop_behavior *" and refeence tree.
26100         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26101         (create_data_ref): Update call accordingly.
26102         * tree-predcom.c (find_looparound_phi): Likewise.
26104 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26106         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26107         fields with dr_wrt_vec_loop.
26108         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26109         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26110         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26111         (vect_dr_behavior): New function.
26112         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26113         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26114         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26115         track whether the step preserves the misalignment.
26116         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26117         Use vect_dr_behavior.
26118         (vect_setup_realignment): Update call accordingly.
26119         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26120         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26121         call to vect_create_addr_base_for_vector_ref.
26122         (vect_create_cond_for_align_checks): Likewise.
26123         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26124         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26125         (vect_recog_mask_conversion_pattern): Likewise.
26126         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26127         (new_stmt_vec_info): Remove redundant zeroing.
26129 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26131         * common/config/arm/arm-common.c (arm_be8_option): New function.
26132         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26133         (ISA_ARMv6): Add isa_bit_be8.
26134         * config/arm/arm.h (arm_be8_option): Add prototype.
26135         (BE8_SPEC_FUNCTION): New define.
26136         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26137         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26138         (mlittle-endian): Similarly.
26139         (mbe8, mbe32): New options.
26140         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26141         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26143 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26145         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26147 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26149         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26150         (cleanup_tree_cfg_bb): Use it.
26151         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26152         New functions.
26153         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26155 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26157         PR bootstrap/81285
26158         * loop-doloop.c (add_test): Update profile.
26160 2017-07-03  Martin Liska  <mliska@suse.cz>
26162         PR sanitize/81040
26163         * sanopt.c (rewrite_usage_of_param): New function.
26164         (sanitize_rewrite_addressable_params): Likewise.
26165         (pass_sanopt::execute): Call rewrite_usage_of_param.
26167 2017-07-03  Richard Biener  <rguenther@suse.de>
26169         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26170         back to using VIEW_CONVERT_EXPR.
26172 2017-07-03  Martin Liska  <mliska@suse.cz>
26174         PR other/78366
26175         * doc/extend.texi: Document when a resolver function is
26176         generated for target_clones.
26178 2017-07-03  Martin Liska  <mliska@suse.cz>
26180         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26181         and shadow memory of auto variables which are subject of
26182         use-after-scope sanitization.
26183         (asan_expand_mark_ifn): Add do set only when is_poison.
26185 2016-07-03  Richard Biener  <rguenther@suse.de>
26187         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26188         reduction PHIs.
26189         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26190         (vectorizable_reduction): Perform reduction PHI creation when
26191         visiting a reduction PHI and adjust and simplify code generation
26192         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26193         (vect_transform_loop): Visit reduction PHIs.
26194         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26195         defs into the SLP tree.
26196         (vect_build_slp_tree): Reduction defs terminate the recursion.
26197         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26198         of reduction defs.
26199         (vect_get_vec_defs_for_stmt_copy): Export.
26200         (vect_get_vec_defs): Likewise.
26201         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26202         purpose.
26203         (vect_get_vec_defs_for_stmt_copy): Declare.
26204         (vect_get_vec_defs): Likewise.
26206 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26208         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26209         parameter with a "loop" parameter and use it instead of the
26210         loop containing DR_STMT.  Don't check simple_iv when doing
26211         BB analysis.  Describe the two analysis modes in the comment.
26213 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26215         PR tree-optimization/69468
26216         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26217         (find_same_succ_bb): Handle ignore_edge_flags.
26219 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26221         PR tree-optimization/81192
26222         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26223         hash.
26224         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26225         differs.
26226         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26228 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26230         PR tree-optimization/81192
26231         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26232         BB_SAME_SUCC (bb) == NULL.
26234 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26236         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26237         consistency.
26239 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26241         * dumpfile.c: Include profile-count.h
26242         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26243         update profile.
26244         (insert_cond_bb): Update profile.
26245         * tree-cfg.h (insert_cond_bb): Update prototype.
26246         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26247         * tree-dump.c: Do not include tree-cfg.
26249 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26251         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26253 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26255         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26256         bb.
26258 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26260         * tree-complex.c (expand_complex_div_wide): update profile.
26262 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26263             Alan Hayward  <alan.hayward@arm.com>
26264             David Sherwood  <david.sherwood@arm.com>
26266         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26267         (CORETYPES_H): New define.
26268         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26269         (insn-modes-inline.h, s-modes-inline-h): New rules.
26270         (generated_files): Add insn-modes-inline.h.
26271         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26272         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26273         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26274         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26275         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26276         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26277         (build/gencodes.o, build/genconditions.o): Likewise.
26278         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26279         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26280         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26281         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26282         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26283         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26284         * coretypes.h: Include everything up to real.h for generators.
26285         Include insn-modes.h first.  Include wide-int-print.h after
26286         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26287         * machmode.h: Don't include insn-modes.h here.
26288         * function-tests.c: Remove includes of signop.h, machmode.h,
26289         double-int.h and wide-int.h.
26290         * rtl.h: Likewise.
26291         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26292         and wide-int.h.
26293         * optc-save-gen.awk: Likewise.
26294         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26295         * godump.c: Remove include of wide-int-print.h.
26296         * pretty-print.h: Likewise.
26297         * wide-int-print.cc: Likewise.
26298         * wide-int.cc: Likewise.
26299         * hash-map-tests.c: Remove include of signop.h.
26300         * hash-set-tests.c: Likewise.
26301         * rtl-tests.c: Likewise.
26302         * mkconfig.sh: Remove include of machmode.h.
26303         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26304         into...
26305         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26306         into an insn-modes-inline.h header file, adding appropriate
26307         include guards and end comments.
26308         (emit_insn_modes_c_header): Remove include of machmode.h.
26309         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26310         machmode.h.
26311         (main): Handle -i flag and call emit_insn_modes_inline_h when
26312         it is passed.
26314 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26316         * tree-ssa-strlen.c (strinfo): Rename the length field to
26317         nonzero_chars.  Add a full_string_p field.
26318         (compare_nonzero_chars, zero_length_string_p): New functions.
26319         (get_addr_stridx): Add an offset_out parameter.
26320         Use compare_nonzero_chars.
26321         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26322         (new_strinfo): Update after above changes to strinfo.
26323         (set_endptr_and_length): Set full_string_p.
26324         (get_string_length): Update after above changes to strinfo.
26325         (unshare_strinfo): Update call to new_strinfo.
26326         (maybe_invalidate): Likewise.
26327         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26328         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26329         as a uhwi instead of an shwi.  Update after above changes to
26330         strinfo and new_strinfo.
26331         (zero_length_string): Assert that chainsi contains full strings.
26332         Use zero_length_string_p.  Update call to new_strinfo.
26333         (adjust_related_strinfos): Update after above changes to strinfo.
26334         Copy full_string_p from origsi.
26335         (adjust_last_stmt): Use zero_length_string_p.
26336         (handle_builtin_strlen): Update after above changes to strinfo and
26337         new_strinfo.  Install the lhs as the string length if the previous
26338         entry didn't describe a full string.
26339         (handle_builtin_strchr): Update after above changes to strinfo
26340         and new_strinfo.
26341         (handle_builtin_strcpy): Likewise.
26342         (handle_builtin_strcat): Likewise.
26343         (handle_builtin_malloc): Likewise.
26344         (handle_pointer_plus): Likewise.
26345         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26346         that aren't necessarily followed by a nul terminator.
26347         (handle_char_store): Likewise.
26349 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26351         PR tree-optimization/80769
26352         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26353         for malloc and calloc.  Document the new invariant that all related
26354         strinfos have delayed lengths or none do.
26355         (verify_related_strinfos): Move earlier in file.
26356         (set_endptr_and_length): New function, split out from...
26357         (get_string_length): ...here.  Also set the lengths of related
26358         strinfos.
26359         (zero_length_string): Assert that chainsi has known (rather than
26360         delayed) lengths.
26361         (adjust_related_strinfos): Likewise.
26363 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26365         PR tree-optimization/81136
26366         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26367         assert that two references with the same misalignment have the same
26368         compile-time misalignment if those compile-time misalignments
26369         are known.
26371 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26373         * print-tree.c (print_node): Print all attributes.
26375 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26377         * cfg.c (scale_bbs_frequencies): New function.
26378         * cfg.h (scale_bbs_frequencies): Declare it.
26379         * cfgloopanal.c (single_likely_exit): Cleanup.
26380         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26381         as parameter.
26382         (scale_loop_profile): Likewise.
26383         (loop_version): Likewise.
26384         (create_empty_loop_on_edge): Update.
26385         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26386         scale_loop_frequencies, scale_loop_profile, loopify,
26387         loop_version): Update prototypes.
26388         * modulo-sched.c (sms_schedule): Update.
26389         * predict.c (unlikely_executed_edge_p): Also check probability.
26390         (probably_never_executed_edge_p): Fix typo.
26391         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26392         * tree-parloops.c (gen_parallel_loop): Update.
26393         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26394         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26395         * tree-ssa-loop-split.c (split_loop): Update.
26396         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26397         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26398         (vect_loop_versioning): Update.
26399         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26401 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26403         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26405 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26407         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26408         to keep profile consistent.
26410 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26412         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26413         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26414         * profile-count.h (max_safe_multiplier): Make unsigned.
26415         (profile_count::guessed_zero): New.
26417 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26419         * bb-reorder.c (fix_up_crossing_landing_pad,
26420         fix_crossing_conditional_branches): Use make_single_succ_edge
26421         to keep profile consistent.
26423 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26425         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26426         to update profile.
26428 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26430         PR sanitizer/81262
26431         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26432         the right scopes, make sure cond_jump isn't preserved between multiple
26433         iterations.  Search for fallthru edge whenever there are 3+ edges and
26434         use find_fallthru_edge for it.
26436 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26438         Patch by Alexander Monakov <amonakov@ispras.ru>
26439         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26440         probabilities consistently.
26442 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26444         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26445         emit_cmp_and_jump_insns.
26447 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26449         PR ipa/81261
26450         * tree-inline.c (expand_call_inline): Combine profile statuses.
26452 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26454         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26455         fold_stmt returned true.
26457 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26459         * ggc.h (empty_string): Delete.
26460         * cfgexpand.c (expand_asm_stmt): Use plain "".
26461         * optabs.c (expand_asm_memory_barrier): Likewise.
26462         * stringpool.c (empty_string): Delete.
26463         (digit_vector, digit_string): Delete.
26464         (ggc_alloc_string): Use plain "", don't optimize single digit
26465         strings.  Use ggc_alloc_atomic.
26467 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26469         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26470         comparison set and one other set, use the cost of the non-comparison
26471         set.
26473 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26475         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26476         some formatting.
26478 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26480         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26481         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26482         just set removed labels to NULL_TREE.
26484 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26486         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26487         (set_range_info): ...here.  Only call set_range_info_raw if domain
26488         is useful.
26489         (set_nonzero_bits): Call set_range_info_raw.
26490         * tree-ssanames.h (set_range_info_raw): New.
26492 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26494         PR target/81225
26495         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26496         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26497         of nonimmediate_operand and <store_mask_constraint> instead of m
26498         for the input operand.  For V8FI iterator, always split if input
26499         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26500         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26501         <store_mask_predicate> instead of register_operand and
26502         <store_mask_constraint> instead of v for the input operand.  Make
26503         sure both operands aren't MEMs for if not <mask_applied>.
26505 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26507         * lto-wrapper.c (copy_file) Close both file descriptors before
26508         exiting normally.
26510 2017-06-30  Martin Liska  <mliska@suse.cz>
26512         PR ipa/81214
26513         * multiple_target.c (create_dispatcher_calls): Make ifunc
26514         also for function that don't have calls or are not referenced.
26516 2017-06-30  Richard Biener  <rguenther@suse.de>
26518         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26519         analyze the first scalar stmt.  Move vector type computation
26520         for the BB case here from ...
26521         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26522         live operation processing in the SLP case properly.
26524 2017-06-30  Richard Biener  <rguenther@suse.de>
26526         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26528 2017-06-30  Martin Liska  <mliska@suse.cz>
26530         PR sanitizer/81021
26531         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26532         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26534 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26536         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26537         ones.
26539 2017-06-30  Richard Biener  <rguenther@suse.de>
26541         PR tree-optimization/81249
26542         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26543         condition reduction result to original scalar type.
26545 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26547         * profile-count.h (enum profile_quality): Fix typos and whitespace
26548         issues.
26550 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26552         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26553         type for branch probabilities.
26555 2017-06-29  Julian Brown  <julian@codesourcery.com>
26556             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26558         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26559         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26560         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26561         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26563 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26565         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26566         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26567         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26568         CC usage from generic code to here.
26569         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26570         CC usage into the target macros.
26572 2017-06-29  Maya Rashish  <coypu@sdf.org>
26574         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26575         objects.
26577 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26579         * arm/arm-builtins.c: Include profile-count.h
26580         * except.c (sjlj_emit_function_enter): Use
26581         profile_probability::unlikely.
26583 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26585         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26586         and tocrel_offset be pointer args rather than implicitly using
26587         static versions.
26588         (legitimate_constant_pool_address_p, rs6000_emit_move,
26589         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26590         tocrel_offset and use in toc_relative_expr_p call.
26591         (print_operand, print_operand_address): Use static tocrel_base_oac
26592         and tocrel_offset_oac.
26593         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26594         tocrel_offset_oac.
26596 2017-06-29  Maya Rashish  <coypu@sdf.org>
26598         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26600 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26602         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26603         objects, take into account only the alignment of 'op0' and 'mode1' if
26604         'op0' is a MEM.
26606 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26608         * ccmp.c (ccmp_tree_comparison_p): New function.
26609         (ccmp_candidate_p): Update to use above function.
26610         (get_compare_parts): New function.
26611         (expand_ccmp_next): Update to use new functions.
26612         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26613         new functions.
26614         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26615         take mode as argument.
26616         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26617         * expr.c (expand_expr_real_1): Pass mode as argument.
26619 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26621         * combine.c (combine_instructions): Print insns to dump_file, together
26622         with their costs.
26624 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26626         * asan.c (asan_emit_stack_protection): Update.
26627         (create_cond_insert_point): Update.
26628         * auto-profile.c (afdo_propagate_circuit): Update.
26629         * basic-block.h (struct edge_def): Turn probability to
26630         profile_probability.
26631         (EDGE_FREQUENCY): Update.
26632         * bb-reorder.c (find_traces_1_round): Update.
26633         (better_edge_p): Update.
26634         (sanitize_hot_paths): Update.
26635         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26636         (make_single_succ_edge): Update.
26637         (check_bb_profile): Update.
26638         (dump_edge_info): Update.
26639         (update_bb_profile_for_threading): Update.
26640         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26641         probabilitycount to 0.
26642         * cfgbuild.c (compute_outgoing_frequencies): Update.
26643         * cfgcleanup.c (try_forward_edges): Update.
26644         (outgoing_edges_match): Update.
26645         (try_crossjump_to_edge): Update.
26646         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26647         (expand_gimple_tailcall): Update.
26648         (construct_init_block): Use make_single_succ_edge.
26649         (construct_exit_block): Use make_single_succ_edge.
26650         * cfghooks.c (verify_flow_info): Update.
26651         (redirect_edge_succ_nodup): Update.
26652         (split_edge): Update.
26653         (account_profile_record): Update.
26654         * cfgloopanal.c (single_likely_exit): Update.
26655         * cfgloopmanip.c (scale_loop_profile): Update.
26656         (set_zero_probability): Remove.
26657         (duplicate_loop_to_header_edge): Update.
26658         * cfgloopmanip.h (loop_version): Update prototype.
26659         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26660         (force_nonfallthru_and_redirect): Update.
26661         (update_br_prob_note): Update.
26662         (rtl_verify_edges): Update.
26663         (purge_dead_edges): Update.
26664         (rtl_lv_add_condition_to_bb): Update.
26665         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26666         * cgraphunit.c (init_lowered_empty_function): Update.
26667         (cgraph_node::expand_thunk): Update.
26668         * cilk-common.c: Include profile-count.h
26669         * dojump.c (inv): Remove.
26670         (jumpifnot): Update.
26671         (jumpifnot_1): Update.
26672         (do_jump_1): Update.
26673         (do_jump): Update.
26674         (do_jump_by_parts_greater_rtx): Update.
26675         (do_compare_rtx_and_jump): Update.
26676         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26677         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26678         * dwarf2cfi.c: Include profile-count.h
26679         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26680         (sjlj_emit_dispatch_table): Likewise.
26681         * explow.c: Include profile-count.h
26682         * expmed.c (emit_store_flag_force): Update.
26683         (do_cmp_and_jump): Update.
26684         * expr.c (compare_by_pieces_d::generate): Update.
26685         (compare_by_pieces_d::finish_mode): Update.
26686         (emit_block_move_via_loop): Update.
26687         (store_expr_with_bounds): Update.
26688         (store_constructor): Update.
26689         (expand_expr_real_2): Update.
26690         (expand_expr_real_1): Update.
26691         * expr.h (try_casesi, try_tablejump): Update prototypes.
26692         * gimple-pretty-print.c (dump_probability): Update.
26693         (dump_profile): New.
26694         (dump_gimple_label): Update.
26695         (dump_gimple_bb_header): Update.
26696         * graph.c (draw_cfg_node_succ_edges): Update.
26697         * hsa-gen.c (convert_switch_statements): Update.
26698         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26699         (find_if_case_1): Update.
26700         (find_if_case_2): Update.
26701         * internal-fn.c (expand_arith_overflow_result_store): Update.
26702         (expand_addsub_overflow): Update.
26703         (expand_neg_overflow): Update.
26704         (expand_mul_overflow): Update.
26705         (expand_vector_ubsan_overflow): Update.
26706         * ipa-cp.c (good_cloning_opportunity_p): Update.
26707         * ipa-split.c (split_function): Use make_single_succ_edge.
26708         * ipa-utils.c (ipa_merge_profiles): Update.
26709         * loop-doloop.c (add_test): Update.
26710         (doloop_modify): Update.
26711         * loop-unroll.c (compare_and_jump_seq): Update.
26712         (unroll_loop_runtime_iterations): Update.
26713         * lra-constraints.c (lra_inheritance): Update.
26714         * lto-streamer-in.c (input_cfg): Update.
26715         * lto-streamer-out.c (output_cfg): Update.
26716         * mcf.c (adjust_cfg_counts): Update.
26717         * modulo-sched.c (sms_schedule): Update.
26718         * omp-expand.c (expand_omp_for_init_counts): Update.
26719         (extract_omp_for_update_vars): Update.
26720         (expand_omp_ordered_sink): Update.
26721         (expand_omp_for_ordered_loops): Update.
26722         (expand_omp_for_generic): Update.
26723         (expand_omp_for_static_nochunk): Update.
26724         (expand_omp_for_static_chunk): Update.
26725         (expand_cilk_for): Update.
26726         (expand_omp_simd): Update.
26727         (expand_omp_taskloop_for_outer): Update.
26728         (expand_omp_taskloop_for_inner): Update.
26729         * omp-simd-clone.c (simd_clone_adjust): Update.
26730         * optabs.c (expand_doubleword_shift): Update.
26731         (expand_abs): Update.
26732         (emit_cmp_and_jump_insn_1): Update.
26733         (expand_compare_and_swap_loop): Update.
26734         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26735         * predict.c (predictable_edge_p): Update.
26736         (edge_probability_reliable_p): Update.
26737         (set_even_probabilities): Update.
26738         (combine_predictions_for_insn): Update.
26739         (combine_predictions_for_bb): Update.
26740         (propagate_freq): Update.
26741         (estimate_bb_frequencies): Update.
26742         (force_edge_cold): Update.
26743         * profile-count.c (profile_count::dump): Add missing space into dump.
26744         (profile_count::debug): Add newline.
26745         (profile_count::differs_from_p): Explicitly convert to unsigned.
26746         (profile_count::stream_in): Update.
26747         (profile_probability::dump): New member function.
26748         (profile_probability::debug): New member function.
26749         (profile_probability::differs_from_p): New member function.
26750         (profile_probability::differs_lot_from_p): New member function.
26751         (profile_probability::stream_in): New member function.
26752         (profile_probability::stream_out): New member function.
26753         * profile-count.h (profile_count_quality): Rename to ...
26754         (profile_quality): ... this one.
26755         (profile_probability): New.
26756         (profile_count): Update.
26757         * profile.c (compute_branch_probabilities): Update.
26758         * recog.c (peep2_attempt): Update.
26759         * sched-ebb.c (schedule_ebbs): Update.
26760         * sched-rgn.c (find_single_block_region): Update.
26761         (compute_dom_prob_ps): Update.
26762         (schedule_region): Update.
26763         * sel-sched-ir.c (compute_succs_info): Update.
26764         * stmt.c (struct case_node): Update.
26765         (do_jump_if_equal): Update.
26766         (get_outgoing_edge_probs): Update.
26767         (conditional_probability): Update.
26768         (emit_case_dispatch_table): Update.
26769         (expand_case): Update.
26770         (expand_sjlj_dispatch_table): Update.
26771         (emit_case_nodes): Update.
26772         * targhooks.c: Update.
26773         * tracer.c (better_p): Update.
26774         (find_best_successor): Update.
26775         * trans-mem.c (expand_transaction): Update.
26776         * tree-call-cdce.c: Update.
26777         * tree-cfg.c (gimple_split_edge): Upate.
26778         (move_sese_region_to_fn): Upate.
26779         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26780         * tree-eh.c (lower_resx): Upate.
26781         (cleanup_empty_eh_move_lp): Upate.
26782         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26783         * tree-inline.c (copy_edges_for_bb): Update.
26784         (copy_cfg_body): Update.
26785         * tree-parloops.c (gen_parallel_loop): Update.
26786         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26787         (gimple_gen_time_profiler): Update.
26788         * tree-ssa-dce.c (remove_dead_stmt): Update.
26789         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26790         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26791         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26792         (unloop_loops): Update.
26793         (try_peel_loop): Update.
26794         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26795         * tree-ssa-loop-split.c (connect_loops): Update.
26796         (split_loop): Update.
26797         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26798         (hoist_guard): Update.
26799         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26800         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26801         (value_replacement): Update.
26802         * tree-ssa-reassoc.c (branch_fixup): Update.
26803         * tree-ssa-tail-merge.c (replace_block_by): Update.
26804         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26805         (create_edge_and_update_destination_phis): Update.
26806         (compute_path_counts): Update.
26807         (recompute_probabilities): Update.
26808         (update_joiner_offpath_counts): Update.
26809         (freqs_to_counts_path): Update.
26810         (duplicate_thread_path): Update.
26811         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26812         (struct switch_conv_info): Update.
26813         (gen_inbound_check): Update.
26814         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26815         (vect_do_peeling): Update.
26816         (vect_loop_versioning): Update.
26817         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26818         (optimize_mask_stores): Update.
26819         * ubsan.c (ubsan_expand_null_ifn): Update.
26820         * value-prof.c (gimple_divmod_fixed_value): Update.
26821         (gimple_divmod_fixed_value_transform): Update.
26822         (gimple_mod_pow2): Update.
26823         (gimple_mod_pow2_value_transform): Update.
26824         (gimple_mod_subtract): Update.
26825         (gimple_mod_subtract_transform): Update.
26826         (gimple_ic): Update.
26827         (gimple_stringop_fixed_value): Update.
26828         (gimple_stringops_transform): Update.
26829         * value-prof.h: Update.
26831 2017-06-29  Carl Love  <cel@us.ibm.com>
26833         * config/rs6000/rs6000-c.c: Add support for built-in functions
26834         vector signed int vec_signed (vector float);
26835         vector signed long long vec_signed (vector double);
26836         vector signed int vec_signed2 (vector double, vector double);
26837         vector signed int vec_signede (vector double);
26838         vector signed int vec_signedo (vector double);
26839         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26840         instruction generator.
26841         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26842         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26843         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26844         Add define_insn.
26845         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26846         vunsignede_v2df): Add define_expands.
26847         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26848         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26849         VEC_UNSIGNEDO): Add definitions.
26850         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26851         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26852         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26853         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26854         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26855         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26856         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26857         vec_unsignede, vec_unsignedo): Add builtin defines.
26858         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26859         declaration.
26860         * doc/extend.texi: Update the built-in documentation file for the
26861         new built-in functions.
26863 2017-06-29  Richard Biener  <rguenther@suse.de>
26865         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26866         reduction chains to LOOP_VINFO_REDUCTIONS.
26867         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26868         SLP reductions after processing reduction chains.
26870 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26872         * builtins.c (fold_builtin_FUNCTION): Use
26873         lang_hooks.decl_printable_name.
26875 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26877         PR middle-end/81194
26878         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26879         with only one label.
26880         * stmt.c (expand_case): Assert NCASES is greater than one.
26882 2017-06-29  Richard Biener  <rguenther@suse.de>
26884         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26885         anything.
26886         (group_case_labels): Likewise.
26887         (find_taken_edge): Push sanity checking on val to workers...
26888         (find_taken_edge_cond_expr): ... here
26889         (find_taken_edge_switch_expr): ... and here, handle cases
26890         with just a default label.
26891         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26892         (group_case_labels): Likewise.
26893         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26894         group_case_labels does anything cleanup the CFG again.
26896 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26898         PR tree-optimization/81196
26899         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26900         exit condition comparing two IVs.
26902 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26904         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26905         profile to the dummy entry at the end of the list of architectures.
26906         * config/arm/arm-cpu-cdata.h: Regenerated.
26908 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26909             Michael Collison <michael.collison@arm.com>
26911         PR target/70119
26912         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26913         New pattern.
26914         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26915         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26916         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26917         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26918         of shift when the shift amount is masked with constant equal to
26919         the size of the mode.
26920         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26921         predicate.
26923 2017-06-29  Martin Liska  <mliska@suse.cz>
26925         * config/i386/i386.opt: Change range from [1,5] to [0,5].
26927 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
26929         PR bootstrap/80565
26930         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26931         code.
26932         * ipa-inline.h
26933         (edge_growth_cache_entry::edge_growth_cache_entry): New
26934         function.
26935         (reset_edge_growth_cache): Update to use constructor.
26937 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26939         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26940         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26941         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26943 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
26945         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
26946         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
26948 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
26950         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
26951         (*-linux-uclibc*): Add t-uclibc tmake_file.
26952         * config/t-musl: New.
26953         * config/t-uclibc: New.
26955 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
26957         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
26958         context.
26959         (gen_comm_data): Emit architectural setting of arch_prof.
26960         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
26961         profile.
26962         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
26963         (armv8-m.base, armv8-m.main): Likewise.
26964         * arm-protos.h (arm_build_target): Add profile field.
26965         (arch_option): Likewise.
26966         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
26967         the active target.
26968         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
26969         arm_active_target.profile.
26971 2017-06-28  Richard Biener  <rguenther@suse.de>
26973         PR middle-end/81227
26974         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
26975         TYPE_OVERFLOW_WRAPS.
26976         * match.pd (negate_expr_p): Likewise.
26977         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
26978         fold_build2, not fold_binary.
26980 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26982         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
26983         Convert memory address to Pmode.
26984         (aarch64_print_operand): Assert MEM operands are always Pmode.
26986 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26988         PR target/79665
26989         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
26990         Remove redundant if.
26991         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
26992         * config/arm/aarch-common-protos.h
26993         (aarch_forward_to_shift_is_not_shifted_re): Remove.
26994         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
26996 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
26998         PR ipa/81238
26999         * multiple_target.c (create_dispatcher_calls): Set the default
27000         clone to be static, not public.
27002 2017-06-28  Richard Biener  <rguenther@suse.de>
27004         * tree-vect-loop.c (vectorizable_reduction): Move special
27005         cond reduction IV var creation ...
27006         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
27007         parameter.  Use STMT_VINFO_VECTYPE.
27008         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
27009         constant_p.
27011 2017-06-28  Martin Liska  <mliska@suse.cz>
27013         PR ipa/81128
27014         * ipa-visibility.c (non_local_p): Handle visibility.
27016 2017-06-28  Martin Liska  <mliska@suse.cz>
27018         PR driver/79659
27019         * common.opt: Add IntegerRange to various options.
27020         * opt-functions.awk (integer_range_info): New function.
27021         * optc-gen.awk: Add integer_range_info to cl_options struct.
27022         * opts-common.c (decode_cmdline_option): Handle
27023         CL_ERR_INT_RANGE_ARG.
27024         (cmdline_handle_error): Likewise.
27025         * opts.c (print_filtered_help): Show valid interval in
27026         when --help is provided.
27027         * opts.h (struct cl_option): Add range_min and range_max fields.
27028         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27030 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27032         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27033         (x * C EQ/NE y * C): New transformation.
27035 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27037         * genmultilib (combination_space): Accept '+' in option names.
27039 2017-06-28  Martin Liska  <mliska@suse.cz>
27041         PR sanitizer/81224
27042         * asan.c (instrument_derefs): Bail out inner references
27043         that are hard register variables.
27045 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27047         PR target/81175
27048         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27049         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27051 2017-06-28  Richard Biener  <rguenther@suse.de>
27053         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27054         (vect_get_slp_defs): Adjust.
27055         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27056         out from ...
27057         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27058         simplify.
27059         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27060         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27061         (vectorizable_reduction): Adjust.
27062         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27063         handling.
27064         (vect_get_slp_defs): Likewise.
27065         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27066         (vectorizable_bswap): Adjust.
27067         (vectorizable_call): Likewise.
27068         (vectorizable_conversion): Likewise.
27069         (vectorizable_assignment): Likewise.
27070         (vectorizable_shift): Likewise.
27071         (vectorizable_operation): Likewise.
27072         (vectorizable_store): Likewise.
27073         (vectorizable_condition): Likewise.
27074         (vectorizable_comparison): Likewise.
27076 2017-06-28  Michael Collison  <michael.collison@arm.com>
27078         PR target/68535
27079         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27080         set of base_reg
27081         (arm_gen_movmemqi): Removed unused variable 'i'.
27082         Convert 'for' loop into 'while' loop.
27083         (arm_expand_prologue): Remove last unnecessary set of insn.
27084         (thumb_pop): Remove unused variable 'pushed_words'.
27085         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27087 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27089         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27090         * config/s390/s390-protos.h: Add prototype of
27091         s390_rel_address_ok_p.
27092         * config/s390/s390.c (s390_got_symbol): New function.
27093         (s390_rel_address_ok_p): New function.
27094         (legitimize_pic_address): Use s390_rel_address_ok_p.
27095         (s390_load_got): Use s390_got_symbol.
27096         (s390_option_override): Issue error if
27097         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27098         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27099         New macro.
27100         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27102 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27104         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27105         (X * copysign (1.0, X)): New pattern.
27106         (X * copysign (1.0, -X)): New pattern.
27107         (copysign (-1.0, CST)): New pattern.
27109 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27111         * genmultilib (combination_space): Remove variable.
27112         Validate reuse rules against regular expression for any sequence
27113         of multilib options in any order.
27115 2017-06-27  Michael Collison  <michael.collison@arm.com>
27117         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27118         call aarch64_split_simd_combine.
27119         * (aarch64_combine_internal<mode>): Delete pattern.
27120         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27121         Allow register and subreg operands.
27123 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27125         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27126         specific need, just fallback on defaults.
27127         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27129 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27130             Olivier Hainque  <hainque@adacore.com>
27132         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27133         map for 64bits.
27134         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27135         targets. Pick a default if no particular attempt applied.
27136         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27137         larger contexts.
27139 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27141         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27142         (x86_64-wrs-vxworks7): Likewise.
27144 2017-06-27  Marek Polacek  <polacek@redhat.com>
27146         PR sanitizer/81223
27147         * ubsan.c (instrument_null): Check get_base_address's result for null.
27149 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27151         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27153 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27155         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27156         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27157         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27158         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27159         New function types.
27160         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27161         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27162         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27163         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27164         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27165         BUILT_IN_FEUPDATEENV): New builtins.
27166         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27167         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27168         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27169         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27170         macros.
27171         (builtin_structptr_types): Adjust size.
27172         * tree.c (builtin_structptr_types): Add four entries.
27174 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27175             Olivier Hainque  <hainque@adacore.com>
27177         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27178         (TLS_SYM): New local macro, forcing reference to __tls__ on
27179         link command lines for VxWorks 7 RTPs, triggering initialization
27180         of tlsLib.
27181         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27182         OS features TLS support, true for RTPs on VxWorks 7.
27183         * config/vxworks.c (vxworks_override_options): Setup emutls
27184         accordingly.
27186 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27188         * predict.c (test_prediction_value_range): Use -1U instead of -1
27189         to avoid narrowing conversion warning.
27190         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27191         to avoid narrowing conversion warning.
27192         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27193         -1.
27194         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27196 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27198         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27199         64bit configurations.
27200         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27201         (SIZE_TYPE): Likewise.
27202         * config/vxworks.c (vxworks_emutls_var_fields): Use
27203         long_unsigned_type_node instead of unsigned_type_node as the offset
27204         field type, which is "pointer" mode in emutls.c.
27206 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27208         PR sanitizer/81209
27209         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27211         PR middle-end/81207
27212         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27213         gimple_vuse copying separately from gimple_vdef copying.
27215 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27217         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27218         condition.
27220 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27221             Olivier Hainque  <hainque@adacore.com>
27223         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27224         for all vxworks7 targets.
27225         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27226         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27227         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27228         variations for VX6/VX7 and 32/64bits later on in ...
27229         (VXWORKS_LIB_SPEC): Leverage new macros.
27230         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27231         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27233 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27235         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27236         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27238 2017-06-26  Carl Love  <cel@us.ibm.com>
27240         * config/rs6000/rs6000-c.c: Add support for built-in functions
27241         vector bool char vec_reve (vector bool char);
27242         vector signed char vec_reve (vector signed char);
27243         vector unsigned char vec_reve (vector unsigned char);
27244         vector bool int vec_reve (vector bool int);
27245         vector signed int vec_reve (vector signed int);
27246         vector unsigned int vec_reve (vector unsigned int);
27247         vector bool long long vec_reve (vector bool long long);
27248         vector signed long long vec_reve (vector signed long long);
27249         vector unsigned long long vec_reve (vector unsigned long long);
27250         vector bool short vec_reve (vector bool short);
27251         vector signed short vec_reve (vector signed short);
27252         vector double vec_reve (vector double);
27253         vector float vec_reve (vector float);
27254         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27255         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27256         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27257         (altivec_vreve): New pattern.
27258         * config/rs6000/altivec.h (vec_reve): New define.
27259         * doc/extend.texi (vec_rev): Update the built-in documentation file
27260         for the new built-in functions.
27262 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27264         PR tree-optimization/71815
27265         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27266         function.
27267         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27268         has_single_use.
27269         (slsr_process_phi): Likewise.
27270         (replace_uncond_cands_and_profitable_phis): Don't replace a
27271         multiply candidate with a stride of 1 (copy or cast).
27272         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27273         has_single_use.
27274         (lowest_cost_path): Likewise.
27275         (total_savings): Likewise.
27277 2017-06-26  Richard Biener  <rguenther@suse.de>
27279         PR target/81175
27280         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27281         Use def_builtin_pure for all gather builtins.
27283 2017-06-26  Richard Biener  <rguenther@suse.de>
27285         PR tree-optimization/81203
27286         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27287         non-dominating BBs.
27289 2017-06-26  Marek Polacek  <polacek@redhat.com>
27291         PR c/80116
27292         * doc/invoke.texi: Document -Wmultistatement-macros.
27294 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27296         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27297         arm_neon_ok_no_float_abi effective target.
27299 2017-06-26  Richard Biener  <rguenther@suse.de>
27301         PR tree-optimization/80928
27302         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27303         (copy_bbs): Set BB_DUPLICATED flag early.
27304         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27305         marked blocks.
27306         (execute_on_shrinking_pred): Likewise.
27307         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27308         BB_DUPLICATED blocks.
27309         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27310         iterate over all PHIs considering removal of *gsi.
27312 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27314         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27315         qdf24xx.
27317 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27319         * config/rs6000/rs6000-string.c: (expand_block_clear,
27320         do_load_for_compare, select_block_compare_mode,
27321         compute_current_alignment, expand_block_compare,
27322         expand_strncmp_align_check, expand_strn_compare,
27323         expand_block_move, rs6000_output_load_multiple)
27324         Move functions related to string/block move/compare
27325         to a separate file.
27326         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27327         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27328         for this function which is now used in two files.
27329         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27330         * config.gcc: Add rs6000-string.o to extra_objs for
27331         targets powerpc*-*-* and rs6000*-*-*.
27333 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27335         PR target/80510
27336         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27337         32-bit, since indexed is not valid for DImode.
27338         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27339         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27340         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27341         (define_peephole2 for Altivec d-form store): Likewise.
27343         PR ipa/81185
27344         * multiple_target.c (create_dispatcher_calls): Only create the
27345         dispatcher call if the function is the default clone of a
27346         versioned function.
27348 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27350         PR middle-end/80902
27351         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27352         a call, force the call to not be a tail call.
27354 2017-06-23  Jeff Law  <law@redhat.com>
27356         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27357         enquire.
27359 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27361         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27362         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27363         handling for early expansion of vector shifts (sl,sr,sra,rl).
27364         (builtin_function_type): Add vector shift right instructions
27365         to the unsigned argument list.
27367 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27369         rtl-optimizatoin/79286
27370         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27371         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27372         trap.  PIC register plus a const unspec without offset can never trap.
27374 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27376         * tree.h (builtin_structptr_type): New type.
27377         (builtin_structptr_types): Declare new array.
27378         * tree.c (builtin_structptr_types): New array.
27379         (free_lang_data, build_common_tree_nodes): Use it.
27381 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27383         PR c++/81187
27384         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27385         -Wnoexcept.
27387 2017-06-22  Matt Turner  <mattst88@gmail.com>
27389         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27390         Lake models to skylake case.  Assume skylake for unknown
27391         models with clflushopt.
27393 2017-06-22  Jeff Law  <law@redhat.com>
27395         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27396         frame sizes that do not satisfy aarch64_uimm12_shift.
27398 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27400         * profile-count.h (apply_probability,
27401         apply_scale, probability_in): Fix checks for zero.
27403 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27405         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27406         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27408 2016-06-22  Richard Biener  <rguenther@suse.de>
27410         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27411         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27412         REDUC_MAX_EXPR support.
27413         (vectorizable_reduction): Likewise.
27414         (vect_create_epilog_for_reduction): Likewise.
27416 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27418         * match.pd (A / (1 << B) -> A >> B): New.
27419         * generic-match-head.c: Include optabs-tree.h.
27420         * gimple-match-head.c: Likewise.
27421         * optabs-tree.h (target_supports_op_p): New.
27422         * optabs-tree.c (target_supports_op_p): New.
27424 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27426         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27427         $gcc_cv_ld --help output.
27428         (gcc_cv_ld_demangle): Likewise.
27429         (gcc_cv_ld_eh_frame_hdr): Likewise.
27430         (gcc_cv_ld_pie): Likewise.
27431         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27432         (gcc_cv_ld_buildid): Likewise.
27433         (gcc_cv_ld_sysroot): Likewise.
27434         (ld_bndplt_support): Likewise.
27435         (ld_pushpopstate_support): Likewise.
27436         * configure: Regenerate.
27437         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27439 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27441         PR target/81151
27442         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27443         operands indexes to avoid gap between operands and match_dups.
27445 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27447         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27448         Increment Arith_shift and Arith_shift_reg by 1.
27449         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27450         New tuning flag.
27451         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27452         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27453         (aarch64_strip_extend): Add new argument and test for it.
27454         (aarch64_cheap_mult_shift_p): New function.
27455         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27456         add a cost if it is true.
27457         Update calls to aarch64_strip_extend.
27458         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27460 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27462         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27463         tunings.
27464         (thunderxt88): Likewise.
27465         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27466         (thunderx_prefetch_tune): New variable.
27467         (thunderx2t99_prefetch_tune): Update for the correct values.
27468         (thunderxt88_tunings): New variable.
27469         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27470         generic_prefetch_tune.
27471         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27473 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27475         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27476         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27477         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27478         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27479         (aarch64_atomic_cas<mode>, GPI): Likewise.
27481 2017-06-21  Martin Liska  <mliska@suse.cz>
27483         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27484         statements on cold and hot labels.
27485         * predict.c (tree_estimate_probability_bb): Remove the
27486         prediction from this place.
27488 2017-06-21  Martin Liska  <mliska@suse.cz>
27490         PR tree-optimization/79489
27491         * gimplify.c (maybe_add_early_return_predict_stmt): New
27492         function.
27493         (gimplify_return_expr): Call the function.
27494         * predict.c (tree_estimate_probability_bb): Remove handling
27495         of early return.
27496         * predict.def: Update comment about early return predictor.
27497         * gimple-predict.h (is_gimple_predict): New function.
27498         * predict.def: Change default value of early return to 66.
27499         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27500         statements.
27501         * passes.def: Put pass_strip_predict_hints to the beginning of
27502         IPA passes.
27504 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27506         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27507         FUNCTION_DECL declarations.
27508         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27509         declarations.
27510         (dwaf2out_decl): Likewise.
27511         * godump.c (go_early_global_decl): Skip call to the real debug hook
27512         for FUNCTION_DECL declarations.
27513         * passes.c (rest_of_decl_compilation): Skip call to the
27514         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27515         -fdump-go-spec is passed.
27517 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27519         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27520         Reorder for compactness.
27521         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27522         (def_builtin_pure, def_builtin_pure2): New functions.
27523         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27525 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27527         * match.pd (nop_convert): New predicate.
27528         ((A +- CST1) +- CST2): Allow some NOP conversions.
27530 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27532         PR c++/81130
27533         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27534         with ctors/dtors if GOVD_SHARED is set.
27536 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27538         * config/aarch64/aarch64.md (movti_aarch64):
27539         Emit mov rather than orr.
27540         (movtf_aarch64): Likewise.
27541         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27542         Emit mov rather than orr.
27544 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27546         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27547         Swap alternatives, make integer dup more expensive.
27549 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27551         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27552         Return true for non-tls symbols.
27554 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27556         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27557         (cortex-a75): Likewise.
27558         (cortex-a75.cortex-a55): Likewise.
27559         * config/aarch64/aarch64-tune.md: Regenerate.
27560         * doc/invoke.texi (-mtune): Document new values for -mtune.
27562 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27564         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27565         stack_size feature.
27566         (Effective-Target Keywords, Other attributes): Suggest using
27567         dg-add-options stack_size feature to get stack limit in stack_size
27568         effective target documentation.
27570 2017-06-21  Julian Brown  <julian@codesourcery.com>
27571             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27573         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27574         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27575         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27576         reservation.
27577         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27578         attribute type list for neon_multiply.
27579         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27580         attribute type list for neon_multiply.
27581         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27582         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27583         attribute type list for neon_multiply.
27584         * config/arm/types.md (crypto_pmull): Add.
27585         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27586         attribute type list.
27588 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27590         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27591         arm1176jzf-s.
27593 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27595         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27596         to make sure not to dereference a NULL cost_classes_ptr pointer.
27598 2017-06-20  Carl Love  <cel@us.ibm.com>
27600         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27601         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27602         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27603         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27604         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27605         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27606         VMULOSW): New enum "unspec" values.
27607         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27608         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27609         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27610         altivec_vmulosw): New patterns.
27611         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27612         VMULOSW): Add definitions.
27614 2017-06-20  Julia Koval  <julia.koval@intel.com>
27616         * config/i386/i386.c: Fix rounding expand for new pattern.
27617         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27619 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27621         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27622         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27624 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27626         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27627         feature string.
27629 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27631         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27632         architecture, then by implementer ID.
27633         * config/aarch64/aarch64-tune.md: Regenerate.
27635 2017-06-20  Richard Biener  <rguenther@suse.de>
27637         PR middle-end/81097
27638         * fold-const.c (split_tree): Fold to type before negating.
27640 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27642         * diagnostic-show-locus.c
27643         (selftest::test_fixit_deletion_affecting_newline): New function.
27644         (selftest::diagnostic_show_locus_c_tests): Call it.
27646 2017-06-20  Andreas Schwab  <schwab@suse.de>
27648         PR target/80970
27649         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27650         instead of "+d".
27652 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27654         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27655         __ARM_FEATURE_COPROC according to support.
27657 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27659         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27660         Rewritten to avoid overflow for > 32-bit pointers.
27662         PR sanitizer/81125
27663         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27664         by removing enum keyword.
27665         (ubsan_type_descriptor): Likewise.  Formatting fix.
27667         PR target/81121
27668         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27669         splitter): Require TARGET_SSE2 in the condition.
27671 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27673         PR target/79799
27674         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27675         for doing vector set of SFmode on ISA 3.0.
27676         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27677         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27678         element.
27679         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27680         SFmode value into a V4SF variable that was extracted from another
27681         V4SF variable without converting the element to double precision
27682         and back to single precision vector format.
27683         (vsx_insert_extract_v4sf_p9_2): Likewise.
27685 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27687         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27688         in UWHI to avoid undefined overflow.
27690         PR sanitizer/81125
27691         * ubsan.h (enum ubsan_encode_value_phase): New.
27692         (ubsan_encode_value): Change second argument to
27693         enum ubsan_encode_value_phase with default value of
27694         UBSAN_ENCODE_VALUE_GENERIC.
27695         * ubsan.c (ubsan_encode_value): Change second argument to
27696         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27697         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27698         create_tmp_var_raw instead of create_tmp_var and use a
27699         TARGET_EXPR.
27700         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27701         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27702         ubsan_encode_value callers.
27704         PR sanitizer/81111
27705         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27706         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27707         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27709 2017-06-19  Richard Biener  <rguenther@suse.de>
27711         PR middle-end/81118
27712         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27713         estimates if we changed anything.
27715 2017-06-19  Richard Biener  <rguenther@suse.de>
27717         PR tree-optimization/80887
27718         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27719         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27720         simplified lookups, then reset mprts_hook.
27721         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27722         simplifying.
27723         (try_to_simplify): Likewise.
27725 2017-06-19  Martin Liska  <mliska@suse.cz>
27727         PR sanitizer/80879
27728         * gimplify.c (gimplify_switch_expr):
27729         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27731 2017-06-19  Martin Liska  <mliska@suse.cz>
27733         * doc/install.texi: Document that PGO runs in 4 stages.
27735 2017-06-19  Martin Liska  <mliska@suse.cz>
27737         PR ipa/80732
27738         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27739         to dispatcher function name.
27740         * multiple_target.c (replace_function_decl): New function.
27741         (create_dispatcher_calls): Redirect both edges and references.
27743 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27745         * profile-count.c (profile_count::dump): Dump quality.
27746         (profile_count::differs_from_p): Update for unsigned val.
27747         * profile-count.h (profile_count_quality): New enum.
27748         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27750 2017-06-19  Richard Biener  <rguenther@suse.de>
27752         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27753         struct function as arg.
27754         (estimate_numbers_of_iterations): Export overload with loop arg.
27755         (free_numbers_of_iterations_estimates_loop): Use an overload of
27756         free_numbers_of_iterations_estimates instead.
27757         * tree-cfg.c (remove_bb): Adjust.
27758         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27759         * tree-parloops.c (gen_parallel_loop): Likewise.
27760         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27761         Likewise.
27762         (tree_unroll_loops_completely): Likewise.
27763         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27764         Use an overload instead and export.
27765         (estimated_loop_iterations): Adjust.
27766         (max_loop_iterations): Likewise.
27767         (likely_max_loop_iterations): Likewise.
27768         (estimate_numbers_of_iterations): Take struct function as arg
27769         and adjust.
27770         (loop_exits_before_overflow): Adjust.
27771         (free_numbers_of_iterations_estimates_loop): Use an overload.
27772         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27773         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27775 2017-06-19  Richard Biener  <rguenther@suse.de>
27777         PR ipa/81112
27778         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27779         RANGE_EXPR conservatively.
27781 2017-06-16  Carl Love  <cel@us.ibm.com>
27783         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27784         definitions for vec_float, vec_float2, vec_floato,
27785         vec_floate built-ins.
27786         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27787         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27788         floate.
27789         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27790         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27791         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27792         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27793         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27794         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27795         vec_floato): Add builtin defines.
27796         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27797         Update the built-in documentation file for the new built-in
27798         functions.
27800 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27802         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27803         (mthumb): Mark as the negative of -marm.
27805 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27807         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27808         extension options.
27809         (ARM Options, -mtune): Document that this accepts the same
27810         extension options as -mcpu.
27811         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27813 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27815         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27816         permitted extensions.
27818 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27820         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27821         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27822         (armv8-m.main): Add option +nodsp.
27823         * config/arm/arm-cpu-cdata.h: Regenerated.
27825 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27827         * config/arm/t-fuchsia: New file.
27828         * config.gcc (arm*-*-fuchsia*): Use it.
27830 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27832         * config/arm/t-symbian: Rewrite for new option infrastructure.
27834 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27836         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27837         (MULTILIB_REQUIRED): Likewise.
27839 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27841         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27842         (MULTILIB_RESUE): Likewise.
27843         (MULTILIB_MATCHES): Likewise.
27844         (MULTLIB_REQUIRED): Likewise.
27846 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27848         * config/arm/t-rtems: Rewrite for new option framework.
27850 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27852         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27853         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27854         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27855         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27856         * config/arm/t-multilib: ... here.
27857         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27858         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27859         armv7-a and armv8*-a when A-profile libraries have not been built.
27860         * config/arm/t-rmprofile: Rewrite.
27862 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27864         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27865         with a backslash.  Remove the backslash after substituting unescaped
27866         periods.
27867         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27869 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27871         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27872         the driver to pass through the default setting of -mfloat-abi.
27873         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27874         rather than NULL.
27875         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27876         (all_feat_combs): New rule.
27877         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27878         default libraries.
27879         * config/arm/t-aprofile: Rewrite.
27881 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27883         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27884         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27885         fpu is not specified by the user/command-line.
27886         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27887         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27888         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27889         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27890         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27891         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27893 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27895         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27896         * config/arm/t-arm-elf: Rewritten.
27898 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27900         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27901         have some floating-point instructions.
27902         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27903         (TARGET_MAYBE_HARD_FLOAT): New macro.
27904         * config/arm/arm-builtins.c (arm_init_builtins): Use
27905         TARGET_MAYBE_HARD_FLOAT.
27906         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27908 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27910         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27911         (configargs.h): Include it.
27912         (arm_print_hint_for_fpu_option): New function.
27913         (arm_parse_fpu_option): New function.
27914         (candidate_extension): New class.
27915         (arm_canon_for_multilib): New function.
27916         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27917         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27918         (ARCH_CANONICAL_SPECS): New macro.
27919         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27921 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27923         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27924         are set after handling multilib fragments.  Set target_cpu_default2
27925         from with_cpu.
27927 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27929         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27930         cpu name.
27931         (arm*-*-*): Set target_cpu_default2 to a quoted string.
27932         * config/arm/parsecpu.awk (check_cpu): Validate any extension
27933         options.
27934         (check_arch): Likewise.
27935         * config/arm/arm.c (arm_configure_build_target): Handle
27936         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
27937         options in the default.
27939 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27941         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
27942         when an option is an alias of another.
27943         * config/arm/parsecpu.awk (optalias): New parser token.
27944         (gen_comm_data): Mark non-alias options as such.  Emit entries
27945         for extension aliases.
27946         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
27947         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
27948         (armv6kz, armv6zk, armv6t2): Likewise.
27949         (armv7): Make vfpv3-d16 an alias.
27950         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
27951         canonical order.
27952         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
27953         Sort in canonical order.
27954         (armv8-a): Sort in canonical order.
27955         (armv8.1-a, armv8.2-a):  Likewise.
27956         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
27957         canonical order.
27958         (cortex-a9): Sort in canonical order.
27959         * config/arm/arm.c (selftests.h): Include it.
27960         (arm_test_cpu_arch_data): New function.
27961         (arm_run_self_tests): New function.
27962         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
27963         (targetm): Move declaration to the end of the file.
27964         * arm-cpu-cdata.h: Regenerated.
27966 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27968         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
27969         call to target_mode_check describing the type of option passed.
27970         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
27971         (arm_target_thumb_only): Use arm_parse_arch_option_name or
27972         arm_parse_cpu_option_name to match parameters against list of
27973         available targets.
27974         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
27975         arm_arch_core_flags data structure.
27976         * config/arm/arm-cpu_cdata.h: Regenerated.
27978 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27980         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
27981         config/arm/arm.c.
27982         (arm_print_hint_for_cpu_option): Likewise.
27983         (arm_print_hint_for_arch_option): Likewise.
27984         (arm_parse_cpu_option_name): Likewise.
27985         (arm_parse_arch_option_name): Likewise.
27986         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
27987         of entries in the all_fpus list.
27988         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
27989         (arm_parse_cpu_option_name): Declare.
27990         (arm_parse_arch_option_name): Declare.
27991         (arm_parse_option_features): Declare.
27992         (arm_intialize_isa): Declare.
27993         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
27994         data tables to ...
27995         (gen_comm_data): ... here.  Make definitions non-static.
27996         * config/arm/arm-cpu-data.h: Regenerated.
27997         * config/arm/arm-cpu-cdata.h: Regenerated.
27999 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28001         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
28002         (cpu_arch_extension): New structure.
28003         (cpu_arch_option, arch_option, cpu_option): New structures.
28004         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
28005         architecture types.
28006         (gen_data): Generate new format data tables.
28007         * config/arm/arm.c (cpu_tune): New structure.
28008         (cpu_option, processors): Delete.
28009         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
28010         (arm_print_hint_for_cpu_option): ... this and ...
28011         (arm_print_hint_for_arch_option): ... this.
28012         (arm_parse_arch_cpu_name): Delete.  Replace with ...
28013         (arm_parse_cpu_option_name): ... this and ...
28014         (arm_parse_arch_option_name): ... this.
28015         (arm_unrecognized_feature): Change type of target parameter to
28016         cpu_arch_option.
28017         (arm_parse_arch_cpu_features): Delete.  Replace with ...
28018         (arm_parse_option_features): ... this.
28019         (arm_configure_build_target): Rework to use new configuration data
28020         tables.
28021         (arm_print_tune_info): Rework for new configuration data tables.
28022         * config/arm/arm-cpu-data.h: Regenerated.
28023         * config/arm/arm-cpu.h: Regenerated.
28025 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28027         * Makefile.in (OBJS): Move sbitmap.o from here ...
28028         (OBJS-libcommon): ... to here.
28030 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28032         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28033         (ISA_ALL_CRYPTO): New macro.
28034         (ISA_ALL_SIMD): New macro
28035         (ISA_ALL_FP): New macro.
28036         * config/arm/arm.c (fpu_bitlist): Update initializer.
28037         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28038         simd or fp.
28039         (arm9e): Add fpu.  Add option for nofp
28040         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28041         (arm926ej-s, arm1026ej-s): Likewise.
28042         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28043         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28044         neon-fp16, neon-vfpv4, nofp and nosimd.
28045         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28046         (cortex-a8): Add fpu.  Add option for nofp.
28047         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28048         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28049         (cortex-r4f): Add fpu.
28050         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28051         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28052         for nofp.
28053         (cortex-r8): Likewise.
28054         (cortex-m4): Add fpu.  Add option for nofp.
28055         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28056         (cortex-a17.cortex-a7): Likewise.
28057         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28058         (cortex-a35, cortex-a53): Likewise.
28059         (cortex-a57): Add fpu.  Add option for crypto.
28060         (cortex-a72, cortex-a73): Likewise.
28061         (exynos-m1): Likewise.
28062         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28063         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28064         (cortex-m33): Add fpu.  Add option for nofp.
28065         * config/arm/arm-cpu-cdata.h: Regenerated
28066         * config/arm/arm-cpu-data.h: Regenerated.
28068 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28070         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28071         (armv5te, armv5tej): Likewise.
28072         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28073         (armv7): Add options fp and vfpv3-d16.
28074         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28075         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28076         nofp and nosimd.
28077         (armv7ve): Likewise.
28078         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28079         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28080         (armv8-a): Add nocrypto option.
28081         (armv8.1-a, armv8.2-a): Likewise.
28082         (armv8-m.main): add options fp, fp.dp and nofp.
28084 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28086         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28087         nofp.
28088         (armv8-a+crc): Delete.
28089         (armv8.1-a): Add options simd, crypto and nofp.
28090         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28091         (armv8.2-a+fp16): Delete.
28092         (armv8-m.main): Add option dsp.
28093         (armv8-m.main+dsp): Delete.
28094         (cortex-a8): Add fpu.  Add nofp option.
28095         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28096         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28097         link to main cpu and architecture data structures.
28098         (gen_comm_data): Only put isa attributes from the main architecture
28099         in common tables.
28100         (option): New statement for architecture and CPU entries.
28101         * arm.c (struct cpu_option): New structure.
28102         (struct processors): Add entry for options.
28103         (arm_unrecognized_feature): New function.
28104         (arm_parse_arch_cpu_name): Ignore any characters after the first
28105         '+' character.
28106         (arm_parse_arch_cpu_feature): New function.
28107         (arm_configure_build_target): Separate out any CPU and architecture
28108         features and parse separately.  Don't error out if -mfpu=auto is
28109         used with only an architecture string.
28110         (arm_print_asm_arch_directives): New function.
28111         (arm_file_start): Call it.
28112         * config/arm/arm-cpu-cdata.h: Regenerated.
28113         * config/arm/arm-cpu-data.h: Likewise.
28114         * config/arm/arm-tables.opt: Likewise.
28116 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28118         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28119         assembler when it is not -mfpu=auto.
28121 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28123         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28124         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28125         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28126         (ASM_CPU_SPEC): Rewrite.
28127         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28128         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28129         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28130         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28131         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28132         copied string is NUL-terminated.  Also strip any characters prefixed
28133         by '+'.
28134         (arm_rewrite_selected_arch): New function.
28135         (arm_rewrite_march): New function.
28137 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28139         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28140         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28141         (march, mcpu, mtune): Convert to string-based options.
28142         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28143         (arm_parse_arch_cpu_name): New function.
28144         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28145         identify selected architecture or CPU.
28146         (arm_option_save): New function.
28147         (TARGET_OPTION_SAVE): Redefine.
28148         (arm_option_restore): Restore string options.
28149         (arm_option_print): Print string options.
28151 2017-06-16  Martin Sebor  <msebor@redhat.com>
28153         PR tree-optimization/80933
28154         PR tree-optimization/80934
28155         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28156         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28157         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28158         (gimple_fold_builtin): Call them.
28160 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28162         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28163         as unlikely; update profile.
28165 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28167         * predict.c (force_edge_cold): Handle declaring edges impossible
28168         more aggresively.
28170 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28172         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28173         profile.
28174         (try_unroll_loop_completely): Fix reporting.
28176 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28178         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28180 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28182         PR target/71778
28183         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28184         if given a non-constant argument for an intrinsic which requires a
28185         constant.
28187 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28189         * profile.c (compare_freqs): New function.
28190         (branch_prob): Sort edge list.
28191         (find_spanning_tree): Assume that the list is priority sorted.
28193 2017-06-16  Richard Biener  <rguenther@suse.de>
28195         PR tree-optimization/81090
28196         * passes.def (pass_record_bounds): Remove.
28197         * tree-pass.h (make_pass_record_bounds): Likewise.
28198         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28199         make_pass_record_bounds): Likewise.
28200         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28201         not free niter estimates at the beginning but at the end.
28202         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28204 2017-06-16  Richard Biener  <rguenther@suse.de>
28206         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28207         initializer to workaround ICE in host GCC 4.8.
28209 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28211         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28212         counts.
28213         (clone_inlined_nodes): Update.
28215 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28217         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28218         prefetch settings, and enable prefetching by default at -O3.
28220 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28222         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28223         Set flag_prefetch_loop_arrays according to tuning data.
28225 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28227         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28228         New tune structure.
28229         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28230         [Unrelated to main purpose of the patch] Place the pointer field last
28231         to enable type checking errors when tune structure are wrongly merged.
28232         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28233         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28234         (thunderx2t99_prefetch_tune): New tune constants.
28235         (tune_params *_tunings): Update all tunings (no functional change).
28236         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28237         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28238         from tunings structures.
28240 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28242         PR sanitizer/81094
28243         * ubsan.c (instrument_null): Add T argument, use it instead
28244         of computing it based on IS_LHS.
28245         (instrument_object_size): Likewise.
28246         (pass_ubsan::execute): Adjust instrument_null and
28247         instrument_object_size callers to pass gimple_get_lhs or
28248         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28249         calling get_base_address and instrument_mem_ref.  Handle
28250         aggregate call arguments for object-size sanitization.
28252 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28254         PR tree-optimization/81089
28255         * tree-vrp.c (is_masked_range_test): Validate operands of
28256         subexpression.
28258 2017-06-15  Martin Sebor  <msebor@redhat.com>
28260         PR c++/80560
28261         * dumpfile.c (dump_register): Avoid calling memset to initialize
28262         a class with a default ctor.
28263         * gcc.c (struct compiler): Remove const qualification.
28264         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28265         * hash-table.h: Ditto.
28266         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28267           assignment.
28268         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28269         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28270         default ctor.
28271         * params.h (struct param_info): Make struct members non-const.
28272         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28273         with default initialization.
28274         * vec.h (vec_copy_construct, vec_default_construct): New helper
28275         functions.
28276         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28277         with vec_copy_construct.
28278         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28279         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28280         * doc/invoke.texi (-Wclass-memaccess): Document.
28282 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28284         * emit-rtl.h (is_leaf): Update comment about local
28285         register allocator.
28287 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28289         PR target/78818
28290         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28291         for a variable to have a section before checking if the section has a
28292         name.
28293         Set section to.persistent if persistent attribute is set.
28294         Warn if .persistent attribute is used on an automatic variable.
28296 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28298         PR rtl-optimization/80474
28299         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28301 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28303         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28304         of REGNO.
28306 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28308         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28309         (casesi): Emit bounds checking as RTL.
28310         (casesi_internal_mips16_<mode>): Remove bounds checking.
28312 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28314         * config/xtensa/xtensa.c (xtensa_option_override): Append
28315         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28316         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28317          xtensa_doloop_hooks): Define unconditionally.
28318         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28319         TARGET_LOOPS.
28320         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28321         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28322         for it in xtensa_option_override.
28323         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28324          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28326 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28328         * doc/cppopts.texi: Document '-' special value to -MF.
28330 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28332         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28333         (cortex_a53_fconst): Likewise.
28334         (cortex_a53_fpmul): Likewise.
28335         (cortex_a53_f_load_64): Likewise.
28336         (cortex_a53_f_load_many): Likewise.
28337         (cortex_a53_advsimd_alu): Likewise.
28338         (cortex_a53_advsimd_alu_q): Likewise.
28339         (cortex_a53_advsimd_mul): Likewise.
28340         (cortex_a53_advsimd_mul_q): Likewise.
28341         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28342         Add missing fmul, r2f_cvt and fconst cases.
28344 2017-06-14  Richard Biener  <rguenther@suse.de>
28346         PR middle-end/81088
28347         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28348         literal constants.
28349         (fold_binary_loc): When associating do not treat pre-existing
28350         TREE_OVERFLOW on literal constants as a reason to allow
28351         TREE_OVERFLOW on associated literal constants.
28353 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28355         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28356         (MASK_FEATURES): New macro.
28357         * config/sparc/sparc.c (sparc_option_override): Remove the special
28358         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28360 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28362         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28363         a division of 0 if non-call exceptions are enabled.
28365 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28366             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28368         PR target/71663
28369         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28370         Improve vector initialization code gen for only variable case.
28372 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28374         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28376 2017-06-14  Richard Biener  <rguenther@suse.de>
28378         PR tree-optimization/81083
28379         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28380         as values.
28382 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28384         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28385         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28386         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28387         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28388         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28389         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28391 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28393         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28394         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28396 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28398         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28400 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28402         * config/rs6000/t-rtems: Don't handle SPE.
28404 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28406         * config/rs6000/t-linux: Don't handle SPE.
28408 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28410         * config/rs6000/eabispe.h: Delete file.
28412 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28414         * config/rs6000/t-spe: Delete file.
28416 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28418         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28419         (rs6000_legitimate_offset_address_p): Return false for anything in
28420         V2SImode or V2SFmode.
28422 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28424         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28425         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28426         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28427         and V4HImode.
28428         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28429         (rs6000_legitimate_offset_address_p): Ditto.
28430         (rs6000_emit_move): Ditto.
28431         (rs6000_init_builtins): Remove V4HI_type_node.
28433 2017-06-13  Martin Liska  <mliska@suse.cz>
28435         PR sanitize/78204
28436         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28437         (gate_asan): Likewise.
28438         * asan.h (asan_no_sanitize_address_p): Remove the function.
28439         (sanitize_flags_p): New function.
28440         * builtins.def: Fix coding style.
28441         * common.opt: Use renamed enum value.
28442         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28443         * doc/extend.texi: Document no_sanitize attribute.
28444         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28445         to SANITIZE_UNDEFINED_NONDEFAULT.
28446         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28447         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28448         Use sanitize_flags_p.
28449         * gimplify.c (gimplify_function_tree): Likewise.
28450         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28451         * opts.c (parse_no_sanitize_attribute): New function.
28452         (common_handle_option): Use renamed enum value.
28453         * opts.h (parse_no_sanitize_attribute): Declare.
28454         * tree.c (sanitize_flags_p): New function.
28455         * tree.h: Declared here.
28456         * tsan.c: Use sanitize_flags_p.
28457         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28458         (instrument_mem_ref): Likewise.
28459         (instrument_bool_enum_load): Likewise.
28460         (do_ubsan_in_current_function): Remove the function.
28461         (pass_ubsan::execute): Use sanitize_flags_p.
28462         * ubsan.h: Remove do_ubsan_in_current_function
28463         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28464         (dump_function_to_file): Use it here.
28466 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28468         PR tree-optimization/80803
28469         PR tree-optimization/81063
28470         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28471         (propagate_subaccesses_across_link): Enqueue subtree whenever
28472         necessary instead of relying on the caller.
28474 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28476         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28477         that have a first_link.
28478         (sort_and_splice_var_accesses): Do not check first_link before
28479         enquing.
28480         (subtree_mark_written_and_enqueue): Likewise.
28481         (propagate_all_subaccesses): Likewise and do not stop at first
28482         parent with a first_link.
28484 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28486         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28487         instead of f.
28489 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28491         * match.pd: New pattern.
28493 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28495         * tree-vrp.c (is_masked_range_test): New function.
28496         (register_edge_assert_for): Determine ranges for
28497         some bit tests.
28499 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28501         PR tree-optimization/67328
28502         * fold-const.c (maskable_range_p): New function.
28503         (build_range_check): Generate bittests if possible.
28505 2017-06-13  Martin Liska  <mliska@suse.cz>
28507         * gimple-pretty-print.c (dump_probability): Add new argument.
28508         (dump_edge_probability): Dump both probability and count.
28509         (dump_gimple_label): Likewise.
28510         (dump_gimple_bb_header): Likewise.
28512 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28514         PR target/81072
28515         * config/avr/avr-devices.c: Fix indentation.
28516         * config/avr/gen-avr-mmcu-specs.c: Dito.
28518 2017-06-13  Richard Biener  <rguenther@suse.de>
28520         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28521         instead get vector type from stmt_info.
28522         (vectorizable_reduction): Adjust.  Remove dead code.
28524 2017-06-13  Richard Biener  <rguenther@suse.de>
28526         PR middle-end/81065
28527         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28528         case of C * (x * C2 + C3).
28529         (fold_addr_of_array_ref_difference): Properly fold index difference.
28531 2017-06-12  David S. Miller  <davem@davemloft.net>
28533         PR target/80968
28534         * config/sparc/sparc.md (return expander): Emit frame blockage if
28535         function uses alloca.
28537 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28539         * combine.c (make_field_assignment): Check len rather than the mode
28540         precision when calling force_to_mode.
28542 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28544         Support multilibs and devices that see flash in RAM address range.
28546         PR target/81072
28547         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28548         (avr_mcu_t) <flash_pm_offset>: New field.
28549         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28550         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28551         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28552         (AVR_TINY_PM_OFFSET): Remove macro.
28553         * config/avr/avr.opt (-mshort-calls): New option.
28554         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28555         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28556         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28557         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28558         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28559         instead of avr_arch->have_jmp_call.
28560         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28561         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28562         avr_arch->flash_pm_offset to define.
28563         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28564         new field flash_pm_offset.  Add entry for avrxmega3.
28565         (avr_texinfo): Add entry for avrxmega3.
28566         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28567         attiny212, attiny214,
28568         attiny412, attiny414, attiny416, attiny417,
28569         attiny814, attiny816, attiny817,
28570         attiny1614, attiny1616, attiny1617,
28571         attiny3214, attiny3216, attiny3217.
28572         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28573         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28574         (avr_print_operand_address) [AVR_TINY]: Same.
28575         (avr_asm_init_sections) <readonly_data_section>: Only patch
28576         callback if avr_arch->flash_pm_offset = 0.
28577         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28578         for rodata if avr_arch->flash_pm_offset != 0.
28579         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28580         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28581         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28582         (m_options): Append opt_rcall.
28583         (m_dirnames): Append dir_rcall.
28584         * config/avr/t-multilib: Regenerate.
28586         * configure.ac [target=avr]: Check whether avrxmega3 default
28587         linker description file works as needed.
28588         * configure: Regenerate.
28589         * doc/avr-mmcu.texi: Regenerate.
28590         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28591         <__AVR_ARCH__>: Document avrxmega3 and 103.
28592         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28593         <__AVR_SHORT_CALLS__>: Document it.
28594         <__AVR_PM_BASE_ADDRESS__>: Document it.
28595         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28596         (AVR Variable Attributes) <progmem>: Document this is
28597         not needed for avrxmega3.
28598         (AVR Named Address Spaces) <__flash>: Dito.
28600 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28602         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28604 2017-06-12  Doug Rupp  <rupp@adacore.com>
28606         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28607         Append vxworks-stdint.h to the tm_file list.
28608         * config/vxworks-stdint.h: New file.
28610 2017-06-12  Martin Liska  <mliska@suse.cz>
28612         PR tree-optimization/81041
28613         * tree-profile.c (gimple_gen_ic_func_profiler):
28614         Create an extra BB in profile-generate
28615         (gimple_gen_time_profiler): Likewise.
28617 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28619         PR tree-optimization/81003
28620         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28621         (update_range_test): Use it instead of force_gimple_operand_gsi.
28623 2017-06-12  Richard Biener  <rguenther@suse.de>
28625         PR tree-optimization/81053
28626         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28627         with backedge value not defined in loop.  Simplify def stmt
28628         compute.
28630 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28632         PR target/79939
28633         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28634         Return true.
28635         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28636         nvptx_cannot_force_const_mem.
28638 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28640         * opts.c (finish_options): Move test for flag_split_stack after
28641         it has been initialized.
28643 2017-06-11  Jason Merrill  <jason@redhat.com>
28645         * tree.h (id_equal): New.
28646         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28647         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28648         instead of strcmp of IDENTIFIER_POINTER.
28650 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28652         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28653         (mark_all_inlined_calls_cdtor): Fix formating.
28654         (inline_transform): Rescale profile before inlining.
28656 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28658         * cgraph.h (cgraph_edge::clone): Update prototype.
28659         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28660         (cgraph_node::create_clone): Update.
28661         (cgraph_node::create_version_clone): Update.
28662         * tree-inline.c (copy_bb): Update.
28663         (expand_call_inline): Update.
28665 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28667         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28668         factored out from ...
28669         (rs6000_emit_prologue): ... here.
28671 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28673         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28674         factored out from ...
28675         (rs6000_emit_prologue): ... here.
28677 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28679         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28680         edge counts.
28681         (handle_missing_profiles): Fix computation of tp_first_run.
28682         (counts_to_freqs): Do not touch freqs when count is 0.
28684 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28686         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28687         profile.
28689 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28691         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28692         attributes): Document signal effective target.
28694 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28696         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28697         Document effective target stack_size.
28699 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28701         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28702         to the edit_context if they can be auto-applied.
28704 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28706         * opts.c (finish_options): If -fsplit-stack, disable implicit
28707         -forder-blocks-and-partition.
28708         * doc/invoke.texi (Optimize Options): Document that when using
28709         -fsplit-stack -forder-blocks-and-partition is not implicitly
28710         enabled.
28712 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28714         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28715         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28716         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28717         * builtins.def (abort, trap, unreachable): Declare cold.
28718         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28719         * tree-core.h (ECF_COLD): New.
28720         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28721         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28723 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28725         * predict.c (unlikely_executed_stmt_p): Cleanup.
28727 2017-06-09  Richard Biener  <rguenther@suse.de>
28729         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28730         model if the ref is always written to.
28732 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28734         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28736 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28738         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28739         than udiv.
28741 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28743         PR target/80855
28744         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28745         "target cannot support label values" when encountering LABEL_REF.
28747 2017-06-09  Martin Liska  <mliska@suse.cz>
28749         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28750         (gimple_gen_ic_func_profiler): Emit direct comparison
28751         of __gcov_indirect_call_callee with NULL.
28752         (gimple_gen_time_profiler): Change probability from
28753         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28755 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28757         * profile.c (edge_gcov_counts): Turn to pointer.
28758         (compute_branch_probabilities, compute_branch_probabilities): Update.
28759         (branch_prob): Do not clear edge_gcov_count.
28760         * profile.h (edge_gcov_counts): Turn to pointer.
28761         (edge_gcov_count): Update.
28763 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28765         * gimple.h (gimple_check_failed): Mark cold.
28767 2017-06-09  Richard Biener  <rguenther@suse.de>
28769         PR tree-optimization/66623
28770         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28771         refactor check_reduction into two parts, properly computing
28772         whether we have to check reduction validity for outer loop
28773         vectorization.
28775 2017-06-09  Richard Biener  <rguenther@suse.de>
28777         PR tree-optimization/79483
28778         * graphite-scop-detection.c (order): New global.
28779         (get_order): Compute bb to order mapping that satisfies code
28780         generation constraints.
28781         (cmp_pbbs): New helper.
28782         (build_scops): Start domwalk at entry block, sort generated
28783         pbbs.
28785 2017-06-09  Richard Biener  <rguenther@suse.de>
28787         PR middle-end/81007
28788         * ipa-polymorphic-call.c
28789         (ipa_polymorphic_call_context::restrict_to_inner_class):
28790         Skip FIELD_DECLs with error_mark_node type.
28791         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28792         last again.
28794 2017-06-09  Martin Liska  <mliska@suse.cz>
28796         * predict.c (struct branch_predictor): New struct.
28797         (test_prediction_value_range): New test.
28798         (predict_c_tests): New function.
28799         * selftest-run-tests.c (selftest::run_tests): Run the function.
28800         * selftest.h: Declare new tests.
28802 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28804         PR target/80966
28805         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28806         gen_add3_insn did not fail.
28807         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28808         r0, construct that number in a temporary reg and add that reg to r0.
28809         If asked to put the result in r0 as well, fail.
28811 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28813         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28814         for early expansion of vec_eqv.
28816 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28818         PR middle-end/81005
28819         * ubsan.c (instrument_null): Avoid pointless code temporary.
28820         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28822 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28824         PR target/81015
28825         Revert:
28826         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28828         PR target/59874
28829         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28830         (*clzhi2): Ditto.
28832 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28834         * predict.c (unlikely_executed_edge_p): Move ahead.
28835         (probably_never_executed_edge_p): Use it.
28837 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28839         PR middle-end/79988
28840         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28841         gimple_call_builtin_p call.
28843 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28845         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28846         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28847         rtl_check_failed_type2, rtl_check_failed_code1,
28848         rtl_check_failed_code2, rtl_check_failed_code_mode,
28849         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28850         rtvec_check_failed_bounds, rtl_check_failed_flag,
28851         _fatal_insn_not_found, _fatal_insn): Likewise.
28852         * tree.h (tree_contains_struct_check_failed,
28853         tree_check_failed, tree_not_check_failed,
28854         tree_class_check_failed, tree_range_check_failed,
28855         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28856         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28857         tree_operand_check_failed, omp_clause_check_failed,
28858         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28859         Likewise.
28861 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28863         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28864         flag_branch_probabilities.
28865         * ipa-inline.c (edge_badness): Likewise.
28866         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28867         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28868         * predict.c (maybe_hot_frequency_p): Likewise.
28869         (probably_never_executed): Likewise.
28870         * sched-ebb.c (schedule_ebbs): Likewise.
28871         * sched-rgn.c (find_single_block_region): Likewise.
28872         * tracer.c (tail_duplicate): Likewise.
28874 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28876         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28877         longer requires x_flag_profile_use.
28879 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28881         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28882         instead of flag_reorder_blocks_and_partition.
28883         * dbxout.c (dbxout_function_end): Likewise.
28884         * dwarf2out.c (gen_subprogram_die): Likewise.
28885         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28886         * hw-doloop.c (reorg_loops): Likewise.
28887         * varasm.c (assemble_start_function,
28888         assemble_end_function): Likewise.
28889         (decide_function_section): Do not check for
28890         flag_reorder_blocks_and_partition.
28892 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28894         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28895         New function.
28896         (chkp_get_hard_register_fake_addr_expr): Ditto.
28897         (chkp_build_addr_expr): Add check for hard reg case.
28898         (chkp_parse_array_and_component_ref): Ditto.
28899         (chkp_find_bounds_1): Ditto.
28900         (chkp_process_stmt): Don't generate bounds store for
28901         hard reg case.
28903 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28905         * predict.c (maybe_hot_bb_p): Do not check profile status.
28906         (maybe_hot_edge_p): Likewise.
28907         (probably_never_executed): Check for zero counts even if profile
28908         is not read.
28909         (unlikely_executed_edge_p): New function.
28910         (unlikely_executed_stmt_p): New function.
28911         (unlikely_executed_bb_p): New function.
28912         (set_even_probabilities): Use unlikely predicates.
28913         (combine_predictions_for_bb): Likewise.
28914         (predict_paths_for_bb): Likewise.
28915         (predict_paths_leading_to_edge): Likewise.
28916         (determine_unlikely_bbs): New function.
28917         (estimate_bb_frequencies): Use it.
28918         (compute_function_frequency): Use zero counts even if profile is
28919         not read.
28920         * profile-count.h: Fix typo.
28922 2017-08-08  Julia Koval  <julia.koval@intel.com>
28924         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28925         _mm512_mask_cvtsepi16_storeu_epi8,
28926         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28927         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28928         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28929         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28930         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28931         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28932         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28933         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28934         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28935         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28936         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28937         __builtin_ia32_pmovuswb256mem_mask,
28938         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28939         __builtin_ia32_pmovwb512mem_mask): New builtins.
28941 2017-08-08  Julia Koval  <julia.koval@intel.com>
28943         PR target/73350,80862
28944         * config/i386/subst.md (round): Fix round pattern.
28945         * config/i386/i386.c (ix86_erase_embedded_rounding):
28946         Fix erasing rounding for the fixed pattern.
28948 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28950         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
28952 2017-06-08  Martin Liska  <mliska@suse.cz>
28954         PR gcov-profile/80911
28955         * gcov.c (block_info::block_info): New constructor.
28957 2017-06-07  Carl Love  <cel@us.ibm.com>
28959         * config/rs6000/rs6000-c: The return type of the following
28960         built-in functions was implemented as int not long long.  Fix sign
28961         of return value for the unsigned version of vec_mulo and vec_mule.
28962         vector unsigned long long vec_bperm (vector unsigned long long,
28963                                              vector unsigned char)
28964         vector signed long long vec_mule (vector signed int,
28965                                           vector signed int)
28966         vector unsigned long long vec_mule (vector unsigned int,
28967                                             vector unsigned int)
28968         vector signed long long vec_mulo (vector signed int,
28969                                           vector signed int)
28970         vector unsigned long long vec_mulo (vector unsigned int,
28971                                             vector unsigned int)
28972         * doc/extend.texi: Fix the documentation for the built-in
28973         functions.
28975 2017-06-07  Carl Love  <cel@us.ibm.com>
28977         PR target/80982
28978         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
28979         for BE.
28981 2017-06-07  Carl Love  <cel@us.ibm.com>
28983         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
28984         support, Generate       doublehv for signed int/float for BE case only.
28986 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
28988         * doc/invoke.texi (mcx16): Rewrite.
28990 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28992         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
28993         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
28994         *mov<mode>_softfloat, and an anonymous splitter): Use
28995         nonimmediate_operand instead of rs6000_nonimmediate_operand.
28997 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28999         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
29000         SPEFSCR registers.
29001         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29002         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
29003         (rs6000_debug_reg_global): Adjust.
29004         (rs6000_init_hard_regno_mode_ok): Adjust.
29005         (rs6000_dbx_register_number): Adjust.
29006         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
29007         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29008         Remove SPE_ACC and SPEFSCR.
29009         (REG_ALLOC_ORDER): Ditto.
29010         (FRAME_POINTER_REGNUM): Change to 111.
29011         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
29012         (REG_CLASS_NAMES): Ditto.
29013         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
29014         (REGISTER_NAMES): Ditto.
29015         (ADDITIONAL_REG_NAMES): Ditto.
29016         (rs6000_reg_names): Ditto.
29017         * config/rs6000/rs6000.md: Renumber some register number
29018         define_constants.
29020 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29022         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
29023         registers.
29024         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29025         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29026         to 117.
29027         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29028         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29029         Delete the SPE high registers.
29030         (REG_ALLOC_ORDER): Ditto.
29031         (enum reg_class): Remove SPE_HIGH_REGS.
29032         (REG_CLASS_NAMES): Ditto.
29033         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29034         (REGISTER_NAMES): Ditto.
29035         (rs6000_reg_names): Ditto.
29036         * doc/tm.texi.in: Remove SPE as example.
29037         * doc/tm.texi: Regenerate.
29039 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29041         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29042         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29043         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29044         * config/rs6000/rs6000.md (type): Remove "brinc".
29046 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29048         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29049         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29050         * config/rs6000/linuxspe.h: Delete file.
29051         * config/rs6000/rs6000.md: Don't include spe.md.
29052         * config/rs6000/spe.h: Delete file.
29053         * config/rs6000/spe.md: Delete file.
29054         * config/rs6000/t-rs6000: Remove spe.md.
29056 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29058         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29059         (reg_or_none500mem_operand): Delete.
29060         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29061         instead of reg_or_none500mem_operand.
29063 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29065         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29066         handling of SPE flags.
29067         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29069 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29071         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29072         SPE ABI handling.
29073         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29074         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29075         paired_divv2sf3): Similar.
29076         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29077         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29078         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29079         RS6000_BUILTIN_S.
29080         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29081         Rename the paired_* instruction patterns.
29082         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29083         define __SPE__.
29084         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29085         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29086         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29087         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29088         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29089         PAIRED_VECTOR_MODE.
29090         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29091         (spe_func_has_64bit_regs_p): Delete.
29092         (spe_expand_predicate_builtin): Delete.
29093         (spe_expand_evsel_builtin): Delete.
29094         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29095         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29096         (invalid_e500_subreg): Delete.
29097         (rs6000_legitimize_address): Always force_reg op2 as well, for
29098         paired single memory accesses.
29099         (rs6000_member_type_forces_blk): Delete.
29100         (rs6000_spe_function_arg): Delete.
29101         (rs6000_expand_unop_builtin): Delete SPE handling.
29102         (rs6000_expand_binop_builtin): Ditto.
29103         (spe_expand_stv_builtin): Delete.
29104         (bdesc_2arg_spe): Delete.
29105         (spe_expand_builtin): Delete.
29106         (spe_expand_predicate_builtin): Delete.
29107         (spe_expand_evsel_builtin): Delete.
29108         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29109         (spe_init_builtins): Delete.
29110         (spe_func_has_64bit_regs_p): Delete.
29111         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29112         (rs6000_emit_stack_reset): Ditto.
29113         (rs6000_dwarf_register_span): Delete.
29114         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29115         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29116         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29117         Delete.
29118         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29119         Delete.
29120         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29121         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29122         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29123         mulv2sf3, divv2sf3): Delete expanders.
29125 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29127         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29129 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29131         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29132         * config/rs6000/rs6000.c: Ditto.
29134 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29136         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29137         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29138         comparison_operator.
29140 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29142         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29143         * config/rs6000/rs6000.opt: Ditto.
29144         * config/rs6000/t-rtems: Ditto.
29146 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29148         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29149         TARGET_E500_SINGLE by 0, simplify.
29150         * config/rs6000/rs6000.c: Ditto.
29151         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29152         (spe_build_register_parallel): Delete.
29153         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29154         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29155         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29156         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29157         (E500_CONVERT): Delete.
29158         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29160 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29162         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29163         * config/rs6000/dfp.md: Ditto.
29164         (negdd2, *negdd2_fpr): Merge.
29165         (absdd2, *absdd2_fpr): Merge.
29166         (negtd2, *negtd2_fpr): Merge.
29167         (abstd2, *abstd2_fpr): Merge.
29168         * config/rs6000/e500.h: Delete file.
29169         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29170         TARGET_FPRS by 1 and simplify.
29171         * config/rs6000/rs6000-c.c: Ditto.
29172         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29173         TARGET_DF_SPE by 0.
29174         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29175         TARGET_DF_SPE.
29176         * config/rs6000/rs6000.md: Ditto.
29177         (floatdidf2, *floatdidf2_fpr): Merge.
29178         (move_from_CR_gt_bit): Delete.
29179         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29180         (E500_CR_IOR_COMPARE): Delete.
29181         (All patterns that require !TARGET_FPRS): Delete.
29182         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29184 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29186         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29188 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29190         * graphds.c (add_edge): Intitialize edge's attached data.
29191         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29192         pointer parameter.  Call pointed function on each edge during
29193         graph traversing.  Skip traversing the edge when the function
29194         returns true.
29195         (graphds_dfs, graphds_scc): Ditto.
29196         (for_each_edge): New parameter.  Pass the new parameter to callback
29197         function.
29198         * graphds.h (skip_edge_callback): New function pointer type.
29199         (graphds_dfs, graphds_scc): New function pointer parameter.
29200         (graphds_edge_callback, for_each_edge): New parameter.
29202 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29204         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29205         out code checking if runtime alias check is possible to below ...
29206         Call the new function.
29207         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29208         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29210 2017-06-07  Marek Polacek  <polacek@redhat.com>
29212         PR sanitizer/80932
29213         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29214         TYPE_OVERFLOW_WRAPS check.
29216 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29218         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29219         if versioning is required.
29220         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29221         peeling with the check for versioning.
29223 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29225         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29226         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29227         Set true to new parameter if new ssa variable is defined.
29228         (vect_gen_vector_loop_niters): Refactor.  Set range information
29229         for the new vector loop bound variable.
29230         (vect_do_peeling): Ditto.
29232 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29234         * tree-affine.c (ssa.h): Include header file.
29235         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29236         has wrapping overflow behavior.
29238 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29240         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29242 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29244         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29245         (tree_to_aff_combination): ... here.
29247 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29249         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29250         reg_pressure model function.
29251         (ivopts_global_cost_for_size): Delete.
29252         (determine_set_costs, iv_ca_recount_cost): Call new model function
29253         ivopts_estimate_reg_pressure.
29255 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29257         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29258         expensive than udiv.  Remove floating point cases from mod.
29260 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29262         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29263         Increase idiv cost.
29265 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29267         * config/aarch64/aarch64.md
29268         (copysignsf3): Fix mask generation.
29270 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29272         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29273         TDI_gimple.
29274         (class dump_manager): Add register_dumps method.
29275         * dumpfile.c: Include langhooks.h.
29276         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29277         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29278         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29279         (dump_manager::dump_register): Start with 512 entries instead of 32.
29280         (dump_manager::register_dumps): New method.
29281         * toplev.c (general_init): Instead of invoking register_dumps
29282         langhook, invoke register_dumps method on the dump manager.
29283         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29284         TDI_generic.
29286 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29288         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29289         Say that # requires an associated define_split to exist, and that
29290         the define_split must be suitable for use after register allocation.
29292 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29294         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29295         (compute_outgoing_frequencies): Also initialize zero counts.
29296         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29297         around loops; preserve more of profile when nothing changes.
29299 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29301         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29302         here.
29303         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29304         * config/arm/arm-cpu-cdata.h: Regenerate.
29305         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29306         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29307         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29308         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29309         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29310         support.
29311         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29312         support.
29313         * config/arm/t-rmprofile: Likewise.
29314         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29316 2017-06-06  David S. Miller  <davem@davemloft.net>
29318         PR target/80968
29319         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29320         blockage if function uses alloca.
29322 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29324         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29325         New "uid" fields to hold pretty-print IDs of group and ref.
29326         Memory references are now identified as <group_id>:<ref_id>
29327         instead of using [random] addresses.
29328         (dump_mem_details): Simplify, no functional change.
29329         (dump_mem_ref): Simplify and make output more concise.
29330         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29331         (find_or_create_group): Initialize group uid.
29332         (record_ref): Initialize ref uid.  Improve debug output.
29333         (prune_group_by_reuse, should_issue_prefetch_p,)
29334         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29335         (mark_nontemporal_store, determine_loop_nest_reuse):
29336         Improve debug output.
29338 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29340         * dbgcnt.def (prefetch): New debug counter.
29341         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29342         (schedule_prefetches): Stop issueing prefetches if debug counter
29343         tripped.
29345 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29347         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29348         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29350 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29352         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29353         Use aarch64_reg_or_zero predicate for operand 4.
29354         (aarch64_compare_and_swap<mode> define_insn_and_split):
29355         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29356         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29358 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29360         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29361         (arm_compute_save_core_reg_mask): This.
29362         (thumb1_compute_save_reg_mask): Rename into ...
29363         (thumb1_compute_save_core_reg_mask): This.
29364         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29365         (arm_compute_frame_layout): Likewise.
29367 2017-06-06  Richard Biener  <rguenther@suse.de>
29369         PR tree-optimization/80974
29370         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29371         keep or clear leaders SSA info.
29373 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29375         * config/nvptx/nvptx.c (split_mode_p): New function.
29376         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29378 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29380         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29382 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29384         PR bootstrap/80978
29385         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29386         profile.
29388 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29390         * shrink-wrap.c (handle_simple_exit): Update profile.
29391         (try_shrink_wrapping): Upate profile.
29393 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29395         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29396         (tree_guess_outgoing_edge_probabilities): New.
29397         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29398         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29400 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29402         * ipa-split.c (split_function): Initialize return bb profile.
29404 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29406         * profile.c (compute_branch_probabilities): Also initialize
29407         EXIT_BLOCK profile.
29409 2017-06-06  Richard Biener  <rguenther@suse.de>
29411         PR tree-optimization/80928
29412         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29413         (vect_analyze_loop_operations): Properly guard analysis for
29414         pure SLP case.
29415         (vect_transform_loop): Likewise.
29416         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29417         (vect_model_induction_cost): Do not cost for pure SLP.
29418         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29419         of induction in inner loop vectorization.
29420         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29421         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29422         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29423         recursion.
29424         (vect_analyze_slp_cost_1): Cost induction.
29425         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29426         (vect_get_slp_vect_defs): Likewise.
29427         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29428         (vect_transform_stmt): Handle SLP reductions.
29429         * tree-vectorizer.h (vectorizable_induction): Adjust.
29431 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29433         * config/rs6000/rs6000.c (make_resolver_func): Update
29434         init_lowered_empty_function call.
29436 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29438         * doc/invoke.texi: Document the -fprofile-abs-path option.
29439         * common.opt (fprofile-abs-path): New option.
29440         * gcov-io.h (gcov_write_filename): Declare.
29441         * gcov-io.c (gcov_write_filename): New function.
29442         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29443         * profile.c (output_location): Likewise.
29445 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29447         * shring-wrap.c: Revert accidental commit.
29449 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29451         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29453 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29455         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29456         new edge.
29457         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29458         profile in callgraph edge.
29459         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29460         (apply_scale): Likewise.
29461         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29462         Also scale profile when inlining function with zero profile.
29463         (initialize_cfun): Update exit block profile even when it is zero.
29464         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29465         when profile is read.
29467 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29469         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29470         (CLONE_*): New constants to define the processors we can generate
29471         code for with the target_clone attribute.
29472         (rs6000_clone_map): New array to identify which clone processors
29473         the current program is running on.
29474         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29475         target_clone attribute.
29476         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29477         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29478         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29479         (cpu_expand_builtin): Add support for target_clone attribute.
29480         (rs6000_valid_attribute_p): Allow "default" attribute.
29481         (get_decl_name): New debug function to simplify printing the
29482         current function name in debugging statements.
29483         (rs6000_clone_priority): New functions to support the target_clone
29484         attribute, and be able to generate code to switch between ISA 2.05
29485         through ISA 3.0 (power6 through power9).
29486         (rs6000_compare_version_priority): Likewise.
29487         (rs6000_get_function_versions_dispatcher): Likewise.
29488         (make_resolver_func): Likewise.
29489         (add_condition_to_bb): Likewise.
29490         (dispatch_function_versions): Likewise.
29491         (rs6000_generate_version_dispatcher_body): Likewise.
29492         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29493         (fusion_gpr_load_p): Fix a spacing issue.
29494         * doc/extend.texi (Common Function Attributes): Document that the
29495         PowerPC supports the target_clone attribute.
29497 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29499         * config/arm/arm.h: explain F symbol found in description of ARM
29500         register allocation in its legend.
29502 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29504         * config/mips/frame-header-opt.c: Include profile-count.h.
29505         * config/riscv/riscv.c: Include profile-count.h
29507 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29509         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29510         update profile.
29511         (sm_set_flag_if_changed): Add bbs field.
29512         (execute_sm_if_changed_flag_set): Pass BBS.
29513         (execute_sm): Update.
29515 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29517         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29518         New pattern.
29520 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29522         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29523         (peephole2): New peephole2 to emit the above.
29524         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29526 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29528         * config/aarch64/aarch64.c (define_peephole2 above
29529         *sub_<shift>_<mode>): New peephole.
29531 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29533         * config/i386/i386.c (make_resolver_func): Update.
29534         * Makefile.in: Add profile-count.h and profile-count.o
29535         * auto-profile.c (afdo_indirect_call): Update to new API.
29536         (afdo_set_bb_count): Update.
29537         (afdo_propagate_edge): Update.
29538         (afdo_propagate_circuit): Update.
29539         (afdo_calculate_branch_prob): Update.
29540         (afdo_annotate_cfg): Update.
29541         * basic-block.h: Include profile-count.h
29542         (struct edge_def): Turn count to profile_count.
29543         (struct basic_block_def): Likewie.
29544         (REG_BR_PROB_BASE): Move to profile-count.h
29545         (RDIV): Move to profile-count.h
29546         * bb-reorder.c (max_entry_count): Turn to profile_count.
29547         (find_traces): Update.
29548         (rotate_loop):Update.
29549         (connect_traces):Update.
29550         (sanitize_hot_paths):Update.
29551         * bt-load.c (migrate_btr_defs): Update.
29552         * cfg.c (RDIV): Remove.
29553         (init_flow): Use alloc_block.
29554         (alloc_block): Uninitialize count.
29555         (unchecked_make_edge): Uninitialize count.
29556         (check_bb_profile): Update.
29557         (dump_edge_info): Update.
29558         (dump_bb_info): Update.
29559         (update_bb_profile_for_threading): Update.
29560         (scale_bbs_frequencies_int): Update.
29561         (scale_bbs_frequencies_gcov_type): Update.
29562         (scale_bbs_frequencies_profile_count): New.
29563         * cfg.h (update_bb_profile_for_threading): Update.
29564         (scale_bbs_frequencies_profile_count): Declare.
29565         * cfgbuild.c (compute_outgoing_frequencies): Update.
29566         (find_many_sub_basic_blocks): Update.
29567         * cfgcleanup.c (try_forward_edges): Update.
29568         (try_crossjump_to_edge): Update.
29569         * cfgexpand.c (expand_gimple_tailcall): Update.
29570         (construct_exit_block): Update.
29571         * cfghooks.c (verify_flow_info): Update.
29572         (dump_bb_for_graph): Update.
29573         (split_edge): Update.
29574         (make_forwarder_block): Update.
29575         (duplicate_block): Update.
29576         (account_profile_record): Update.
29577         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29578         (get_estimated_loop_iterations): Update.
29579         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29580         (single_likely_exit): Update.
29581         * cfgloopmanip.c (scale_loop_profile): Update.
29582         (loopify): Update.
29583         (set_zero_probability): Update.
29584         (lv_adjust_loop_entry_edge): Update.
29585         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29586         (purge_dead_edges): Update.
29587         (rtl_account_profile_record): Update.
29588         * cgraph.c (cgraph_node::create): Uninitialize count.
29589         (symbol_table::create_edge): Uninitialize count.
29590         (cgraph_update_edges_for_call_stmt_node): Update.
29591         (cgraph_edge::dump_edge_flags): Update.
29592         (cgraph_node::dump): Update.
29593         (cgraph_edge::maybe_hot_p): Update.
29594         * cgraph.h: Include profile-count.h
29595         (create_clone), create_edge, create_indirect_edge): Update.
29596         (cgraph_node): Turn count to profile_count.
29597         (cgraph_edge0: Likewise.
29598         (make_speculative, clone): Update.
29599         (create_edge): Update.
29600         (init_lowered_empty_function): Update.
29601         * cgraphclones.c (cgraph_edge::clone): Update.
29602         (duplicate_thunk_for_node): Update.
29603         (cgraph_node::create_clone): Update.
29604         * cgraphunit.c (cgraph_node::analyze): Update.
29605         (cgraph_node::expand_thunk): Update.
29606         * final.c (dump_basic_block_info): Update.
29607         * gimple-streamer-in.c (input_bb): Update.
29608         * gimple-streamer-out.c (output_bb): Update.
29609         * graphite.c (print_global_statistics): Update.
29610         (print_graphite_scop_statistics): Update.
29611         * hsa-brig.c: Include basic-block.h.
29612         * hsa-dump.c: Include basic-block.h.
29613         * hsa-gen.c (T sum_slice): Update.
29614         (convert_switch_statements):Update.
29615         * hsa-regalloc.c: Include basic-block.h.
29616         * ipa-chkp.c (chkp_produce_thunks): Update.
29617         * ipa-cp.c (struct caller_statistics): Update.
29618         (init_caller_stats): Update.
29619         (gather_caller_stats): Update.
29620         (ipcp_cloning_candidate_p): Update.
29621         (good_cloning_opportunity_p): Update.
29622         (get_info_about_necessary_edges): Update.
29623         (dump_profile_updates): Update.
29624         (update_profiling_info): Update.
29625         (update_specialized_profile): Update.
29626         (perhaps_add_new_callers): Update.
29627         (decide_about_value): Update.
29628         (ipa_cp_c_finalize): Update.
29629         * ipa-devirt.c (struct odr_type_warn_count): Update.
29630         (struct decl_warn_count): Update.
29631         (struct final_warning_record): Update.
29632         (possible_polymorphic_call_targets): Update.
29633         (ipa_devirt): Update.
29634         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29635         * ipa-icf.c (sem_function::merge): Update.
29636         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29637         * ipa-inline.c (compute_uninlined_call_time): Update.
29638         (compute_inlined_call_time): Update.
29639         (want_inline_small_function_p): Update.
29640         (want_inline_self_recursive_call_p): Update.
29641         (edge_badness): Update.
29642         (lookup_recursive_calls): Update.
29643         (recursive_inlining): Update.
29644         (inline_small_functions): Update.
29645         (dump_overall_stats): Update.
29646         (dump_inline_stats): Update.
29647         * ipa-profile.c (ipa_profile_generate_summary): Update.
29648         (ipa_propagate_frequency): Update.
29649         (ipa_profile): Update.
29650         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29651         * ipa-utils.c (ipa_merge_profiles): Update.
29652         * loop-doloop.c (doloop_modify): Update.
29653         * loop-unroll.c (report_unroll): Update.
29654         (unroll_loop_runtime_iterations): Update.
29655         * lto-cgraph.c (lto_output_edge): Update.
29656         (lto_output_node): Update.
29657         (input_node): Update.
29658         (input_edge): Update.
29659         (merge_profile_summaries): Update.
29660         * lto-streamer-in.c (input_cfg): Update.
29661         * lto-streamer-out.c (output_cfg): Update.
29662         * mcf.c (create_fixup_graph): Update.
29663         (adjust_cfg_counts): Update.
29664         (sum_edge_counts): Update.
29665         * modulo-sched.c (sms_schedule): Update.
29666         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29667         * predict.c (maybe_hot_count_p): Update.
29668         (probably_never_executed): Update.
29669         (dump_prediction): Update.
29670         (combine_predictions_for_bb): Update.
29671         (propagate_freq): Update.
29672         (handle_missing_profiles): Update.
29673         (counts_to_freqs): Update.
29674         (rebuild_frequencies): Update.
29675         (force_edge_cold): Update.
29676         * predict.h: Include profile-count.h
29677         (maybe_hot_count_p, counts_to_freqs): UPdate.
29678         * print-rtl-function.c: Do not include cfg.h
29679         * print-rtl.c: Include basic-block.h
29680         * profile-count.c: New file.
29681         * profile-count.h: New file.
29682         * profile.c (is_edge_inconsistent): Update.
29683         (correct_negative_edge_counts): Update.
29684         (is_inconsistent): Update.
29685         (set_bb_counts): Update.
29686         (read_profile_edge_counts): Update.
29687         (compute_frequency_overlap): Update.
29688         (compute_branch_probabilities): Update; Initialize and deinitialize
29689         gcov_count tables.
29690         (branch_prob): Update.
29691         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29692         (edge_gcov_count): New.
29693         (bb_gcov_count): New.
29694         * shrink-wrap.c (try_shrink_wrapping): Update.
29695         * tracer.c (better_p): Update.
29696         * trans-mem.c (expand_transaction): Update.
29697         (ipa_tm_insert_irr_call): Update.
29698         (ipa_tm_insert_gettmclone_call): Update.
29699         * tree-call-cdce.c: Update.
29700         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29701         (gimple_duplicate_sese_tail): Update.
29702         (gimple_account_profile_record): Update.
29703         (execute_fixup_cfg): Update.
29704         * tree-inline.c (copy_bb): Update.
29705         (copy_edges_for_bb): Update.
29706         (initialize_cfun): Update.
29707         (freqs_to_counts): Update.
29708         (copy_cfg_body): Update.
29709         (expand_call_inline): Update.
29710         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29711         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29712         (try_unroll_loop_completely): Update.
29713         (try_peel_loop): Update.
29714         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29715         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29716         * tree-ssa-loop-split.c (connect_loops): Update.
29717         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29718         * tree-ssa-reassoc.c (branch_fixup): Update.
29719         * tree-ssa-tail-merge.c (replace_block_by): Update.
29720         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29721         (compute_path_counts): Update.
29722         (update_profile): Update.
29723         (recompute_probabilities): Update.
29724         (update_joiner_offpath_counts): Update.
29725         (estimated_freqs_path): Update.
29726         (freqs_to_counts_path): Update.
29727         (clear_counts_path): Update.
29728         (ssa_fix_duplicate_block_edges): Update.
29729         (duplicate_thread_path): Update.
29730         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29731         (struct switch_conv_info): Update.
29732         * tree-tailcall.c (decrease_profile): Update.
29733         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29734         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29735         * value-prof.c (check_counter): Update.
29736         (gimple_divmod_fixed_value): Update.
29737         (gimple_mod_pow2): Update.
29738         (gimple_mod_subtract): Update.
29739         (gimple_ic_transform): Update.
29740         (gimple_stringop_fixed_value): Update.
29741         * value-prof.h (gimple_ic): Update.
29743 2017-06-02  Carl Love  <cel@us.ibm.com>
29745         * config/rs6000/rs6000-c: Add support for built-in functions
29746         vector double vec_doublee (vector signed int);
29747         vector double vec_doublee (vector unsigned int);
29748         vector double vec_doublee (vector float);
29749         vector double vec_doubleh (vector signed int);
29750         vector double vec_doubleh (vector unsigned int);
29751         vector double vec_doubleh (vector float);
29752         vector double vec_doublel (vector signed int);
29753         vector double vec_doublel (vector unsigned int);
29754         vector double vec_doublel (vector float);
29755         vector double vec_doubleo (vector signed int);
29756         vector double vec_doubleo (vector unsigned int);
29757         vector double vec_doubleo (vector float);.
29758         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29759         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29760         UNS_DOUBLEL.
29761         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29762         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29763         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29764         VS_sxwsp.
29765         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29766         vec_doublel, vec_doubleh.
29767         * doc/extend.texi: Update the built-in documentation file for the
29768         new built-in functions.
29770 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29772         PR jit/80954
29773         * ipa-inline-analysis.c (free_growth_caches): Set
29774         edge_removal_hook_holder to NULL after removing it.
29776 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29778         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29779         comparision with zero.
29781 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29782         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29783         for early expansion of vec_min and vec_max builtins.
29784         (builtin_function_type): Add min/max unsigned variants to those
29785         identified as having unsigned arguments.
29787 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29789         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29791 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29793         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29794         Use VALL_F16 iterator rather than VALL.
29796 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29798         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29799         Emit CBNZ inside loop when doing a strong exchange and comparing
29800         against zero.  Generate the CC flags after the loop.
29802 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29804         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29805         (dl_section_ref): New.
29806         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29807         On AIX, append an expression to subtract the size of the
29808         section length to dl_section_ref.
29810 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29812         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29813         for early expansion of vector absolute builtins.
29815 2017-06-02  Richard Biener  <rguenther@suse.de>
29817         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29818         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29820 2017-06-02  Richard Biener  <rguenther@suse.de>
29822         PR tree-optimization/80948
29823         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29824         stmt order as well.
29826 2017-06-02  Richard Biener  <rguenther@suse.de>
29828         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29829         PHIs are ok.
29830         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29831         for inductions as relevant.
29833 2017-06-02  Richard Biener  <rguenther@suse.de>
29835         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29836         (vectorizable_induction): ... this.  Remove dead code.
29838 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29840         * builtins. (expand_builtin_alloca): Remove second parameter and
29841         infer its value from the first parameter instead.
29842         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29844 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29846         PR rtl-optimization/80903
29847         * loop-doloop.c (add_test): Unshare sequence.
29849 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29851         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29853 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29855         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29856         static.
29857         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29858         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29859         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29860         (xlogue_layout::get_stub_rtx): Make static.
29861         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29862         (xlogue_layout::compute_stub_managed_regs): Rename to...
29863         (xlogue_layout::count_stub_managed_regs): ...this.
29864         (xlogue_layout::is_stub_managed_reg): New function.
29865         (xlogue_layout::m_stub_names): Rename to...
29866         (xlogue_layout::s_stub_names): ...this, make static.
29867         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29868         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29869         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29870         xlogue_layout::s_stub_names): Instantiate statics.
29871         (stub_managed_regs): Remove.
29872         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29873         (disable_call_ms2sysv_xlogues): Rename to...
29874         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29875         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29876         warning logic.
29877         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29878         change after reload_completed.
29879         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29880         directly.
29881         (ix86_expand_prologue): Likewise.
29882         (ix86_expand_epilogue): Likewise.
29883         (ix86_expand_split_stack_prologue): Likewise.
29884         (ix86_compute_frame_layout): Remove frame parameter ...
29885         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29886         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29887         only if necessary.
29888         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29889         (ix86_frame): Move from here ...
29890         * config/i386/i386.h (ix86_frame): ... to here.
29891         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29892         complete ix86_frame data structure instead.  Remove some_ld_name.
29894 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29896         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29897         symbols that hold a DECL_VALUE_EXPR.
29899 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29901         PR tree-optimization/80898
29902         * tree-sra.c (process_subtree_disqualification): Removed.
29903         (disqualify_candidate): Do not acll
29904         process_subtree_disqualification.
29905         (subtree_mark_written_and_enqueue): New function.
29906         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29907         RHS has been disqualified and re-queue LHS if necessary.  Apart
29908         from that, ignore disqualified RHS.
29910 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29912         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29913         address fetch for z10 or later.
29915 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29917         * config/arc/arc.md (tst_movb): Add guard when splitting.
29919 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29921         * config/arc/arc.c (arc_can_eliminate): Test against
29922         arc_frame_pointer_needed.
29924 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29926         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29927         to prevent store reordering.
29928         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29929         (type): Add block type.
29930         (stack_tie): Define special instruction to be used in
29931         expand_prologue.
29933 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29935         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29936         constraint. It is not valid for the pattern.
29937         (noncommutative_binary_comparison): Likewise.
29939 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29941         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
29942         scaled addresses.
29944 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29946         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
29947         be used by the reg-alloc.
29949 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29951         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
29952         reg-alloc when having mul64 or mul32x16 instructions.
29953         (mulsidi3): Likewise.
29954         (umulsidi3): Likewise.
29955         (mulsi32x16): New pattern.
29956         (mulsi64): Likewise.
29957         (mulsidi64): Likewise.
29958         (umulsidi64): Likewise.
29959         (MUL32x16_REG): Define.
29960         (mul64_600): Use MUL32x16_REG.
29961         (mac64_600): Likewise.
29962         (umul64_600): Likewise.
29963         (umac64_600): Likewise.
29965 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29967         * config/arc/arc.md (mulsi3_700): Make it commutative.
29969 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
29971         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
29972         type for movstouw.
29973         (*sign_extendsidi2_insn): Likewise for movstosw.
29975 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29977         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
29978         the type of the input discriminant value.  Convert the
29979         discriminant value of signedness vary.
29981 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
29983         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
29984         Add to -Wall section.
29986 2017-06-01  Richard Biener  <rguenther@suse.de>
29988         PR middle-end/66313
29989         * fold-const.c (fold_plusminus_mult_expr): If the factored
29990         factor may be zero use a wrapping type for the inner operation.
29991         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
29992         and handle moved defs.
29993         (process_assignment): Properly guard the unary op case.  Return a
29994         tri-state indicating that moving the stmt before the call may allow
29995         to continue.  Pass through to_move.
29996         (find_tail_calls): Handle moving unrelated defs before
29997         the call.
29999 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
30001         PR target/80618
30002         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
30003         splitter result in the canonical way.
30005 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
30007         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
30008         also for 32bit target.  Update insn attributes.
30009         (zero-extendsidi2 splitter): Allow all registers for operand 1.
30011 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
30013         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
30014         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
30015         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
30016         (_mm_maskz_min_ss): New intrinsics.
30018 2017-05-31  Martin Liska  <mliska@suse.cz>
30020         * tree-vect-loop.c (vect_create_epilog_for_reduction):
30021         Change comment style to one we normally use.
30022         (vectorizable_reduction): Likewise.
30023         (vectorizable_induction): Likewise.
30024         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30025         (vectorizable_call): Likewise.
30026         (vectorizable_simd_clone_call): Likewise.
30027         (vectorizable_conversion): Likewise.
30028         (vectorizable_assignment): Likewise.
30029         (vectorizable_shift): Likewise.
30030         (vectorizable_operation): Likewise.
30031         (vectorizable_store): Likewise.
30032         (vectorizable_load): Likewise.
30033         * tree-vectorizer.h: Likewise.
30035 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30037         * passes.c (emergency_dump_function): New.
30038         * tree-pass.h (emergency_dump_function): Declare.
30039         * plugin.c (plugins_internal_error_function): Remove.
30040         * plugin.h (plugins_internal_error_function): Remove declaration.
30041         * toplev.c (internal_error_function): New static function.  Use it...
30042         (general_init): ...here.
30044 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30046         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30047         (arc_rtx_costs): Add costs for new patterns.
30048         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30049         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30051 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30053         * tree-ssa-strlen.c (get_next_strinfo): New function.
30054         (get_stridx_plus_constant): Use it.
30055         (zero_length_string): Likewise.
30056         (adjust_related_strinfos): Likewise.
30057         (adjust_last_stmt): Likewise.
30059 2017-05-31  Richard Biener  <rguenther@suse.de>
30061         PR target/80880
30062         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30063         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30065 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30067         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30068         loop_vinfo argument and use of dependence distance vectors.
30069         Check instead whether the two references differ only in their
30070         initial value and assume that they have the same alignment if the
30071         difference is a multiple of the vector alignment.
30072         (vect_analyze_data_refs_alignment): Update call accordingly.
30074 2017-05-31  Martin Liska  <mliska@suse.cz>
30076         PR target/79155
30077         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30079 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30081         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30082         (create_intersect_range_checks): Move from ...
30083         * tree-data-ref.c (create_intersect_range_checks_index)
30084         (create_intersect_range_checks): ... to here.
30085         (create_runtime_alias_checks): New function factored from ...
30086         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30087         here.  Call above function.
30088         * tree-data-ref.h (create_runtime_alias_checks): New function.
30090 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30092         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30093         segment length for dr_b and compute it in wide_int.
30095 2017-05-31  Richard Biener  <rguenther@suse.de>
30097         PR tree-optimization/80906
30098         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30099         and pass through iv_map.
30100         (copy_bb_and_scalar_dependences): Adjust.
30101         (translate_pending_phi_nodes): Likewise.
30102         (copy_loop_close_phi_args): Handle code-generating IVs instead
30103         of ICEing.
30105 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30107         * diagnostic-color.c (color_dict): Add "type-diff".
30108         (parse_gcc_colors): Update comment.
30109         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30110         -fdiagnostics-show-template-tree and -fno-elide-type.
30111         (GCC_COLORS): Add type-diff to example.
30112         (type-diff=): New.
30113         (-fdiagnostics-show-template-tree): New.
30114         (-fno-elide-type): New.
30115         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30116         the pp_format_decoder callback.  Call any m_format_postprocessor's
30117         "handle" method.
30118         (pretty_printer::pretty_printer): Initialize
30119         m_format_postprocessor.
30120         (pretty_printer::~pretty_printer): Delete any
30121         m_format_postprocessor.
30122         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30123         (class format_postprocessor): New class.
30124         (struct pretty_printer::format_decoder): Document the new parameters.
30125         (struct pretty_printer::m_format_postprocessor): New field.
30126         * tree-diagnostic.c (default_tree_printer): Update for new
30127         bool and const char ** params.
30128         * tree-diagnostic.h (default_tree_printer): Likewise.
30130 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30132         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30133         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30134         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30135         rs6000_gen_cell_microcode code.
30136         (rs6000_final_prescan_insn): Delete.
30137         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30138         "warn-cell-microcode" entries.
30139         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30140         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30141         throughout.  Change cc_reg_not_micro_cr0_operand to
30142         cc_reg_not_cr0_operand throughout.
30143         (*extendhi<mode>2_noload): Delete.
30144         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30145         (mwarn-cell-microcode): Delete.
30146         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30147         -mgen-cell-microcode and -mwarn-cell-microcode.
30149 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30151         PR target/80833
30152         * config/i386/constraints.md (Yd): New constraint.
30153         (Ye): Ditto.
30154         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30155         and (?Yd, r) alternatives.  Update insn attributes.
30156         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30157         and (?*Yd, r) alternatives.  Update insn attributes.
30158         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30160 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30162         * gimplify.c (gimplify_modify_expr): Don't create a
30163         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30164         function.
30166 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30168         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30170 2017-05-30  Richard Biener  <rguenther@suse.de>
30172         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30173         and reduc_def fields.
30174         (STMT_VINFO_REDUC_TYPE): New define.
30175         (STMT_VINFO_REDUC_DEF): Likewise.
30176         (vect_force_simple_reduction): Adjust prototype.
30177         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30178         (vect_is_simple_reduction): Remove check_reduction argument.
30179         (vect_force_simple_reduction): Adjust and set
30180         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30181         (vectorizable_reduction): Do not re-do reduction analysis
30182         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30183         * tree-parloops.c (gather_scalar_reductions): Adjust.
30185 2017-05-30  Richard Biener  <rguenther@suse.de>
30187         PR middle-end/80901
30188         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30189         split_edge code.
30191 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30193         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30194         Introduce unknown_misalignment parameter and remove vf.
30195         (vect_peeling_hash_get_lowest_cost):
30196         Pass unknown_misalignment parameter.
30197         (vect_enhance_data_refs_alignment):
30198         Fix unsupportable data ref treatment.
30200 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30202         * tree-vect-data-refs.c (vect_get_data_access_cost):
30203         Workaround for SLP handling.
30204         (vect_enhance_data_refs_alignment):
30205         Compute costs for doing no peeling at all, compare to the best
30206         peeling costs so far and avoid peeling if cheaper.
30208 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30210         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30211         Return peeling info and set costs to zero for unlimited cost
30212         model.
30213         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30214         with unknown misalignment. Compute and costs for unknown
30215         misalignment, compare them to the costs for known misalignment
30216         and choose the cheapest for peeling.
30218 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30220         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30221         (vect_get_peeling_costs_all_drs): Create function.
30222         (vect_peeling_hash_get_lowest_cost):
30223         Use vect_get_peeling_costs_all_drs.
30224         (vect_peeling_supportable): Create function.
30225         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30227 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30229         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30230         DR_HAS_NEGATIVE_STEP.
30231         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30232         (vect_enhance_data_refs_alignment): Use.
30233         (vect_duplicate_ssa_name_ptr_info): Use.
30234         * tree-vectorizer.h (dr_misalignment): Use.
30235         (known_alignment_for_access_p): Use.
30237 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30239         PR target/78838
30240         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30241         .lowtext.
30242         (has_section_name): New function.
30244 2017-05-30  Martin Liska  <mliska@suse.cz>
30246         PR other/80909
30247         * auto-profile.c (get_function_decl_from_block): Fix
30248         parenthesis.
30250 2017-05-30  Richard Biener  <rguenther@suse.de>
30252         PR middle-end/80876
30253         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30255 2017-05-30  Martin Liska  <mliska@suse.cz>
30257         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30258         * dumpfile.h (struct dump_file_info): Remove ctors.
30260 2017-05-30  Martin Liska  <mliska@suse.cz>
30262         * predict.def: Fix GNU coding style.
30264 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30266         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30267         Mark 'to' argument with ATTRIBUTE_UNUSED.
30269 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30271         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30272         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30273         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30274         format string.
30276 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30278         * doc/install.texi (Options specification): Restore entry of
30279         --enable-sjlj-exceptions.
30281 2017-05-27  Michael Eager  <eager@eagercon.com>
30283         Revert:
30284         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30286         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30288         * config/microblaze/microblaze.h
30289         (FIXED_REGISTERS): Update in macro.
30290         (CALL_USED_REGISTERS): Update in macro.
30292 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30294         * doc/install.texi: Add links to macOS binary distributions.
30296 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30298         PR bootstrap/80887
30299         Revert:
30300         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30302         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30304 2017-05-26  Martin Liska  <mliska@suse.cz>
30306         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30308 2017-05-26  Martin Liska  <mliska@suse.cz>
30310         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30311         always leading ';; '.
30312         (dump_bb_info): Likewise.
30313         (brief_dump_cfg): Likewise.
30314         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30315         * dumpfile.c: Remove usage of TDF_VERBOSE.
30316         * dumpfile.h (enum dump_kind): Likewise.
30317         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30318         * print-tree.c (debug_verbose): Remove.
30319         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30320         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30321         * tree-diagnostic.c (default_tree_printer): Replace
30322         TDF_DIAGNOSTIC with TDF_SLIM.
30324 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30326         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30327         in parameter loop, rather than loop_vinfo.
30328         (create_intersect_range_checks): Ditto.
30329         (vect_create_cond_for_alias_checks): Update call to above functions.
30331 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30333         PR tree-optimization/80815
30334         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30335         for merging runtime alias checks.  Handle negative DR_STEPs.
30337 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30339         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30340         Move from ...
30341         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30342         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30343         out code pruning runtime alias checks.
30344         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30345         factored out from above.
30346         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30347         Move from ...
30348         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30349         ... to here.
30350         (prune_runtime_alias_test_list): New decalaration.
30352 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30354         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30355         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30356         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30357         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30358         (operator==, comp_dr_with_seg_len_pair): Ditto.
30359         (vect_prune_runtime_alias_test_list): Ditto.
30361 2017-05-26  Martin Liska  <mliska@suse.cz>
30363         PR ipa/80663
30364         * params.def: Bound partial-inlining-entry-probability param.
30366 2017-05-26  Marek Polacek  <polacek@redhat.com>
30368         PR sanitizer/80875
30369         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30370         can be negated.
30372 2017-05-26  Richard Biener  <rguenther@suse.de>
30374         PR tree-optimization/80842
30375         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30376         value.
30378 2017-05-26  Richard Biener  <rguenther@suse.de>
30380         PR tree-optimization/80844
30381         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30383 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30385         * doc/md.texi (Machine Constraints): Update x86 family
30386         machine constraints section to match 'config/i386/constraints.md'.
30388 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30390         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30392 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30394         * doc/invoke.texi (--enable-languages): Update documentation.
30396 2017-05-25  Martin Liska  <mliska@suse.cz>
30398         * dumpfile.c: Add TDF_FOLDING.
30399         * dumpfile.h (enum dump_kind): Likewise.
30400         * genmatch.c (dt_simplify::gen_1): Use it.
30402 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30404         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30406 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30408         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30409         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30411 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30413         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30414         * match.pd (X == C): Rewrite it here.
30415         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30416         with_certain_nonzero_bits2): New predicates.
30417         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30419 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30421         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30422         avoid warning.
30424         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30425         warning.
30427 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30429         * config/powerpcspe: New port.  Files are copied from the rs6000
30430         port, with "rs6000" in filenames replaced by "powerpcspe".
30432 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30434         PR rtl-optimization/80754
30435         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30437 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30439         * config/sparc/sparc.md (length): Return the correct value for -mflat
30440         sibcalls to match output_sibcall.
30442 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30444         PR bootstrap/80860
30445         PR bootstrap/80843
30446         * config/rs6000/rs6000.c (struct machine_function): Add new field
30447         n_components.
30448         (rs6000_get_separate_components): Init that field, use it.
30449         (rs6000_components_for_bb): Use the field.
30451 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30453         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30455 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30457         PR middle-end/80823
30458         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30460 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30462         PR target/80725
30463         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30464         address against address_operand predicate.
30465         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30467 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30469         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30470         parameters passed indirectly.
30472 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30474         * config/i386/i386.md (*movdi_internal): Remove SSE4
30475         alternative 18 (?r, *v).  Update insn attributes.
30476         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30477         Update insn attributes.
30478         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30479         Update insn attributes.
30480         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30481         alternative 1 (r, v). Remove isa attribute.
30482         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30483         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30484         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30486 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30488         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30489         dg-line directive.
30491 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30493         * cgraphunit.c (symbol_table::process_new_functions): Update.
30494         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30495         (inline_generate_summary): Rename to ...
30496         (ipa_fn_summary_generate): ... this one.
30497         (inline_read_summary): Rename to ...
30498         (ipa_fn_summary_read): ... this one.
30499         (inline_write_summary): Rename to ...
30500         (ipa_fn_summary_write): ... this one.
30501         (inline_free_summary): Rename to ...
30502         (ipa_free_fn_summary): ... this one.
30503         (pass_data_local_fn_summary, pass_local_fn_summary,
30504         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30505         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30506         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30507         make_pass_ipa_fn_summary): New.
30508         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30509         inline_write_summary, inline_free_summary): Remove.
30510         (ipa_free_fn_summary) : New.
30511         * ipa-inline.c (ipa_inline): Update.
30512         (pass_ipa_inline): Do not generate summaries.
30513         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30514         Remove.
30515         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30516         and add pass_ipa_fn_summary.
30517         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30518         New.
30519         (make_pass_inline_parameters): Remove.
30521 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30523         * omp-low.c (struct omp_context): Remove "default_kind" member.
30524         Adjust all users.
30526         * omp-offload.c (execute_oacc_device_lower): Remove the
30527         parallelism dimensions function attributes for unparallelized
30528         OpenACC kernels constructs.
30530 2017-05-23  Martin Liska  <mliska@suse.cz>
30532         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30533         functions.
30534         (cgraph_edge::make_speculative): Likewise.
30535         (cgraph_edge::resolve_speculation): Likewise.
30536         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30537         (cgraph_node::dump): Likewise.
30538         * cgraph.h: Likewise.
30539         * cgraphunit.c (analyze_functions): Likewise.
30540         (symbol_table::compile): Likewise.
30541         * ipa-cp.c (print_all_lattices): Likewise.
30542         (determine_versionability): Likewise.
30543         (initialize_node_lattices): Likewise.
30544         (ipcp_verify_propagated_values): Likewise.
30545         (estimate_local_effects): Likewise.
30546         (update_profiling_info): Likewise.
30547         (create_specialized_node): Likewise.
30548         (perhaps_add_new_callers): Likewise.
30549         (decide_about_value): Likewise.
30550         (decide_whether_version_node): Likewise.
30551         (identify_dead_nodes): Likewise.
30552         (ipcp_store_bits_results): Likewise.
30553         * ipa-devirt.c (dump_targets): Likewise.
30554         (ipa_devirt): Likewise.
30555         * ipa-icf.c (sem_item::dump): Likewise.
30556         (sem_function::equals): Likewise.
30557         (sem_variable::equals): Likewise.
30558         (sem_item_optimizer::read_section): Likewise.
30559         (sem_item_optimizer::execute): Likewise.
30560         (congruence_class::dump): Likewise.
30561         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30562         (dump_inline_summary): Likewise.
30563         (estimate_node_size_and_time): Likewise.
30564         (inline_analyze_function): Likewise.
30565         * ipa-inline-transform.c (inline_call): Likewise.
30566         * ipa-inline.c (report_inline_failed_reason): Likewise.
30567         (want_early_inline_function_p): Likewise.
30568         (edge_badness): Likewise.
30569         (update_edge_key): Likewise.
30570         (inline_small_functions): Likewise.
30571         * ipa-profile.c (ipa_profile): Likewise.
30572         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30573         (ipa_make_edge_direct_to_target): Likewise.
30574         (remove_described_reference): Likewise.
30575         (ipa_impossible_devirt_target): Likewise.
30576         (propagate_controlled_uses): Likewise.
30577         (ipa_print_node_params): Likewise.
30578         (ipcp_transform_function): Likewise.
30579         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30580         (propagate_pure_const): Likewise.
30581         * ipa-reference.c (generate_summary): Likewise.
30582         (read_write_all_from_decl): Likewise.
30583         (propagate): Likewise.
30584         (ipa_reference_read_optimization_summary): Likewise.
30585         * ipa-utils.c (ipa_merge_profiles): Likewise.
30586         * ipa.c (walk_polymorphic_call_targets): Likewise.
30587         (symbol_table::remove_unreachable_nodes): Likewise.
30588         (ipa_single_use): Likewise.
30589         * passes.c (execute_todo): Likewise.
30590         * predict.c (drop_profile): Likewise.
30591         * symtab.c (symtab_node::get_dump_name): New function.
30592         (symtab_node::dump_name): Likewise.
30593         (symtab_node::dump_asm_name): Likewise.
30594         (symtab_node::dump_references): Likewise.
30595         (symtab_node::dump_referring): Likewise.
30596         (symtab_node::dump_base): Likewise.
30597         (symtab_node::debug_symtab): Likewise.
30598         * tree-sra.c (convert_callers_for_node): Likewise.
30599         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30600         * value-prof.c (init_node_map): Likewise.
30602 2017-05-23  Martin Liska  <mliska@suse.cz>
30604         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30605         and symtab_node::debug_symtab to symbol_table::debug.
30606         * cgraphunit.c (analyze_functions): Use the renamed function.
30607         (symbol_table::compile): Likewise.
30608         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30609         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30610         * passes.c (execute_todo): Likewise.
30611         * symtab.c (symbol_table::dump): New function.
30612         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30614 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30616         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30617         that nonconst implies exec.
30619 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30621         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30622         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30623         (inline_edge_summary_vec): Turn into ...
30624         (ipa_call_summaries): ... this one.
30625         (redirect_to_unreachable, edge_set_predicate,
30626         evaluate_properties_for_edge, inline_summary_alloc,
30627         reset_ipa_call_summary, reset_inline_summary,
30628         inline_summary_t::duplicate): Update.
30629         (inline_edge_duplication_hook): Turn to ...
30630         (ipa_call_summary_t::duplicate): ... this one.
30631         (inline_edge_removal_hook): Turn to ...
30632         (ipa_call_summary_t::remove): ... this one.
30633         (dump_inline_edge_summary): Turn to ...
30634         (dump_ipa_call_summary): ... this one.
30635         (estimate_function_body_sizes): Update.
30636         (inline_update_callee_summaries): Update.
30637         (remap_edge_change_prob): Update.
30638         (remap_edge_summaries): Update.
30639         (inline_merge_summary): Update.
30640         (do_estimate_edge_time): Update.
30641         (inline_generate_summary): Update.
30642         (inline_read_section): Update.
30643         (inline_read_summary): Update.
30644         (inline_free_summary): Update.
30645         * ipa-inline.c (can_inline_edge_p): Update.
30646         (compute_inlined_call_time): Update.
30647         (want_inline_small_function_p): Update.
30648         (edge_badness): Update.
30649         (early_inliner): Update.
30650         * ipa-inline.h (inline_edge_summary): Turn to ...
30651         (ipa_call_summary): ... this one.
30652         (ipa_call_summary_t): New class.
30653         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30654         (ipa_call_summaries): New.
30655         (inline_edge_summary): Remove.
30656         (estimate_edge_growth): Update.
30657         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30658         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30659         * ipa-split.c (execute_split_functions): Update.
30660         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30662 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30664         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30665         attributes): Document rdrand effective target.
30667 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30669         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30670         attributes): Sort alphabetically.
30672 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30674         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30676 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30678         PR target/80718
30679         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30680         V2DF/V2DI splat into two separate patterns, one that handles
30681         registers, and the other that only handles memory.  Drop support
30682         for splatting from a GPR on ISA 2.07 and then splitting the
30683         splat into direct move and splat.
30684         (vsx_splat_<mode>_reg): Likewise.
30685         (vsx_splat_<mode>_mem): Likewise.
30687 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30689         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30691 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30693         PR middle-end/80809
30694         * omp-low.c (finish_taskreg_remap): New function.
30695         (finish_taskreg_scan): If unit size of ctx->record_type
30696         is non-constant, unshare the size expression and replace
30697         decls in it with possible outer var refs.
30699         PR middle-end/80809
30700         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30701         GOVD_SHARED rather than GOVD_PRIVATE with it.
30702         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30703         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30705         PR middle-end/80853
30706         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30707         as last argument to build_outer_var_ref for pointer bases of array
30708         section reductions.
30710 2017-05-19  Martin Sebor  <msebor@redhat.com>
30712         * print-tree.c (print_node): Print DECL_READ_P flag.
30714 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30716         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30717         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30718         * cgraph.c: Likewise.
30719         * cgraphunit.c: Likewise.
30720         * gengtype.c: Likewise.
30721         * ipa-cp.c: Likewise.
30722         * ipa-devirt.c: Likewise.
30723         * ipa-icf.c: Likewise.
30724         * ipa-predicate.c: Likewise.
30725         * ipa-profile.c: Likewise.
30726         * ipa-prop.c: Likewise.
30727         * ipa-split.c: Likewise.
30728         * ipa.c: Likewise.
30729         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30730         edge_predicate_pool, dump_inline_hints,
30731         inline_summary::account_size_time, redirect_to_unreachable,
30732         edge_set_predicate, set_hint_predicate,
30733         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30734         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30735         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30736         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30737         ipa_call_summary_t::remove, initialize_growth_caches,
30738         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30739         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30740         mark_modified, unmodified_parm_1, unmodified_parm,
30741         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30742         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30743         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30744         will_be_nonconstant_predicate, record_modified_bb_info,
30745         get_minimal_bb, record_modified, param_change_prob,
30746         phi_result_unknown_predicate, predicate_for_phi_result,
30747         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30748         estimate_function_body_sizes, compute_inline_parameters,
30749         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30750         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30751         inline_update_callee_summaries, remap_edge_change_prob,
30752         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30753         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30754         inline_analyze_function, inline_summary_t::insert,
30755         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30756         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30757         inline_free_summary): Move to ipa-fnsummary.h
30758         (predicate_t): Remove.
30759         * ipa-fnsummary.c: New file.
30760         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30761         (enum inline_hints_vals, inline_hints, agg_position_info,
30762         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30763         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30764         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30765         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30766         inline_read_summary, inline_write_summary, inline_free_summary,
30767         inline_analyze_function, initialize_inline_failed,
30768         inline_merge_summary, inline_update_overall_summary,
30769         compute_inline_parameters): Move to ipa-fnsummary.h
30770         * ipa-fnsummary.h: New file.
30771         * ipa-inline-transform.h: Include ipa-inline.h.
30772         * ipa-inline.c: LIkewise.
30774 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30776         * ipa-inline.c (edge_badness): Use inlined_time instead of
30777         inline_summaries->get.
30779 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30781         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30783 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30785         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30786         (fdump-lang): Document 'raw' option.
30787         * dumpfile.h (TDI_tu): Delete.
30788         * dumpfile.c (dump_files): Remove translation-unit.
30789         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30791 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30793         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30794         command option from $(AWK) call.
30795         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30796         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30797         [FORMAT]: Remove handling of variable.
30798         * config/avr/t-multilib: Regenerate.
30800 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30802         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30803         self_time.
30804         (dump_inline_summary): Do not print self_time.
30805         (estimate_function_body_sizes): Do not set self_time.
30806         (compute_inline_parameters): Likewise.
30807         (inline_read_section, inline_write_summary): Do not stream self_time.
30808         * ipa-inline.h (inline_summary): Drop self_time.
30810 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30812         * ipa-inline-analysis.c (account_size_time): Rename to ...
30813         (inline_summary::account_size_time): ... this one.
30814         (reset_ipa_call_summary): Turn to ...
30815         (ipa_call_summary::reset): ... this one.
30816         (reset_inline_summary): Turn to ...
30817         (inline_summary::reset): ... this one.
30818         (inline_summary_t::remove): Update.
30819         (inline_summary_t::duplicate): Update.
30820         (ipa_call_summary_t::remove): Update.
30821         (dump_inline_summary): Update.
30822         (estimate_function_body_sizes): Update.
30823         (compute_inline_parameters): Update.
30824         (estimate_node_size_and_time): Update.
30825         (inline_merge_summary): Update.
30826         (inline_update_overall_summary): Update.
30827         (inline_read_section): Update.
30828         (inline_write_summary): Update.
30829         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30830         add account_size_time and reset member functions.
30831         (ipa_call_summary): Add reset function.
30832         * ipa-predicate.h (predicate::operator &): Constify.
30834 2017-05-22  Richard Biener  <rguenther@suse.de>
30836         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30838 2017-05-19  Jason Merrill  <jason@redhat.com>
30840         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30842 2017-05-19  Marek Polacek  <polacek@redhat.com>
30844         PR sanitizer/80800
30845         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30846         TYPE_OVERFLOW_WRAPS checks.
30848 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30850         * tree-core.h (enum omp_clause_default_kind): Add
30851         "OMP_CLAUSE_DEFAULT_PRESENT".
30852         * tree-pretty-print.c (dump_omp_clause): Handle it.
30853         * gimplify.c (enum gimplify_omp_var_data): Add
30854         "GOVD_MAP_FORCE_PRESENT".
30855         (gimplify_adjust_omp_clauses_1): Map it to
30856         "GOMP_MAP_FORCE_PRESENT".
30857         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30859         * gimplify.c (oacc_default_clause): Clarify.
30861 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30863         LANG_HOOK_REGISTER_DUMPS
30864         * toplev.c (general_init): Call register dump lang hook.
30865         * doc/invoke.texi: Document -fdump-lang option family.
30866         * dumpfile.c (dump_files): Remove class dump here.
30867         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30868         * dumpfile.h (tree_dump_index): Remove TDI_class.
30869         * langhooks-def.h (lhd_register_dumps): Declare.
30870         (LANG_HOOKS_REGISTER_DUMPS): Define.
30871         (LANG_HOOKS_INITIALIZER): Add it.
30872         * langhooks.c (lhd_register_dumps): Define.
30873         * langhooks.h (struct lang_hooks): Add register_dumps.
30875 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30877         * context.h (context::set_passes): New.
30878         * context.c (context::context): Do not create pass manager.
30879         * toplev.c (general_init): Create pass manager here.
30881 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30883         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30884         use this splitter if two add or or instructions would also work for
30885         the constant we want to generate.
30887 2017-05-19  Richard Biener  <rguenther@suse.de>
30889         PR build/80821
30890         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30891         predicate evaluation.
30893 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30895         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30896         add ctor.
30897         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30898         unsigned.
30900 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30902         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30903         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30904         (inline_edge_summary_vec): Turn into ...
30905         (ipa_call_summaries): ... this one.
30906         (redirect_to_unreachable, edge_set_predicate,
30907         evaluate_properties_for_edge, inline_summary_alloc,
30908         reset_ipa_call_summary, reset_inline_summary,
30909         inline_summary_t::duplicate): Update.
30910         (inline_edge_duplication_hook): Turn to ...
30911         (ipa_call_summary_t::duplicate): ... this one.
30912         (inline_edge_removal_hook): Turn to ...
30913         (ipa_call_summary_t::remove): ... this one.
30914         (dump_inline_edge_summary): Turn to ...
30915         (dump_ipa_call_summary): ... this one.
30916         (estimate_function_body_sizes): Update.
30917         (inline_update_callee_summaries): Update.
30918         (remap_edge_change_prob): Update.
30919         (remap_edge_summaries): Update.
30920         (inline_merge_summary): Update.
30921         (do_estimate_edge_time): Update.
30922         (inline_generate_summary): Update.
30923         (inline_read_section): Update.
30924         (inline_read_summary): Update.
30925         (inline_free_summary): Update.
30926         * ipa-inline.c (can_inline_edge_p): Update.
30927         (compute_inlined_call_time): Update.
30928         (want_inline_small_function_p): Update.
30929         (edge_badness): Update.
30930         (early_inliner): Update.
30931         * ipa-inline.h (inline_edge_summary): Turn to ...
30932         (ipa_call_summary): ... this one.
30933         (ipa_call_summary_t): New class.
30934         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30935         (ipa_call_summaries): New.
30936         (inline_edge_summary): Remove.
30937         (estimate_edge_growth): Update.
30938         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30939         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30940         * ipa-split.c (execute_split_functions): Update.
30941         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30943 2017-05-19  Richard Biener  <rguenther@suse.de>
30945         PR middle-end/80764
30946         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
30948 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
30950         * config/rs6000/rs6000.c (struct machine_function): Add field
30951         fpr_is_wrapped_separately.
30952         (rs6000_get_separate_components): Use 64 components.  Handle the
30953         new FPR components.
30954         (rs6000_components_for_bb): Handle the FPR components.
30955         (rs6000_emit_prologue_components): Handle the FPR components.
30956         (rs6000_emit_epilogue_components): Handle the FPR components.
30957         (rs6000_set_handled_components): Handle the FPR components.
30958         (rs6000_emit_prologue): Don't output prologue code for those FPRs
30959         that are already separately shrink-wrapped.
30960         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
30961         that are already separately shrink-wrapped.
30963 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
30965         PR target/80510
30966         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
30967         New predicate.
30969         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
30970         (define_peephole2 for Altivec d-form load): Add peepholes to catch
30971         cases where the register allocator uses a move and an offsettable
30972         memory operation to/from a FPR register on ISA 2.06/2.07.
30973         (define_peephole2 for Altivec d-form store): Likewise.
30975 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
30977         PR target/80799
30978         * config/i386/mmx.md (*mov<mode>_internal): Enable
30979         alternatives 11, 12, 13 and 14 also for 32bit targets.
30980         Remove alternatives 15, 16, 17 and 18.
30981         * config/i386/sse.md (vec_concatv2di): Change
30982         alternative (!x, *y) to (x, ?!*Yn).
30984 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
30986         * dumpfile.h (enum dump_kind): Remove stray comma.
30988 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30990         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
30991         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
30992         predicate::num_conditions
30993         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
30994         (CHANGED): turn into predicate::changed.
30995         (agg_position_info): Move to ipa-predicate.h
30996         (add_condition, predicate::add_clause, predicate::operator &=,
30997         predicate::or_with, predicate::evaluate, predicate::probability,
30998         dump_condition, dump_clause, predicate::dump,
30999         predicate::remap_after_duplication, predicate::remap_after_inlining,
31000         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
31001         (evaluate_conditions_for_known_args): Update.
31002         (set_cond_stmt_execution_predicate): Update.
31003         * ipa-inline.h: Include ipa-predicate.h
31004         (condition, inline_param_summary, conditions, agg_position_info,
31005         predicate): Move to ipa-predicate.h
31006         * ipa-predicate.c: New file.
31007         * ipa-predicate.h: New file.
31009 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
31011         * final.c (leaf_function_p): Check we are not in a sequence.
31013 2017-05-18  Martin Liska  <mliska@suse.cz>
31015         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
31016         * dumpfile.c (dump_register): Use new enum dump_kind.
31017         (get_dump_file_name): Likewise.
31018         (dump_enable_all): Likewise.
31019         (dump_switch_p_1): Likewise.
31020         (enable_rtl_dump_file): Remove usage of TDF_RTL.
31021         * dumpfile.h (enum dump_kind): New enum type.
31022         (struct dump_file_info): Create constructor and
31023         format fields and comments.
31024         * passes.c (pass_manager::register_one_dump_file):
31025         Use num dump_kind.
31026         * statistics.c (statistics_early_init): Likewise.
31027         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31028         TDF_TREE with TDF_SLIM.
31029         (gather_memory_references_ref): Likewise.
31031 2017-05-18  Martin Liska  <mliska@suse.cz>
31033         * vec.h (struct vnull): Use it.
31035 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31037         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31038         (true_predicate, false_predicate, true_predicate_p,
31039         false_predicate_p): Remove.
31040         (single_cond_predicate, not_inlined_predicate): Turn to member function
31041         in ipa-inline.h
31042         (add_condition): Update.
31043         (add_clause): Turn to...
31044         (predicate::add_clause): ... this one; update; allow passing NULL
31045         as parameter.
31046         (and_predicates): Turn to ...
31047         (predicate::operator &=): ... this one.
31048         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31049         (or_predicates): Turn to ...
31050         (predicate::or_with): ... this one.
31051         (evaluate_predicate): Turn to ...
31052         (predicate::evaluate): ... this one.
31053         (predicate_probability): Turn to ...
31054         (predicate::probability): ... this one.
31055         (dump_condition): Update.
31056         (dump_predicate): Turn to ...
31057         (predicate::dump): ... this one.
31058         (account_size_time): Update.
31059         (edge_set_predicate): Update.
31060         (set_hint_predicate): UPdate.
31061         (evaluate_conditions_for_known_args): Update.
31062         (evaluate_properties_for_edge): Update.
31063         (remap_predicate_after_duplication): Turn to...
31064         (predicate::remap_after_duplication): ... this one.
31065         (remap_hint_predicate_after_duplication): Update.
31066         (inline_summary_t::duplicate): UPdate.
31067         (dump_inline_edge_summary): Update.
31068         (dump_inline_summary): Update.
31069         (set_cond_stmt_execution_predicate): Update.
31070         (set_switch_stmt_execution_predicate): Update.
31071         (compute_bb_predicates): Update.
31072         (will_be_nonconstant_expr_predicate): Update.
31073         (will_be_nonconstant_predicate): Update.
31074         (phi_result_unknown_predicate): Update.
31075         (predicate_for_phi_result): Update.
31076         (array_index_predicate): Update.
31077         (estimate_function_body_sizes): Update.
31078         (estimate_node_size_and_time): Update.
31079         (estimate_ipcp_clone_size_and_time): Update.
31080         (remap_predicate): Rename to ...
31081         (predicate::remap_after_inlining): ... this one.
31082         (remap_hint_predicate): Update.
31083         (inline_merge_summary): Update.
31084         (inline_update_overall_summary): Update.
31085         (estimate_size_after_inlining): Update.
31086         (read_predicate): Rename to ...
31087         (predicate::stream_in): ... this one.
31088         (read_inline_edge_summary): Update.
31089         (write_predicate): Rename to ...
31090         (predicate::stream_out): ... this one.
31091         (write_inline_edge_summary): Update.
31092         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31093         (clause_t): Turn to uint32_t
31094         (predicate): Turn to class; implement constructor and operators
31095         ==, !=, &
31096         (size_time_entry): Update.
31097         (inline_summary): Update.
31098         (inline_edge_summary): Update.
31100 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31102         * fold-const.c (fold_binary_loc): Move transformation...
31103         * match.pd (C - X CMP X): ... here.
31105 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31107         * config/sparc/sparc.c (sparc_option_override): Set function
31108         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31109         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31110         latency to 1.
31111         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31112         latency to 2.
31113         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31115 2017-05-18  Marek Polacek  <polacek@redhat.com>
31117         PR sanitizer/80797
31118         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31119         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31120         gimple_assign_load_p.
31122 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31124         PR middle-end/80692
31125         * real.c (do_compare): Give decimal_do_compare preference over
31126         comparing just the signs.
31128 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31130         * doc/md.texi (Canonicalization of Instructions): Describe the
31131         canonical form of instructions that inherently set a condition
31132         code register.
31134 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31136         PR middle-end/80775
31137         * tree-cfg.c: Move deletion of unreachable case statements to after
31138         the merging of consecutive case labels.
31140 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31142         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31143         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31144         restoring of callee-saved registers.
31146 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31148         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31149         * config/visium/visium.c (single_set_and_flags): Likewise.
31150         * config/visium/visium.md (Substitutions): Likewise.
31152 2017-05-17  Martin Liska  <mliska@suse.cz>
31154         * cfg.c: Introduce dump_flags_t type and
31155         use it instead of int type.
31156         * cfg.h: Likewise.
31157         * cfghooks.c: Likewise.
31158         * cfghooks.h (struct cfg_hooks): Likewise.
31159         * cfgrtl.c: Likewise.
31160         * cfgrtl.h: Likewise.
31161         * cgraph.c (cgraph_node::get_body): Likewise.
31162         * coretypes.h: Likewise.
31163         * domwalk.c: Likewise.
31164         * domwalk.h: Likewise.
31165         * dumpfile.c (struct dump_option_value_info): Likewise.
31166         (dump_enable_all): Likewise.
31167         (dump_switch_p_1): Likewise.
31168         (opt_info_switch_p): Likewise.
31169         * dumpfile.h (enum tree_dump_index): Likewise.
31170         (struct dump_file_info): Likewise.
31171         * genemit.c: Likewise.
31172         * generic-match-head.c: Likewise.
31173         * gengtype.c (open_base_files): Likewise.
31174         * gimple-pretty-print.c: Likewise.
31175         * gimple-pretty-print.h: Likewise.
31176         * graph.c (print_graph_cfg): Likewise.
31177         * graphite-scop-detection.c (dot_all_sese): Likewise.
31178         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31179         * loop-unroll.c (report_unroll): Likewise.
31180         * passes.c (pass_manager::register_one_dump_file): Likewise.
31181         * print-tree.c: Likewise.
31182         * statistics.c: Likewise.
31183         * tree-cfg.c: Likewise.
31184         * tree-cfg.h: Likewise.
31185         * tree-dfa.c: Likewise.
31186         * tree-dfa.h: Likewise.
31187         * tree-dump.c (dump_function): Likewise.
31188         * tree-dump.h (struct dump_info): Likewise.
31189         * tree-pretty-print.c: Likewise.
31190         * tree-pretty-print.h: Likewise.
31191         * tree-ssa-live.c: Likewise.
31192         * tree-ssa-live.h: Likewise.
31193         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31194         * tree-vect-loop.c: Likewise.
31195         * tree-vect-slp.c: Likewise.
31197 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31198             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31200         PR tree-optimization/80457
31201         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31202         of all arguments to a statement as scalar_to_vec operations.
31203         (vectorizable_call): Adjust call to vect_model_simple_cost for
31204         new parameter.
31205         (vectorizable_conversion): Likewise.
31206         (vectorizable_assignment): Likewise.
31207         (vectorizable_shift): Likewise.
31208         (vectorizable_operation): Likewise.
31209         (vectorizable_comparison): Likewise.
31210         (vect_is_simple_cond): Record the def types for operands.
31211         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31212         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31213         for statement argument count.
31215 2017-05-16  Carl Love  <cel@us.ibm.com>
31217         * config/rs6000/rs6000-c: Add support for built-in functions
31218         vector unsigned long long vec_bperm (vector unsigned long long,
31219                                              vector unsigned char)
31220         vector signed long long vec_mule (vector signed int,
31221                                           vector signed int)
31222         vector unsigned long long vec_mule (vector unsigned int,
31223                                             vector unsigned int)
31224         vector signed long long vec_mulo (vector signed int,
31225                                           vector signed int)
31226         vector unsigned long long vec_mulo (vector unsigned int,
31227                                             vector unsigned int)
31228         vector signed char vec_sldw (vector signed char,
31229                                      vector signed char,
31230                                      const int)
31231         vector unsigned char vec_sldw (vector unsigned char,
31232                                        vector unsigned char,
31233                                        const int)
31234         vector signed short vec_sldw (vector signed short,
31235                                       vector signed short,
31236                                       const int)
31237         vector unsigned short vec_sldw (vector unsigned short,
31238                                         vector unsigned short,
31239                                         const int)
31240         vector signed int vec_sldw (vector signed int,
31241                                     vector signed int,
31242                                     const int)
31243         vector unsigned int vec_sldw (vector unsigned int,
31244                                       vector unsigned int,
31245                                       const int)
31246         vector signed long long vec_sldw (vector signed long long,
31247                                           vector signed long long,
31248                                           const int)
31249         vector unsigned long long vec_sldw (vector unsigned long long,
31250                                             vector unsigned long long,
31251                                             const int)
31252         * config/rs6000/rs6000-c: Add support for built-in functions
31253         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31254         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31255         * doc/extend.texi: Update the built-in documentation for the
31256         new built-in functions.
31258 2017-05-16  Marek Polacek  <polacek@redhat.com>
31260         PR sanitizer/80536
31261         PR sanitizer/80386
31262         * tree.c (save_expr): Don't fold the expression.
31264 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31266         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31267         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31268         and (?*y,m).  Update insn attributes.
31270 2017-05-16  Martin Liska  <mliska@suse.cz>
31272         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31273         flags argument of print_gimple_stmt, print_gimple_expr,
31274         print_generic_stmt and print_generic_expr.
31275         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31276         * coretypes.h: Likewise.
31277         * except.c (dump_eh_tree): Likewise.
31278         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31279         * gimple-pretty-print.h: Likewise.
31280         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31281         (backprop::push_to_worklist): Likewise.
31282         (backprop::pop_from_worklist): Likewise.
31283         (backprop::process_use): Likewise.
31284         (backprop::intersect_uses): Likewise.
31285         (note_replacement): Likewise.
31286         * gimple-ssa-store-merging.c
31287         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31288         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31289         (pass_store_merging::execute): Likewise.
31290         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31291         (ssa_base_cand_dump_callback): Likewise.
31292         (dump_incr_vec): Likewise.
31293         (replace_refs): Likewise.
31294         (replace_mult_candidate): Likewise.
31295         (create_add_on_incoming_edge): Likewise.
31296         (create_phi_basis): Likewise.
31297         (insert_initializers): Likewise.
31298         (all_phi_incrs_profitable): Likewise.
31299         (introduce_cast_before_cand): Likewise.
31300         (replace_one_candidate): Likewise.
31301         * gimplify.c (gimplify_expr): Likewise.
31302         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31303         (set_rename): Likewise.
31304         (rename_uses): Likewise.
31305         (copy_loop_phi_nodes): Likewise.
31306         (add_close_phis_to_merge_points): Likewise.
31307         (copy_loop_close_phi_args): Likewise.
31308         (copy_cond_phi_args): Likewise.
31309         (graphite_copy_stmts_from_block): Likewise.
31310         (translate_pending_phi_nodes): Likewise.
31311         * graphite-poly.c (print_pdr): Likewise.
31312         (dump_gbb_cases): Likewise.
31313         (dump_gbb_conditions): Likewise.
31314         (print_scop_params): Likewise.
31315         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31316         (build_cross_bb_scalars_use): Likewise.
31317         (gather_bbs::before_dom_children): Likewise.
31318         * hsa-dump.c (dump_hsa_immed): Likewise.
31319         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31320         (get_replacement_map): Likewise.
31321         * ipa-inline-analysis.c (dump_condition): Likewise.
31322         (estimate_function_body_sizes): Likewise.
31323         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31324         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31325         * ipa-prop.c (ipa_dump_param): Likewise.
31326         (ipa_print_node_jump_functions_for_edge): Likewise.
31327         (ipa_modify_call_arguments): Likewise.
31328         (ipa_modify_expr): Likewise.
31329         (ipa_dump_param_adjustments): Likewise.
31330         (ipa_dump_agg_replacement_values): Likewise.
31331         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31332         * ipa-pure-const.c (check_stmt): Likewise.
31333         (pass_nothrow::execute): Likewise.
31334         * ipa-split.c (execute_split_functions): Likewise.
31335         * omp-offload.c (dump_oacc_loop_part): Likewise.
31336         (dump_oacc_loop): Likewise.
31337         * trans-mem.c (tm_log_emit): Likewise.
31338         (tm_memopt_accumulate_memops): Likewise.
31339         (dump_tm_memopt_set): Likewise.
31340         (dump_tm_memopt_transform): Likewise.
31341         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31342         (print_loop): Likewise.
31343         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31344         (chkp_gather_checks_info): Likewise.
31345         (chkp_get_check_result): Likewise.
31346         (chkp_remove_check_if_pass): Likewise.
31347         (chkp_use_outer_bounds_if_possible): Likewise.
31348         (chkp_reduce_bounds_lifetime): Likewise.
31349         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31350         (chkp_mark_completed_bounds): Likewise.
31351         (chkp_register_incomplete_bounds): Likewise.
31352         (chkp_mark_invalid_bounds): Likewise.
31353         (chkp_maybe_copy_and_register_bounds): Likewise.
31354         (chkp_build_returned_bound): Likewise.
31355         (chkp_get_bound_for_parm): Likewise.
31356         (chkp_build_bndldx): Likewise.
31357         (chkp_get_bounds_by_definition): Likewise.
31358         (chkp_generate_extern_var_bounds): Likewise.
31359         (chkp_get_bounds_for_decl_addr): Likewise.
31360         * tree-chrec.c (chrec_apply): Likewise.
31361         * tree-data-ref.c (dump_data_reference): Likewise.
31362         (dump_subscript): Likewise.
31363         (dump_data_dependence_relation): Likewise.
31364         (analyze_overlapping_iterations): Likewise.
31365         * tree-inline.c (expand_call_inline): Likewise.
31366         (tree_function_versioning): Likewise.
31367         * tree-into-ssa.c (dump_defs_stack): Likewise.
31368         (dump_currdefs): Likewise.
31369         (dump_names_replaced_by): Likewise.
31370         (dump_update_ssa): Likewise.
31371         (update_ssa): Likewise.
31372         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31373         * tree-parloops.c (build_new_reduction): Likewise.
31374         (try_create_reduction_list): Likewise.
31375         (ref_conflicts_with_region): Likewise.
31376         (oacc_entry_exit_ok_1): Likewise.
31377         (oacc_entry_exit_single_gang): Likewise.
31378         * tree-pretty-print.h: Likewise.
31379         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31380         (get_scalar_evolution): Likewise.
31381         (add_to_evolution): Likewise.
31382         (get_loop_exit_condition): Likewise.
31383         (analyze_evolution_in_loop): Likewise.
31384         (analyze_initial_condition): Likewise.
31385         (analyze_scalar_evolution): Likewise.
31386         (instantiate_scev): Likewise.
31387         (number_of_latch_executions): Likewise.
31388         (gather_chrec_stats): Likewise.
31389         (final_value_replacement_loop): Likewise.
31390         (scev_const_prop): Likewise.
31391         * tree-sra.c (dump_access): Likewise.
31392         (disqualify_candidate): Likewise.
31393         (create_access): Likewise.
31394         (reject): Likewise.
31395         (maybe_add_sra_candidate): Likewise.
31396         (create_access_replacement): Likewise.
31397         (analyze_access_subtree): Likewise.
31398         (analyze_all_variable_accesses): Likewise.
31399         (sra_modify_assign): Likewise.
31400         (initialize_constant_pool_replacements): Likewise.
31401         (find_param_candidates): Likewise.
31402         (decide_one_param_reduction): Likewise.
31403         (replace_removed_params_ssa_names): Likewise.
31404         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31405         * tree-ssa-copy.c (dump_copy_of): Likewise.
31406         (copy_prop_visit_cond_stmt): Likewise.
31407         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31408         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31409         (record_equivalences_from_stmt): Likewise.
31410         * tree-ssa-dse.c (compute_trims): Likewise.
31411         (delete_dead_call): Likewise.
31412         (delete_dead_assignment): Likewise.
31413         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31414         (forward_propagate_into_cond): Likewise.
31415         (pass_forwprop::execute): Likewise.
31416         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31417         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31418         Likewise.
31419         (move_computations_worker): Likewise.
31420         (execute_sm): Likewise.
31421         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31422         (remove_exits_and_undefined_stmts): Likewise.
31423         (remove_redundant_iv_tests): Likewise.
31424         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31425         (adjust_iv_update_pos): Likewise.
31426         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31427         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31428         (value_replacement): Likewise.
31429         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31430         * tree-ssa-pre.c (print_pre_expr): Likewise.
31431         (get_representative_for): Likewise.
31432         (create_expression_by_pieces): Likewise.
31433         (insert_into_preds_of_block): Likewise.
31434         (eliminate_insert): Likewise.
31435         (eliminate_dom_walker::before_dom_children): Likewise.
31436         (eliminate): Likewise.
31437         (remove_dead_inserted_code): Likewise.
31438         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31439         * tree-ssa-reassoc.c (get_rank): Likewise.
31440         (eliminate_duplicate_pair): Likewise.
31441         (eliminate_plus_minus_pair): Likewise.
31442         (eliminate_not_pairs): Likewise.
31443         (undistribute_ops_list): Likewise.
31444         (eliminate_redundant_comparison): Likewise.
31445         (update_range_test): Likewise.
31446         (optimize_range_tests_var_bound): Likewise.
31447         (optimize_vec_cond_expr): Likewise.
31448         (rewrite_expr_tree): Likewise.
31449         (rewrite_expr_tree_parallel): Likewise.
31450         (linearize_expr): Likewise.
31451         (break_up_subtract): Likewise.
31452         (linearize_expr_tree): Likewise.
31453         (attempt_builtin_powi): Likewise.
31454         (attempt_builtin_copysign): Likewise.
31455         (transform_stmt_to_copy): Likewise.
31456         (transform_stmt_to_multiply): Likewise.
31457         (dump_ops_vector): Likewise.
31458         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31459         (print_scc): Likewise.
31460         (set_ssa_val_to): Likewise.
31461         (visit_reference_op_store): Likewise.
31462         (visit_use): Likewise.
31463         (sccvn_dom_walker::before_dom_children): Likewise.
31464         (run_scc_vn): Likewise.
31465         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31466         Likewise.
31467         (expr_hash_elt::print): Likewise.
31468         (const_and_copies::pop_to_marker): Likewise.
31469         (const_and_copies::record_const_or_copy_raw): Likewise.
31470         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31471         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31472         (dump_predicates): Likewise.
31473         (find_uninit_use): Likewise.
31474         (warn_uninitialized_phi): Likewise.
31475         (pass_late_warn_uninitialized::execute): Likewise.
31476         * tree-ssa.c (verify_vssa): Likewise.
31477         (verify_ssa): Likewise.
31478         (maybe_optimize_var): Likewise.
31479         * tree-vrp.c (dump_value_range): Likewise.
31480         (dump_all_value_ranges): Likewise.
31481         (dump_asserts_for): Likewise.
31482         (register_edge_assert_for_2): Likewise.
31483         (vrp_visit_cond_stmt): Likewise.
31484         (vrp_visit_switch_stmt): Likewise.
31485         (vrp_visit_stmt): Likewise.
31486         (vrp_visit_phi_node): Likewise.
31487         (simplify_cond_using_ranges_1): Likewise.
31488         (fold_predicate_in): Likewise.
31489         (evrp_dom_walker::before_dom_children): Likewise.
31490         (evrp_dom_walker::push_value_range): Likewise.
31491         (evrp_dom_walker::pop_value_range): Likewise.
31492         (execute_early_vrp): Likewise.
31494 2017-05-16  Richard Biener  <rguenther@suse.de>
31496         * dwarf2out.c (loc_list_from_tree_1): Do not create
31497         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31499 2017-05-16  Richard Biener  <rguenther@suse.de>
31501         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31502         just generated.
31503         (note_variable_value_in_expr): If we resolved the decl ref
31504         do not push to the stack.
31506 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31508         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31509         operations in fast-math mode.
31510         (vaddq_f16): Likewise.
31511         (vmul_f16): Likewise.
31512         (vmulq_f16): Likewise.
31513         (vsub_f16): Likewise.
31514         (vsubq_f16): Likewise.
31515         * config/arm/neon.md (add<mode>3): New.
31516         (sub<mode>3): New.
31517         (fma:<VH:mode>3): New.  Also remove outdated comment.
31518         (mul<mode>3): New.
31520 2017-05-16  Martin Liska  <mliska@suse.cz>
31522         PR ipa/79849.
31523         PR ipa/79850.
31524         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31525         (odr_types_equivalent_p): Likewise.
31527 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31529         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31531 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31533         PR target/80425
31534         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31535         non-interunit SSE move alternatives with '?'.
31536         (zero-extendsidi peephole2): New peephole to skip intermediate
31537         general register in SSE zero-extend sequence.
31539 2017-05-15  Jeff Law  <law@redhat.com>
31541         * reorg.c (relax_delay_slots): Create a new variable to hold
31542         the temporary target rather than clobbering TARGET_LABEL.
31544         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31545         missing argument to extract_bit_field call.
31546         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31548 2017-05-15  Martin Liska  <mliska@suse.cz>
31550         PR driver/31468
31551         * gcc.c (process_command): Do not allow empty argument of -o option.
31553 2017-05-15  Renlin Li  <renlin.li@arm.com>
31555         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31556         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31557         * config/aarch64/constraints.md (Usf): Add long call check.
31558         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31559         (call_value): Likewise.
31560         (sibcall): Likewise.
31561         (sibcall_value): Likewise.
31562         (call_insn): New.
31563         (call_value_insn): New.
31564         (sibcall_insn): Update rtx pattern.
31565         (sibcall_value_insn): Likewise.
31566         (call_internal): Remove.
31567         (call_value_internal): Likewise.
31568         (sibcall_internal): Likewise.
31569         (sibcall_value_internal): Likewise.
31570         (call_reg): Likewise.
31571         (call_symbol): Likewise.
31572         (call_value_reg): Likewise.
31573         (call_value_symbol): Likewise.
31575 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31577         PR target/80600
31578         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31580 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31582         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31583         compatible with CCGOCmode and with CCZmode.
31585 2017-05-14  Martin Sebor  <msebor@redhat.com>
31587         PR middle-end/77671
31588         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31589         (gimple_fold_builtin_snprintf): Same.
31590         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31591         (gimple_fold_builtin_snprintf): Same.
31592         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31593         of character types.
31594         (is_call_safe): New function.
31595         (try_substitute_return_value): Call it.
31596         (try_simplify_call): New function.
31597         (pass_sprintf_length::handle_gimple_call): Call it.
31599 2017-05-14  Martin Sebor  <msebor@redhat.com>
31601         PR middle-end/80669
31602         * builtins.c (expand_builtin_stpncpy): Simplify.
31604 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31606         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31607         * config/i386/i386.h
31608         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31609         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31610         (struct machine_function): Add new members call_ms2sysv,
31611         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31612         (struct machine_frame_state): New fields sp_realigned and
31613         sp_realigned_offset.
31614         * config/i386/i386.c
31615         (enum xlogue_stub): New enum.
31616         (enum xlogue_stub_sets): New enum.
31617         (class xlogue_layout): New class.
31618         (struct ix86_frame): New fields stack_realign_allocate_offset,
31619         stack_realign_offset and outlined_save_offset.  Modify comments to
31620         detail stack layout when using out-of-line stubs.
31621         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31622         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31623         -mcall-ms2sysv-xlogues.
31624         (stub_managed_regs): New static variable.
31625         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31626         registers managed by out-of-line stub.
31627         (disable_call_ms2sysv_xlogues): New function.
31628         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31629         m->call_ms2sysv when appropriate and compute frame layout for
31630         out-of-line stubs.
31631         (sp_valid_at, fp_valid_at): New inline functions.
31632         (choose_basereg): New function.
31633         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31634         all callers.
31635         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31636         Use align parameter of choose_baseaddr to generated aligned SSE movs
31637         when possible.
31638         (pro_epilogue_adjust_stack): Modify to track
31639         machine_frame_state::sp_realigned.
31640         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31641         (ix86_nsaved_sseregs): Likewise.
31642         (ix86_emit_save_regs): Likewise.
31643         (ix86_emit_save_regs_using_mov): Likewise.
31644         (ix86_emit_save_sse_regs_using_mov): Likewise.
31645         (get_scratch_register_on_entry): Likewise.
31646         (gen_frame_set): New function.
31647         (gen_frame_load): Likewise.
31648         (gen_frame_store): Likewise.
31649         (emit_outlined_ms2sysv_save): Likewise.
31650         (emit_outlined_ms2sysv_restore): Likewise.
31651         (ix86_expand_prologue): Modify stack re-alignment code and call
31652         emit_outlined_ms2sysv_save when appropriate.
31653         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31654         parameter rtx_insn *insn, which allows the function to be used to only
31655         generate the notes.
31656         (ix86_expand_epilogue): Modify validity checks of frame and stack
31657         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31658         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31659         * config/i386/predicates.md
31660         (save_multiple): New predicate.
31661         (restore_multiple): Likewise.
31662         * config/i386/sse.md
31663         (save_multiple<mode>): New pattern.
31664         (save_multiple_realign<mode>): Likewise.
31665         (restore_multiple<mode>): Likewise.
31666         (restore_multiple_and_return<mode>): Likewise.
31667         (restore_multiple_leave_return<mode>): Likewise.
31668         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31670 2017-05-14  Julia Koval  <julia.koval@intel.com>
31672         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31673         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31674         (__builtin_ia32_xsetbv): New builtins.
31675         * config/i386/i386.c (ix86_expand_special_args_builtin):
31676         Process new types.
31677         (ix86_expand_builtin): Special expand for new intrinsics.
31678         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31679         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31680         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31682 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31684         * cfganal.c (inverted_post_order_compute): Change argument type
31685         to vec *.
31686         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31687         * df-core.c (rest_of_handle_df_initialize): Adjust.
31688         (rest_of_handle_df_finish): Likewise.
31689         (df_analyze_1): Likewise.
31690         (df_analyze): Likewise.
31691         (loop_inverted_post_order_compute): Change argument to be a vec *.
31692         (df_analyze_loop): Adjust.
31693         (df_get_n_blocks): Likewise.
31694         (df_get_postorder): Likewise.
31695         * df.h (struct df_d): Change field to be a vec.
31696         * lcm.c (compute_laterin): Adjust.
31697         (compute_available): Likewise.
31698         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31699         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31700         * tree-ssa-pre.c (compute_antic): Likewise.
31702 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31704         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31705         (depth_first_search::depth_first_search): Change structure init
31706         function to this constructor.
31707         (depth_first_search::add_bb): Rename function to this member.
31708         (depth_first_search::execute): Likewise.
31709         (flow_dfs_compute_reverse_finish): Adjust.
31711 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31713         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31714         (longest_simple_path): Likewise.
31715         * shrink-wrap.c (spread_components): Likewise.
31716         (disqualify_problematic_components): Likewise.
31717         (emit_common_heads_for_components): Likewise.
31718         (emit_common_tails_for_components): Likewise.
31719         (insert_prologue_epilogue_for_components): Likewise.
31721 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31723         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31724         auto_sbitmap.
31726 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31728         * df-core.c (df_set_blocks): Start using auto_bitmap.
31729         (df_compact_blocks): Likewise.
31730         * df-problems.c (df_rd_confluence_n): Likewise.
31731         * df-scan.c (df_insn_rescan_all): Likewise.
31732         (df_process_deferred_rescans): Likewise.
31733         (df_update_entry_block_defs): Likewise.
31734         (df_update_exit_block_uses): Likewise.
31735         (df_entry_block_bitmap_verify): Likewise.
31736         (df_exit_block_bitmap_verify): Likewise.
31737         (df_scan_verify): Likewise.
31738         * lra-constraints.c (lra_constraints): Likewise.
31739         (undo_optional_reloads): Likewise.
31740         (lra_undo_inheritance): Likewise.
31741         * lra-remat.c (calculate_gen_cands): Likewise.
31742         (do_remat): Likewise.
31743         * lra-spills.c (assign_spill_hard_regs): Likewise.
31744         (spill_pseudos): Likewise.
31745         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31746         (bitmap_set_subtract_values): Likewise.
31748 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31750         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31751         management with auto_bitmap.
31752         (fix_inter_tick): Likewise.
31753         (fix_recovery_deps): Likewise.
31754         * ira.c (add_store_equivs): Likewise.
31755         (find_moveable_pseudos): Likewise.
31756         (split_live_ranges_for_shrink_wrap): Likewise.
31757         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31758         (rtx_reuse_manager::seen_def_p): Likewise.
31759         (rtx_reuse_manager::set_seen_def): Likewise.
31760         * print-rtl.h (class rtx_reuse_manager): Likewise.
31762 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31764         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31765         lifetime.
31766         (migrate_btr_def): Likewise.
31767         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31768         * df-core.c (loop_post_order_compute): Likewise.
31769         (loop_inverted_post_order_compute): Likewise.
31770         * hsa-common.h: Likewise.
31771         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31772         * init-regs.c (initialize_uninitialized_regs): Likewise.
31773         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31774         (inline_small_functions): Likewise.
31775         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31776         * ira.c (combine_and_move_insns): Likewise.
31777         (build_insn_chain): Likewise.
31778         * loop-invariant.c (find_invariants): Likewise.
31779         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31780         * predict.c (tree_predict_by_opcode): Likewise.
31781         (predict_paths_leading_to): Likewise.
31782         (predict_paths_leading_to_edge): Likewise.
31783         (estimate_loops_at_level): Likewise.
31784         (estimate_loops): Likewise.
31785         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31786         (spread_components): Likewise.
31787         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31788         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31789         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31790         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31791         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31792         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31793         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31794         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31795         (mark_threaded_blocks): Likewise.
31796         (thread_through_all_blocks): Likewise.
31797         * tree-ssa.c (verify_ssa): Likewise.
31798         (execute_update_addresses_taken): Likewise.
31799         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31801 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31803         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31804         auto_vec.
31805         (post_order_compute): Likewise.
31806         (inverted_post_order_compute): Likewise.
31807         (pre_and_rev_post_order_compute_fn): Likewise.
31809 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31811         * genrecog.c (int_set::int_set): Explicitly construct our
31812         auto_vec base class.
31813         * vec.h (auto_vec::auto_vec): New constructor.
31815 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31817         * bitmap.h (class auto_bitmap): New constructor taking
31818         bitmap_obstack * argument.
31820 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31822         * bitmap.h (class auto_bitmap): Change type of m_bits to
31823         bitmap_head, and adjust ctor / dtor and member operators.
31825 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31827         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31828         when returned register mode doesn't match original mode.
31830 2017-05-12  Jeff Law  <law@redhat.com>
31831             Jakub Jelinek  <jakub@redhat.com>
31833         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31834         we look for cc setter after the compare-elim changes.
31835         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31836         within the vector to match what compare-elim now expects.
31837         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31838         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31840         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31841         after the compare-elim changes.
31842         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31843         the vector to match what compare-elim now expects.
31844         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31845         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31846         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31847         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31848         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31850         * config/visium/visium.c (single_set_and_flags): Fix where
31851         we look for cc setter after the compare-elim changes.
31852         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31853         with the vector to match what compare-elim now expects.
31854         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31855         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31856         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31857         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31858         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31859         (neg<mode>2_insn_set_overflow): Likewise.
31861 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31863         PR middle-end/79794
31864         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31865         maybe_expand_insn call, set ops[0].target.  If still set after call,
31866         set alt_rtl.  Add extra arg to recursive calls.
31867         (extract_bit_field): Add alt_rtl argument.  Pass to
31868         extract_bit_field.
31869         * expmed.h (extract_bit_field): Fix prototype.
31870         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31871         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31872         to extract_bit_field_calls.
31873         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31874         Pass alt_rtl to extract_bit_field calls.
31875         * calls.c (store_unaligned_arguments_into_psuedos)
31876         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31877         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31878         gen_reg_rtx.
31879         * optabs.h (struct expand_operand): Add target bitfield.
31881 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31883         * compare-elim.c (try_eliminate_compare): Canonicalize
31884         operation with embedded compare to
31885         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31886          (set (reg) (operation)].
31888         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31890 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31892         PR target/80723
31893         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31894         cost of adding a carry flag for ADC instruction.
31895         [case MINUS]: Ignore the cost of subtracting a carry flag
31896         for SBB instruction.
31898 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31900         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31901         and x86intrin.h
31902         * config/rs6000/bmiintrin.h: New file.
31903         * config/rs6000/bmi2intrin.h: New file.
31904         * config/rs6000/x86intrin.h: New file.
31906 2017-05-12  Jeff Law  <law@redhat.com>
31908         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31909         markers.
31911 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31913         PR middle-end/80707
31914         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31916 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31918         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31919         early expansion of vector divide builtins.
31920         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31921         builtins identified as having unsigned arguments.
31923 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31925         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31926         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31927         expansion of vector logical operations (and, andc, or, xor,
31928         nor, orc, nand).
31930 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31932         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31933         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31935 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31937         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31938         early GIMPLE expansion of vector multiplies.
31940 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
31942         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
31943         TARGET_HAVE_MOVT conditional.
31944         (movt splitter): Likewise.
31946 2017-05-12  Richard Biener  <rguenther@suse.de>
31948         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
31949         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
31950         Fold all stmts not inplace.
31952 2017-05-12  Richard Biener  <rguenther@suse.de>
31954         PR tree-optimization/80713
31955         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
31956         inserted_exprs bit for not removed stmts.
31958 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
31960         PR middle-end/69921
31961         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
31962         parallelized" attribute for parallelized OpenACC kernels.
31963         * omp-offload.c (execute_oacc_device_lower): Use it.
31965         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
31966         Set "oacc kernels" attribute.
31967         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
31968         parameter.  Adjust all users.
31969         (oacc_fn_attrib_kernels_p): Remove function.
31970         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
31971         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
31972         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
31973         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
31974         assert "oacc kernels" attribute is set.
31976 2017-05-11  Carl Love  <cel@us.ibm.com>
31978         * config/rs6000/rs6000-c: Add support for built-in functions
31979         vector unsigned char vec_popcnt (vector signed char)
31980         vector unsigned char vec_popcnt (vector unsigned char)
31981         vector unsigned short vec_popcnt (vector signed short)
31982         vector unsigned short vec_popcnt (vector unsigned short)
31983         vector unsigned int vec_popcnt (vector signed int)
31984         vector unsigned int vec_popcnt (vector unsigned int)
31985         vector unsigned long long vec_popcnt (vector signed long long)
31986         vector unsigned long long vec_popcnt (vector unsigned long long)
31987         vector signed long long vec_slo (vector signed long long,
31988                                          vector signed char)
31989         vector signed long long vec_slo (vector signed long long,
31990                                          vector unsigned char)
31991         vector unsigned long long vec_slo (vector unsigned long long,
31992                                            vector signed char)
31993         vector unsigned long long vec_slo (vector unsigned long long,
31994                                            vector unsigned char)
31995         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
31996         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
31997         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
31998         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
31999         * doc/extend.texi: Update the built-in documentation file for the
32000         new built-in functions.
32002 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
32004         * attribs.h (sorted_attr_string): Move machine independent
32005         functions for target clone support from the i386 port to common
32006         code.  Rename ix86_function_versions to common_function_versions.
32007         Rename make_name to make_unique_name.
32008         (common_function_versions): Likewise.
32009         (make_unique_name): Likewise.
32010         (make_dispatcher_decl): Likewise.
32011         (is_function_default_version): Likewise.
32012         * attribs.c (attr_strcmp): Likewise.
32013         (sorted_attr_string): Likewise.
32014         (common_function_versions): Likewise.
32015         (make_unique_name): Likewise.
32016         (make_dispatcher_decl): Likewise.
32017         (is_function_default_version): Likewise.
32018         * config/i386/i386.c (attr_strcmp): Likewise.
32019         (sorted_attr_string): Likewise.
32020         (ix86_function_versions): Likewise.
32021         (make_name): Likewise.
32022         (make_dispatcher_decl): Likewise.
32023         (is_function_default_version): Likewise.
32024         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32026 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32028         PR target/80695
32029         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32030         Account for direct move costs for vec_construct of integer
32031         vectors.
32033 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32035         PR target/80706
32036         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32037         (UNSPEC_STX_ATOMIC): Ditto.
32038         (loaddi_via_sse): New insn.
32039         (storedi_via_sse): Ditto.
32040         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32041         Update corresponding peephole2 patterns.
32042         (atomic_storedi_fpu): Ditto.
32044 2017-05-11  Julia Koval  <julia.koval@intel.com>
32046         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32047         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32048         New intrinsics.
32049         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32050         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32051         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32053 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32055         * graphite-poly.c: Include dumpfile.h.
32057         * dumpfle.h (dump_function): Declare here ...
32058         * tree-dump.h (dump_function): ... not here.
32059         * dumpfile.c: #include tree-cfg.h.
32060         (dump_function): Move here from ...
32061         * tree-dump.c (dump_function): ... here.
32062         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32063         * graphite-poly.c: Don't include tree-dump.h.
32064         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32065         * print-tree.c: Likewise.
32066         * stor-layout.c: Likewise.
32067         * tree-nested.c: Likewise.
32069         * dumpfile.c (dump_start): Use TDF_FLAGS.
32070         (dump_enable_all): Fix TDF_KIND check thinko.
32072 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32074         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32075         array entries to represent two legal parameterizations of the
32076         overloaded __builtin_cmpb function, as represented by the
32077         P6_OV_BUILTIN_CMPB constant.
32078         (altivec_resolve_overloaded_builtin): Add special case handling
32079         for the __builtin_cmpb function, as represented by the
32080         P6_OV_BUILTIN_CMPB constant.
32081         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32082         (BU_P6_64BIT_2): New macro.
32083         (BU_P6_OVERLOAD_2): New macro
32084         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32085         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32086         (CMPB): Add overload support to represent both 32-bit and 64-bit
32087         compare-bytes function.
32088         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32089         support for TARGET_CMPB.
32090         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32091         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32092         documentation of the __builtin_cmpb overloaded built-in function.
32094 2017-05-11  Richard Biener  <rguenther@suse.de>
32096         PR tree-optimization/80705
32097         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32098         bases are not vectorizable.
32100 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32102         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32103         when counting register pressure.
32105 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32107         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32108         (iv_ca_more_deps): Renamed to ...
32109         (iv_ca_compare_deps): ... this.
32110         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32112 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32114         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32115         to ...
32116         (determine_group_iv_costs): ... here.
32117         (find_inv_vars_cb): Record inv var if it's not recorded before.
32119 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32121         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32122         (get_shiftadd_cost): Ditto.
32124 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32126         * tree-ssa-address.c: Include header file.
32127         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32128         address.
32129         (add_to_parts): Refactor.
32130         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32131         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32132         in new order.
32134 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32136         PR tree-optimization/53090
32137         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32138         COMP_IV_EXPR_2.
32139         (extract_cond_operands): Detect condition with IV on both sides
32140         and return COMP_IV_EXPR_2.
32141         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32142         (rewrite_use_compare): Simplify by removing call to function
32143         extract_cond_operands.
32145 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32147         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32148         (extract_cond_operands): Detect condition comparing against non-
32149         invariant bound and return appropriate enum value.
32150         (find_interesting_uses_cond): Update use of extract_cond_operands.
32151         Handle its return value accordingly.
32152         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32154 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32156         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32157         nonlinear iv_use computation in loop invariant sensitive way.
32159 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32161         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32162         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32164 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32166         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32167         (dump_cand): Support iv_cand.inv_exprs.
32168         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32169         for candidates.
32170         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32171         iv_cand.inv_exprs.
32173 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32175         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32176         from ...
32177         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32178         as local function.  Include necessary header files.
32179         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32181 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32183         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32185 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32187         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32188         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32189         RSHIFT_EXPR and BIT_NOT_EXPR.
32191 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32193         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32194         (adjust_setup_cost): New parameter supporting round up adjustment.
32195         (struct address_cost_data): Delete.
32196         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32197         (split_address_cost, ptr_difference_cost): Delete.
32198         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32199         (struct ainc_cost_data): New struct.
32200         (get_address_cost_ainc): New function.
32201         (get_address_cost, get_computation_cost): Reimplement.
32202         (determine_group_iv_cost_address): Record inv_expr for all uses of
32203         a group.
32204         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32205         (iv_ca_has_deps): Reimplemented to ...
32206         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32207         than OLD_CP.
32208         (iv_ca_extend): Call iv_ca_more_deps.
32210 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32212         * tree-ssa-address.c (struct mem_address): Move to header file.
32213         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32214         * tree-ssa-address.h (struct mem_address): Move from C file.
32215         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32217 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32219         * tree-affine.h (aff_combination_type): New interface.
32220         (aff_combination_zero_p): Remove static.
32221         (aff_combination_const_p): New interface.
32222         (aff_combination_singleton_var_p): New interfaces.
32224 2017-05-11  Richard Biener  <rguenther@suse.de>
32226         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32227         Skip unreachable blocks and destinations.
32228         (eliminate): Move stmt removal and fixup ...
32229         (fini_eliminate): ... here.  Skip inserted exprs.
32230         (pass_pre::execute): Move fini_pre after fini_eliminate.
32231         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32232         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32233         PRE to get rid of dead code that has invalid SSA form and
32234         split critical edges again.
32236 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32238         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32240 2017-05-11  Richard Biener  <rguenther@suse.de>
32242         * passes.c (execute_function_todo): Verify loops if they are
32243         said to be up-to-date.
32244         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32245         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32247 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32249         PR target/80090
32250         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32251         handle calling assemble_external ourself.
32253         PR target/79027
32254         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32255         modes with zero size.  Enhance comment.
32257 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32259         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32260         built-ins for vec_xl and vec_xst with short and char pointer
32261         arguments.
32263 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32265         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32266         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32267         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32268         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32269         (_mm_maskz_min_round_ss): New intrinsics.
32270         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32271         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32272         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32273         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32274         (__builtin_ia32_minss_mask_round): New builtins.
32275         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32276         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32277         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32278         Rename to ...
32279         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32280         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32281         Change to ...
32282         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32283         ... this.
32285 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32287         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32288         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32289         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32290         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32291         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32292         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32293         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32294         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32295         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32296         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32297         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32298         (__builtin_ia32_mulss_mask_round): New builtins.
32299         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32300         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32301         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32302         Rename to ...
32303         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32304         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32305         Change to ...
32306         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32307         ... this.
32309 2017-05-10  Julia Koval  <julia.koval@intel.com>
32311         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32312         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32313         (_mm256_setr_m128i): New intrinsics.
32315 2017-05-10  Julia Koval  <julia.koval@intel.com>
32317         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32318         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32319         (_mm_maskz_rcp14_ss): New intrinsics.
32320         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32321         (__builtin_ia32_rcp14ss_mask): New builtins.
32322         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32324 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32326         PR tree-optimization/51513
32327         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32328         (assert_unreachable_fallthru_edge_p): Use it.
32329         (group_case_labels_stmt): Likewise.
32330         * tree-cfg.h: Prototype it.
32331         * stmt.c: Include cfghooks.h and tree-cfg.h.
32332         (emit_case_dispatch_table) <gap_label>: New local variable.
32333         Use it to fill dispatch table gaps.
32334         Test for default_label before updating probabilities.
32335         (expand_case) <default_label>: Remove unneeded initialization.
32336         Test for unreachable default case statement and remove its edge.
32337         Set default_label accordingly.
32338         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32340 2017-05-10  Carl Love  <cel@us.ibm.com>
32342         * config/rs6000/rs6000-c: Add support for built-in functions
32343         vector signed char      vec_neg (vector signed char)
32344         vector signed short int vec_neg (vector short int)
32345         vector signed int       vec_neg (vector signed int)
32346         vector signed long long vec_neg (vector signed long long)
32347         vector float            vec_neg (vector float)
32348         vector double           vec_neg (vector double)
32349         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32350         overload.
32351         * config/rs6000/altivec.h: Add define for vec_neg
32352         * doc/extend.texi: Update the built-in documentation for the
32353         new built-in functions.
32355 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32357         PR tree-optimization/77644
32358         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32360 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32362         * dumpfile.h (TDI_lang_all): New.
32363         (TDF_KIND): New. Renumber others
32364         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32365         than bits.
32366         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32367         lang-all.
32368         (get_dump_file_name): Adjust suffix generation.
32369         (dump_enable_all): Use TDF_KIND.
32370         * doc/invoke.texi (-fdump-lang-all): Document.
32372         * dumpfile.h: Tabify.
32374 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32376         PR target/80671
32377         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32378         Move member access before delete.
32380 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32382         * tree-inline.c (expand_call_inline): Split block at stmt
32383         before the call.
32385 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32387         PR target/68163
32388         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32389         are now unused after splitting mov{sf,sd}_hardfloat.
32390         (f32_lr2): Likewise.
32391         (f32_lm): Likewise.
32392         (f32_lm2): Likewise.
32393         (f32_li): Likewise.
32394         (f32_li2): Likewise.
32395         (f32_lv): Likewise.
32396         (f32_sr): Likewise.
32397         (f32_sr2): Likewise.
32398         (f32_sm): Likewise.
32399         (f32_sm2): Likewise.
32400         (f32_si): Likewise.
32401         (f32_si2): Likewise.
32402         (f32_sv): Likewise.
32403         (f32_dm): Likewise.
32404         (f32_vsx): Likewise.
32405         (f32_av): Likewise.
32406         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32407         For movsf, order stores so the VSX stores occur before the GPR
32408         store which encourages the register allocator to use a traditional
32409         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32410         store comes before the VSX stores to allow the power6 to work.
32411         This is due to the power6 not having a 32-bit integer store
32412         instruction from a FPR.
32413         (movsf_hardfloat): Likewise.
32414         (movsd_hardfloat): Likewise.
32416 2017-05-09  Martin Sebor  <msebor@redhat.com>
32418         PR translation/80280
32419         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32420         added in r247778.
32422         PR translation/80280
32423         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32424         data member added in r247778.
32425         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32427 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32429         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32431         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32432         typedefs.
32434 2017-05-09  Marek Polacek  <polacek@redhat.com>
32436         * doc/invoke.texi: Fix typo.
32438 2017-05-09  Richard Biener  <rguenther@suse.de>
32440         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32441         (vrp_val_is_min): Likewise.
32442         (set_value_range_to_value): Likewise.
32443         (set_value_range_to_nonnegative): Likewise.
32444         (gimple_assign_nonzero_p): Likewise.
32445         (gimple_stmt_nonzero_p): Likewise.
32446         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32447         (adjust_range_with_scev): Adjust comments.
32448         (compare_range_with_value): Likewise.
32449         (extract_range_from_phi_node): Likewise.
32450         (test_for_singularity): Likewise.
32452 2017-05-09  Richard Biener  <rguenther@suse.de>
32454         * tree-vrp.c (get_single_symbol): Add assert that we don't
32455         get overflowed constants as invariant part.
32456         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32457         checks.  Use wi::cmp instead of recursing for integer constants.
32458         (compare_values): Just ignore whether we assumed undefined
32459         overflow instead of failing the compare.
32460         (extract_range_for_var_from_comparison_expr): Add comment before the
32461         TREE_NO_WARNING sets.
32462         (test_for_singularity): Likewise.
32463         (extract_range_from_comparison): Do not disable optimization
32464         when we assumed undefined overflow.
32465         (extract_range_basic): Remove init of unused var.
32467 2017-05-09  Richard Biener  <rguenther@suse.de>
32469         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32470         (extract_range_from_multiplicative_op_1): Adjust.
32471         (extract_range_from_binary_expr_1): Use int_const_binop.
32473 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32475         PR target/80101
32476         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32477         rs6000_store_data_bypass_p in seven define_bypass directives and
32478         in several comments.
32479         * config/rs6000/rs6000-protos.h: Add prototype for
32480         rs6000_store_data_bypass_p function.
32481         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32482         function implements slightly different (rs6000-specific) semantics
32483         than store_data_bypass_p, returning false rather than aborting
32484         with assertion error when arguments do not satisfy the
32485         requirements of store data bypass.
32486         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32487         rs6000_store_data_bypass_p.
32489 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32491         * config/xtensa/xtensa-protos.h
32492         (xtensa_initial_elimination_offset): New declaration.
32493         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32494         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32495         macro definition, add case for FRAME_POINTER_REGNUM when
32496         FRAME_GROWS_DOWNWARD.
32497         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32498         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32499         xtensa_initial_elimination_offset.
32501 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32503         * doc/invoke.texi: Alphabetize -fdump options.
32505 2017-05-08  Martin Sebor  <msebor@redhat.com>
32507         PR translation/80280
32508         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32510 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32512         * target.def (compute_frame_layout): New optional target hook.
32513         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32514         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32515         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32516         target hook.
32517         * reload1.c (verify_initial_elim_offsets): Likewise.
32518         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32519         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32520         (arm_get_frame_offsets): Split up into this ...
32521         (arm_compute_frame_layout): ... and this function.
32523 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32525         * config/aarch64/constraints.md (Usa): New constraint.
32526         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32528 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32530         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32531         with_multilib_list after it has been checked.
32533 2017-05-08  Richard Biener  <rguenther@suse.de>
32535         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32536         (bitmap_set_subtract_values): Likewise.
32538 2017-05-08  Richard Biener  <rguenther@suse.de>
32540         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32541         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32542         argument.
32543         (gimple_stmt_nonzero_warnv_p): Rename to ...
32544         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32545         argument.
32546         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32547         (extract_range_basic): Adjust, do not disable propagation on
32548         strict overflow sensitive simplification.
32549         (vrp_visit_cond_stmt): Likewise.
32551 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32553         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32554         body size unconditionally.
32556 2017-05-07  Jeff Law  <law@redhat.com>
32558         Revert:
32559         2017-05-06  Jeff Law  <law@redhat.com>
32560         PR tree-optimization/78496
32561         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32562         code.
32564         PR tree-optimization/78496
32565         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32566         (simplify_stmt_using_ranges): Call it.
32567         (vrp_dom_walker::before_dom_children): Extract equivalences
32568         from an ASSERT_EXPR with an equality comparison against a
32569         constant.
32571 2017-05-06  Jeff Law  <law@redhat.com>
32573         PR tree-optimization/78496
32574         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32575         code.
32577         PR tree-optimization/78496
32578         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32579         (simplify_stmt_using_ranges): Call it.
32580         (vrp_dom_walker::before_dom_children): Extract equivalences
32581         from an ASSERT_EXPR with an equality comparison against a
32582         constant.
32584 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32586         * lra-constraints.c (lra_copy_reg_equiv): New function.
32587         (split_reg): Use it to copy equivalence information from the
32588         original register to the spill register.
32590 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32592         PR rtl-optimization/75964
32593         * simplify-rtx.c (simplify_const_relational_operation): Remove
32594         invalid handling of comparisons of integer ABS.
32596 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32598         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32599         initialize to zero.
32600         (init_regs): Remove declaration.
32601         (function_arg_advance_32): Initialize error_p as boolean variable.
32603 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32605         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32606         lines.  Use for (;;).
32608 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32610         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32611         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32612         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32613         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32614         VF=2 that require versioning.
32616 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32618         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32619         int.
32621 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32623         * diagnostic.h (diagnostic_override_option_index): Convert from
32624         macro to inline function.
32626 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32628         * diagnostic.c (last_module_changed_p): New function.
32629         (set_last_module): New function.
32630         (diagnostic_report_current_module): Convert macro usage to
32631         the above functions.
32632         * diagnostic.h (diagnostic_context::last_module): Strengthen
32633         from const line_map * to const line_map_ordinary *.
32634         (diagnostic_last_module_changed): Delete macro.
32635         (diagnostic_set_last_module): Delete macro.
32637 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32639         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32640         with diagnostic_report_diagnostic.
32641         (diagnostic_n_impl_richloc): Likewise.
32642         * diagnostic.h (report_diagnostic): Delete macro.
32643         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32644         with diagnostic_report_diagnostic.
32645         * substring-locations.c (format_warning_va): Likewise.
32647 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32649         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32650         save/restor of format_spec.  Move option-printing code to...
32651         (print_option_information): ...this new function, and
32652         reimplement by simply printing to the pretty_printer,
32653         rather than appending to the format string.
32655 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32657         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32658         handling logic into...
32659         (update_effective_level_from_pragmas): ...this new function.
32661 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32663         * config/riscv/riscv.opt (mstrict-align): New option.
32664         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32665         (SLOW_UNALIGNED_ACCESS): Define.
32666         (riscv_slow_unaligned_access): Declare.
32667         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32668         field.
32669         (riscv_slow_unaligned_access): New variable.
32670         (rocket_tune_info): Set slow_unaligned_access to true.
32671         (optimize_size_tune_info): Set slow_unaligned_access to false.
32672         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32673         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32674         (riscv_option_override): Set riscv_slow_unaligned_access.
32675         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32677 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32679         * config/riscv/riscv.md: Unify indentation.
32681 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32683         PR target/79038
32684         PR target/79202
32685         PR target/79203
32686         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32687         UNSIGNED_FIX.
32688         (extendsi<mode>2): Add support for doing sign extension via
32689         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32690         don't have ISA 3.0 instructions.
32691         (extendsi<mode>2 splitter): Likewise.
32692         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32693         generate the normal insns since SImode can now go in vector
32694         registers.  Disallow the special UNSPECs needed for previous
32695         machines to hide SImode being used.  Add new insns
32696         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32697         (fix_trunc<mode>si2_stfiwx): Likewise.
32698         (fix_trunc<mode>si2_internal): Likewise.
32699         (fixuns_trunc<mode>si2): Likewise.
32700         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32701         (fctiw<u>z_<mode>_smallint): Likewise.
32702         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32703         of floating point to 32-bit integer from doing a direct move to
32704         the GPR registers to do a store.
32705         (fctiwz_<mode>): Break long line.
32707 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32709         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32710         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32711         (addr_list, addr_offset_valid_p): New.
32712         (split_address_groups): Check offset validity with above function.
32713         (gt-tree-ssa-loop-ivopts.h): Include header file.
32715 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32717         * config.gcc (arm*-*-*): Add missing 'fi'.
32719 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32721         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32722         included in -fopt-info does not matter.
32723         * doc/optinfo.texi (-fopt-info): Fix description of default
32724         behavour. Explicitly say order of options included in -fopt-info
32725         does not matter.
32727 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32729         * config.gcc: Allow combinations of aprofile and rmprofile values for
32730         --with-multilib-list.
32731         * config/arm/t-multilib: New file.
32732         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32733         variables.  Remove setting of ISA and floating-point ABI in
32734         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32735         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32736         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32737         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32738         CPU options.
32739         * config/arm/t-rmprofile: Likewise except for the matches changes.
32740         * doc/install.texi (--with-multilib-list): Document the combination of
32741         aprofile and rmprofile values and warn about pitfalls in doing that.
32743 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32745         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32746         (movdi_aarch64): Likewise.
32748 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32750         PR tree-optimization/80632
32751         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32752         field.
32753         (build_arrays): Initialize it for virtual phis.
32754         (fix_phi_nodes): Use it for virtual phis.
32756         PR tree-optimization/80558
32757         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32758         [x, y] op z into [x op, y op z] for op & or | if conditions
32759         are met.
32761 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32762             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32764         PR target/71607
32765         * config/arm/arm.md (use_literal_pool): Remove.
32766         (64-bit immediate split): No longer takes cost into consideration
32767         if arm_disable_literal_pool is enabled.
32768         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32769         used when arm_disable_literal_pool is enabled.
32770         (arm_max_const_double_inline_cost): Remove use of
32771         arm_disable_literal_pool.
32772         (push_minipool_fix): Add assert.
32773         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32774         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32775         (no_literal_pool_sf_immediate): New.
32777 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32779         PR tree-optimization/80613
32780         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32781         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32783 2017-05-05  Richard Biener  <rguenther@suse.de>
32785         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32787 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32789         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32790         of this flag from insn conditions due to removal from r247495.
32792 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32794         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32795         New function.
32796         (arm_early_store_addr_dep_ptr): Likewise.
32797         * config/arm/aarch-common-protos.h
32798         (arm_early_load_addr_dep_ptr): Add prototype.
32799         (arm_early_store_addr_dep_ptr): Likewise.
32800         * config/arm/cortex-a53.md: Add new bypasses.
32802 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32804         * tree.c (next_type_uid): Change type to unsigned.
32805         (type_hash_canon): Decrement back next_type_uid if
32806         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32807         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32808         if possible.
32810 2017-05-04  Martin Sebor  <msebor@redhat.com>
32812         * builtins.c: Fix a trivial typo in a comment.
32814         PR middle-end/79234
32815         * builtins.c (check_sizes): Adjust to handle reading past the end.
32816         Avoid printing excessive upper bound of ranges.  Use %E to print
32817         tree nodes instead of converting them to %wu.
32818         (expand_builtin_memchr): New function.
32819         (compute_dest_size): Rename...
32820         (compute_objsize): ...to this.
32821         (expand_builtin_memcpy): Adjust.
32822         (expand_builtin_mempcpy): Adjust.
32823         (expand_builtin_strcat): Adjust.
32824         (expand_builtin_strcpy): Adjust.
32825         (check_strncat_sizes): Adjust.
32826         (expand_builtin_strncat): Adjust.
32827         (expand_builtin_strncpy): Adjust and simplify.
32828         (expand_builtin_memset): Adjust.
32829         (expand_builtin_bzero): Adjust.
32830         (expand_builtin_memcmp): Adjust.
32831         (expand_builtin): Handle memcmp.
32832         (maybe_emit_chk_warning): Check strncat just once.
32834 2017-05-04  Martin Sebor  <msebor@redhat.com>
32836         PR preprocessor/79214
32837         PR middle-end/79222
32838         PR middle-end/79223
32839         * builtins.c (check_sizes): Add inlining context and issue
32840         warnings even when -Wno-system-headers is set.
32841         (check_strncat_sizes): Same.
32842         (expand_builtin_strncat): Same.
32843         (expand_builtin_memmove): New function.
32844         (expand_builtin_stpncpy): Same.
32845         (expand_builtin): Handle memmove and stpncpy.
32847 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32849         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32850         which is not used any more.
32852 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32854         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32856 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32858         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32859         (cortexa53_tunings): Likewise.
32860         (cortexa57_tunings): Likewise.
32861         (cortexa72_tunings): Likewise.
32862         (cortexa73_tunings): Likewise.
32864 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32866         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32867         Set loop alignment to 8.
32869 2017-05-04  Martin Sebor  <msebor@redhat.com>
32871         PR translation/80280
32872         * builtins.c (expand_builtin_object_size): Add missing quoting to
32873         %D and like directives.
32874         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32875         (hsa_type_for_tree_type): Same.
32876         (verify_function_arguments): Same.
32877         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32878         * varasm.c (get_section): Same.
32879         (mark_weak): Same.
32881 2017-05-04  Martin Sebor  <msebor@redhat.com>
32883         PR translation/80280
32884         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32886 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32888         * config/aarch64/aarch64.c (generic_addrcost_table):
32889         Change HI/TI mode setting.
32891 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32893         PR tree-optimization/80622
32894         * tree-sra.c (comes_initialized_p): New function.
32895         (build_accesses_from_assign): Only set write lazily when
32896         comes_initialized_p is false.
32897         (analyze_access_subtree): Use comes_initialized_p.
32898         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32899         instead of testing for PARM_DECL.
32901 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32903         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32904         constraint on operand 0 to allow more general addressing modes.
32905         Adjust output template.
32906         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32907         New function.
32908         * config/aarch64/aarch64-protos.h
32909         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32910         * config/aarch64/constraints.md (Dp): New address constraint.
32911         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32912         predicate.
32914 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32916         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32917         update use of estimate_ipcp_clone_size_and_time.
32918         (estimate_local_effects): Update use of
32919         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32920         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32921         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32922         Return nonspecialized time.
32924 2017-05-04  Richard Biener  <rguenther@suse.de>
32926         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32927         for the last VUSE which def dominates the PHI.  Directly call
32928         maybe_skip_until.
32929         (get_continuation_for_phi_1): Remove.
32931 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
32933         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32934         to explain the use of truncating division.  Cap the number of
32935         iterations to the maximum given by nb_iterations_upper_bound,
32936         if defined.
32938 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32940         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
32941         * configure: Regenerate.
32942         * config.in: Regenerate.
32943         * config/i386/driver-mingw32.c: new file.
32944         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
32945         * config.host: Link driver-mingw32.o on MinGW host.
32946         * doc/install.texi: Document new --enable-mingw-wildcard configure
32947         option.
32949 2017-05-04  Marek Polacek  <polacek@redhat.com>
32951         PR tree-optimization/80612
32952         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
32954 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32955             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
32957         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
32958         (movt splitter): Likewise.
32959         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
32960         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
32961         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
32962         block for Thumb-1 with MOVT.
32963         (thumb2_legitimate_address_p): Move code block ...
32964         (can_avoid_literal_pool_for_label_p): ... into this new function.
32965         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
32966         literal pool.
32967         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
32968         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
32969         "M-profile targets with the MOVT instruction".
32971 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32973         * config/arm/arm-builtins.c (arm_init_builtins): Rename
32974         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
32975         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
32977 2017-05-04  Martin Liska  <mliska@suse.cz>
32979         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
32980         variable cond_code.
32982 2017-05-04  Richard Biener  <rguenther@suse.de>
32984         * tree.c (array_at_struct_end_p): Handle arrays at struct
32985         end with flexarrays more conservatively.  Refactor and treat
32986         arrays of arrays or aggregates more strict.  Fix
32987         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
32988         * tree.c (array_at_struct_end_p): Adjust prototype.
32989         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
32990         * gimple-fold.c (get_range_strlen): Likewise.
32991         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
32993 2017-05-04  Richard Biener  <rguenther@suse.de>
32995         PR tree-optimization/31130
32996         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
32997         false.
32998         (supports_overflow_infinity): Likewise.
32999         (is_negative_overflow_infinity): Likewise.
33000         (is_positive_overflow_infinity): Likewise.
33001         (is_overflow_infinity): Likewise.
33002         (stmt_overflow_infinity): Likewise.
33003         (overflow_infinity_range_p): Likewise.
33004         (usable_range_p): Remove as always returning true.
33005         (make_overflow_infinity): Remove.
33006         (negative_overflow_infinity): Likewise.
33007         (positive_overflow_infinity): Likewise.
33008         (avoid_overflow_infinity): Likewise.
33009         (set_value_range): Adjust accordingly.
33010         (set_value_range_to_nonnegative): Likewise, remove now unused
33011         overflow_infinity arg.
33012         (vrp_operand_equal_p): Adjust.
33013         (update_value_range): Likewise.
33014         (range_int_cst_singleton_p): Likewise.
33015         (operand_less_p): Likewise.
33016         (compare_values_warnv): Likewise.
33017         (extract_range_for_var_from_comparison_expr): Likewise.
33018         (vrp_int_const_binop): Likewise.
33019         (zero_nonzero_bits_from_vr): Likewise.
33020         (extract_range_from_multiplicative_op_1): Likewise.
33021         (extract_range_from_binary_expr_1): Likewise.
33022         (extract_range_from_unary_expr): Likewise.
33023         (extract_range_from_comparison): Likewise.
33024         (extract_range_basic): Likewise.
33025         (adjust_range_with_scev): Likewise.
33026         (compare_ranges): Likewise.
33027         (compare_range_with_value): Likewise.
33028         (dump_value_range): Likewise.
33029         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33030         never used.
33031         (simplify_cond_using_ranges): Adjust.
33033 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33035         * brig-builtins.def: Added a builtin for class_f64.
33036         * builtin-types.def: Added a builtin type needed by class_f64.
33038 2017-05-03  Jason Merrill  <jason@redhat.com>
33040         * timevar.def: Add TV_CONSTEXPR.
33042 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33044         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33046 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33048         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33049         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33050         * ipa-cp.c (ipcp_propagate_stage): Do not call
33051         ipa_update_after_lto_read.
33052         * ipa-inline.c (ipa_inline): Likewise.
33054 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33056         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33057         tag.  Added a default constructor and a destructor.
33058         (ipa_edge_args_sum_t): New class;
33059         (ipa_edge_args_sum): Declare.
33060         (ipa_edge_args_vector): Remove declaration.
33061         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33062         (ipa_free_edge_args_substructures): Remove declaration.
33063         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33064         (ipa_edge_args_info_available_for_edge_p): Likewise.
33065         * ipa-prop.c (ipa_edge_args_vector): Removed.
33066         (edge_removal_hook_holder): Likewise.
33067         (edge_duplication_hook_holder): Likewise.
33068         (ipa_edge_args_sum): New variable.
33069         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33070         ipa_edge_args_vector.
33071         (ipa_free_edge_args_substructures): Likewise.
33072         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33073         ipa_edge_args_vector.
33074         (ipa_edge_removal_hook): Turned into method
33075         ipa_edge_args_sum_t::remove.
33076         (ipa_edge_duplication_hook): Turned into method
33077         ipa_edge_args_sum_t::duplicate.
33078         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33079         registering edge hooks.
33080         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33081         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33082         ipa_edge_args_sum instead of ipa_edge_args_vector.
33083         * ipa-profile.c (ipa_profile): Likewise.
33085 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33087         * symbol-summary.h (function_summary): New method exists.
33088         (function_summary::symtab_removal): Deallocate through release.
33089         (call_summary): New class.
33090         (gt_ggc_mx): New overload.
33091         (gt_pch_nx): Likewise.
33092         (gt_pch_nx): Likewise.
33094 2017-05-03  Jeff Law  <law@redhat.com>
33096         PR tree-optimization/78496
33097         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33098         from simplify_cond_using_ranges.  Split off code to walk
33099         backwards through casts into ...
33100         (simplify_cond_using_ranges_2): New function.
33101         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33102         (execute_vrp): After identifying jump threads, call
33103         simplify_cond_using_ranges_2.
33105 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33107         PR bootstrap/80609
33108         * ipa-inline.h (inline_summary): Add ctor.
33109         (create_ggc): Do not use ggc_cleared_alloc.
33111 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33112             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33114         * gcc.c (handle_braces): Support escaping in switch matching
33115         text.
33116         * doc/invoke.texi (Spec Files): Document it.
33117         Remove superfluous @code markup in items.
33119 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33121         * diagnostic-show-locus.c (struct column_range): New struct.
33122         (get_affected_columns): New function.
33123         (get_printed_columns): New function.
33124         (struct correction): New struct.
33125         (correction::ensure_capacity): New function.
33126         (correction::ensure_terminated): New function.
33127         (struct line_corrections): New struct.
33128         (line_corrections::~line_corrections): New dtor.
33129         (line_corrections::add_hint): New function.
33130         (layout::print_trailing_fixits): Reimplement in terms of the new
33131         classes.
33132         (selftest::test_overlapped_fixit_printing): New function.
33133         (selftest::diagnostic_show_locus_c_tests): Call it.
33135 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33137         Canonicalize canonical type hashing
33138         * tree.h (type_hash_canon_hash): Declare.
33139         * tree.c (type_hash_list, attribute_hash_list): Move into
33140         type_hash_canon_hash.
33141         (build_type_attribute_qual_variant): Break out hash code calc into
33142         type_hash_canon_hash.
33143         (type_hash_canon_hash): New.  Generic type hash computation.
33144         (build_range_type_1, build_array_type_1, build_function_type,
33145         build_method_type_directly, build_offset_type, build_complex_type,
33146         make_vector_type): Call it.
33148 2017-05-03  Richard Biener  <rguenther@suse.de>
33150         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33151         When all DRs have unknown misaligned do not always peel
33152         when there is a store but apply the same costing model as if
33153         there were only loads.
33155 2017-05-03  Richard Biener  <rguenther@suse.de>
33157         Revert
33158         PR tree-optimization/80492
33159         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33160         compare_base_decls returning dont-know properly.
33162 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33164         * config/arm/iterators.md (CCSI): New mode iterator.
33165         (arch): New mode attribute.
33166         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33167         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33168         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33169         code iterator for success result mode.
33170         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33171         the corresponding new insn generators.
33173 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33175         Revert r247509
33176         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33177         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33179 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33181         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33182         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33183         (DDR_A): Wrap DDR argument in brackets.
33184         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33185         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33186         (DDR_REVERSED_P): Likewise.
33188 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33190         PR tree-optimization/79472
33191         * tree-switch-conversion.c (struct switch_conv_info): Add
33192         contiguous_range and default_case_nonstandard fields.
33193         (collect_switch_conv_info): Compute contiguous_range and
33194         default_case_nonstandard fields, don't clear final_bb if
33195         contiguous_range and only the default case doesn't have the required
33196         structure.
33197         (check_all_empty_except_final): Set default_case_nonstandard instead
33198         of failing if contiguous_range and the default case doesn't have empty
33199         block.
33200         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33201         and only the default case doesn't have the required constants.  Skip
33202         virtual phis.
33203         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33204         if default_case_nonstandard.
33205         (build_constructors): Build constant 1 just once.  Assert that default
33206         values aren't inserted in between cases if contiguous_range.  Skip
33207         virtual phis.
33208         (build_arrays): Skip virtual phis.
33209         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33210         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33211         Handle virtual phis.
33212         (gen_inbound_check): Handle default_case_nonstandard case.
33213         (process_switch): Adjust check_final_bb caller.  Call
33214         gather_default_values with the first non-default case instead of
33215         default case if default_case_nonstandard.
33217 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33219         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33220         check.  Fix formatting.
33222 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33224         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33225         errors when comparing specialized and unspecialized times.
33227 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33229         * diagnostic-show-locus.c
33230         (layout::should_print_annotation_line_p): Make private.
33231         (layout::print_annotation_line): Make private.
33232         (layout::annotation_line_showed_range_p): Make private.
33233         (layout::show_ruler): Make private.
33234         (layout::print_source_line): Make private.  Pass in line and
33235         line_width, rather than calling location_get_source_line.  Drop
33236         returned value.
33237         (layout::print_leading_fixits): New method.
33238         (layout::print_any_fixits): Rename to...
33239         (layout::print_trailing_fixits): ...this, and make private.
33240         Don't print newline fixits.
33241         (diagnostic_show_locus): Move logic for printing one row into...
33242         (layout::print_line): ...this new function.  Move the
33243         location_get_source_line call and error-handling from
33244         print_source_line to here.  Call print_leading_fixits, and rename
33245         print_any_fixits to print_trailing_fixits.
33246         (selftest::test_fixit_insert_containing_newline): Update now that
33247         newlines are partially supported.
33248         (selftest::test_fixit_insert_containing_newline_2): New test.
33249         (selftest::test_fixit_replace_containing_newline): Update comments.
33250         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33251         * edit-context.c (class added_line): New class.
33252         (class edited_line): Describe newline handling in comment.
33253         (edited_line::actually_edited_p): New method.
33254         (edited_line::print_content): Delete redundant decl.
33255         (edited_line::m_predecessors): New field.
33256         (edited_file::print_content): Call edited_line::print_content.
33257         (edited_file::print_diff): Update to support newlines.
33258         (edited_file::print_diff_hunk): Likewise.
33259         (edited_file::print_run_of_changed_lines): New function.
33260         (edited_file::print_diff_line): Convert to...
33261         (print_diff_line): ...this.
33262         (edited_file::get_effective_line_count): New function.
33263         (edited_line::edited_line): Initialize new field m_predecessors.
33264         (edited_line::~edited_line): Clean up m_predecessors.
33265         (edited_line::apply_fixit): Handle newlines.
33266         (edited_line::get_effective_line_count): New function.
33267         (edited_line::print_content): New function.
33268         (edited_line::print_diff_lines): New function.
33269         (selftest::test_applying_fixits_insert_containing_newline): New
33270         test.
33271         (selftest::test_applying_fixits_replace_containing_newline): New
33272         test.
33273         (selftest::insert_line): New function.
33274         (selftest::test_applying_fixits_multiple_lines): Add example of
33275         inserting a line.
33276         (selftest::edit_context_c_tests): Call the new tests.
33278 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33280         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33281         parameter cand.  Update dump information.
33282         (get_computation_cost): Update uses.
33284 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33286         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33287         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33288         (get_computation_at, rewrite_use_address): Update use of
33289         get_computation_aff.
33291 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33293         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33294         (get_computation): Delete.
33295         (get_computation_cost): Implement like get_computation_cost_at.
33296         Use get_computation_at.
33297         (get_computation_cost_at): Delete.
33298         (rewrite_use_nonlinear_expr): Use get_computation_at.
33299         (rewrite_use_compare, remove_unused_ivs): Ditto.
33301 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33303         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33305 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33307         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33308         (ivopts_global_cost_for_size): Rename parameter and update uses.
33309         (iv_ca_recount_cost): Update uses.
33310         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33311         candidates seperately in n_invs and n_cands.
33312         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33314 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33316         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33317         (find_inv_vars_cb): New.
33318         (find_depends): Renamed to ...
33319         (find_inv_vars): ... this.
33320         (add_candidate_1, force_var_cost): Call find_inv_vars.
33321         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33323 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33325         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33326         inv_vars.  Add inv_exprs.
33327         (struct iv_cand): Rename depends_on to inv_vars.
33328         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33329         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33330         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33331         (dump_cand): Dump inv_vars.
33332         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33333         (record_invariant, find_depends, add_candidate_1): Ditto.
33334         (set_group_iv_cost, force_var_cost): Ditto.
33335         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33336         (get_computation_cost_at, get_computation_cost): Ditto.
33337         (determine_group_iv_cost_generic): Ditto.
33338         (determine_group_iv_cost_address): Ditto.
33339         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33340         (determine_group_iv_costs): Ditto.
33341         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33342         (iv_ca_set_remove_invariants): Renamed to ...
33343         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33344         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33345         (iv_ca_set_add_invariants):  Renamed to ...
33346         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33347         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33348         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33349         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33350         (create_new_ivs): Remove useless dump.
33352 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33354         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33355         iv_cand code.
33356         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33357         (iv_ca_set_no_cp, create_new_iv): Ditto.
33359 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33361         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33363 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33365         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33366         function tree_check2.
33368 2017-05-02  Martin Liska  <mliska@suse.cz>
33370         * doc/gcov.texi: Add missing preposition.
33371         * gcov.c (function_info::function_info): Properly fill up
33372         all member variables.
33374 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33376         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33378 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33380         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33382 2017-05-02  Martin Liska  <mliska@suse.cz>
33384         PR lto/77954.
33385         * lto-streamer-in.c (lto_read_tree_1): Remove
33386         LTO_STREAMER_DEBUG.
33387         * lto-streamer.c (struct tree_hash_entry): Likewise.
33388         (struct tree_entry_hasher): Likewise.
33389         (tree_entry_hasher::hash): Likewise.
33390         (tree_entry_hasher::equal): Likewise.
33391         (lto_streamer_init): Likewise.
33392         (lto_orig_address_map): Likewise.
33393         (lto_orig_address_get): Likewise.
33394         (lto_orig_address_remove): Likewise.
33395         * lto-streamer.h: Likewise.
33396         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33397         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33399 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33401         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33402         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33403         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33404         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33405         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33406         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33407         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33408         (mm_maskz_sub_ss): New intrinsics.
33409         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33410         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33411         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33412         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33413         (__builtin_ia32_subss_mask_round): New builtins.
33414         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33415         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33416         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33417         Renamed to ...
33418         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33419         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33420         Changed to ...
33421         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33422         ... this.
33424 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33426         PR tree-optimization/78687
33427         * tree-sra.c (access): New field parent.
33428         (process_subtree_disqualification): New function.
33429         (disqualify_candidate): Call it.
33430         (build_accesses_from_assign): Reset write flag if creating an
33431         assighnment link.
33432         (build_access_subtree): Fill in parent field and also prpagate
33433         down grp_write flag.
33434         (create_artificial_child_access): New parameter set_grp_write, set
33435         grp_write to its value.
33436         (propagate_subaccesses_across_link): Also propagate grp_write flag
33437         values.
33438         (propagate_all_subaccesses): Push the closest parent back to work
33439         queue if add_access_to_work_queue returned true.
33441 2017-05-02  Richard Biener  <rguenther@suse.de>
33443         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33444         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33445         -fstrict-overflow documentation.
33446         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33447         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33448         flag_strict_overflow.
33449         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33450         * lto-opts.c (lto_write_options): Do not stream it.
33451         * lto-wrapper.c (merge_and_complain): Do not handle it.
33452         * opts.c (default_options_table): Do not set -fstrict-overflow.
33453         (finish_options): Likewise do not clear it when sanitizing.
33454         * simplify-rtx.c (simplify_const_relational_operation): Do not
33455         test flag_strict_overflow.
33457 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33459         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33460         using enabled attribute.
33461         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33462         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33463         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33464         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33465         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33466         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33467         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33468         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33469         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33470         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33472 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33474         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33476 2017-05-02  Richard Biener  <rguenther@suse.de>
33478         PR tree-optimization/80591
33479         Revert
33480         2017-04-10  Richard Biener  <rguenther@suse.de>
33482         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33483         asm inputs.
33485 2017-05-02  Richard Biener  <rguenther@suse.de>
33487         PR tree-optimization/80549
33488         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33489         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33490         headers if they do not have a preheader.
33492 2017-05-02  Martin Liska  <mliska@suse.cz>
33494         PR other/80589
33495         * common.opt: Fix typo.
33496         * doc/invoke.texi: Likewise.
33498 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33500         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33501         swapping, add (x,x,m,x,n) alternative.
33503 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33505         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33506         unnecessary unadjusted_alignment check.
33508 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33510         PR c++/80038
33511         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33512         operations here.
33513         * gimplify.c (gimplify_cilk_detach): New function.
33514         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33515         * tree-core.h: Document EXPR_CILK_SPAWN.
33516         * tree.h (EXPR_CILK_SPAWN): Define.
33518 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33520         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33521         to use new fixit_hint representation, using the "replace" logic.
33522         (get_line_span_for_fixit_hint): Likewise.
33523         (layout::print_any_fixits): Likewise.
33524         (selftest::test_one_liner_many_fixits): Rename to...
33525         (selftest::test_one_liner_many_fixits_1): ...this, and update
33526         comment and expected output to reflect that the multiple fix-it
33527         hints are now consolidated into one insertion.
33528         (selftest::test_one_liner_many_fixits_2): New test.
33529         (selftest::test_diagnostic_show_locus_one_liner): Update for
33530         above.
33531         (selftest::test_fixit_consolidation): Update for fix-it API
33532         change.
33533         * diagnostic.c (print_parseable_fixits): Likewise.
33534         * edit-context.c (edited_line::m_line_events): Convert from
33535         auto_vec <line_event *> to auto_vec <line_event>.
33536         (class line_event): Convert from abstract base class to a concrete
33537         class, taking over the role of replace_event.
33538         (class insert_event): Delete.
33539         (class replace_event): Rename to class line_event.  Convert to
33540         half-open range.
33541         (edit_context::add_fixits): Reimplement.
33542         (edit_context::apply_insert): Delete.
33543         (edit_context::apply_replace): Rename to...
33544         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33545         (edited_file::apply_insert): Delete.
33546         (edited_file::apply_replace): Rename to...
33547         (edited_file::apply_fixit): ...this.
33548         (edited_line::~edited_line): Drop deletion of events.
33549         (edited_line::apply_insert): Delete.
33550         (edited_line::apply_replace): Rename to...
33551         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33552         Update for change to type of m_line_events.
33553         * edit-context.h (edit_context::apply_insert): Delete.
33554         (edit_context::apply_replace): Rename to...
33555         (edit_context::apply_fixit): ...this.
33557 2017-05-01  Martin Sebor  <msebor@redhat.com>
33559         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33560         known.
33562 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33564         PR target/68491
33565         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33566         __get_cpuid_max returns 0.
33567         (__get_cpuid_count): Ditto.
33569 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33571         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33572         replacement expression is another instance of one of its arguments.
33574 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33576         PR target/79430
33577         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33578         check for stack push/pop autoinc.
33579         * config/i386/i386.c (ix86_agi_dependent): Return false
33580         if the only reason why modified_in_p returned true is that
33581         addr is SP based and set_insn is a push or pop.
33583 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33585         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33586         overflow check.
33588 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33590         PR ipa/79224
33591         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33592         (account_size_time): Use two predicates - exec_pred and
33593         nonconst_pred_ptr.
33594         (evaluate_conditions_for_known_args): Compute both clause and
33595         nonspec_clause.
33596         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33597         (inline_summary_t::duplicate): Update.
33598         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33599         separately.
33600         (compute_inline_parameters): Likewise.
33601         (estimate_edge_size_and_time): Update caluclation of time.
33602         (estimate_node_size_and_time): Compute both time and nonspecialized
33603         time.
33604         (estimate_ipcp_clone_size_and_time): Update.
33605         (inline_merge_summary): Update.
33606         (do_estimate_edge_time): Update.
33607         (do_estimate_edge_size): Update.
33608         (do_estimate_edge_hints): Update.
33609         (inline_read_section, inline_write_summary): Stream both new predicates.
33610         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33611         as argument.
33612         (compute_inlined_call_time): Cleanup.
33613         (big_speedup_p): Update.
33614         (edge_badness): Update.
33615         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33616         (size_time_entry): Replace predicate by exec_predicate and
33617         nonconst_predicate.
33618         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33619         (estimate_edge_time): Return also nonspec_time.
33620         (reset_edge_growth_cache): Update.
33622 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33624         PR rtl-optimization/80491
33625         * ifcvt.c (noce_process_if_block): When looking for x setter
33626         with missing else_bb, don't check only the insn right before
33627         cond_earliest, but look for the last insn that x is modified in
33628         within the same bb.
33630         PR rtl-optimization/80491
33631         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33633 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33635         PR tree-optimization/80487
33636         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33638 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33640         PR tree-optimization/79697
33641         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33642         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33643         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33644         BUILT_IN_STRNDUP.
33645         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33646         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33648 2017-04-28  Martin Sebor  <msebor@redhat.com>
33650         PR tree-optimization/80523
33651         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33652         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33653         functions.
33654         (maybe_warn, format_directive, parse_directive): Use new functions.
33655         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33657 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33659         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33661 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33663         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33664         target_header_dir): Set correctly.
33665         * configure: Regenerated.
33666         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33667         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33668         instead of SYSTEM_HEADER_DIR.
33670 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33672         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33673         (estimate_local_effects): Likewise.
33674         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33675         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33676         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33677         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33678         do_estimate_edge_time, estimate_edge_time): Likewise.
33679         * ipa-inline-analysis.c (estimate_node_size_and_time,
33680         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33681         (estimate_time_after_inlining): Remove.
33683 2017-04-28  Martin Liska  <mliska@suse.cz>
33685         * doc/gcov.texi: Enhance documentation of gcov.
33687 2017-04-28  Martin Liska  <mliska@suse.cz>
33689         * doc/gcov.texi: Sort options in alphabetic order.
33690         * doc/gcov-dump.texi: Likewise.
33691         * doc/gcov-tool.texi: Likewise.
33692         * gcov.c (print_usage): Likewise.
33693         * gcov-dump.c (print_usage): Likewise.
33694         * gcov-tool.c (print_merge_usage_message): Likewise.
33695         (print_rewrite_usage_message): Likewise.
33696         (print_overlap_usage_message): Likewise.
33698 2017-04-28  Martin Liska  <mliska@suse.cz>
33700         PR gcov-profile/53915
33701         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33703 2017-04-28  Martin Liska  <mliska@suse.cz>
33705         PR gcov-profile/79891
33706         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33707         is marked by compiler as living on a line.
33708         (get_cycles_count): Remove usage of the union.
33709         (output_intermediate_file): Likewise.
33710         (find_source): Fix GNU coding style.
33711         (accumulate_line_counts): Remove old non-all block mode.
33712         (output_lines): Remove usage of the union.
33713         * profile.c (output_location): Include all BBs, even if
33714         belonging to a same line (and file) as a previous BB.
33716 2017-04-28  Martin Liska  <mliska@suse.cz>
33718         * gcov.c (process_args): Handle new argument 'w'.
33719         (read_graph_file): Assign ID to BBs.
33720         (output_branch_count): Display BB # if verbose flag is set.
33721         (output_lines): Likewise for arcs.
33722         (print_usage): Add '--verbose' option help.
33723         * doc/gcov.texi: Document --verbose (-w) option.
33725 2017-04-28  Martin Liska  <mliska@suse.cz>
33727         * gcov.c (struct block_location_info): New struct.
33728         (process_file): Fill up the new structure.
33729         (read_graph_file): Replace usage of encoding by the newly added
33730         struct.
33731         (add_line_counts): Likewise.
33732         (accumulate_line_counts): Remove usage of the union.
33733         (function_info::function_info): New function.
33734         (function_info::~function_info): Likewise.
33735         (process_file): Call delete instead of release_function.
33736         (release_function): Release the function.
33737         (release_structures): Call delete instead of release_function.
33738         (solve_flow_graph): Replace usage of num_blocks.
33739         (find_exception_blocks): Likewise.
33740         (output_lines): Fix GNU coding style.
33742 2017-04-28  Martin Liska  <mliska@suse.cz>
33744         PR driver/56469
33745         * coverage.c (coverage_remove_note_file): New function.
33746         * coverage.h: Declare the function.
33747         * toplev.c (finalize): Clean if an error has been seen.
33749 2017-04-28  Martin Liska  <mliska@suse.cz>
33751         PR gcov-profile/80031
33752         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33753         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33754         * gcov.c (read_graph_file): Read just number of blocks.
33755         * profile.c (branch_prob): Do not stream 0 flags per a basic
33756         block.
33758 2017-04-28  Martin Liska  <mliska@suse.cz>
33760         * gcov-dump.c (tag_*): Add new argument to declarations.
33761         (dump_gcov_file): Likewise.
33762         (tag_blocks): Add and use new argument depth.
33763         (tag_arcs): Likewise.
33764         (tag_lines): Likewise.
33765         (tag_counters): Likewise.
33766         (tag_summary): Likewise.
33767         (dump_working_sets): Use depth to do a proper indentation.
33769 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33771         PR bootstrap/80531
33772         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33773         * symtab.c (symtab_node::debug_symtab): Move definition here.
33775 2017-04-28  Richard Biener  <rguenther@suse.de>
33777         * lto-streamer.h (LTO_major_version): Bump to 7.
33779 2017-04-28  Richard Biener  <rguenther@suse.de>
33781         * tree-vrp.c (assert_info): New struct.
33782         (add_assert_info): New helper.
33783         (register_edge_assert_for_2): Refactor to add asserts to a vector
33784         of assert_info.
33785         (register_edge_assert_for_1): Likewise.
33786         (register_edge_assert_for): Likewise.
33787         (finish_register_edge_assert_for): New helper actually registering
33788         asserts where live on edge.
33789         (find_conditional_asserts): Adjust.
33790         (find_switch_asserts): Likewise.
33791         (evrp_dom_walker::try_find_new_range): Generalize.
33792         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33794 2017-04-27  Marek Polacek  <polacek@redhat.com>
33796         PR sanitizer/80349
33797         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33798         arg10 and arg11 to itype.
33800 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33802         * doc/extend.texi (Object Size Checking): Improve grammar.
33804 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33806         PR target/80530
33807         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33808         that the logic for permitting reciprocal estimates matches that
33809         in use_rsqrt_p.
33811 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33813         PR c++/80534
33814         * tree.c (type_cache_hasher::equal): Only compare
33815         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33816         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33817         non-aggregate element types.
33818         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33819         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33821 2017-04-27  Richard Biener  <rguenther@suse.de>
33823         PR middle-end/80533
33824         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33825         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33826         keeping a reference to a trailing array.
33828 2017-04-27  Richard Biener  <rguenther@suse.de>
33830         PR middle-end/80539
33831         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33832         being in loop-closed SSA form conservatively.
33833         (chrec_fold_multiply_poly_poly): Likewise.
33835 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33837         PR middle-end/79665
33838         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33839         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33841 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33843         PR target/77728
33844         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33845         (aarch64_function_arg_alignment): Return unsigned int again, but still
33846         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33847         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33848         Don't emit -Wpsabi note.
33849         (aarch64_function_arg_boundary): Likewise.
33850         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33851         caller.
33853 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33855         * tree.h (crc32_unsigned_n): Declare.
33856         (crc32_unsigned, crc32_unsigned): Make inline.
33857         * tree.c (crc32_unsigned_bits): Replace with ...
33858         (crc32_unsigned_n): ... this.
33859         (crc32_unsigned, crc32_byte): Remove.
33860         (crc32_string): Remove unnecessary braces.
33862 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33864         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33865         * ipa-inline-analysis.c (MAX_TIME): Remove.
33866         (account_size_time): Use sreal for time.
33867         (dump_inline_summary): Update.
33868         (estimate_function_body_sizes): Update.
33869         (estimate_edge_size_and_time): Update.
33870         (estimate_calls_size_and_time): Update.
33871         (estimate_node_size_and_time): Update.
33872         (inline_merge_summary): Update.
33873         (inline_update_overall_summary): Update.
33874         (estimate_time_after_inlining): Update.
33875         (inline_read_section): Update.
33876         (inline_write_summary): Update.
33877         * ipa-inline.c (compute_uninlined_call_time): Update.
33878         (compute_inlined_call_time): Update.
33879         (recursive_inlining): Update.
33880         (inline_small_functions): Update.
33881         (dump_overall_stats): Update.
33882         * ipa-inline.h: Include sreal.h.
33883         (size_time_entry): Turn time to sreal.
33884         (inline_summary): Turn self_time nad time to sreal.
33886 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33888         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33889         data-streamer.h
33890         (sreal::stream_out, sreal::stream_in): New.
33891         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33893 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33895         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33896         environment.
33898 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33900         PR target/70799
33901         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33902         Handle ASHIFTRT.
33903         (dimode_scalar_chain::compute_convert_gain): Ditto.
33904         (dimode_scalar_chain::make_vector_copies): Ditto.
33905         (dimode_scalar_chain::convert_reg): Ditto.
33906         (dimode_scalar_chain::convert_insn): Ditto.
33907         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33908         (VI248_AVX512BW_1): New mode iterator.
33909         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33910         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33911         mode iterator.
33913 2017-04-25  Martin Sebor  <msebor@redhat.com>
33915         PR tree-optimization/80497
33916         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33917         constants are representable in HOST_WIDE_INT.
33918         (parse_directive): Ditto.
33920 2017-04-25  Martin Sebor  <msebor@redhat.com>
33922         PR bootstrap/80486
33923         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33924         (new_zero_array): Adjust signature.
33925         (dom_info::dom_init): Used unsigned rather that size_t.
33926         (dom_info::dom_info): Same.
33928 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33929             Jakub Jelinek  <jakub@redhat.com>
33931         PR target/77728
33932         * config/arm/arm.c: Include gimple.h.
33933         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33934         returns negative, increment ncrn only if it returned positive.
33935         (arm_needs_doubleword_align): Return int instead of bool,
33936         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33937         members, but if there is any such non-FIELD_DECL
33938         > PARM_BOUNDARY aligned decl, return -1 instead of false.
33939         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33940         returns negative, increment nregs only if it returned positive.
33941         (arm_setup_incoming_varargs): Likewise.
33942         (arm_function_arg_boundary): Emit -Wpsabi note if
33943         arm_needs_doubleword_align returns negative, return
33944         DOUBLEWORD_ALIGNMENT only if it returned positive.
33946 2017-04-25  Marek Polacek  <polacek@redhat.com>
33948         PR sanitizer/80349
33949         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
33950         first argument to type.
33952 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
33954         PR target/80482
33955         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
33956         type checks to test for compatibility instead of equality.
33958 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33959             Jakub Jelinek  <jakub@redhat.com>
33961         PR target/77728
33962         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
33963         type.
33964         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
33965         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
33966         the alignment computation, but return their maximum in warn_alignment.
33967         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33968         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
33969         is smaller.
33970         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
33971         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33972         caller.
33974 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33976         * config/arc/simdext.md (dmpyh): Fix typo.
33978 2017-04-25  Richard Biener  <rguenther@suse.de>
33980         PR tree-optimization/80492
33981         * alias.c (compare_base_decls): Handle registers with asm
33982         specification conservatively.
33983         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33984         compare_base_decls returning dont-know properly.
33986 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33988         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
33989         (legitimate_offset_address_p): New function.
33990         (arc_legitimate_address_p): Use above function.
33992 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33994         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
33996 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33998         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
33999         ACCH registers whenever they are available.
34001 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34003         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
34004         double regs fix when not used.
34006 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34008         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
34009         core registers.
34010         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
34011         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
34013 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34015         * config/arc/arc.c (arc_output_addsi): Check for h-register class
34016         when emitting short ADD instructions.
34018 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34020         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
34021         constraint.
34022         (cmpsi_cc_c_insn): Likewise.
34023         (cbranchsi4_scratch): Compute proper instruction length using
34024         compact_hreg_operand.
34025         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34027 2017-04-25  Richard Biener  <rguenther@suse.de>
34029         PR middle-end/80509
34030         * passes.c (pass_manager::pass_manager): Initialize
34031         m_name_to_pass_map.
34033 2017-04-25  Richard Biener  <rguenther@suse.de>
34035         PR tree-optimization/79201
34036         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34038 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34040         PR target/80464
34041         * config/s390/vector.md: Split MEM->GPR vector moves for
34042         non-s_operand addresses.
34044 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34046         PR target/79895
34047         * config/s390/predicates.md (reload_const_wide_int_operand): New
34048         predicate.
34049         * config/s390/s390.md ("movti"): Remove d/P alternative.
34050         ("movti_bigconst"): New pattern definition.
34052 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34054         PR target/80080
34055         * s390-protos.h (s390_expand_cs_hqi): Removed.
34056         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34057         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34058         modes as well as CCZ1mode and CCZmode.
34059         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34060         signature of s390_emit_compare_and_swap.
34061         (s390_expand_cs_hqi): Likewise, make static.
34062         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34063         compare-and-swap, in some cases.
34064         (s390_expand_cs): Wrapper function.
34065         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34066         atomic_exchange.
34067         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34068         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34069         patterns for small and large integers.  Forbid symref memory operands.
34070         Move expander to s390.c.  Require cc register.
34071         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34072         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34073         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34074         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34075         symref memory operands.  Remove CC mode and call s390_match_ccmode
34076         instead.
34077         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34079 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34081         * config/s390/s390.md (define_peephole2): New peephole to help
34082         combining the load-and-test pattern with volatile memory.
34084 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34086         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34087         with CCZmode for TARGET_Z196.
34089 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34091         PR rtl-optimization/80501
34092         * combine.c (make_compound_operation_int): Set subreg_code to SET
34093         even for AND with mask of the sign bit of mode.
34095         PR rtl-optimization/80500
34096         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34097         sum's initial value.
34099 2017-04-25  Julian Brown  <julian@codesourcery.com>
34100             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34102         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34104 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34106         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34108 2017-04-25  Julian Brown  <julian@codesourcery.com>
34109             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34111         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34112         (thunderx2t99_sha): New Reservation.
34114 2017-04-25  Julian Brown  <julian@codesourcery.com>
34115             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34117         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34118         type for 1-element load.
34120 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34122         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34124 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34126         PR tree-optimization/80293
34127         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34128         char arrays not totally scalarizable if it is false.
34129         (analyze_all_variable_accesses): Pass correct value in the new
34130         parameter.  Add a statistics counter.
34132 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34134         PR middle-end/79931
34135         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34137 2017-04-24  Richard Biener  <rguenther@suse.de>
34139         PR tree-optimization/80494
34140         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34141         out for complex types.
34143 2017-04-24  Richard Biener  <rguenther@suse.de>
34145         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34146         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34147         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34148         (DFS): Adjust and never fail.
34149         (sccvn_dom_walker::fail): Remove.
34150         (sccvn_dom_walker::before_dom_children): Adjust.
34151         (run_scc_vn): Likewise and never fail.
34152         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34153         (pass_fre::execute): Likewise.
34155 2017-04-24  Richard Biener  <rguenther@suse.de>
34157         PR tree-optimization/79725
34158         * tree-ssa-sink.c (statement_sink_location): Return whether
34159         failure reason was zero uses.  Move that check later.
34160         (sink_code_in_bb): Deal with zero uses by removing the stmt
34161         if possible.
34163 2017-04-24  Richard Biener  <rguenther@suse.de>
34165         PR c++/2972
34166         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34167         pointer-based references.
34169 2017-04-24  Richard Biener  <rguenther@suse.de>
34171         PR bootstrap/79814
34172         * pass_manager.h (pass_manager::operator new): Remove.
34173         (pass_manager::operator delete): Likewise.
34174         * passes.c (pass_manager::operator new): Remove.
34175         (pass_manager::operator delete): Likewise.
34176         (pass_manager::pass_manager): Zero individual pass members.
34178 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34180         PR target/70799
34181         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34182         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34183         Check "XEXP (src, 1)" operand here.
34184         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34185         Check "XEXP (src, 1)" operand here.
34186         (dimode_scalar_chain::make_vector_copies): Detect count register
34187         of a shift instruction.  Zero extend count register from QImode
34188         to DImode to satisfy vector shift pattern count operand predicate.
34189         Substitute vector shift count operand with a DImode copy.
34190         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34191         vector register.
34193 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34195         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34196         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34197         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34198         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34199         (UNSPEC_NOREX_MEM): Remove definition.
34201 2017-04-21  Richard Biener  <rguenther@suse.de>
34203         PR tree-optimization/79547
34204         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34205         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34206         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34207         without any constraints.
34209 2017-04-21  Richard Biener  <rguenther@suse.de>
34211         PR tree-optimization/78847
34212         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34214 2017-04-21  Richard Biener  <rguenther@suse.de>
34216         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34217         (build_distinct_type_copy): Likewise.
34218         (build_variant_type_copy): Likewise.
34219         * tree.c (build_qualified_type): Pass down mem-stat info.
34220         (build_distinct_type_copy): Likewise.
34221         (build_variant_type_copy): Likewise.
34223 2017-04-21  Richard Biener  <rguenther@suse.de>
34225         PR tree-optimization/80237
34226         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34227         defaulted to NULL.
34228         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34229         for a simplified result.
34231 2016-04-21  Richard Biener  <rguenther@suse.de>
34233         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34234         sth as strict as a simple_iv but a chrec without symbols and an
34235         operand defined in the loop we are peeling (and not some subloop).
34236         (propagate_constants_for_unrolling): Propagate all constants.
34238 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34240         PR target/79804
34241         * config/i386/i386.c (print_reg): Remove assert for disalowed
34242         regno values, call output_operand_lossage instead.
34244 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34246         PR target/78090
34247         * config/i386/constraints.md (Yc): New register constraint.
34248         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34249         Use Yc constraint for alternative 2 of operand 0.  Remove
34250         preferred_for_speed attribute.
34252 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34254         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34255         lastprivate clauses in SIMT case.
34257 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34259         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34261 2017-04-20  Richard Biener  <rguenther@suse.de>
34263         PR tree-optimization/57796
34264         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34265         as N scalar stores.
34266         (vect_model_load_cost): Cost gathers as N scalar loads.
34268 2017-04-20  Richard Biener  <rguenther@suse.de>
34270         * ggc-page.c (ggc_allocated_p): Rename to ...
34271         (safe_lookup_page_table_entry): ... this and return the lookup
34272         result.
34273         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34275 2017-04-20  Richard Biener  <rguenther@suse.de>
34277         PR tree-optimization/80453
34278         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34279         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34280         from the conditions.
34281         (vn_phi_eq): Pass them down.
34282         (vn_phi_lookup): Record them.
34283         (vn_phi_insert): Likewise.
34285 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34287         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34288         uninitialized variable warning to avoid buffer overrun.
34290 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34292         PR other/71250
34293         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34294         is suppressed for '{ 0 }' in C.
34296 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34298         * BASE-VER: Set to 8.0.0.
34300 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34302         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34303         priority .init_array and .fini_array section with SECTION_NOTYPE
34304         flag.
34306 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34308         PR middle-end/80423
34309         * tree.h (build_array_type): Add typeless_storage default argument.
34310         * tree.c (type_cache_hasher::equal): Also compare
34311         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34312         (build_array_type): Add typeless_storage argument, set
34313         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34314         recursive call.
34315         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34316         (build_array_type): Likewise.  Add typeless_storage argument.
34318 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34319             Jakub Jelinek  <jakub@redhat.com>
34321         PR tree-optimization/80426
34322         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34323         operation on symbolic operands, also compute the overflow for the
34324         invariant part when the operation degenerates into a negation.
34326 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34328         PR debug/80461
34329         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34330         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34332         PR debug/80436
34333         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34335 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34337         PR target/80462
34338         * config/avr/avr.c (tree.h): Include it.
34339         (cgraph.h): Include it.
34340         (avr_encode_section_info): Don't warn for uninitialized progmem
34341         variable if it's just an alias.
34343 2017-04-19  Richard Biener  <rguenther@suse.de>
34345         PR ipa/65972
34346         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34347         when needed by AutoPGO.
34349 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34351         PR lto/50345
34352         * doc/lto.texi: Remove an extra 'that'.
34354 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34356         PR rtl-optimization/80429
34357         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34358         are only used in debug insns.
34360 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34361             Vladimir Makarov  <vmakarov@redhat.com>
34363         * config/sparc/predicates.md (input_operand): Add comment.  Return
34364         true for any memory operand when LRA is in progress.
34365         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34367 2017-04-18  Jeff Law  <law@redhat.com>
34369         PR target/74563
34370         * mips.md ({return,simple_return}_internal): Do not overwrite
34371         operands[0].
34373 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34375         PR tree-optimization/80443
34376         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34377         instead of adding 1, subtract -1 and similarly instead of subtracting
34378         1 add -1.
34380 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34382         PR rtl-optimization/80357
34383         * haifa-sched.c (tmp_bitmap): New variable.
34384         (model_recompute): Handle duplicate use records.
34385         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34386         (free_global_sched_pressure_data): Free it.
34388 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34390         Revert:
34391         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34392         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34393         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34394         instead of SYSTEM_HEADER_DIR.
34396 2017-04-18  Jeff Law  <law@redhat.com>
34398         PR middle-end/80422
34399         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34400         predecessors after walking up the insn chain.
34402 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34404         PR debug/80263
34405         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34406         sizetype type into debug info.
34408 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34410         PR target/80099
34411         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34412         unneeded test for TARGET_UPPER_REGS_SF.
34413         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34415 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34417         PR sanitizer/80444
34418         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34419         instead of gsi_after_labels.
34421 2017-04-18  Jeff Law  <law@redhat.com>
34423         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34424         stack pointer.
34426         Revert:
34427         2017-04-13  Jeff Law  <law@redhat.com>
34428         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34429         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34431 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34433         PR target/79453
34434         * config/avr/avr.c (intl.h): Include it.
34435         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34437 2017-04-18  Martin Liska  <mliska@suse.cz>
34439         PR gcov-profile/78783
34440         * gcov-tool.c (gcov_output_files): Validate that destination
34441         file is either removed by the tool or by a user.
34443 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34444             Guy Benyei  <guybe@mellanox.com>
34446         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34447         block, and do not negate it, the stored id is already negative.
34449 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34451         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34453 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34455         PR target/80098
34456         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34457         masks of options that should be turned off if the VSX vector
34458         options are turned off.
34459         (OTHER_P8_VECTOR_MASKS): Likewise.
34460         (OTHER_VSX_VECTOR_MASKS): Likewise.
34461         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34462         rs6000_disable_incompatible_switches to validate no type switches
34463         like -mvsx.
34464         (rs6000_incompatible_switch): New function to disallow turning on
34465         other vector options if -mno-vsx, -mno-power8-vector, or
34466         -mno-power9-vector are specified.
34468 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34470         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34472 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34474         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34475         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34476         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34477         (ARG_POINTER_CFA_OFFSET): Likewise.
34479 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34481         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34482         conditions to take advantage of various optimizations.
34484 2017-04-13  Jeff Law  <law@redhat.com>
34486         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34487         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34488         (zero_extendsidi2_dext): Likewise.
34490 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34492         PR sanitizer/80403
34493         * fold-const.c (fold_ternary_loc): Revert
34494         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34495         2017-04-12 change.
34497 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34499         PR rtl-optimization/80343
34500         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34501         new scratch pseudo.
34503 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34505         PR sanitizer/80414
34506         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34507         to ubsan_encode_value.
34509 2017-04-13  Jeff Law  <law@redhat.com>
34511         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34512         appearing in DEBUG_INSNs.
34514 2017-04-13  Martin Liska  <mliska@suse.cz>
34516         PR gcov-profile/80413
34517         * gcov-io.c (gcov_write_string): Copy to buffer just when
34518         allocated size is greater than zero.
34520 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34522         PR debug/80321
34523         * dwarf2out.c (decls_for_scope): Ignore declarations of
34524         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34526 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34528         PR lto/69953
34529         * ipa-visibility.c (non_local_p): Fix typos.
34530         (localize_node): When localizing symbol in same comdat group,
34531         dissolve the group only when we know external symbols are going
34532         to be privatized.
34533         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34535 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34537         PR tree-optimization/79390
34538         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34539         order does not result in usable sequence, retry with reversed operand
34540         order.
34542         PR sanitizer/80403
34543         PR sanitizer/80404
34544         PR sanitizer/80405
34545         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34546         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34547         op0 instead of fold_convert_loc (loc, type, arg0).
34549 2017-04-12  Jeff Law  <law@redhat.com>
34551         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34552         has a delay slot in the generated code.
34554         * config/cris/cris.md (cris_preferred_reload_class): Return
34555         GENNONACR_REGS rather than GENERAL_REGS.
34557 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34559         PR c/80163
34560         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34561         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34562         signedness of the result type.
34564 2017-04-12  Richard Biener  <rguenther@suse.de>
34565             Jeff Law  <law@redhat.com>
34567         PR tree-optimization/80359
34568         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34569         trim stores to TARGET_MEM_REFs.
34571 2017-04-12  Richard Biener  <rguenther@suse.de>
34573         PR tree-optimization/79390
34574         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34575         threading case even more.
34577 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34579         PR target/80382
34580         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34581         for quad_address_p for TImode, instead of just not indexed_address.
34583 2017-04-12  Richard Biener  <rguenther@suse.de>
34584             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34586         PR middle-end/79671
34587         * alias.c (component_uses_parent_alias_set_from): Handle
34588         TYPE_TYPELESS_STORAGE.
34589         (get_alias_set): Likewise.
34590         * tree-core.h (tree_type_common): Add typeless_storage flag.
34591         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34592         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34593         for types containing members with TYPE_TYPELESS_STORAGE.
34594         (place_field): Likewise.
34595         (layout_type): Likewise for ARRAY_TYPE.
34596         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34597         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34598         TYPE_TYPELESS_STORAGE.
34599         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34601 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34603         PR sanitizer/80349
34604         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34605         first argument to type.
34607 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34609         PR target/80376
34610         PR target/80315
34611         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34612         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34613         (rs6000_expand_binop_builtin): Likewise.
34614         (rs6000_expand_ternop_builtin): Likewise; also add missing
34615         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34616         vshasigma built-ins.
34617         * doc/extend.texi: Document that vec_xxpermdi's third argument
34618         must be a constant.
34620 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34622         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34623         Use shift_const cost parameter when calculating gain of STV shifts.
34625 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34627         PR rtl-optimization/70478
34628         * lra-constraints.c (process_alt_operands): Check memory for
34629         disfavoring memory insn operand.
34631 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34633         PR middle-end/80100
34634         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34635         left shift in unsigned HOST_WIDE_INT type.
34637         PR rtl-optimization/80385
34638         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34639         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34641         PR libgomp/80394
34642         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34643         if they have any depend clauses.
34645 2017-04-11  Martin Liska  <mliska@suse.cz>
34647         PR ipa/80212
34648         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34649         * ipa-split.c (split_function): Create a local comdat symbol
34650         if caller is in a comdat group.
34652 2017-04-11  Martin Liska  <mliska@suse.cz>
34654         PR ipa/80212
34655         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34656         flags.
34658 2017-04-11  Martin Sebor  <msebor@redhat.com>
34660         PR middle-end/80364
34661         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34662         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34663         for INTEGER_TYPE.
34664         (directive::set_width, directive::set_precision, format_character):
34665         Adjust.
34666         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34667         INTEGER_TYPE.
34669 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34671         PR target/80389
34672         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34673         conflict, set target->arch_name instead of target->cpu_name.
34675 2017-04-11  Richard Biener  <rguenther@suse.de>
34677         PR tree-optimization/80374
34678         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34679         build_zero_cst, remove fold_convertible_p check again.
34681 2017-04-11  Martin Liska  <mliska@suse.cz>
34683         PR sanitizer/70878
34684         * ubsan.c (instrument_object_size): Do not instrument register
34685         variables.
34687 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34689         PR target/80381
34690         * config/i386/i386-builtin-types.def
34691         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34692         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34693         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34694         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34695         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34696         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34697         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34698         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34699         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34700         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34701         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34702         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34703         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34704         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34705         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34706         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34707         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34708         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34709         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34710         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34711         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34712         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34713         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34714         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34715         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34716         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34717         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34718         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34719         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34720         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34721         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34722         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34723         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34724         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34725         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34726         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34727         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34728         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34729         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34730         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34731         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34732         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34733         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34734         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34735         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34736         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34737         aliases.
34738         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34739         flag to second_arg_count, handle 4 argument function type _COUNT
34740         aliases, handle second_arg_count on second argument rather than last.
34742 2017-04-10  Jeff Law  <law@redhat.com>
34744         PR tree-optimization/80374
34745         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34746         record anything if we can not convert integer_zero_node to the
34747         desired type.
34749 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34751         PR target/80108
34752         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34753         Enhance special handling given to the TARGET_P9_MINMAX option in
34754         relation to certain other options.
34756 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34758         PR tree-optimization/80153
34759         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34760         remove POINTER_PLUS_EXPR's base part directly, rather than through
34761         aff_tree.
34763 2017-04-10  Richard Biener  <rguenther@suse.de>
34764             Bin Cheng  <bin.cheng@arm.com>
34766         PR tree-optimization/80153
34767         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34768         the first element of pointer type aff_tree.  Build result expr in
34769         aff_tree's type.
34770         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34771         fold_convert calls.
34772         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34773         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34775 2017-04-10  Richard Biener  <rguenther@suse.de>
34777         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34778         asm inputs.
34780 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34782         PR rtl-optimization/70478
34783         * lra-constraints.c (curr_small_class_check): New.
34784         (update_and_check_small_class_inputs): New.
34785         (process_alt_operands): Update curr_small_class_check.  Disfavor
34786         alternative insn memory operands.  Check available regs for small
34787         class operands.
34789 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34791         PR target/80057
34792         * config/mips/mips.opt (-mvirt): Update description.
34793         * doc/invoke.texi (-mvirt): Likewise.
34795 2017-04-10  Richard Biener  <rguenther@suse.de>
34797         PR middle-end/80362
34798         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34799         looking for NEGATE_EXPR in -A / -B to A / B folding.
34801 2017-04-10  Martin Liska  <mliska@suse.cz>
34803         PR gcov-profile/80224
34804         * gcov.c (print_usage): Fix usage string.
34805         (get_gcov_intermediate_filename): Remove.
34806         (output_gcov_file): Use both for normal and intermediate format.
34807         (generate_results): Do not initialize special file for
34808         intermediate format.
34810 2017-04-10  Richard Biener  <rguenther@suse.de>
34812         PR tree-optimization/80304
34813         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34814         for safelen.
34816 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34818         PR target/79905
34819         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34820         (rs6000_init_builtins): Use it.
34822 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34824         * config/arm/arm.md (<mrc>): Add mode to SET source.
34825         (<mrrc>): Likewise.
34827 2017-04-10  Richard Biener  <rguenther@suse.de>
34829         PR middle-end/80344
34830         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34832 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34834         PR target/80324
34835         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34836         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34837         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34838         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34839         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34840         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34841         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34842         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34843         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34844         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34845         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34846         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34847         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34848         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34849         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34850         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34851         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34852         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34853         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34854         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34855         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34856         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34857         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34859 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34861         PR rtl-optimization/70478
34862         * lra-constraints.c: Reverse the last patch.
34864 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34866         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34867         Add comment for WCHAR_T.
34869 2017-04-08  Martin Liska  <mliska@suse.cz>
34871         Revert:
34872         2017-04-07  Martin Liska  <mliska@suse.cz>
34874         PR ipa/80212
34875         * ipa-split.c (split_function): Add function part to a same comdat
34876         group.
34878 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34880         PR target/80358
34881         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34883 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34885         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34886         (vec_pairing): ...to this.
34887         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34888         (rs6000_sched_init): Adjust for name change.
34889         (struct rs6000_sched_context): Likewise.
34890         (rs6000_init_sched_context): Likewise.
34891         (rs6000_set_sched_context): Likewise.
34893 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34895         PR target/80322
34896         PR target/80323
34897         PR target/80325
34898         PR target/80326
34899         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34900         intrinsics.
34901         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34902         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34903         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34905 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34907         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34909 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34911         PR rtl-optimization/70703
34912         * ira-color.c (update_conflict_hard_regno_costs): Use
34913         int64_t instead of HOST_WIDE_INT.
34915 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34917         PR rtl-optimization/70478
34918         * lra-constraints.c (process_alt_operands): Disfavor alternative
34919         insn memory operands.
34921 2017-04-07  Jeff Law  <law@redhat.com>
34923         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34924         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34926 2017-04-07  Martin Liska  <mliska@suse.cz>
34928         PR target/79889
34929         * config/aarch64/aarch64.c (aarch64_process_target_attr):
34930         Show error message instead of an ICE.
34932 2017-04-07  Martin Liska  <mliska@suse.cz>
34934         PR ipa/80212
34935         * ipa-split.c (split_function): Add function part to a same comdat
34936         group.
34938 2017-04-07  Richard Biener  <rguenther@suse.de>
34940         PR middle-end/80341
34941         * tree.c (get_unwidened): Also handle ! for_type case for
34942         INTEGER_CSTs.
34943         * convert.c (do_narrow): Split out from ...
34944         (convert_to_integer_1): ... here.  Do not pass final truncation
34945         type to get_unwidened for TRUNC_DIV_EXPR.
34947 2017-04-07  Richard Biener  <rguenther@suse.de>
34949         * tree-affine.c (wide_int_ext_for_comb): Take type rather
34950         than aff_tree.
34951         (aff_combination_const): Adjust.
34952         (aff_combination_scale): Likewise.
34953         (aff_combination_add_elt): Likewise.
34954         (aff_combination_add_cst): Likewise.
34955         (aff_combination_convert): Likewise.
34956         (add_elt_to_tree): Likewise.  Remove unused argument.
34957         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
34959 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34961         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
34962         definition.
34963         * config/arm/arm.c (arm_default_short_enums): Use
34964         ARM_DEFAULT_SHORT_ENUMS.
34965         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
34967 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
34969         PR debug/80234
34970         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
34971         members with redundant out-of-class redeclaration.
34973 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34975         PR target/80286
34976         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
34977         * config/i386/i386.md (*zero_extendsidi2):
34978         Add (?*x,*x) and (?*v,*v) alternatives.
34980 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34982         PR target/79733
34983         * config/i386/i386.c (ix86_expand_builtin)
34984         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
34985         mode from insn data. Convert operands to insn operand mode.
34986         Copy operands that don't satisfy insn predicate to a register.
34988 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
34990         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
34991         Update comments.
34993 2017-04-06  Richard Biener  <rguenther@suse.de>
34995         PR tree-optimization/80334
34996         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
34997         preserve alignment of accesses.
34999 2017-04-06  Richard Biener  <rguenther@suse.de>
35001         PR tree-optimization/80262
35002         * tree-sra.c (build_ref_for_offset): Preserve address-space
35003         information.
35004         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
35005         Drop useless address-space information on MEM_REF offsets.
35007 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
35009         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
35011 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35013         PR rtl-optimization/70703
35014         * ira-color.c (update_conflict_hard_regno_costs): Use
35015         HOST_WIDE_INT instead of long.
35017 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
35019         PR target/80298
35020         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
35021         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
35022         is not defined.
35023         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35024         for x86_64 target.  Handle -m3dnowa option.
35026 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35028         PR rtl-optimization/70703
35029         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35030         (update_conflict_hard_regno_costs): Use long instead of unsigned
35031         arithmetic for cost calculation.
35033 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35034             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35036         PR sanitizer/80308
35037         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35038         for big endian.
35040 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35042         PR target/78002
35043         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35044         ptr_mode with Pmode throughout.
35045         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35046         into probe_stack_range and use DImode.
35048 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35050         PR target/79890
35051         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35052         call_eh_return is true.
35054 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35056         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35057         Initialize last_match_fntype_index.
35059 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35061         PR target/80310
35062         * tree-nvr.c: Include internal-fn.h.
35063         (pass_return_slot::execute): Ignore internal calls without
35064         direct optab.
35066 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35067             Richard Biener  <rguenther@suse.de>
35069         PR c++/80297
35070         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35071         captures used multiple times, except for the last use.
35072         * generic-match-head.c: Include gimplify.h.
35074 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35076         PR tree-optimization/79390
35077         * target.h (struct noce_if_info): Declare.
35078         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35079         * target.def (noce_conversion_profitable_p): New target hook.
35080         * ifcvt.h (struct noce_if_info): New type, moved from ...
35081         * ifcvt.c (struct noce_if_info): ... here.
35082         (noce_conversion_profitable_p): Renamed to ...
35083         (default_noce_conversion_profitable_p): ... this.  No longer
35084         static nor inline.
35085         (noce_try_store_flag_constants, noce_try_addcc,
35086         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35087         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35088         instead of noce_conversion_profitable_p.
35089         * config/i386/i386.c: Include ifcvt.h.
35090         (ix86_option_override_internal): Don't override
35091         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35092         (ix86_noce_conversion_profitable_p): New function.
35093         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35094         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35095         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35096         * doc/tm.texi: Regenerated.
35098 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35100         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35101         correction.
35103 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35105         PR target/80307
35106         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35107         instructions for small multiply cores.
35109 2017-04-04  Jeff Law  <law@redhat.com>
35111         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35112         added member.
35113         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35114         that are not set by the loop over the elements.
35116 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35118         PR target/80286
35119         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35120         int mode, convert_modes it to mode as unsigned, otherwise use
35121         lowpart_subreg to mode rather than SImode.
35122         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35123         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35124         Use DImode instead of SImode for the shift count operand.
35125         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35126         Likewise.
35128 2017-04-04  Richard Biener  <rguenther@suse.de>
35130         PR middle-end/80281
35131         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35132         arithmetic done for the negate or the plus.  Simplify.
35133         (A - (-B) -> A + B): Likewise.
35134         * fold-const.c (split_tree): Make sure to not negate pointers.
35136 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35138         PR rtl-optimization/60818
35139         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35140         a compare of comparisons with the thing compared if this results
35141         in a different machine mode.
35143 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35145         * alias.c (base_alias_check): Fix typo in comment.
35146         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35147         * cgraphunit.c (symbol_table::compile): Likewise.
35148         * collect2.c (maybe_run_lto_and_relink): Likewise.
35149         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35150         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35151         * config/avr/avr.c (avr_map_op_t): Likewise.
35152         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35153         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35154         * config/epiphany/epiphany.md (movcc): Likewise.
35155         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35156         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35157         Likewise.
35158         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35159         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35160         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35161         * config/sh/sh.c (sh_rtx_costs): Likewise.
35162         * fold-const.c (fold_truth_andor): Likewise.
35163         * genautomata.c (collapse_flag): Likewise.
35164         * gengtype.h (struct type::u::s): Likewise.
35165         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35166         * input.c (FORMAT_AMOUNT): Likewise.
35167         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35168         (known_aggs_to_agg_replacement_list): Likewise.
35169         * ipa-inline-analysis.c: Likewise.
35170         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35171         * ipa-polymorphic-call.c
35172         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35173         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35174         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35175         Likewise.
35176         * modulo-sched.c (apply_reg_moves): Likewise.
35177         * omp-expand.c (build_omp_regions_1): Likewise.
35178         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35179         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35180         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35181         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35182         * value-prof.c: Likewise.
35183         * var-tracking.c (val_reset): Likewise.
35185 2017-04-03  Richard Biener  <rguenther@suse.de>
35187         PR tree-optimization/80275
35188         * fold-const.c (split_address_to_core_and_offset): Handle
35189         POINTER_PLUS_EXPR.
35191 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35193         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35194         descriptors is at least equal to that of functions.
35196 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35198         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35200 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35202         PR target/80250
35203         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35204         (mov<IMOD4:mode>): New expander.
35205         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35207 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35209         PR rtl-optimization/79405
35210         * fwprop.c (propagations_left): New variable.
35211         (forward_propagate_into): Decrement it.
35212         (fwprop_init): Initialize it.
35213         (fw_prop): If the variable has reached zero, stop propagating.
35214         (fwprop_addr): Ditto.
35216 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35218         PR debug/79255
35219         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35220         a FUNCTION_DECL, pass it as decl instead of origin to
35221         process_scope_var.
35223 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35225         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35226         string.
35228 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35230         PR target/80107
35231         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35232         TARGET_VSX_SMALL_INTEGER.
35234 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35236         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35237         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35239 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35241         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35242         extraction from odd-numbered MSA register.
35244 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35246         PR middle-end/80173
35247         * expmed.c (store_bit_field_1): Don't attempt to create
35248         a word subreg out of hard registers wider than word if they
35249         have HARD_REGNO_NREGS of 1 for their mode.
35251         PR middle-end/80163
35252         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35253         conversions to integer types wider than word and pointer.
35255         PR debug/80025
35256         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35257         (rtx_equal_for_cselib_p): Pass 0 to it.
35258         * cselib.c (cselib_hasher::equal): Likewise.
35259         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35260         is 128, don't look up VALUE locs and punt.  Increment
35261         depth in recursive calls when walking VALUE locs.
35263 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35265         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35266         (make_gcov_file_name): Use the canonical path name for generating
35267         the MD5 value.
35268         (read_line): Fix handling of files with ascii null bytes.
35270 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35272         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35273         to initialise a vector register instead
35274         of using a const_int.
35276 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35278         PR translation/80189
35279         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35280         diagnostic messages.
35282 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35284         PR target/80246
35285         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35286         (dfp_diex_<mode>): Update mode of operand 1.
35287         * doc/extend.texi (dxex, dxexq): Document change to return type.
35288         (diex, diexq): Document change to argument type.
35290 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35292         PR ipa/77333
35293         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35294         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35295         it reflects the signature changes performed at the callee side.
35296         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35297         to cgraph_build_function_type_skip_args.
35298         (build_function_decl_skip_args): Adjust call to the above function.
35300 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35302         PR target/80206
35303         * config/i386/sse.md
35304         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35305         register as dest whenever it is a MEM not rtx_equal_p to the
35306         corresponding dup operand, and when forcing into reg move the
35307         reg into the memory afterwards.
35308         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35309         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35310         for the force_reg mode.
35311         (avx512vl_vextractf128<mode>): Use register as dest either
35312         always when a MEM, or when it is a MEM not rtx_equal_p to the
35313         corresponding dup operand, or even not when it is a CONST_VECTOR
35314         depending on the mode and lo vs. hi.
35315         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35316         parens.
35317         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35318         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35319         Likewise.  Require that operands[2] is even.
35320         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35321         Remove extraneous parens.  Require that operands[2] is a multiple
35322         of 4.
35323         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35324         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35325         disallow memory then.
35327 2017-03-30  Richard Biener  <rguenther@suse.de>
35329         PR tree-optimization/77498
35330         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35331         to non-constants over backedges.
35333 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35335         PR rtl-optimization/80233
35336         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35337         as last_combined_insn.  Do not test for BARRIER_P separately.
35339 2017-03-29  Andreas Schwab  <schwab@suse.de>
35341         PR ada/80146
35342         * calls.c (prepare_call_address): Convert funexp to Pmode before
35343         copying to temp reg.
35345 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35347         PR tree-optimization/80158
35348         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35349         Handle possible future case of more than one alternate
35350         interpretation.
35351         (replace_rhs_if_not_dup): Likewise.
35352         (replace_one_candidate): Likewise.
35354 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35356         PR rtl-optimization/80193
35357         * ira.c (ira): Do not check allocation for LRA.
35359 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35361         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35362         (nvptx_output_simt_exit): Declare.
35363         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35364         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35365         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35366         (nvptx_declare_function_name): ...here.  Emit declaration of local
35367         memory space buffer for omp_simt_enter insn.
35368         (nvptx_output_unisimt_switch): New.
35369         (nvptx_output_softstack_switch): New.
35370         (nvptx_output_simt_enter): New.
35371         (nvptx_output_simt_exit): New.
35372         * config/nvptx/nvptx.h (struct machine_function): New fields
35373         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35374         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35375         (UNSPECV_SIMT_EXIT): Ditto.
35376         (omp_simt_enter_insn): New insn.
35377         (omp_simt_enter): New expansion.
35378         (omp_simt_exit): New insn.
35379         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35381         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35382         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35383         (expand_GOMP_SIMT_EXIT): New.
35384         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35385         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35386         (GOMP_SIMT_EXIT): Ditto.
35387         * target-insns.def (omp_simt_enter): New insn.
35388         (omp_simt_exit): Ditto.
35389         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35390         simt_dlist.
35391         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35392         (lower_rec_input_clauses): Likewise.
35393         (lower_lastprivate_clauses): Handle SIMT privatization.
35395         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35396         (ompdevlow_adjust_simt_enter): New.
35397         (find_simtpriv_var_op): New.
35398         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35399         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35401         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35402         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35403         (copy_decl_for_dup_finish): Ditto.
35405         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35407 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35409         PR target/53383
35410         * config/i386/i386.c (ix86_option_override_internal): Always
35411         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35413 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35415         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35417 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35419         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35420         mark new edge's irreducible flag accordign to it.
35421         (vect_do_peeling): Check loop preheader edge's irreducible flag
35422         and pass it to function slpeel_add_loop_guard.
35424 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35426         PR tree-optimization/80218
35427         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35428         Update block frequencies and counts.
35430 2017-03-28  Richard Biener  <rguenther@suse.de>
35432         PR tree-optimization/78644
35433         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35434         of a simplification result we may not use it at all.
35436 2017-03-28  Richard Biener  <rguenther@suse.de>
35438         PR ipa/80205
35439         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35440         without arguments, generate default definition of a SSA name.
35442 2017-03-28  Richard Biener  <rguenther@suse.de>
35444         PR middle-end/80222
35445         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35446         TYPE_REF_CAN_ALIAS_ALL references.
35447         * fold-const.c (fold_indirect_ref_1): Likewise.
35449 2017-03-28  Martin Liska  <mliska@suse.cz>
35451         PR ipa/80104
35452         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35453         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35455 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35456             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35458         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35459         (EXTRA_SPECS): Define.
35460         (SUBTARGET_EXTRA_SPECS): Likewise.
35461         (SUBTARGET_CPP_SPEC): Likewise.
35462         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35463         SUBTARGET_EXTRA_SPECS.
35464         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35466 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35468         * config/arc/simdext.md (vst64_insn): Update pattern.
35469         (vld32wh_insn): Likewise.
35470         (vld32wl_insn): Likewise.
35471         (vld64_insn): Likewise.
35472         (vld32_insn): Likewise.
35474 2017-03-28  Marek Polacek  <polacek@redhat.com>
35476         PR sanitizer/80067
35477         * fold-const.c (fold_comparison): Use protected_set_expr_location
35478         instead of SET_EXPR_LOCATION.
35480 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35482         * tree.c (add_expr): Avoid name lookup warning.
35484 2017-03-27  Jeff Law  <law@redhat.com>
35486         PR tree-optimization/80216
35487         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35488         function name.  Limit recursion depth.
35489         (record_temporary_equivalences): Corresponding changes.
35491 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35493         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35494         covered first.
35496 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35498         PR target/80102
35499         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35500         notes.
35501         * cfgcleanup.c (reg_note_cfa_p): New array.
35502         (insns_have_identical_cfa_notes): New function.
35503         (old_insns_match_p): Don't cross-jump in between /f
35504         and non-/f instructions.  If both i1 and i2 are frame related,
35505         verify all CFA notes, their order and content.
35507 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35509         PR target/78543
35510         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35511         HImode and SImode with zero extend to DImode to one insn.
35512         (bswap<mode>2_extenddi): Likewise.
35513         (bswapsi2_extenddi): Likewise.
35514         (bswaphi2_extendsi): Likewise.
35515         (bswaphi2): Combine bswap HImode and SImode into one insn.
35516         Separate memory insns from swapping register.
35517         (bswapsi2): Likewise.
35518         (bswap<mode>2): Likewise.
35519         (bswaphi2_internal): Delete, no longer used.
35520         (bswapsi2_internal): Likewise.
35521         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35522         store, and gpr<-gpr swap insns.
35523         (bswap<mode>2_store): Likewise.
35524         (bswaphi2_reg): Register only splitter, combine with the splitter.
35525         (bswaphi2 splitter): Likewise.
35526         (bswapsi2_reg): Likewise.
35527         (bswapsi2 splitter): Likewise.
35528         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35529         the insns into load, store, and register/register insns.
35530         (bswapdi2_ldbrx): Likewise.
35531         (bswapdi2_load): Likewise.
35532         (bswapdi2_store): Likewise.
35533         (bswapdi2_reg): Likewise.
35535 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35537         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35538         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35540 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35542         PR target/80103
35543         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35544         add comments.
35545         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35546         special handling for target option conflicts between dform
35547         options (-mpower9-dform, -mpower9-dform-vector,
35548         -mpower9-dform-scalar) and -mno-direct-move.
35550 2017-03-27  Richard Biener  <rguenther@suse.de>
35552         PR tree-optimization/80181
35553         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35555 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35557         * config/arc/predicates.md (move_double_src_operand): Replace the
35558         call to move_double_src_operand with a call to address_operand.
35560 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35562         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35563         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35564         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35566 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35568         * config/arc/predicates.md (long_immediate_loadstore_operand):
35569         Consider scaled addresses cases.
35571 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35573         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35574         restored when in interrupt.
35575         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35576         doesn't have delay slot.
35578 2017-03-27  Richard Biener  <rguenther@suse.de>
35580         PR ipa/79776
35581         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35582         inlined thunk clones.
35584 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35586         PR sanitizer/80168
35587         * asan.c (instrument_derefs): Copy over last operand from
35588         original COMPONENT_REF to the new COMPONENT_REF with
35589         DECL_BIT_FIELD_REPRESENTATIVE.
35590         * ubsan.c (instrument_object_size): Likewise.
35592 2017-03-27  Richard Biener  <rguenther@suse.de>
35594         PR tree-optimization/80170
35595         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35596         sure DR/SCEV didnt fold in constants we do not see when looking
35597         at the reference base alignment.
35599 2017-03-27  Richard Biener  <rguenther@suse.de>
35601         PR middle-end/80171
35602         * gimple-fold.c (fold_ctor_reference): Properly guard against
35603         NULL return value from canonicalize_constructor_val.
35605 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35607         PR target/80180
35608         * config/i386/i386.c (ix86_expand_builtin)
35609         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35610         flags reg setting and flags reg using instructions.
35611         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35612         clobbering instructions to zero extend op2.
35614 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35616         * doc/install.texi (Configuration) <--with-aix-soname>:
35617         Update link to AIX ld.
35619 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35621         PR rtl-optimization/80160
35622         PR rtl-optimization/80159
35623         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35624         reg_alternate_class into account.
35626 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35628         PR target/80148
35629         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35630         to consider in curr_insn_transform.
35632 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35634         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35635         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35636         and emit_mode_inner.
35638 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35640         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35641         argument to the overloaded builtin variants.  Use the new flag to
35642         deprecate certain builtin variants.
35643         * config/s390/s390-builtin-types.def: Add new builtin types.
35644         * config/s390/s390-builtins.h: Support new flags field for
35645         overloaded builtins.
35646         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35647         (s390_macro_to_expand): Enable vector float data type.
35648         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35649         builtins by incrementing the __VEC__ version number.
35650         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35651         vec_xst.
35652         (s390_resolve_overloaded_builtin): Emit error messages depending
35653         on the builtin flags.
35654         * config/s390/s390.c (s390_expand_builtin): Support additional
35655         flags argument.  Change error message to match the messages
35656         emitted in s390-c.c.
35657         * config/s390/s390.md: New UNSPEC_* constants.
35658         (op_type): Add new instruction types.
35659         * config/s390/vecintrin.h: Add new builtins and test data class
35660         constants.
35661         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35662         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35663         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35664         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35665         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35666         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35668         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35669         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35670         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35671         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35673         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35674         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35675         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35676         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35678         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35679         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35680         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35681         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35682         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35683         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35684         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35686         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35687         ("vec_scatter_element<V_HW_4:mode>_DI")
35688         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35689         ("vec_fpint<mode>", "vflls")
35690         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35691         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35692         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35693         ("*vec_cmphe<mode>_cc"): ... these.
35695         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35696         mode constant instead of magic value.
35698 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35700         * config/s390/s390.c (s390_expand_vec_compare): Support other
35701         vector floating point modes than just V2DF.
35702         (s390_expand_vcond): Likewise.
35703         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35704         (s390_cannot_change_mode_class): Prevent mode changes between TF
35705         and V1TF in vector registers.
35706         * config/s390/s390.md (DF, SF): New mode attributes.
35707         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35708         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35709         SFmode support for VRs.
35710         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35711         vector fp modes.
35712         (VFT, VF_HW): New mode iterators.
35713         (vw, sdx): New mode attributes.
35714         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35715         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35716         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35717         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35718         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35719         also the new vector floating point modes.  Renaming to ...
35721         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35722         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35723         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35724         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35725         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35726         ("vec_unordered<mode>"): ... these.
35728         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35729         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35730         ("*vec_extendv2df"): New insn definitions.
35732 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35734         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35735         ("mulditi3_2", "*muldi3_sign"): New patterns.
35736         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35737         rename the pattern definition.
35739 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35741         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35742         expander.
35743         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35745 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35747         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35748         instruction if possible.
35749         * config/s390/vector.md (vec_halfnumelts): New mode
35750         attribute.
35751         ("*vec_vllezlf<mode>"): New pattern.
35753 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35755         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35756         ("popcountv4si2", "popcountv2di2"): Rename to ...
35757         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35758         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35759         condition.
35760         ("popcount<mode>2_vxe"): New pattern.
35762 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35764         * common/config/s390/s390-common.c (processor_flags_table): Add
35765         arch12.
35766         * config.gcc: Add arch12.
35767         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35768         Default to arch12 for unknown CPU model numbers.
35769         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35770         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35771         PROCESSOR_max sanity check.
35772         * config/s390/s390-opts.h (enum processor_type): Add
35773         PROCESSOR_ARCH12.
35774         * config/s390/s390.c (processor_table): Add arch12.
35775         (s390_expand_builtin): Add check for B_VXE flag.
35776         (s390_issue_rate): Add PROCESSOR_ARCH12.
35777         (s390_get_sched_attrmask): Likewise.
35778         (s390_get_unit_mask): Likewise.
35779         (s390_sched_score): Enable z13 scheduling for arch12.
35780         (s390_sched_reorder): Likewise.
35781         (s390_sched_variable_issue): Likewise.
35782         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35783         PF_VXE.
35784         (s390_tune_attr): Use z13 scheduling also for arch12.
35785         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35786         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35787         (TARGET_VXE_P): New macros.
35788         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35789         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35790         * config/s390/s390.opt: Add arch12 as processor_type.
35792 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35794         * config/s390/s390.md
35795         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35796         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35797         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35799         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35800         Rename expanders to ...
35802         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35803         ("fixuns_truncdddi2_emu"): ... these.
35805         ("fixuns_trunc<mode>si2_emu"): New expander.
35807         ("*fixuns_truncdfdi2_z13"): Rename to ...
35808         ("*fixuns_truncdfdi2_vx"): ... this.
35810 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35812         * config/s390/2964.md: Remove the single element vector compare
35813         instructions which are no longer used.
35814         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35815         vector CCmodes.
35816         (s390_canonicalize_comparison): Remove handling of DFmode
35817         compares.
35818         (s390_expand_vec_compare_scalar): Remove function.
35819         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35820         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35821         pattern.
35822         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35824 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35826         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35827         FP zero.
35828         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35829         will anyway by matched by mov<mode>_64dfp.
35831 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35833         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35834         vlef/vstef.  Add missing operand to vleif.
35836 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35838         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35839         pair for all vector types with 64 bit elements.
35840         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35841         * config/s390/vector.md (V_HW_64): ... here.
35842         (V_128_NOSINGLE): New mode iterator.
35843         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35844         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35845         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35846         ("*vec_load_pairv2di"): Change to ...
35847         ("*vec_load_pair<mode>"): ... this one.
35849 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35851         * config/s390/constraints.md: Add comments.
35852         (jKK): Reject element sizes > 8 bytes.
35853         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35854         s_operands.
35855         * config/s390/s390.md: Add the s_operand checks formerly in
35856         s390_split_ok_p to various splitters where they are still
35857         required.
35858         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35859         for 128 bit vectors.  Plus two splitters.
35861 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35863         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35864         the file.
35866 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35868         PR target/79893
35869         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35870         error if the boundary argument is not constant.
35872 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35874         PR rtl-optimization/80112
35875         * loop-doloop.c (doloop_condition_get): Don't check condition
35876         if cmp isn't SET with IF_THEN_ELSE src.
35878 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35880         PR tree-optimization/80158
35881         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35882         replacing a candidate statement, also replace it for the
35883         candidate's alternate interpretation.
35884         (replace_rhs_if_not_dup): Likewise.
35885         (replace_one_candidate): Likewise.
35887 2017-03-24  Richard Biener  <rguenther@suse.de>
35889         PR tree-optimization/80167
35890         * graphite-isl-ast-to-gimple.c
35891         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35892         properly.
35893         (translate_isl_ast_to_gimple::get_rename): Likewise.
35895 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35897         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35898         handling of certain combinations of target options, including the
35899         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35900         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35902 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35904         PR target/71436
35905         * config/arm/arm.md (*load_multiple): Add reload_completed to
35906         matching condition.
35908 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35909             Richard Biener  <rguenth@suse.de>
35911         PR tree-optimization/79908
35912         PR tree-optimization/80136
35913         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35914         been cast away, gimplify_and_add suffices.
35916 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35918         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35920 2017-03-23  Richard Biener  <rguenther@suse.de>
35922         PR tree-optimization/80032
35923         * gimplify.c (gimple_push_cleanup): Forced unconditional
35924         cleanups still have to go to the conditional_cleanups
35925         sequence.
35927 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
35929         PR tree-optimization/80072
35930         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35931         to unsigned int.
35932         (next_operand_entry_id): Change type to unsigned int.
35933         (sort_by_operand_rank): Make sure to return the right return value
35934         even if unsigned fields are bigger than INT_MAX.
35935         (struct oecount): Change cnt and id type to unsigned int.
35936         (oecount_hasher::equal): Formatting fix.
35937         (oecount_cmp): Make sure to return the right return value
35938         even if unsigned fields are bigger than INT_MAX.
35939         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
35941         PR c++/80129
35942         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
35943         TREE_READONLY on result if writing it more than once.
35945         PR sanitizer/80110
35946         * doc/invoke.texi (-fsanitize=thread): Document that with
35947         -fnon-call-exceptions atomics are not able to throw
35948         exceptions.
35950         PR sanitizer/80110
35951         * tsan.c: Include tree-eh.h.
35952         (instrument_builtin_call): Call maybe_clean_eh_stmt or
35953         maybe_clean_or_replace_eh_stmt where needed.
35954         (instrument_memory_accesses): Add cfg_changed argument.
35955         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
35956         if it returned true.
35957         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
35959         PR rtl-optimization/63191
35960         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
35961         wrapper function, moved the whole old content into ...
35962         (ix86_delegitimize_address_1): ... this.  New inline function.
35963         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
35964         true as last argument instead of ix86_delegitimize_address.
35966 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35968         * config/aarch64/aarch64.c (generic_branch_cost): Copy
35969         cortexa57_branch_cost.
35971 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35973         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
35975 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35977         PR target/80123
35978         * doc/md.texi (Constraints): Document wA constraint.
35979         * config/rs6000/constraints.md (wA): New.
35980         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
35981         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
35982         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
35983         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
35985 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
35987         PR c++/80029
35988         * gimplify.c (is_oacc_declared): New function.
35989         (oacc_default_clause): Use it to set default flags for acc declared
35990         variables inside parallel regions.
35991         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
35992         declared variables.
35993         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
35994         declare attribute to any decl as necessary.
35996 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35998         PR target/80082
35999         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
36000         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
36001         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
36002         (arm_arch_lpae): This.
36003         * config/arm/arm.c (arm_arch7ve): Rename into ...
36004         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
36005         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
36006         arm_arch_lpae.
36008 2017-03-22  Martin Liska  <mliska@suse.cz>
36010         PR target/79906
36011         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
36012         error message instead of an ICE.
36014 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36016         * doc/extend.texi (6.11 Additional Floating Types): Revise.
36018 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36020         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
36021         comments.
36022         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
36023         comments.
36025 2017-03-21  Martin Sebor  <msebor@redhat.com>
36027         * doc/extend.texi: Use "cannot" instead of "can't."
36028         * doc/hostconfig.texi: Same.
36029         * doc/install.texi: Same.
36030         * doc/invoke.texi: Same.
36031         * doc/loop.texi: Same.
36032         * doc/md.texi: Same.
36033         * doc/objc.texi: Same.
36034         * doc/rtl.texi: Same.
36035         * doc/tm.texi: Same.
36036         * doc/tm.texi.in: Same.
36037         * doc/trouble.texi: Same.
36039 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36041         PR debug/63238
36042         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36043         (collect_checksum_attributes): Set it.
36044         (die_checksum_ordered): Use it.
36046 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36048         PR tree-optimization/79908
36049         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36050         change: For a VA_ARG whose LHS has been cast away, use
36051         force_gimple_operand to construct the side effects.
36053 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36055         PR translation/80001
36056         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36057         more amenable to translation.
36058         (oacc_loop_auto_partitions): Likewise.
36060 2017-03-21  Marek Polacek  <polacek@redhat.com>
36061             Martin Sebor  <msebor@redhat.com>
36063         PR tree-optimization/80109
36064         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36065         on INTEGRAL_TYPE_P.
36067 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36068             Segher Boessenkool  <segher@kernel.crashing.org>
36070         PR target/80125
36071         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36072         check reg_used_between_p between insn and one of succ or succ2
36073         depending on if succ is artificial insn not inserted into insn
36074         stream.
36076 2017-03-21  Martin Liska  <mliska@suse.cz>
36078         PR gcov-profile/80081
36079         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36080         * doc/gcc.texi: Include gcov-dump stuff.
36081         * doc/gcov-dump.texi: New file.
36083 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36085         PR rtl-optimization/79150
36086         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36087         conditional jump, if the jump is the last insn of the loop.
36089 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36090             Richard Biener  <rguenth@suse.de>
36092         PR tree-optimization/79908
36093         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36094         been cast away, use force_gimple_operand to construct the side
36095         effects.
36097 2017-03-21  Martin Liska  <mliska@suse.cz>
36099         PR libfortran/79956
36100         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36101         to NULL.
36103 2017-03-21  Brad Spengler <spender@grsecurity.net>
36105         PR plugins/80094
36106         * plugin.c (htab_hash_plugin): New function.
36107         (add_new_plugin): Use it and adjust.
36108         (parse_plugin_arg_opt): Adjust.
36109         (init_one_plugin): Likewise.
36111 2017-03-21  Richard Biener  <rguenther@suse.de>
36113         PR tree-optimization/80032
36114         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36115         if set force the cleanup to happen unconditionally.
36116         (gimplify_target_expr): Push inserted clobbers with force_uncond
36117         to avoid them being removed by control-dependent DCE.
36119 2017-03-21  Richard Biener  <rguenther@suse.de>
36121         PR tree-optimization/80122
36122         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36123         va_arg_pack_len when the inlined call stmt requires pack
36124         expansion itself.
36125         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36127 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36129         PR sanitizer/78158
36130         * tsan.c (instrument_builtin_call): If the memory model argument
36131         is not a constant, assume it is valid.
36133         PR c/67338
36134         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36135         avoid UB.
36137 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36139         PR rtl-optimization/79910
36140         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36141         if its dest is used by an insn before I2 (other than the combined
36142         insns themselves, which are properly handled already).
36144 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36146         Revert:
36147         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36149         * combine.c (record_used_regs): New static function.
36150         (try_combine): Handle situations where there is an additional
36151         instruction between I2 and I3 which needs to have a LOG_LINK
36152         updated.
36154         Revert:
36155         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36157         * combine.c (try_combine): Delete redundant i1 test.  Call
36158         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36160 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36162         PR target/80083
36163         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36164         alternatives 13/14.
36166 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36168         PR tree-optimization/80054
36169         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36170         the optimization if a PHI or any of its arguments is not dominated
36171         by the candidate's basis.  Use gphi* rather than gimple* as
36172         appropriate.
36173         (replace_profitable_candidates): Clean up a gimple* variable that
36174         should be a gphi* variable.
36176 2017-03-20  Martin Sebor  <msebor@redhat.com>
36178         PR c++/52477
36179         * doc/extend.texi (attribute constructor): Document present limitation.
36181 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36183         PR target/79963
36184         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36185         __POWER9_VECTOR__ #ifdef control, change template definition to
36186         use Power9-specific built-in function.
36187         (vec_any_eq): Likewise.
36188         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36189         to control outcomes from this test.
36190         (vector_ae_<mode>p): For VEC_F modes, likewise.
36192 2017-03-20  Ian Lance Taylor  <iant@google.com>
36194         * config/i386/i386.c (ix86_function_regparm): Save an extra
36195         register for -fsplit-stack with DECL_STATIC_CHAIN.
36197 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36199         PR target/79912
36200         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36201         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36203 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36205         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36206         iorw,ow".
36207         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36208         iorw,iorw".
36210 2017-03-20  Marek Polacek  <polacek@redhat.com>
36212         PR sanitizer/80063
36213         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36215 2017-03-20  Richard Biener  <rguenther@suse.de>
36217         PR tree-optimization/80113
36218         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36219         allocate extra SSA name for PHI def.
36220         (add_close_phis_to_outer_loops): Likewise.
36221         (add_close_phis_to_merge_points): Likewise.
36222         (copy_loop_close_phi_args): Likewise.
36223         (copy_cond_phi_nodes): Likewise.
36225 2017-03-20  Martin Liska  <mliska@suse.cz>
36227         PR middle-end/79753
36228         * tree-chkp.c (chkp_build_returned_bound): Do not build
36229         returned bounds for a LHS that's not a BOUNDED_P type.
36231 2017-03-20  Martin Liska  <mliska@suse.cz>
36233         PR target/79769
36234         PR target/79770
36235         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36236         COMPLEX_CST and VECTOR_CST.
36238 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36240         PR target/78857
36241         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36242         target operand.  A new splitter adds the clobber statement in case
36243         the target operand is dead anyway.
36245 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36247         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36248         to age-old versions of binutils and glibc.
36250 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36252         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36254 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36256         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36258 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36260         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36261         requirement for binutils 2.13.
36263 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36265         * combine.c (try_combine): Delete redundant i1 test.  Call
36266         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36268 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36270         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36271         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36272         contents.
36273         <riscv64-*-elf>: Re-arrange section
36274         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36275         <riscv32-*-linux>: Likewise.
36276         <riscv64-*-elf>: Likewise
36277         <riscv64-*-linux>: Likewise.
36279 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36281         PR target/80052
36282         * aarch64.opt(verbose-cost-dump): Fix typo.
36284 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36286         PR target/79951
36287         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36288         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36290 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36292         * reload.c (find_reloads): When reloading a nonoffsettable address,
36293         use RELOAD_OTHER for it and its address reloads.
36295         PR rtl-optimization/79910
36296         * combine.c (record_used_regs): New static function.
36297         (try_combine): Handle situations where there is an additional
36298         instruction between I2 and I3 which needs to have a LOG_LINK
36299         updated.
36301 2017-03-17  Jeff Law  <law@redhat.com>
36303         PR tree-optimization/71437
36304         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36305         conditional in the hash table first.
36306         (vrp_dom_walker::before_dom_children): Extract condition from
36307         ASSERT_EXPR.  Record condition, its inverion and any implied
36308         conditions as well.
36310 2017-03-17  Marek Polacek  <polacek@redhat.com>
36311             Markus Trippelsdorf  <markus@trippelsdorf.de>
36313         PR tree-optimization/80079
36314         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36315         m_stores_head.
36317 2017-03-17  Richard Biener  <rguenther@suse.de>
36319         PR middle-end/80075
36320         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36321         Properly verify the LHS before the RHS possibly claims to be
36322         handled.
36323         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36324         do not throw.
36326 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36328         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36329         (List of -O2 options): Likewise.
36330         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36331         (-fipa-vrp) New.
36333 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36335         * gcov-dump.c (print_usage): Print bug_report_url.
36337 2017-03-17  Richard Biener  <rguenther@suse.de>
36339         PR middle-end/80050
36340         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36341         (parser::peek): Likewise.
36343 2017-03-17  Richard Biener  <rguenther@suse.de>
36345         PR tree-optimization/80048
36346         * sese.c (free_sese_info): Properly release rename_map and
36347         copied_bb_map elements.
36349 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36351         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36352         Add linked-list forward and backlinks.  Insert on
36353         construction, remove on destruction.
36354         (class pass_store_merging): Add m_stores_head field.
36355         (pass_store_merging::terminate_and_process_all_chains):
36356         Iterate over m_stores_head list.
36357         (pass_store_merging::terminate_all_aliasing_chains):
36358         Likewise.
36359         (pass_store_merging::execute): Check for debug stmts first.
36360         Push new chains onto the m_stores_head stack.
36362 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36364         PR target/71294
36365         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36366         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36367         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36369 2017-03-16  Jeff Law  <law@redhat.com>
36371         PR tree-optimization/71437
36372         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36373         member function.  Implementation moved into after_dom_children
36374         member function and into the threader's thread_outgoing_edges
36375         function.
36376         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36377         some code into new thread_outgoing_edges.
36378         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36379         definition.  Simplify marker handling (do it here).   Assume we always
36380         have the available expression and the const/copies tables.
36381         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36382         and tree-vrp.c
36383         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36384         * tree-vrp.c (equiv_stack): No longer file scoped.
36385         (vrp_dom_walker): New class.
36386         (vrp_dom_walker::before_dom_children): New member function.
36387         (vrp_dom_walker::after_dom_children): Likewise.
36388         (identify_jump_threads):  Setup domwalker.  Use it rather than
36389         walking edges in a random order by hand.  Simplify setup/finalization.
36390         (finalize_jump_threads): Remove.
36391         (vrp_finalize): Do not call identify_jump_threads here.
36392         (execute_vrp): Do it here instead and call thread_through_all_blocks
36393         here too.
36395         PR tree-optimization/71437
36396         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36397         callers changed.
36398         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36399         callers changed.
36400         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36401         (dom_opt_dom_walker::thread_across_edge): Remove
36402         handle_dominating_asserts argument.  All callers changed.
36403         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36404         changes.  Remove calls to lhs_of_dominating_assert.  Other
36405         uses of handle_dominating_asserts turn into unconditional code
36406         (simplify_control_stmt_condition_1): Likewise.
36407         (simplify_control_stmt_condition): Likewise.
36408         (thread_through_normal_block, thread_across_edge): Likewise.
36409         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36410         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36411         object if it is not an SSA_NAME.
36412         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36413         before calling into the VRP specific simplifiers.
36414         (identify_jump_threads): Remove handle_dominating_asserts
36415         argument.
36417 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36419         PR fortran/79886
36420         * tree-diagnostic.c (default_tree_printer): No longer static.
36421         * tree-diagnostic.h (default_tree_printer): New prototype.
36423 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36425         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36426         Change ins into fmov.
36428 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36430         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36431         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36432         Use h_con constraint for operand 1.
36433         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36434         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36436 2017-03-15  Jeff Law  <law@redhat.com>
36438         PR tree-optimization/71437
36439         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36440         (record_temporary_equivalences): Use it.
36442         PR tree-optimization/71437
36443         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36444         tree-ssa-scopedtables.
36445         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36446         (record_conditions, record_cond, vuse_eq): Likewise.
36447         (record_edge_info): Adjust to API tweak of record_conditions.
36448         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36449         (record_temporary_equivalences, optimize_stmt): Likewise.
36450         (eliminate_redundant_computations): Likewise.
36451         (record_equivalences_from_stmt): Likewise.
36452         * tree-ssa-scopedtables.c: Include options.h and params.h.
36453         (vuse_eq): New function, moved from tree-ssa-dom.c
36454         (build_and_record_new_cond): Likewise.
36455         (record_conditions): Likewise.  Accept vector of conditions rather
36456         than edge_equivalence structure for first argument.
36457         for the first argument.
36458         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36459         from tree-ssa-dom.c.
36460         (avail_exprs_stack::record_cond): Likewise.
36461         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36462         from tree-ssa-dom.c.
36463         (avail_exprs_stack): Add new member functions lookup_avail_expr
36464         and record_cond.
36465         (record_conditions): Declare.
36467 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36469         PR target/80017
36470         * lra-constraints.c (process_alt_operands): Increase reject for
36471         reloading an input/output operand.
36473 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36475         PR target/79038
36476         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36477         insns to convert from signed/unsigned char/short to IEEE 128-bit
36478         floating point.
36479         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36481 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36483         PR target/80019
36484         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36485         subreg of inner mode for values already in registers.
36487 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36489         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36490         iteration reg is used after the loop.
36492 2017-03-14  Martin Sebor  <msebor@redhat.com>
36494         PR tree-optimization/79800
36495         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36496         precision in negative-positive range.
36497         (format_floating): Call non-const overload with adjusted precision.
36499 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36501         PR target/79947
36502         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36503         -mpowerpc-gfxopt.
36505 2017-03-14  Martin Sebor  <msebor@redhat.com>
36507         PR middle-end/80020
36508         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36509         * builtins.def (aligned_alloc): Use it.
36511         PR c/79936
36512         * Makefile.in (GTFILES): Add calls.c.
36513         * calls.c: Include "gt-calls.h".
36515 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36517         PR rtl-optimization/79728
36518         * regs.h (struct target_regs): New field
36519         x_contains_allocatable_regs_of_mode.
36520         (contains_allocatable_regs_of_mode): New macro.
36521         * reginfo.c (init_reg_sets_1): Initialize it, and change
36522         contains_reg_of_mode so it includes global regs as well.
36523         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36524         rather than contains_regs_of_mode.
36526 2017-03-14  Martin Liska  <mliska@suse.cz>
36528         * doc/invoke.texi: Document options that can't be combined with
36529         -fcheck-pointer-bounds.
36531 2017-03-14  Martin Liska  <mliska@suse.cz>
36533         PR middle-end/79831
36534         * doc/invoke.texi (-Wchkp): Document the option.
36536 2017-03-14  Martin Liska  <mliska@suse.cz>
36538         * Makefile.in: Install gcov-dump.
36540 2017-03-14  Martin Liska  <mliska@suse.cz>
36542         * multiple_target.c (expand_target_clones): Bail out for
36543         an invalid attribute.
36545 2017-03-14  Richard Biener  <rguenther@suse.de>
36547         * alias.c (struct alias_set_entry): Pack properly.
36548         * cfgloop.h (struct loop): Likewise.
36549         * cse.c (struct set): Likewise.
36550         * ipa-utils.c (struct searchc_env): Likewise.
36551         * loop-invariant.c (struct invariant): Likewise.
36552         * lra-remat.c (struct cand): Likewise.
36553         * recog.c (struct change_t): Likewise.
36554         * rtl.h (struct address_info): Likewise.
36555         * symbol-summary.h (function_summary): Likewise.
36556         * tree-loop-distribution.c (struct partition): Likewise.
36557         * tree-object-size.c (struct object_size_info): Likewise.
36558         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36559         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36560         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36561         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36562         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36563         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36564         (struct _stmt_vec_info): Likewise.
36566 2017-03-14  Martin Liska  <mliska@suse.cz>
36568         PR target/79892
36569         * multiple_target.c (create_dispatcher_calls): Check that
36570         a target can create a function dispatcher.
36572 2017-03-14  Martin Liska  <mliska@suse.cz>
36574         PR lto/66295
36575         * multiple_target.c (expand_target_clones): Drop local.local
36576         flag for default implementation.
36578 2017-03-14  Richard Biener  <rguenther@suse.de>
36580         PR tree-optimization/80030
36581         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36583 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36585         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36586         gcc_fallthrough() instead of __attribute__((fallthrough));
36588 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36590         * doc/gcc.texi: Remove "up" link to (DIR).
36591         * doc/gccint.texi: Ditto.
36593 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36595         * doc/install.texi (Specific) <avr>: Remove reference to
36596         binutils 2.13.
36598 2017-03-13  Jeff Law  <law@redhat.com>
36600         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36601         attribute rather than comments.
36603         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36604         match_scratch operand is highest.
36606 2017-03-13  Martin Liska  <mliska@suse.cz>
36608         PR middle-end/78339
36609         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36610         is a CHKP clone, use original declaration.
36612 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36614         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36615         (arc_conditional_register_usage): Use a different allocation order
36616         when optimizing for size.
36617         * common/config/arc/arc-common.c (arc_option_optimization_table):
36618         Section anchors default on when optimizing for size.
36620 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36622         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36624 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36626         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36627         * config/arc/arc.md (cpu_facility): Add cd variant.
36628         (*movqi_insn): Add code density variant.
36629         (*movhi_insn): Likewise.
36630         (*movqi_insn): Likewise.
36631         (*addsi3_mixed): Likewise.
36632         (subsi3_insn): Likewise.
36634 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36636         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36638 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36640         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36641         expressions with MINUS and UNARY ops.
36643 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36645         PR target/79911
36646         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36647         Rename to...
36648         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36649         between vec_select and vector argument.
36650         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36651         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36652         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36653         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36654         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36655         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36657 2017-03-13  Richard Biener  <rguenther@suse.de>
36659         PR other/79991
36660         * params.def (vect-max-peeling-for-alignment): Fix typo.
36662 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36664         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36665         issue that only occurred with binutils below 2.18.
36667 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36669         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36670         refer to binutils 2.11/2.12 minimum.
36672 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36674         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36675         ftp.kernel.org and simplify binutils requirement.
36677 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36679         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36680         optimization.
36681         (Optimize Options): Ditto.  Also remove redundancy.
36683 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36685         PR translation/79848
36686         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36687         "%qs".
36688         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36689         to G_ to avoid double translation.
36691 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36693         PR translation/79923
36694         * auto-profile.c (get_combined_location): Convert leading
36695         character of diagnostics to lower case and remove trailing period.
36696         (read_profile): Likewise for various diagnostics.
36697         * config/arm/arm.c (arm_option_override): Remove trailing period
36698         from various diagnostics.
36699         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36700         (msp430_expand_delay_cycles): Likewise.
36702 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36704         PR target/79925
36705         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36706         full command-line argument, rather than just "str".
36707         (aarch64_validate_march): Likewise.
36708         (aarch64_validate_mtune): Likewise.
36710 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36712         PR rtl-optimization/78911
36713         * lra-assigns.c (must_not_spill_p): New function.
36714         (spill_for): Use it.
36716 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36718         PR tree-optimization/79981
36719         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36720         ATOMIC_COMPARE_EXCHANGE ifn result.
36721         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36722         IFN_ATOMIC_COMPARE_EXCHANGE.
36724 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36726         PR driver/79875
36727         * opts.c (parse_sanitizer_options): Add missing question mark to
36728         "did you mean" message.
36730 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36732         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36733         built-in.
36734         (VMULEUH_UNS): Likewise.
36735         (VMULOUB_UNS): Likewise.
36736         (VMULOUH_UNS): Likewise.
36737         * config/rs6000/rs6000.c (builtin_function_type): Remove
36738         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36740 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36742         PR bootstrap/79952
36743         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36744         x with result of extra_parsing_for_operand_code_0.
36745         (function_reader::extra_parsing_for_operand_code_0): Convert
36746         return type from void to rtx, returning x.  When reading
36747         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36748         larger size containing struct block_symbol.
36750 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36752         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36753         -mfloat128-hardware without -m64.
36755 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36757         PR target/79941
36758         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36759         entries to the case statement that marks unsigned arguments to
36760         overloaded functions.
36762 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36764         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36765         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36767 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36769         PR target/79907
36770         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36771         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36773 2017-03-10  Martin Liska  <mliska@suse.cz>
36775         PR target/65705
36776         PR target/69804
36777         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36778         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36779         FIELD != NULL.
36781 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36783         * tree-switch-conversion (array_value_type): Start by resetting
36784         candidate type to it's main variant.
36786 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36788         PR rtl-optimization/79909
36789         * combine.c (try_combine): Use simplify_replace_rtx on individual
36790         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36791         of the whole CALL_INSN_FUNCTION_USAGE.
36793         PR tree-optimization/79972
36794         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36795         get_range_info on SSA_NAMEs.  Formatting fixes.
36797 2017-03-10  Richard Biener  <rguenther@suse.de>
36798             Jakub Jelinek  <jakub@redhat.com>
36800         PR tree-optimization/77975
36801         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36802         edge to be constant.
36803         (get_val_for): For constant x return it.  Formatting fix.
36804         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36805         would use the same bases as the current one.
36807 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36809         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36810         instead of vec_select for V1TImode.
36811         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36812         longer needed.
36813         (VSX_LE_128): Add V1TI to this mode iterator.
36814         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36815         (*vsx_le_perm_store_<mode>): Likewise.
36816         (pre-reload splitter for VSX stores): Likewise.
36817         (post-reload splitter for VSX stores): Likewise.
36818         (*vsx_xxpermdi2_le_<mode>): Likewise.
36819         (*vsx_lxvd2x2_le_<mode>): Likewise.
36820         (*vsx_stxvd2x2_le_<mode>): Likewise.
36822 2017-03-09  Michael Eager  <eager@eagercon.com>
36824         Correct failures with --enable-checking=yes,rtl.
36826         * config/microblaze/microblaze.c (microblaze_expand_shift):
36827         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36828         test for const0_rtx.
36829         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36830         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36832 2017-03-09  Richard Biener  <rguenther@suse.de>
36834         PR tree-optimization/79977
36835         * graphite-scop-detection.c (scop_detection::merge_sese):
36836         Handle the case of extra exits to blocks dominating the entry.
36838 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36840         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36841         Document rdynamic.
36843 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36845         PR rtl-optimization/79949
36846         * lra-constraints.c (process_alt_operands): Check memory when
36847         trying to predict a cycle.  Print about the overall increase.
36849 2017-03-09  Richard Biener  <rguenther@suse.de>
36851         PR middle-end/79971
36852         * gimple-expr.c (useless_type_conversion_p): Preserve
36853         TYPE_SATURATING for fixed-point types.
36855 2017-03-09  Richard Biener  <rguenther@suse.de>
36857         PR ipa/79970
36858         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36859         alignment of BLKmode params.
36861 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36863         PR target/79913
36864         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36865         (VALL_NO_V2Q): Likewise.
36866         (VDQF_DF): Delete.
36867         * config/aarch64/aarch64-simd.md
36868         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36869         iterator.
36870         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36871         VALL_NO_V2Q mode iterator.
36872         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36874 2017-03-09  Martin Liska  <mliska@suse.cz>
36876         PR tree-optimization/79631
36877         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36878         tree_int_cst_sign_bit just for INTEGER constants.
36880 2017-03-09  Martin Liska  <mliska@suse.cz>
36882         PR target/65705
36883         PR target/69804
36884         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36885         sanitizers.
36887 2017-03-09  Marek Polacek  <polacek@redhat.com>
36889         PR c++/79672
36890         * tree.c (inchash::add_expr): Handle TREE_VEC.
36892 2017-03-09  Martin Liska  <mliska@suse.cz>
36894         PR ipa/79764
36895         (chkp_narrow_size_and_offset): New function.
36896         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36897         (void chkp_parse_bit_field_ref): New function.
36898         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36899         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36901 2017-03-09  Martin Liska  <mliska@suse.cz>
36903         PR ipa/79761
36904         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36905         (chkp_find_bounds_1): Remove gcc_unreachable.
36907 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36909         PR sanitizer/79944
36910         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36911         BUILT_IN_SYNC*, determine the access type from the size suffix and
36912         always build a MEM_REF with that type.  Handle forgotten
36913         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36915         PR target/79932
36916         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36917         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36918         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36919         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36920         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36921         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36922         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36923         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36924         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36925         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36926         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36927         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36928         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36929         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36930         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36931         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36932         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36933         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36934         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36935         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36936         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36937         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36938         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36939         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36940         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
36941         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
36942         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
36943         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
36944         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
36945         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
36946         definitions outside of __OPTIMIZE__ guarded section.
36948         PR target/79932
36949         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
36950         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
36951         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
36952         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
36953         guarded section.
36955 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36957         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
36958         ("vfenez<mode>"): Add missing constraints.
36960 2017-03-08  Martin Sebor  <msebor@redhat.com>
36962         PR target/79928
36963         * config/nds32/nds32.c (nds32_option_override):
36964         Fix misspelled diagnostic.
36966 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36968         PR c/79940
36969         * gimplify.c (gimplify_omp_for): Replace index var in outer
36970         taskloop statement with an artificial variable and add
36971         OMP_CLAUSE_PRIVATE clause for it.
36973 2017-03-08  Richard Biener  <rguenther@suse.de>
36975         PR tree-optimization/79955
36976         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
36977         for accesses that are completely outside of the variable.
36979 2017-03-08  Andrew Haley  <aph@redhat.com>
36981         PR tree-optimization/79943
36982         * tree-ssa-loop-split.c (compute_new_first_bound): When
36983         calculating the new upper bound, (END-BEG) should be added, not
36984         subtracted.
36986 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36988         * config/avr/avr.md (setmemhi): Make sure match_dup
36989         operand number comes before match_scratch.
36991 2017-03-08  Richard Biener  <rguenther@suse.de>
36993         PR tree-optimization/79920
36994         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
36995         with ncopies == 1 to ...
36996         (vect_transform_slp_perm_load): ... here.  Properly compute
36997         all element loads by iterating VF times over the group.  Do
36998         not handle ncopies (computed in a broken way) in
36999         vect_create_mask_and_perm.
37001 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37003         PR sanitizer/79904
37004         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
37005         is a uniform vector, use uniform_vector_p return value instead of
37006         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
37008 2017-03-07  Marek Polacek  <polacek@redhat.com>
37010         PR middle-end/79809
37011         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
37012         (alloca_call_type): Likewise.
37014 2017-03-07  Martin Liska  <mliska@suse.cz>
37016         * gcov.c (process_args): Put comment to correct location.
37018 2017-03-07  Martin Liska  <mliska@suse.cz>
37020         PR middle-end/68270
37021         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
37022         Use array_at_struct_end_p instead of DECL_CHAIN (field).
37023         (chkp_narrow_bounds_for_field): Likewise.
37024         (chkp_parse_array_and_component_ref): Pass one more argument to
37025         call.
37027 2017-03-07  Richard Biener  <rguenther@suse.de>
37029         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37030         preheaders.
37032 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37034         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37035         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37037 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37039         PR c/79855
37040         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37041         to end of description.
37042         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37044 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37046         PR rtl-optimization/79901
37047         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37048         ...
37049         (*avx512f_<code><mode>3<mask_name>): ... this.
37050         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37051         iterator instead of VI8_AVX2_AVX512BW.
37053         PR rtl-optimization/79901
37054         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37055         min/max expander, expand it using expand_vec_cond_expr.
37057         PR sanitizer/79897
37058         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37059         temporary.
37061 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37063         PR c++/79821
37064         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37065         to void * for PCH reasons.
37066         * dwarf2out.c (output_loc_operands, output_die): Cast
37067         v.val_vec.array to unsigned char *.
37069 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37071         PR target/77850
37072         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37073         vector types.
37075 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37077         PR rtl-optimization/79571
37078         * lra-constraints.c (process_alt_operands): Calculate static
37079         reject and subtract it from overall when only addresses will be
37080         reloaded.
37082 2017-03-06  Julia Koval  <julia.koval@intel.com>
37084         PR target/79793
37085         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37086         incoming stack boundary to 128 for 64-bit targets.
37088 2017-03-06  Richard Biener  <rguenther@suse.de>
37090         PR tree-optimization/79894
37091         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37092         to NULL after folding it.
37094 2017-03-06  Richard Biener  <rguenther@suse.de>
37096         PR tree-optimization/79824
37097         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37098         check disabling peeling for gaps.
37100 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37102         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37103         attributes): Document gettimeofday.
37105 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37107         * config/s390/s390.c (s390_option_override_internal): Set
37108         PARAM_MIN_VECT_LOOP_BOUND
37110 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37112         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37113         * config/s390/s390.md: Likewise.
37115 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37117         PR target/79812
37118         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37119         (<avx2_avx512>_perm<mode>): Rename to ...
37120         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37121         of VI8F_256_512.
37122         (<avx512>_perm<mode>_mask): Rename to ...
37123         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37124         of VI8F_256_512.
37125         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37126         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37127         instead of VI8F_256_512.
37128         (avx512f_perm<mode>): New define_expand.
37129         (avx512f_perm<mode>_mask): Likewise.
37130         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37131         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37133 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37135         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37136         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37137         if_then_else.
37138         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37140 2017-03-06  Martin Liska  <mliska@suse.cz>
37142         PR sanitize/79783
37143         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37144         when having a SSA NAME w/o VAR_DECL assigned to it.
37146 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37148         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37149         msa_dpsub_<su>_d): Fix MODE for vec_select.
37151 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37153         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37154         argument.
37155         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37157 2017-03-06  Richard Biener  <rguenther@suse.de>
37159         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37160         * plugin.c (register_plugin_info): Likewise.
37161         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37163 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37165         * config/i386/sse.md (sse_storehps, sse_storelps,
37166         avx_<castmode><avxsizesuffix>_<castmode>,
37167         avx512f_<castmode><avxsizesuffix>_<castmode>,
37168         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37169         in condition that at least one operand is not a MEM.
37171 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37173         PR middle-end/79805
37174         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37175         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37176         ECF_NOTHROW.
37177         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37178         gimple_call_nothrow_p flag based on whether original builtin can throw.
37179         If it can, emit following stmts on the fallthrough edge.
37180         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37181         don't create new bb if inserting just debug stmts on the edge, try to
37182         insert them on the fallthru bb or just reset debug stmts.
37184 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37186         PR target/43763
37187         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37188         restore recog_data (including the operand rtxes inside it) around
37189         the call to get_insn_template.
37191 2017-03-03  Martin Sebor  <msebor@redhat.com>
37193         PR tree-optimization/79699
37194         * context.c (context::~context): Free MPFR caches to avoid
37195         a memory leak on program exit.
37197 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37199         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37200         Use wide_int::ulow () instead of .elt (0).
37202 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37204         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37205         (*pushxf): Limit oF constraint to 32bit targets and add oC
37206         constraint for 64bit targets.
37207         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37208         (*pushdf): Change rmF constraint to rmC.
37210 2017-03-03  Martin Liska  <mliska@suse.cz>
37212         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37213         Remove unused variable.
37215 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37217         PR target/79807
37218         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37219         is a memory operand, increase num_memory.
37220         (ix86_expand_args_builtin): Likewise.
37222 2017-03-03  Jan Hubicka  <jh@suse.cz>
37224         PR lto/79760
37225         * ipa-devirt.c (maybe_record_node): Properly handle
37226         __cxa_pure_virtual visibility.
37228 2017-03-03  Martin Liska  <mliska@suse.cz>
37230         PR tree-optimization/79803
37231         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37232         assert.
37233         (pass_loop_prefetch::execute): Disabled optimization if an
37234         assumption about L1 cache size is not met.
37236 2017-03-03  Martin Liska  <mliska@suse.cz>
37238         PR rtl-optimization/79574
37239         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37240         (hash_scan_set): Likewise.
37241         (dump_hash_table): Likewise.
37242         (hoist_code): Likewise.
37244 2017-03-03  Richard Biener  <rguenther@suse.de>
37246         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37247         in place of uhigh ().
37248         (fixed_convert_from_real): Likewise.
37250 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37252         PR target/79514
37253         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37255 2017-03-03  Richard Biener  <rguenther@suse.de>
37257         PR middle-end/79818
37258         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37259         TYPE_OVERFLOW_UNDEFINED check.
37261 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37263         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37264         numbers.
37265         (vector_ae_<mode>_p): Likewise.
37266         (vector_nez_<mode>_p): Likewise.
37267         (vector_ne_v2di_p): Likewise.
37268         (vector_ae_v2di_p): Likewise.
37269         (vector_ne_<mode>_p): Likewise.
37270         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37271         numbers.
37272         (vsx_tsqrt<mode>2_fe): Likewise.
37274 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37276         PR target/79514
37277         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37279 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37281         PR rtl-optimization/79780
37282         * cprop.c (one_cprop_pass): When second and further conditional trap
37283         in a single basic block is turned into an unconditional trap, turn it
37284         into a deleted note to avoid RTL verification failures.
37286 2017-03-02  Richard Biener  <rguenther@suse.de>
37288         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37290 2017-03-02  Richard Biener  <rguenther@suse.de>
37292         PR tree-optimization/79345
37293         PR c++/42000
37294         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37295         param and abort the walk, returning -1 if it is hit.
37296         (walk_aliased_vdefs): Take a limit param and pass it on.
37297         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37298         defaulting to 0 and return a signed int.
37299         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37300         (check_defs): New helper.
37301         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37302         about uninitialized memory.
37303         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37304         bogus uninitialized warning.
37305         (fixed_convert_from_real): Likewise.
37307 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37309         PR tree-optimization/66768
37310         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37311         iv_use if base object can't be determined.
37313 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37315         PR tree-optimization/79345
37316         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37317         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37318         (get_pattern_stats): Initialize it.
37319         * genemit.c (gen_expand): Verify match_scratch numbers come after
37320         match_operand/match_dup numbers.
37321         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37322         match_scratch numbers.
37323         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37324         Likewise.
37325         * config/s390/s390.md (trunctdsd2): Likewise.
37327 2017-03-02  Richard Biener  <rguenther@suse.de>
37329         * wide-int.h (wide_int_storage::operator=): Implement in terms
37330         of wi::copy.
37332 2017-03-02  Richard Biener  <rguenther@suse.de>
37334         PR tree-optimization/79777
37335         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37336         the to insert expression to sth existing.
37338 2017-03-01  Martin Sebor  <msebor@redhat.com>
37340         PR middle-end/79692
37341         * gimple-ssa-sprintf.c
37342         (directive::known_width_and_precision): New function.
37343         (format_integer): Use it.
37344         (get_mpfr_format_length): Consider the full range of precision
37345         when computing %g output with the # flag.  Set the likely byte
37346         count to 3 rather than 1 when precision is indeterminate.
37347         (format_floating): Correct the lower bound of precision.
37349 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37351         * doc/invoke.texi: Document default code model for 64-bit Linux.
37353 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37355         PR target/79752
37356         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37357         udiv rather than div since input pattern is unsigned.
37359 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37361         * config/i386/i386.c (print_reg): Warn for values of
37362         unsupported size in integer register.
37364 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37366         PR target/79439
37367         * config/rs6000/predicates.md (current_file_function_operand): Do
37368         not allow self calls to be local if the function is replaceable.
37370 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37372         PR target/79395
37373         * config/rs6000/altivec.h (vec_ctz and others): Change the
37374         preprocessor macro that controls conditional compilation from
37375         _ARCH_PWR9 to __POWER9_VECTOR__.
37376         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37377         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37378         control (instead of _ARCH_PWR9 control) so that template
37379         definition uses power9-specific function.
37380         (vec_any_eq): Likewise.
37381         (vec_all_ne): Change macro definition to use a power9-specific
37382         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37383         _ARCH_PWR9 control).
37384         (vec_any_eq) Likewise.
37385         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37386         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37387         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37388         support for xvcmpnedp instruction.
37389         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37390         macro expansion so that Power9 implementation of vec_all_ne does
37391         not use the AltiVec predicate framework.
37392         (VCMPNEH_P): Likewise.
37393         (VCMPNEW_P): Likewise.
37394         (VCMPNED_P): Likewise.
37395         (VCMPNEFP_P): Likewise.
37396         (VCMPNEDP_P): Likewise.
37397         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37398         implementation of vec_any_eq to not use AltiVec predicate
37399         framework.
37400         (VCMPAEH_P): Likewise.
37401         (VCMPAEW_P): Likewise.
37402         (VCMPAED_P): Likewise.
37403         (VCMPAEFP_P): Likewise.
37404         (VCMPAEDP_P): Likewise.
37405         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37406         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37407         not use the AltiVec predicate framework.
37408         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37409         of vec_any_eq to not use AltiVec predicate framework.
37410         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37411         support for predefined __POWER9_VECTOR__ macro to indicate that
37412         Power9 instruction selection is enabled.
37413         (altivec_overloaded_builtins): Remove extraneous
37414         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37415         function argument types RS6000_BTI_bool_V16QI and
37416         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37417         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37418         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37419         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37420         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37421         Power9 for implementations of vec_cmpne.  Change the signature for
37422         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37423         (representing vec_all_ne) to remove the previously described first
37424         argument of type RS6000_BTI_INTSI, as this was an artifact of
37425         reliance on the AltiVec predicate framework, which is no longer
37426         used in the implementation of these functions.  Add
37427         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37428         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37429         since, unlike the AltiVec predicate framework implementation, we
37430         do not share function descriptors between vec_alle and vec_anyeq.
37431         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37432         set of modes that receive special treatment even when
37433         TARGET_P9_VECTOR is true.  The special treatment emits code that
37434         does not depend on Power9 instructions.
37435         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37436         define_expand to not rely on AltiVec predicate framework.
37437         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37438         function.
37439         (vector_ne_v2di_p): Change this define_expand to not rely on
37440         AltiVec predicate framework.
37441         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37442         function.
37443         (vector_ne_<mode>_p): Change this define_expand to not rely on
37444         AltiVec predicate framework.
37445         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37446         function.
37447         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37448         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37449         define_insn pattern.
37450         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37451         define_insn pattern because the xvcmpne<VSs>. instruction is not
37452         supported.
37453         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37454         instruction is not supported.
37456 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37458         * config/nvptx/nvptx.c: Include intl.h.
37460 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37462         PR lto/78140
37463         * ipa-prop.h (ipa_bits): Removed field known.
37464         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37465         to pointers.  Adjusted their comments to warn about their sharing.
37466         (ipcp_transformation_summary): Change bits to a vector of pointers.
37467         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37468         (ipa_get_ipa_bits_for_value): Declare.
37469         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37470         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37471         (ipa_bits_hash_table): Likewise.
37472         (ipa_vr_ggc_hash_traits): Likewise.
37473         (ipa_vr_hash_table): Likewise.
37474         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37475         being pointers and vr_known being removed.
37476         (ipa_set_jf_unknown): Likewise.
37477         (ipa_get_ipa_bits_for_value): New function.
37478         (ipa_set_jfunc_bits): Likewise.
37479         (ipa_get_value_range): New overloaded functions.
37480         (ipa_set_jfunc_vr): Likewise.
37481         (ipa_compute_jump_functions_for_edge): Use the above functions to
37482         construct bits and vr parts of jump functions.
37483         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37484         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37485         exist.
37486         (ipcp_grow_transformations_if_necessary): Also allocate
37487         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37488         exist.
37489         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37490         them.  Fix too long lines.
37491         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37492         vr_known being removed.
37493         (ipa_read_jump_function): Use new setter functions to construct bits
37494         and vr parts of jump functions or set them to NULL.
37495         (write_ipcp_transformation_info): Adjust for bits being pointers.
37496         (read_ipcp_transformation_info): Likewise.
37497         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37498         space.
37499         Include gt-ipa-prop.h.
37500         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37501         being pointers.
37502         (ipcp_store_bits_results): Likewise.
37503         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37504         Do not write to existing jump functions but use a temporary instead.
37506 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37508         PR c++/79681
37509         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37510         attempt to use its first operand as BIT_FIELD_REF base.
37512 2017-03-01  Richard Biener  <rguenther@suse.de>
37514         PR middle-end/79721
37515         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37516         interpolating formula in wrapping arithmetic.
37517         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37519 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37521         PR tree-optimization/79734
37522         * tree-vect-generic.c (expand_vector_condition): Optimize
37523         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37524         Handle VEC_COND_EXPR where comparison has different inner width from
37525         type's inner width.
37527 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37529         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37530         markup, and similar issues.  Remove @opindex entries for things
37531         that aren't options.  Add missing -mmpy-option entries.
37533 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37535         PR tree-optimization/79737
37536         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37537         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37538         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37539         instead of byte_size.  Formatting fix.
37540         (shift_bytes_in_array_right): Formatting fix.
37542 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37544         PR target/79749
37545         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37546         condition on optimize for the leaf function test.
37548 2017-02-28  Martin Liska  <mliska@suse.cz>
37550         PR lto/79625
37551         * read-rtl-function.c (function_reader::handle_unknown_directive):
37552         Bail out when one uses -flto.
37554 2017-02-28  Martin Liska  <mliska@suse.cz>
37556         * common.opt: Replace space with tabular for options of <number>
37557         type.
37558         * config/i386/i386.opt: Show <number> value for
37559         -mlarge-data-threshold.
37560         * opts.c (print_filtered_help): Do not display number in hexadecimal
37561         format.
37563 2017-02-28  Martin Liska  <mliska@suse.cz>
37565         * common.opt: Fix --help=option -Q for options which are of
37566         an enum type.
37568 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37570         * config/i386/i386.c (print_reg): Error out for values
37571         of 8-bit size in invalid integer register.
37573 2017-02-28  Martin Sebor  <msebor@redhat.com>
37575         PR tree-optimization/79691
37576         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37578 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37580         PR target/79729
37581         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37582         gcc_unreachable with output_operand_lossage.
37584 2017-02-28  Richard Biener  <rguenther@suse.de>
37586         PR tree-optimization/79740
37587         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37588         inserts.
37589         (visit_nary_op): Insert the nary into the hashtable if we
37590         pattern-matched sth.
37591         * tree-ssa-pre.c (eliminate_insert): Robustify.
37593 2017-02-28  Richard Biener  <rguenther@suse.de>
37595         PR middle-end/79731
37596         * fold-const.c (decode_field_reference): Reject out-of-bound
37597         accesses.
37599 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37601         * config/i386/i386.c: Include intl.h.
37602         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37603         instead of just cond ? "..." : "...".
37604         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37605         * coverage.c (read_counts_file): Likewise.
37606         * omp-offload.c: Include intl.h.
37607         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37608         of just cond ? "..." : "...".
37609         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37610         of just cond ? "..." : "...".
37612 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37614         PR target/79742
37615         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37616         entry, if present.
37617         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37618         'tune for' CPU name.
37619         * config/arm/arm-cpu-data.h: Regenerated.
37621 2017-02-28  Richard Biener  <rguenther@suse.de>
37623         PR tree-optimization/79732
37624         * tree-inline.c (expand_call_inline): Do not shadow var.
37626 2017-02-28  Richard Biener  <rguenther@suse.de>
37628         PR tree-optimization/79723
37629         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37630         address-space properly.
37632 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37634         * doc/optinfo.texi (Optimization groups): Fix option used for
37635         OPTGROUP_ALL.
37636         * doc/invoke.texi (-fopt-info): Document "omp".
37637         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37638         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37639         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37640         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37641         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37643         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37644         all users.
37645         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37646         OPTGROUP_OMP with "omp".
37648 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37650         PR target/79544
37651         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37652         for arithmetic shift of unsigned V2DI.
37654 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37656         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37657         arc/linux.h headers.
37658         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37659         (LINK_SPEC): Likewise.
37660         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37661         (EXTRA_SPECS): Likewise.
37662         (STARTFILE_SPEC): Likewise.
37663         (ENDFILE_SPEC): Likewise.
37664         (LIB_SPEC): Likewise.
37665         (TARGET_SDATA_DEFAULT): Likewise.
37666         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37667         (MULTILIB_DEFAULTS): Likewise.
37668         (DWARF2_UNWIND_INFO): Likewise.
37669         * config/arc/big.h: New file.
37670         * config/arc/elf.h: Likewise.
37671         * config/arc/linux.h: Likewise.
37672         * config/arc/t-uClibc: Remove.
37674 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37676         PR tree-optimization/77536
37677         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37678         (tree_transform_and_unroll_loop): Use above function to compute the
37679         estimated niter of unrolled loop and use it when scaling profile.
37680         Also use count info rather than frequency if it's non-zero.
37681         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37682         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37683         (vect_transform_loop): Call above function.
37685 2017-02-27  Richard Biener  <rguenther@suse.de>
37687         PR tree-optimization/45397
37688         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37689         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37690         (visit_nary_op): Add pattern matching for CSEing sign-changed
37691         or truncated operations with wider ones.
37693 2017-02-27  Richard Biener  <rguenther@suse.de>
37695         PR tree-optimization/79690
37696         * tree-vect-stmts.c (vectorizable_store): Use vector type
37697         built from the DR with address-space.
37699 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37701         * doc/invoke.texi (Optimize Options): Refine the description
37702         of asan-use-after-return.
37704 2017-02-25  Alan Modra  <amodra@gmail.com>
37706         PR rtl-optimization/79584
37707         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37708         base, not ad->base_term, the reg within base.  Remove assertion
37709         that ad->base == ad->base_term.  Replace gen_int_mode using
37710         bogus mode with const0_rtx.
37712 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37714         PR middle-end/79396
37715         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37716         FMA_EXPR like tcc_binary or tcc_unary.
37718         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37720         PR debug/77589
37721         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37722         bitfield.
37723         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37724         (output_loc_operands): Handle DW_OP_call_ref and
37725         DW_OP_GNU_variable_value.
37726         (struct variable_value_struct): New type.
37727         (struct variable_value_hasher): Likewise.
37728         (variable_value_hash): New variable.
37729         (string_types): Remove.
37730         (copy_loc_descr): New function.
37731         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37732         (prepend_loc_descr_to_each): New function.
37733         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37734         instead of add_loc_descr_to_each if the first argument is single
37735         location list and the second has multiple.
37736         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37737         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37738         when looking for variable value which doesn't have other location info.
37739         (loc_list_from_tree): Formatting fix.
37740         (gen_array_type_die): Simplify DW_AT_string_length handling.
37741         (adjust_string_types): Remove.
37742         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37743         string_types.  Call resolve_variable_values.
37744         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37745         (resolve_addr_in_expr): Likewise.  Add A argument.
37746         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37747         original expression being DW_OP_GNU_variable_value with optionally
37748         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37749         optionally after it.
37750         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37751         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37752         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37753         (variable_value_hasher::hash, variable_value_hasher::equal): New
37754         methods.
37755         (resolve_variable_value_in_expr, resolve_variable_value,
37756         resolve_variable_values, note_variable_value_in_expr,
37757         note_variable_value): New functions.
37758         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37759         DIEs.
37761 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37763         PR c/79677
37764         * opts.h (handle_generated_option): Add GENERATED_P argument.
37765         * opts-common.c (handle_option): Adjust function comment.
37766         (handle_generated_option): Add GENERATED_P argument, pass it to
37767         handle_option.
37768         (control_warning_option): Pass false to handle_generated_option
37769         GENERATED_P.
37770         * opts.c (maybe_default_option): Pass true to handle_generated_option
37771         GENERATED_P.
37772         * optc-gen.awk: Likewise.
37774 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37776         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37777         a REG, look at the REG it is a SUBREG of.
37778         (splitter for cmpeqsi_t): Ditto.
37780 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37782         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37783         the special USEs with the pattern of the insn, not the insn itself.
37785 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37787         PR target/79473
37788         * doc/invoke.texi: Document -mload-store-pairs.
37790 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37791             Sandra Loosemore  <sandra@codesourcery.com>
37793         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37794         argument isn't a CONST_INT.
37795         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37796         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37797         (nios2_validate_compare): Bypass alternate compare logic if *op2
37798         is not a CONST_INT.
37799         (ldstwm_operation_p): Return false if first_base is not a REG or
37800         if first_offset is not a CONST_INT.
37802 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37804         * config/cris/cris.md: Use correct operand in a define_peephole2.
37806 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37808         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37810 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37812         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37813         this_insn if it is an INSN or JUMP_INSN.
37814         (force_offsettable): Look at base, not at addr.
37815         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37816         on things that aren't necessarily CONST_INTs.
37818 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37820         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37821         -mfpmath=sse is the default also for x86-32 targets with SSE2
37822         instruction set when @option{-ffast-math} is enabled
37824 2017-02-24  Jeff Law  <law@redhat.com>
37826         PR rtl-optimizatoin/79286
37827         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37828         dominance test.
37830 2017-02-24  Richard Biener  <rguenther@suse.de>
37832         PR tree-optimization/79389
37833         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37834         debug insns.
37836 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37838         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37839         function comment to reflect reality.
37840         (loop_exits_before_overflow): Fix typo in function description.
37842 2017-02-24  Richard Biener  <rguenther@suse.de>
37844         PR tree-optimization/79389
37845         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37846         properly that a threading opportunity exists.  Detect conditional
37847         copy/constant propagation opportunities.
37849 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37851         * config/visium/visium.md (type): Add trap.
37852         (b): New mode attribute.
37853         (*btst): Rename into...
37854         (*btst<mode>): ...this and adjust.
37855         (*cbranchsi4_btst_insn): Rename into...
37856         (*cbranch<mode>4_btst_insn): ...this and adjust.
37857         (trap): New define_insn.
37859 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37861         PR tree-optimization/79389
37862         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37863         (noce_reversed_cond_code): New function.
37864         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37865         reversed_comparison_code.  Formatting fix.
37866         (noce_try_store_flag): Test rev_cond != NULL in addition to
37867         reversed_comparison_code.
37868         (noce_try_store_flag_constants): Likewise.
37869         (noce_try_store_flag_mask): Likewise.
37870         (noce_try_addcc): Use rev_cond if non-NULL instead of
37871         reversed_comparison_code.
37872         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37873         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37874         (noce_find_if_block): Initialize rev_cond.
37875         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37876         instead of false as last argument never attempt to reverse it
37877         afterwards.
37879 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37881         PR tree-optimization/79663
37882         * tree-predcom.c (combine_chains): Process refs in reverse order
37883         only for ZERO length chains, and add explaining comment.
37885 2017-02-23  Jeff Law  <law@redhat.com>
37887         PR tree-optimization/79578
37888         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37889         in call to operand_equal_p.
37891 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37893         PR target/71017
37894         * config/i386/cpuid.h: Fix another undefined behavior.
37896 2017-02-23  Richard Biener  <rguenther@suse.de>
37898         PR tree-optimization/79683
37899         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37900         vector types for data-refs.
37902 2017-02-23  Martin Liska  <mliska@suse.cz>
37904         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37906 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37908         PR middle-end/79665
37909         * internal-fn.c (get_range_pos_neg): Moved to ...
37910         * tree.c (get_range_pos_neg): ... here.  No longer static.
37911         * tree.h (get_range_pos_neg): New prototype.
37912         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37913         are known to be in between 0 and signed maximum inclusive, try to
37914         expand both unsigned and signed divmod and use the cheaper one from
37915         those.
37917 2017-02-22  Jeff Law  <law@redhat.com>
37919         PR tree-optimization/79578
37920         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37921         to compare base operands.
37923 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37925         PR target/79211
37926         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37927         gpc_reg_operand instead of fpr_reg_operand.
37929 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
37931         * config/mips/mips.c (mips_return_in_memory): Force FP
37932         vector types to be returned in memory for o32 ABI.
37934 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37936         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37937         instead of DW_TAG_member for static data member declarations and don't
37938         set no_linkage_name for static inline data members.
37939         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37940         to DW_TAG_member.
37942 2017-02-22  Martin Liska  <mliska@suse.cz>
37944         * doc/invoke.texi: Replace inequality signs with square brackets
37945         for -Wnormalized.
37947 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
37949         PR target/78660
37950         * lra-constraints.c (simplify_operand_subreg): Handle
37951         WORD_REGISTER_OPERATIONS targets.
37953 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37955         PR target/70465
37956         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
37957         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
37958         elimination by swapping fld*.
37960 2017-02-22  Richard Biener  <rguenther@suse.de>
37962         PR tree-optimization/79673
37963         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
37964         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
37965         irrelevant address-space qualifiers and avoiding a
37966         ADDR_SPACE_CONVERT_EXPR from fold_convert.
37968 2017-02-22  Richard Biener  <rguenther@suse.de>
37970         PR tree-optimization/79666
37971         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
37972         to not symbolically negate if that may introduce undefined
37973         overflow.
37975 2017-02-22  Martin Liska  <mliska@suse.cz>
37977         PR lto/79587
37978         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
37979         * data-streamer-out.c (streamer_write_gcov_count_stream):
37980         Likewise.
37981         * value-prof.c (stream_out_histogram_value): Make assert more
37982         precise based on type of counter.
37984 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
37986         PR target/79593
37987         * config/i386/i386.md (standard_x87sse_constant_load splitter):
37988         Use nonimmediate_operand instead of memory_operand for operand 1.
37989         (float-extend standard_x87sse_constant_load splitter): Ditto.
37991 2017-02-21  Jeff Law  <law@redhat.com>
37993         PR tree-optimization/79621
37994         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
37995         blocks with edges to themselves.
37997 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37999         PR target/79633
38000         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
38001         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
38002         Use gimple_call_builtin_p.
38004         PR target/79570
38005         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
38006         on temporarily removed DEBUG_INSNs.
38008         PR tree-optimization/79649
38009         * tree-loop-distribution.c (classify_partition): Give up on
38010         non-generic address space loads/stores.
38012 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
38014         * doc/loop.texi (Loop manipulation): Remove nonexistent
38015         tree_ssa_loop_version from the documentation.
38016         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
38018 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38020         PR target/79494
38021         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
38022         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
38023         * config/rs6000/rs6000.c: Include except.h.
38024         (rs6000_expand_split_stack_prologue): Call
38025         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38027 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38029         PR lto/79579
38030         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38031         have been analyzed.
38033 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38035         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38036         for backward compatibility only.
38037         * doc/invoke.texi (Option Summary): Remove all references to
38038         -fipa-cp-alignment.
38040 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38042         PR target/78660
38043         Revert:
38044         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38046         * lra-constraints.c (curr_insn_transform): Handle
38047         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38049 2017-02-21  Martin Liska  <mliska@suse.cz>
38051         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38053 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38055         PR translation/79638
38056         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38058 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38060         PR ada/67205
38061         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38062         (arm_function_ok_for_sibcall): Return false for an indirect call by
38063         descriptor if all the argument registers are used.
38064         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38065         alignment of the function.
38067 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38069         PR tree-optimization/61441
38070         * simplify-rtx.c (simplify_const_unary_operation): For
38071         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38072         the sNaN unmodified.
38074 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38076         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38077         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38078         instead of SYSTEM_HEADER_DIR.
38080 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38081             Martin LiÅ¡ka  <mliska@suse.cz>
38083         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38084         Fix typos and grammar, use active voice, and clarify.
38086 2017-02-20  Marek Polacek  <polacek@redhat.com>
38088         PR middle-end/79537
38089         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38091         PR sanitizer/79558
38092         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38094 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38096         PR target/79568
38097         * config/i386/i386.c (ix86_expand_builtin): Handle
38098         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38099         ix86_builtins_isa[fcode].isa as a requirement of those
38100         flags and any other flag in the bitmask.
38101         (ix86_init_mmx_sse_builtins): Use 0 instead of
38102         ~OPTION_MASK_ISA_64BIT as mask.
38103         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38104         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38105         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38106         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38108 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38110         PR target/78012
38111         * lra-constraints.c (split_reg): Check requested split mode
38112         is supported by the register.
38114 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38116         * lra-constraints.c (simplify_operand_subreg): Remove early
38117         return false.
38119 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38121         PR target/78660
38122         * lra-constraints.c (curr_insn_transform): Tighten condition
38123         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38125 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38127         PR target/78660
38128         * lra-constraints.c (curr_insn_transform): Handle
38129         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38131 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38133         Revert:
38134         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38136         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38138 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38140         PR c++/69523
38141         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38142         description.
38144 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38146         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38147         for FMA_EXPR.
38149 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38151         * final.c (last_columnnum, override_columnnum): New variables.
38152         (final_start_function): Set last_columnnum, pass it to begin_prologue
38153         hook and pass 0 to dwarf2out_begin_prologue.
38154         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38155         to source_line debug hook.
38156         (notice_source_line): Compute last_columnnum and for debug_column_info
38157         return true on column changes.
38158         * debug.h (struct gcc_debug_hooks): Add column argument to
38159         source_line and begin_prologue hooks.
38160         (debug_nothing_int_charstar_int_bool): Remove prototype.
38161         (debug_nothing_int_int_charstar,
38162         debug_nothing_int_int_charstar_int_bool): New prototypes.
38163         (dwarf2out_begin_prologue): Add column argument.
38164         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38165         begin_prologue hooks.
38166         (debug_nothing_int_charstar_int_bool): Remove.
38167         (debug_nothing_int_int_charstar,
38168         debug_nothing_int_int_charstar_int_bool): New functions.
38169         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38170         through to dwarf2out_source_line.
38171         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38172         (dwarf2out_source_line): Add column argument, emit it if requested.
38173         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38174         arguments.
38175         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38176         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38177         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38178         through to dwarf2out_begin_prologue.
38179         (vmsdbgout_source_line): Add column argument, pass it through to
38180         dwarf2out_source_line.
38181         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38182         dbxout_source_line caller.
38183         (dbxout_source_line): Add column argument.
38185         * common.opt (gno-column-info, gcolumn-info): New options.
38186         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38187         (check_die): Also test for multiple DW_AT_decl_column attributes.
38188         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38189         DW_AT_decl_column if requested.
38190         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38191         if requested.
38192         (gen_variable_die): Likewise.
38193         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38194         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38196         PR target/79569
38197         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38198         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38199         (ix86_handle_option): Handle OPT_m3dnowa.
38200         * doc/invoke.texi (-m3dnowa): Document.
38201         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38202         -m3dnowa instead of -m3dnow -march=athlon.
38204         PR target/79559
38205         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38206         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38208 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38210         PR target/79261
38211         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38212         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38213         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38214         generator for vsx_xxpermdi_<mode>_be.
38215         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38216         force big-endian semantics.
38217         (vsx_xxpermdi_<mode>_be): New define_expand with same
38218         implementation as previous version of vsx_xxpermdi_<mode>.
38220 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38222         PR tree-optimization/79327
38223         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38224         variable, its initialization and use.
38226 2017-02-17  Julia Koval  <julia.koval@intel.com>
38228         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38229         (OPTION_MASK_ISA_PKU_UNSET): New.
38230         (ix86_handle_option): Handle -mrdpid.
38231         * config/i386/cpuid.h (bit_RDPID): New.
38232         * config/i386/driver-i386.c (host_detect_local_cpu):
38233         Detect RDPID feature.
38234         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38235         * config/i386/i386-c.c (ix86_target_macros_internal):
38236         Handle RDPID flag.
38237         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38238         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38239         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38240         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38241         * config/i386/i386.md (define_insn "rdpid"): New.
38242         * config/i386/i386.opt Add -mrdpid.
38243         * config/i386/immintrin.h (_rdpid_u32): New.
38245 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38247         PR rtl-optimization/79541
38248         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38249         instead of transforming it into USE.
38251 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38253         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38254         If HONOR_SNANS (SFmode) force the input to a register.
38255         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38256         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38257         an frsp or similar insn.
38259 2017-02-17  Martin Liska  <mliska@suse.cz>
38261         PR rtl-optimization/79577
38262         * params.def (selsched-max-sched-times): Increase minimum to 1.
38264 2017-02-17  Martin Liska  <mliska@suse.cz>
38266         PR rtl-optimization/79574
38267         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38269 2017-02-17  Martin Liska  <mliska@suse.cz>
38271         PR tree-optimization/79529
38272         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38273         ssa_defined_default_def_p to handle cases which are implicitly
38274         defined.
38275         * tree-ssa.c (ssa_defined_default_def_p): New function.
38276         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38277         which are implicitly defined.
38278         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38280 2017-02-17  Richard Biener  <rguenther@suse.de>
38282         PR middle-end/79576
38283         * params.def (max-ssa-name-query-depth): Limit to 10.
38285 2017-02-17  Richard Biener  <rguenther@suse.de>
38287         PR tree-optimization/79552
38288         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38289         default defs.
38291 2017-02-17  Richard Biener  <rguenther@suse.de>
38293         PR bootstrap/79567
38294         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38296 2017-02-17  Marek Polacek  <polacek@redhat.com>
38298         PR middle-end/79536
38299         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38300         (fold_negate_expr): New wrapper.
38302 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38304         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38305         Correct terminology and de-emphasize pre-standard behavior.
38307 2017-02-16  Alan Modra  <amodra@gmail.com>
38309         PR rtl-optimization/79286
38310         * ira.c (def_dominates_uses): New function.
38311         (update_equiv_regs): Don't create an equivalence for insns that
38312         may trap where the register def does not dominate the use.
38314 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38316         PR rtl-optimization/78127
38317         * lra.c (lra): Call lra_eliminate before finish the loop after
38318         lra_constraint.
38320 2017-02-16  Richard Biener  <rguenther@suse.de>
38322         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38323         isl/isl_val.h.
38324         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38325         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38326         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38327         (isl_val_int_from_wi): New function.
38328         (extract_affine_gmp): Rename to ...
38329         (extract_affine_wi): ... this, take a widest_int.
38330         (extract_affine_int): Just wrap extract_affine_wi.
38331         (add_param_constraints): Use isl_val_int_from_wi.
38332         (add_loop_constraints): Likewise, and extract_affine_wi.
38334 2017-02-15  Jeff Law  <law@redhat.com>
38336         PR middle-end/79521
38337         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38338         ira_init_register_move_cost_if_necessary.
38340 2017-02-15  Martin Sebor  <msebor@redhat.com>
38342         PR middle-end/32003
38343         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38344         removed in a prior commit.
38346 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38348         PR tree-optimization/79347
38349         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38350         counters during peeling.
38352 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38354         * Makefile.in (site.exp): Remove "set ISLVER".
38356 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38358         PR target/79487
38359         * real.c (real_from_integer): Call real_convert even for decimal.
38361 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38363         PR target/79421
38364         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38366 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38368         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38369         cores and change the partno/implementer to be correct.
38370         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38371         the 'B" as the implementer.
38372         * config/aarch64/aarch64-tune.md: Regenerate.
38374 2017-02-14  Carl Love  <cel@us.ibm.com>
38376         * config/rs6000/rs6000.c: Add case statement entry to make the
38377         xvcvuxdsp built-in argument unsigned.
38378         * config/rs6000/vsx.md: Fix the source and return operand types so they
38379         match the instruction definitions from the ISA document.  Fix typo
38380         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38381         statement.
38383 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38385         PR target/79282
38386         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38387         member early_clobber_alts.
38388         * lra-lives.c (reg_early_clobber_p): New.
38389         (process_bb_lives): Use it.
38390         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38391         (debug_operand_data): Initialize early_clobber_alts.
38392         (setup_operand_alternative): Set up early_clobber_alts.
38393         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38394         alternatives to new_insn_reg.
38395         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38396         it.
38397         (lra_update_insn_regno_info): Pass the new arg.
38399 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38401         PR middle-end/79505
38402         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38403         (new_oacc_loop_raw): Don't clear already cleared fields.
38405         PR target/79481
38406         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38407         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38408         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38410 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38412         PR target/79495
38413         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38415 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38417         PR target/79498
38418         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38419         the extra instruction to the right place to store 128-bit constant
38420         when needed.
38422 2017-02-14  Martin Sebor  <msebor@redhat.com>
38424         PR middle-end/79448
38425         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38426           warning for strings of unknown length.
38428 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38430         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38432 2017-02-14  Jeff Law  <law@redhat.com>
38434         PR target/79404
38435         * ira-costs.c (scan_one_insn): Initialize register move costs
38436         for pseudos seen in USE/CLOBBER insns.
38438         PR tree-optimization/79095
38439         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38440         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38441         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38442         if the operands are known to be not equal, then the resulting range
38443         is ~[0,0].
38444         (intersect_ranges): If the new range is ~[0,0] and the old range is
38445         wide, then prefer ~[0,0].
38446         * tree-vrp.c (overflow_comparison_p_1): New function.
38447         (overflow_comparison_p): New function.
38448         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38449         if NAME is used in an overflow test.
38450         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38451         overflow check that can be expressed as an equality test, then adjust
38452         ops to be that equality test.
38454 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38456         * config/s390/s390-builtin-types.def: Remove flags argument.
38457         * config/s390/s390.c (s390_init_builtins): Likewise.
38459 2017-02-14  Martin Liska  <mliska@suse.cz>
38461         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38462         vector.  Fix trailing white spaces.
38464 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38466         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38467         HFmode.
38469 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38471         PR rtl-optimization/68664
38472         * config/arm/arm.c (arm_sched_can_speculate_insn):
38473         New function.  Declare prototype.
38474         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38476 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38478         PR rtl-optimization/68664
38479         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38480         New function.
38481         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38483 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38485         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38486         max skip bytes for function, loop and jump.
38488 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38490         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38491         ABS_EXPR for gimple dump.
38493 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38495         PR target/79462
38496         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38498         PR tree-optimization/79408
38499         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38500         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38501         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38502         also if rhs1 is INTEGER_CST.
38504 2017-02-14  Richard Biener  <rguenther@suse.de>
38506         PR middle-end/79432
38507         * tree-into-ssa.c (insert_phi_nodes): When the function can
38508         have abnormal edges rewrite SSA names with broken use-def
38509         dominance out of SSA and register them for PHI insertion.
38511 2017-02-13  Martin Sebor  <msebor@redhat.com>
38513         PR middle-end/79496
38514         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38515         clearing info.nowrite flag when snprintf size argument is a range.
38517 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38519         * cprop.c (cprop_jump): Add missing space in string literal.
38520         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38521         (get_constraint_for_component_ref): Likewise.
38522         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38523         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38524         * lra-constraints.c (process_alt_operands): Likewise.
38525         * ipa-inline.c (inline_small_functions): Likewise.
38526         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38527         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38528         * trans-mem.c (diagnose_tm_1_op): Likewise.
38529         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38530         (grid_parallel_clauses_gridifiable): Likewise.
38532         * config/nvptx/mkoffload.c (process): Add space in between
38533         , and %d.
38535         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38536         "MOD4_SSE_REGS" and "ALL_REGS".
38538         * spellcheck.c (test_data): Add , in between "foo" and "food".
38540 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38542         PR target/79449
38543         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38544         boundary crossing check and subsequent code generation agree.
38546 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38548         * config/aarch64/aarch64.c (has_memory_op): Delete.
38549         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38550         has_memory_op.
38552 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38554         PR rtl-optimization/79388
38555         PR rtl-optimization/79450
38556         * combine.c (distribute_notes): When removing TEM_INSN for which
38557         corresponding dest has last value recorded, invalidate that last
38558         value.
38560 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38562         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38563         of explicit '@'.  Add missing assembly comment marker on branch costs
38564         printout.
38566 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38568         * gengtype-lex.l (<in_struct>): Add '/'.
38570 2017-02-13  Martin Liska  <mliska@suse.cz>
38572         PR c/79471
38573         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38575 2017-02-13  Richard Biener  <rguenther@suse.de>
38577         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38578         Remove.
38579         * configure: Re-generate.
38580         * config.in: Likewise.
38581         * graphite-dependences.c: Simplify as if
38582         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38583         * graphite-isl-ast-to-gimple.c: Likewise.
38584         * graphite-optimize-isl.c: Likewise.
38585         * graphite-poly.c: Likewise.
38586         * graphite-sese-to-poly.c: Likewise.
38587         * graphite.h: Likewise.
38588         * toplev.c: Include isl/version.h and use isl_version () for
38589         printing the ISL version.
38590         * doc/install.texi: Update ISL requirement.
38592 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38594         * doc/standards.texi (Standards): Update reference to
38595         Objective-C 2.0.
38597 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38599         * doc/extend.texi (Named Address Spaces): sourceware.org now
38600         defaults to https.
38601         * doc/install.texi (Binaries): Ditto.
38602         (Specific): Ditto.
38604 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38606         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38607         standard terminology "stringize"/"stringizing" throughout.
38608         * doc/cppinternals.texi: Likewise.
38610 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38612         * doc/extend.texi: Fix some spelling mistakes and typos.
38613         * doc/invoke.texi: Likewise.
38615 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38617         PR ipa/79224
38618         * params.def (inline-min-speedup) Change from 10 to 8.
38620 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38622         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38623         4.5.
38625 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38627         PR ipa/79224
38628         * ipa-inline-analysis.c (get_minimal_bb): New function.
38629         (record_modified): Use it.
38630         (remap_edge_change_prob): Handle also ancestor functions.
38632 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38634         * doc/contrib.texi (Contributors): Remove broken link into
38635         the Mauve CVS repository.
38637 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38639         PR middle-end/79454
38640         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38641         result computation whenever lhs doesn't have vector mode, not
38642         just when it has BLKmode.
38644 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38646         * doc/makefile.texi (profiledbootstrap): Refer to the
38647         installation instructions only in textual form.
38649 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38651         PR target/79295
38652         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38654 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38656         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38657         (Specific): Update mingw-w64 reference.
38658         (Binaries): Ditto.
38659         (Specific): Remove broken link to Renesas RX processor.
38661 2017-02-10  Richard Biener  <rguenther@suse.de>
38663         * toplev.c (process_options): Do not mention obsolete graphite
38664         options when printing sorry message about missing graphite support.
38665         Mention -floop-nest-optimize.
38667 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38669         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38670         (vtst_p16): Likewise.
38671         (vtstq_p8): Likewise.
38672         (vtstq_p16): Likewise.
38673         (vtst_p64): New.
38674         (vtstq_p64): Likewise.
38675         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38676         (vset_lane_p64): New.
38677         (vsetq_lane_p64): New.
38679 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38681         PR tree-optimization/79411
38682         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38683         stmt operands are SSA_NAMEs used in abnormal phis.
38684         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38685         phis.
38687 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38689         PR ipa/70795
38690         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38691         flag if needed.
38693 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38695         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38697 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38699         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38700         to avoid warning.
38702         PR c/79413
38703         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38704         not arbitrary TREE_CONSTANT.
38706         PR c/79431
38707         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38708         "omp declare target link" attribute unless is_global_var.
38709         * omp-offload.c (find_link_var_op): Likewise.
38711 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38712             Chung-Lin Tang  <cltang@codesourcery.com>
38714         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38715         OMP_CLAUSE_TILE.
38716         (gimplify_adjust_omp_clauses): Don't delete TILE.
38717         (gimplify_omp_for): Deal with TILE.
38718         * internal-fn.c (expand_GOACC_TILE): New function.
38719         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38720         (GOACC_TILE): New.
38721         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38722         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38723         element fields.
38724         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38725         avoid DIV for outermost collapse var.
38726         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38727         Remove out of date comments, fix whitespace.
38728         * omp-general.c (omp_extract_for_data): Deal with tiling.
38729         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38730         adjust OLF_DIM_BASE value.
38731         (struct omp_for_data): Add tiling field.
38732         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38733         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38734         for auto loops.  Remove default auto determining, moved to
38735         oacc_loop_fixed_partitions.
38736         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38737         stmts, add e_mask field.
38738         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38739         (oacc_thread_numbers): Use oacc_dim_call.
38740         (oacc_xform_tile): New.
38741         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38742         (finish_oacc_loop): Adjust for ifns vector.
38743         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38744         add case for GOACC_TILE fns.
38745         (oacc_loop_xform_loop): Delete.
38746         (oacc_loop_process): Iterate over call list directly, and add
38747         handling for GOACC_TILE fns.
38748         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38749         dump partitioning.
38750         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38751         vector partitioning to outer loops.  Assign 2 partitions to loops
38752         when available. Add TILE handling.
38753         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38754         (execite_oacc_device_lower): Process GOACC_TILE fns,
38755         ignore unknown specs.
38756         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38757         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38758         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38760 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38762         * configure.ac (ACX_BUGURL): Update.
38763         * configure: Regenerate.
38765 2017-02-09  Richard Biener  <rguenther@suse.de>
38767         PR tree-optimization/69823
38768         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38769         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38771 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38773         * config/arc/arc-c.def: Add __NPS400__ definition.
38774         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38775         (TARGET_NPS400): Define.
38777 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38779         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38780         file.
38781         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38782         pointer, arch_info.
38783         (arc_cpu_types): Fill the arch_info field with a pointer into the
38784         arc_arch_types table.
38785         (arc_selected_cpu): Declare.
38786         * config/arc/arc.c (arc_selected_cpu): Make global.
38787         (arc_selected_arch): Delete.
38788         (arc_base_cpu): Delete.
38789         (arc_override_options): Remove references to deleted variables,
38790         update access to arch information.
38791         (ARC_OPT): Update access to arch information.
38792         (ARC_OPTX): Likewise.
38793         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38794         (TARGET_ARC600): Update access to arch information.
38795         (TARGET_ARC601): Likewise.
38796         (TARGET_ARC700): Likewise.
38797         (TARGET_EM): Likewise.
38798         (TARGET_HS): Likewise.
38799         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38800         information.
38802 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38804         PR target/78604
38805         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38806         condition/operands for integer GE/LE/GEU/LEU operations.
38808 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38810         PR translation/79397
38811         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38812         of AltiVec.
38814 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38816         PR ipa/79375
38817         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38818         whether allocation happened.
38819         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38820         nothing was allocated.
38822 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38824         PR tree-optimization/79408
38825         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38826         constant, but SSA_NAME with a known integer range, use the minimum
38827         of that range instead of op1 to determine if modulo can be replaced
38828         with its first operand.
38830 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38832         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38834 2017-02-08  Richard Biener  <rguenther@suse.de>
38836         PR tree-optimization/71824
38837         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38838         Check all loops contained in the merged region.
38840 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38842         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38844 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38846         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38847         (thunderxt88): Likewise.
38848         (thunderxt81): Disable LSE and change v8.1 to v8.
38849         (thunderxt83): Likewise.
38851 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38852             Richard Biener  <rguenther@suse.de>
38854         PR middle-end/79399
38855         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38856         type from int to size_t.
38857         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38859 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38861         PR rtl-optimization/79386
38862         * cprop.c (bypass_conditional_jumps): Initialize
38863         bypass_last_basic_block already before splitting bbs after
38864         unconditional traps...
38865         (bypass_conditional_jumps): ... rather than here.
38867         PR target/79299
38868         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38869         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38870         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38871         fix -masm=intel patterns.
38873 2017-02-07  Richard Biener  <rguenther@suse.de>
38875         PR tree-optimization/79256
38876         PR middle-end/79278
38877         * builtins.c (get_object_alignment_2): Use min_align_of_type
38878         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38879         and ADJUST_FIELD_ALIGN.
38881         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38882         type parameter.
38883         * doc/tm.texi: Regenerate.
38884         * stor-layout.c (layout_decl): Adjust.
38885         (update_alignment_for_field): Likewise.
38886         (place_field): Likewise.
38887         (min_align_of_type): Likewise.
38888         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38889         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38890         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38891         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38892         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38893         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38894         * config/i386/i386.c (x86_field_alignment): Likewise.
38895         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38896         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38897         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38898         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38899         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38900         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38901          Likewise.
38903         Revert
38904         2017-01-30  Richard Biener  <rguenther@suse.de>
38906         PR tree-optimization/79256
38907         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38908         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38909         alignment on TYPE.
38911 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38913         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38914         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38915         builtins to SImode and emit a zero-extend, if necessary.
38917 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38919         * docs/invoke.texi (RISC-V Options): Alphabetize.
38921 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38923         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38924         options.
38926 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38928         * config/riscv/riscv.c: New file.
38929         * common/config/riscv/riscv-common.c: Likewise.
38930         * config.gcc: Likewise.
38931         * config/riscv/constraints.md: Likewise.
38932         * config/riscv/elf.h: Likewise.
38933         * config/riscv/generic.md: Likewise.
38934         * config/riscv/linux.h: Likewise.
38935         * config/riscv/multilib-generator: Likewise.
38936         * config/riscv/peephole.md: Likewise.
38937         * config/riscv/pic.md: Likewise.
38938         * config/riscv/predicates.md: Likewise.
38939         * config/riscv/riscv-builtins.c: Likewise.
38940         * config/riscv/riscv-c.c: Likewise.
38941         * config/riscv/riscv-ftypes.def: Likewise.
38942         * config/riscv/riscv-modes.def: Likewise.
38943         * config/riscv/riscv-opts.h: Likewise.
38944         * config/riscv/riscv-protos.h: Likewise.
38945         * config/riscv/riscv.h: Likewise.
38946         * config/riscv/riscv.md: Likewise.
38947         * config/riscv/riscv.opt: Likewise.
38948         * config/riscv/sync.md: Likewise.
38949         * config/riscv/t-elf-multilib: Likewise.
38950         * config/riscv/t-linux: Likewise.
38951         * config/riscv/t-linux-multilib: Likewise.
38952         * config/riscv/t-riscv: Likewise.
38953         * configure.ac: Likewise.
38954         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
38955         Waterman as RISC-V maintainers.
38956         * doc/install.texi: Add RISC-V entries.
38957         * doc/invoke.texi: Add RISC-V options section.
38958         * doc/md.texi: Add RISC-V constraints section.
38959         * configure: Regenerated.
38961 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
38963         PR target/66144
38964         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
38965         false values to be constant vectors with all 0 or all 1 bits set.
38966         (vcondu<mode><mode>): Likewise.
38967         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
38968         predicate.
38969         (fpmask_comparison_operator): Update comment.
38970         (vecint_comparison_operator): New predicate.
38971         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
38972         vector conditionals when the true and false values are constant
38973         vectors with all 0 bits or all 1 bits set.
38975 2017-02-06  Martin Sebor  <msebor@redhat.com>
38977         PR  tree-optimization/79376
38978         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
38980 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
38982         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
38983         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
38984         to simplify split condition.
38986 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38988         * omp-expand.c (oxpand_omp_atomic_fetch_op,
38989         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
38990         false.
38992 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
38994         PR rtl-optimization/68664
38995         * target.def (can_speculate_insn): New hook.
38996         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
38997         * doc/tm.texi: Regenerate.
38998         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
38999         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
39000         (rs6000_sched_can_speculate_insn): New function.
39002 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39004         PR tree-optimization/79284
39005         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
39006         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
39007         vectorizable_mask_load_store, vectorizable_operation,
39008         vect_is_simple_cond, get_same_sized_vectype): Use it instead
39009         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
39010         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
39011         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
39012         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
39013         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
39014         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
39015         is_gimple_assign (stmt).  Replace another such test with
39016         is_gimple_assign (stmt).
39018 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39020         PR target/78883
39021         * config/avr/avr.c (rtl-iter.h): Include it.
39022         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
39023         (avr_legitimate_combined_insn): ...and implementation.
39025 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39027         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39028         * config/s390/s390.c (s390_const_operand_ok)
39029         (s390_canonicalize_comparison, s390_extract_part)
39030         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39031         (s390_contiguous_bitmask_p, s390_rtx_costs)
39032         (legitimize_pic_address): Likewise.
39033         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39034         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39035         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39036         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39037         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39039 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39041         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39042         REGNO($0) == REGNO($1).
39044 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39046         * config/s390/linux.h(SIZE_TYPE): Add comment.
39048 2017-02-06  Julian Brown  <julian@codesourcery.com>
39049             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39050             Virendra Pathak  <virendra.pathak@broadcom.com>
39052         * config/aarch64/aarch64-cores.def: Change the scheduler
39053         to Thunderx2t99.
39054         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39055         * config/aarch64/thunderx2t99.md: New file.
39057 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39059         * doc/standards.texi (Go Language): Update link to language
39060         standard.
39062 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39064         * tree-eh.c (lower_resx): Sanitize profile.
39065         (cleanup_empty_eh_move_lp): Likewise.
39067 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39069         PR tree-ssa/79347
39070         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39071         ELSE_PROB.
39072         * cfgloopmanip.h (loop_version): Update prototype.
39073         * modulo-sched.c (sms_schedule): Update call of loop_version.
39074         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39075         * tree-parloops.c (gen_parallel_loop): Likewise.
39076         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39077         * tree-ssa-loop-split.c (split_loop): Likewise.
39078         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39079         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39081 2017-02-05  Martin Liska  <mliska@suse.cz>
39083         PR bootstrap/78985
39084         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39085         variable to NULL.
39086         (print_operand_address): Initialize a struct to zero.
39088 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39090         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39091         garbage collector only in textual form.
39093 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39095         * doc/extend.texi (x86 specific memory model extensions for
39096         transactional memory): Simplify a phrase.
39098 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39100         PR target/79353
39101         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39102         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39103         (atomic_storedi_1): Likewise.
39105 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39107         PR tree-optimization/79338
39108         * tree-parloops.c (gather_scalar_reductions): Don't call
39109         vect_analyze_loop_form for loop->inner before destroying loop's
39110         loop_vinfo.
39112 2017-02-03  Martin Sebor  <msebor@redhat.com>
39114         PR tree-optimization/79327
39115         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39116         when precision has resulted in leading zeros.
39117         (format_integer): Adjust the likely counter to assume an unknown
39118         argument that may be zero is non-zero.
39120 2017-02-03  Jason Merrill  <jason@redhat.com>
39122         PR c++/78689
39123         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39124         avoid copying non-taken branch.
39126 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39128         PR tree-optimization/79340
39129         * tree-vect-loop.c (vectorizable_reduction): Release
39130         vec_defs elements after safe_splicing them into other vectors.
39131         Formatting fixes.
39133         PR tree-optimization/79327
39134         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39135         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39136         dirtype.
39137         (format_integer): Use wide_int_to_tree instead of build_int_cst
39138         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39139         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39140         of shortest and longest sequence.
39142 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39144         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39145         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39147 2017-02-03  Walter Lee  <walt@tilera.com>
39149         PR target/78862
39150         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39151         after initial stackframe link reg save.
39152         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39154 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39156         PR target/79354
39157         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39158         wu for stxssp alternative.
39160 2017-02-03  Martin Sebor  <msebor@redhat.com>
39162         PR tree-optimization/79352
39163         * gimple-fold.c (get_range_strlen): Add argument.
39164         (get_range_strlen): Change return type to bool.
39165         (get_maxval_strlen): Pass in a dummy argument.
39166         * gimple-fold.h (get_range_strlen): Change return type to bool.
39167         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39168         * tree.h (array_at_struct_end_p): Add argument.
39169         * tree.c (array_at_struct_end_p): Handle it.
39171 2017-02-03  Martin Liska  <mliska@suse.cz>
39173         PR lto/66295
39174         * multiple_target.c (create_dispatcher_calls): Redirect edge
39175         from a caller of a dispatcher.
39176         (expand_target_clones): Make the clones local.
39177         (ipa_target_clone): Do both target clones and resolvers.
39178         (ipa_dispatcher_calls): Remove the pass.
39179         (pass_dispatcher_calls::gate): Likewise.
39180         (make_pass_dispatcher_calls): Likewise.
39181         * passes.def (pass_target_clone): Put as very first IPA early
39182         pass.
39184 2017-02-03  Martin Liska  <mliska@suse.cz>
39186         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39187         in case of a function with ifunc attribute.
39189 2017-02-03  Martin Liska  <mliska@suse.cz>
39191         * cgraph.c (cgraph_node::dump): Dump function version info.
39192         * symtab.c (symtab_node::dump_base): Add missing new line.
39194 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39196         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39197         (ifcombine_ifandif): Use it.
39199 2017-02-03  Martin Liska  <mliska@suse.cz>
39201         * doc/invoke.texi: Document default value for
39202         use-after-scope-direct-emission-threshold.
39204 2017-02-03  Martin Liska  <mliska@suse.cz>
39206         PR tree-optimization/79339
39207         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39208         (format_floating): Likewise.
39210 2017-02-03  Martin Liska  <mliska@suse.cz>
39212         PR ipa/79337
39213         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39214         implementation.
39215         (ipa_node_params_t::remove): Likewise.
39216         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39217         initialization from removed ipa_node_params_t::insert.
39218         (ipa_node_params::~ipa_node_params): Move from removed
39219         ipa_node_params_t::release.
39220         * symbol-summary.h (symbol_summary::m_released): New member.
39221         Do not release a summary twice.  Do not allow to call finalizer
39222         for types of a summary that live in GGC memory.
39224 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39226         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39227         cmp_branch fusion.
39229 2017-02-02  Martin Sebor  <msebor@redhat.com>
39231         PR middle-end/79275
39232         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39233         (format_string): Tighten up the range of output for non-constant
39234         strings and correct the expected range for wide non-constant strings.
39236 2017-02-02  Martin Sebor  <msebor@redhat.com>
39238         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39240         PR middle-end/32003
39241         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39242         index.
39243         (-fdump-tree-@var): Add to index and document how to come up
39244         with pass-specific option and dump file names.
39245         (-fdump-passes): Clarify where to look for output.
39247 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39249         PR middle-end/77445
39250         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39251         statistics of the analyzed path; allow threading for speed when
39252         any of BBs along the path are optimized for speed.
39254 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39256         PR middle-end/78468
39257         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39258         settings of the virtual registers.
39260         Revert again
39261         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39263         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39264         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39265         needed.
39267 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39269         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39270         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39272 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39274         * config/s390/s390.md: Add missing comments with the expanded
39275         mnemonics.
39276         * config/s390/vector.md: Likewise.
39277         * config/s390/vx-builtins.md: Likewise.
39279 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39281         PR target/79197
39282         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39283         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39284         conditions on a single line.
39286 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39288         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39289         __S390_VX__ to __VX__.
39291 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39293         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39294         stmt_info to record_stmt_cost.
39295         (vect_get_known_peeling_cost): Pass stmt_info if known to
39296         record_stmt_cost.
39297         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39298         cpu_vector_cost field into
39299         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39300         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39301         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39302         splitting of scalar_stmt_cost and vec_stmt_cost.
39303         (thunderx_vector_cost): Likewise.
39304         (cortexa57_vector_cost): LIkewise.
39305         (exynosm1_vector_cost): Likewise.
39306         (xgene1_vector_cost): Likewise.
39307         (thunderx2t99_vector_cost): Improve after the splitting of the two
39308         fields.
39309         (aarch64_builtin_vectorization_cost): Update for the splitting of
39310         scalar_stmt_cost and vec_stmt_cost.
39312 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39313             Richard Henderson  <rth@redhat.com>
39315         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39316         conditional on existance of a fast atomic load.
39317         * optabs-query.c (can_atomic_load_p): New function.
39318         * optabs-query.h (can_atomic_load_p): Declare it.
39319         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39320         no fast atomic load is available for the particular size of access.
39321         (expand_atomic_compare_and_swap): Likewise.
39322         (expand_atomic_load): Likewise.
39323         (expand_atomic_store): Likewise.
39324         (expand_atomic_fetch_op): Likewise.
39325         * testsuite/lib/target-supports.exp
39326         (check_effective_target_sync_int_128): Remove x86 because it provides
39327         no fast atomic load.
39328         (check_effective_target_sync_int_128_runtime): Likewise.
39330 2017-02-01  Richard Biener  <rguenther@suse.de>
39332         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39333         (graphite_transform_loops): Provide opt-info for optimized nests.
39334         * tree-parloop.c (parallelize_loops): Provide opt-info for
39335         parallelized loops.
39337 2017-02-01  Richard Biener  <rguenther@suse.de>
39339         PR middle-end/79315
39340         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39341         was not set before.
39343 2017-02-01  Richard Biener  <rguenther@suse.de>
39345         PR tree-optimization/71824
39346         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39347         Verify the loops are valid in the merged SESE region.
39348         (scop_detection::can_represent_loop_1): Check analyzing the
39349         evolution of the number of iterations in the region succeeds.
39351 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39353         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39354         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39356 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39358         PR preprocessor/79210
39359         * input.c (get_substring_ranges_for_loc): Replace line_width
39360         assertion with error-handling.
39362 2017-01-31  Richard Biener  <rguenther@suse.de>
39364         PR tree-optimization/77318
39365         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39366         (create_pw_aff_from_tree): Take loop parameter.
39367         (add_condition_to_pbb): Pass loop of the condition to
39368         create_pw_aff_from_tree.
39370 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39372         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39373         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39375 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39377         PR target/78597
39378         PR target/79038
39379         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39380         no longer used.
39381         (convert_int_to_float128): Likewise.
39382         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39383         (convert_int_to_float128): Likewise.
39384         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39385         (UNSPEC_IEEE128_CONVERT): Likewise.
39386         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39387         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39388         Use local variables for IBM extended format.
39389         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39390         (fix_trunc<mode>si2_fprs): Likewise.
39391         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39392         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39393         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39394         to know that we can now have integers of all sizes in vector
39395         registers.
39396         (fix<uns>_<mode>di2_hw): Likewise.
39397         (float<uns>_<mode>si2_hw): Likewise.
39398         (fix_<mode>si2_hw): Likewise.
39399         (fixuns_<mode>si2_hw): Likewise.
39400         (float<uns>_<mode>di2_hw): Likewise.
39401         (float_<mode>di2_hw): Likewise.
39402         (float_<mode>si2_hw): Likewise.
39403         (floatuns_<mode>di2_hw): Likewise.
39404         (floatuns_<mode>si2_hw): Likewise.
39405         (xscvqp<su>wz_<mode>): Delete, no longer used.
39406         (xscvqp<su>dz_<mode>): Likewise.
39407         (xscv<su>dqp_<mode>): Likewise.
39408         (ieee128_mfvsrd_64bit): Likewise.
39409         (ieee128_mfvsrd_32bit): Likewise.
39410         (ieee128_mfvsrwz): Likewise.
39411         (ieee128_mtvsrw): Likewise.
39412         (ieee128_mtvsrd_64bit): Likewise.
39413         (ieee128_mtvsrd_32bit): Likewise.
39415 2017-01-31  Martin Liska  <mliska@suse.cz>
39417         PR ipa/79285
39418         * ipa-prop.c (ipa_free_all_node_params): Call release method
39419         instead of ~sumbol_summary to not to trigger double times
39420         dtor of hash_map.
39422 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39424         PR tree-optimization/71691
39425         * bitmap.h (class auto_bitmap): New.
39426         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39427         is_maybe_undefined instead of ssa_undefined_value_p.
39429 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39431         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39432         __S390_ARCH_LEVEL__ to __ARCH__.
39434 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39436         PR tree-optimization/79267
39437         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39438         if should_remove_lhs_p is true.
39440 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39442         PR debug/63238
39443         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39444         (add_alignment_attribute): New.
39445         (base_type_die): Add alignment attribute.
39446         (subrange_type_die): Likewise.
39447         (modified_type_die): Likewise.
39448         (gen_array_type_die): Likewise.
39449         (gen_descr_array_type_die: Likewise.
39450         (gen_enumeration_type_die): Likewise.
39451         (gen_subprogram_die): Likewise.
39452         (gen_variable_die): Likewise.
39453         (gen_field_die): Likewise.
39454         (gen_ptr_to_mbr_type_die): Likewise.
39455         (gen_struct_or_union_type_die): Likewise.
39456         (gen_subroutine_type_die): Likewise.
39457         (gen_typedef_die): Likewise.
39458         (base_type_cmp): Compare alignment attribute.
39460 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39462         PR target/79170
39463         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39464         (setb_unsigned) New pattern for setb with CCUNS.
39465         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39466         subfc./subfe sequence to avoid overflow problems.  Generate a
39467         shorter sequence with cmpld/setb for power9.
39468         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39469         for generating subfc. instruction.
39470         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39471         now uses this instruction.
39473 2017-01-30  Ian Lance Taylor  <iant@google.com>
39475         PR debug/79289
39476         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39477         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39479 2017-01-30  Martin Sebor  <msebor@redhat.com>
39481         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39482         Move constant to the right of a relational operator.
39483         (get_mpfr_format_length, format_character, format_string): Ditto.
39484         (should_warn_p, maybe_warn): Same.
39486         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39488 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39490         PR lto/79061
39491         * asan.c (get_translation_unit_decl): Remove function.
39492         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39494 2017-01-30  Martin Liska  <mliska@suse.cz>
39496         PR gcov-profile/79259
39497         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39498         -fprofile-generate.
39500 2017-01-30  Martin Liska  <mliska@suse.cz>
39502         PR bootstrap/78985
39503         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39504         Initialize variables with NULL value.
39506 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39508         PR target/79260
39509         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39510         tm_p_file.
39511         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39513 2017-01-30  Richard Biener  <rguenther@suse.de>
39515         PR tree-optimization/79276
39516         * tree-vrp.c (process_assert_insertions): Properly adjust common
39517         when removing a duplicate.
39519 2017-01-30  Richard Biener  <rguenther@suse.de>
39521         PR tree-optimization/79256
39522         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39523         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39524         alignment on TYPE.
39525         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39527 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39529         PR target/79240
39530         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39531         ("*r<noxa>sbg_<mode>_sll_bitmask")
39532         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39533         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39534         Use contiguous_bitmask_nowrap_operand.
39536 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39538         PR target/79268
39539         * config/rs6000/altivec.h (vec_xl): Revise #define.
39540         (vec_xst): Likewise.
39542 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39544         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39546 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39548         PR rtl-optimization/79194
39549         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39550         traps before call to bypass_conditional_jumps.
39552 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39554         PR tree-optimization/71374
39555         * lra-constraints.c (check_conflict_input_operands): New.
39556         (match_reload): Use it.
39558 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39560         PR target/79131
39561         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39562         account to calculate conflict_set.
39564 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39566         PR rtl-optimization/78559
39567         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39568         other_insn in combine.
39570 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39572         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39573         uint16_type_node for BT_UINT16.
39575 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39577         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39578         "RTL Tests" to menu.
39579         (GIMPLE Tests): New node.
39580         (RTL Tests): New node.
39582 2017-01-27  Richard Biener  <rguenther@suse.de>
39584         PR tree-optimization/79245
39585         * tree-loop-distribution.c (distribute_loop): Apply cost
39586         modeling also to detected patterns.
39588 2017-01-27  Richard Biener  <rguenther@suse.de>
39590         PR tree-optimization/71433
39591         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39592         (compare_assert_loc): New function.
39593         (process_assert_insertions): Sort and optimize assert locations
39594         to remove duplicates and push down identical assertions on
39595         edges to their destination block.
39597 2017-01-27  Richard Biener  <rguenther@suse.de>
39599         PR tree-optimization/79244
39600         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39601         out SSA names even if abnormal.
39603 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39605         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39606         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39607         instead of MPFR_RNDN.
39609 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39611         PR target/79239
39612         * arm.c (arm_option_override): Don't call build_target_option_node
39613         until after doing all option overrides.
39614         (arm_valid_target_attribute_tree): Likewise.
39616 2017-01-27  Martin Liska  <mliska@suse.cz>
39618         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39619         for {cd}tors and C++ {cd}tors.
39621 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39623         * config/s390/s390.md ("*setmem_long_and")
39624         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39626 2017-01-26  Martin Sebor  <msebor@redhat.com>
39628         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39629         of precision.
39631 2017-01-26  Martin Sebor  <msebor@redhat.com>
39633         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39634         HAVE_DFmode before using XFmode or DFmode.
39635         (parse_directive): Avoid using the z length modifier to avoid
39636         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39638         PR middle-end/78703
39639         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39640         to accept adjustment as an array.
39641         (get_int_range): New function.
39642         (struct directive): Make width and prec arrays.
39643         (directive::set_width, directive::set_precision): Call get_int_range.
39644         (format_integer, format_floating): Handle width and precision ranges.
39645         (format_string, parse_directive): Same.
39647 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39649         PR debug/79129
39650         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39651         comdat_type_p set, just clone them, but keep the children in the
39652         original DIE.
39654         PR debug/78835
39655         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39656         which have direct callers with -fvar-tracking-assignments enabled
39657         in the current TU.
39658         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39659         inside of type units.
39661 2017-01-26  Martin Sebor  <msebor@redhat.com>
39663         PR middle-end/78703
39664         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39665         unlikely counters.
39666         (struct format_result): Replace number_chars, number_chars_min,
39667         and number_chars_max with a single member of struct result_range.
39668         Remove bounded.
39669         (format_result::operator+=): Adjust.
39670         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39671         counters.
39672         (fmtresult::adjust_for_width_or_precision): New function.
39673         (fmtresult:type_max_digits): New function.
39674         (bytes_remaining): Handle likely and unlikely counters.
39675         (min_bytes_remaining): Remove.
39676         (format_percent): Simplify.
39677         (format_integer, format_floating): Set likely and unlikely counters.
39678         (get_string_length, format_character, format_string): Same.
39679         (format_plain, should_warn_p): New function.
39680         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39681         and handle those for all directives, including plain strings.
39682         (format_directive): Handle likely and unlikely counters.
39683         Remove unnecessary quoting from diagnostics.  Add an informational
39684         note.
39685         (add_bytes): Remove.
39686         (pass_sprintf_length::compute_format_length): Simplify.
39687         (try_substitute_return_value): Handle likely and unlikely counters.
39689 2017-01-26  Carl Love  <cel@us.ibm.com>
39691         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39692         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39694 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39696         PR target/79131
39697         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39698         endianess for subregs into account.
39699         * lra-constraints.c (lra_constraints): Do risky transformations
39700         always on the first iteration.
39701         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39702         last_call_used_reg_set.
39703         (process_bb_lives): Define and use last_call_used_reg_set.
39704         * lra.c (lra): Always continue after lra_constraints on the first
39705         iteration.
39707 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39709         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39710         constant.
39711         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39713 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39715         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39716         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39717         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39718         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39719         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39720         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39721         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39722         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39723         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39725 2017-01-26  Marek Polacek  <polacek@redhat.com>
39727         PR c/79199
39728         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39729         for the third operand.
39731 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39733         PR middle-end/79236
39734         * omp-low.c (struct omp_context): Add simt_stmt field.
39735         (scan_omp_for): Return omp_context *.
39736         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39737         context to the _simt_ SIMD stmt.
39738         (lower_omp_for): For combined SIMD with sibling _simt_
39739         SIMD, make sure to use the same decls in _looptemp_
39740         clauses as in the sibling.
39742 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39744         PR middle-end/79212
39745         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39746         all contexts.
39748 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39750         PR target/70465
39751         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39752         emit fld b; fld a; if possible.
39754         * brig-builtins.def: Update copyright years.
39755         * config/arm/arm_acle_builtins.def: Update copyright years.
39757 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39759         PR target/79179
39760         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39761         constraint instead of o for the stxsd instruction.
39763 2017-01-25  Carl Love  <cel@us.ibm.com>
39765         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39766         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39768 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39770         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39772 2017-01-25  Richard Biener  <rguenther@suse.de>
39774         PR tree-optimization/69264
39775         * target.def (vector_alignment_reachable): Improve documentation.
39776         * doc/tm.texi: Regenerate.
39777         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39778         and add a comment.
39779         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39780         earlier changes with respect to TYPE_USER_ALIGN.
39781         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39783 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39785         PR target/79145
39786         * config/arm/arm.md (xordi3): Force constant operand into a register
39787         for TARGET_IWMMXT.
39789 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39791         * doc/invoke.texi (-fstore-merging): Correct default optimization
39792         levels at which it is enabled.
39793         (-O): Move -fstore-merging from list to...
39794         (-O2): ... Here.
39796 2017-01-25  Richard Biener  <rguenther@suse.de>
39798         PR debug/78363
39799         * omp-expand.c: Include debug.h.
39800         (expand_omp_taskreg): Make sure to generate early debug before
39801         outlining anything from a function.
39802         (expand_omp_target): Likewise.
39803         (grid_expand_target_grid_body): Likewise.
39805 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39807         PR lto/79061
39808         * asan.c (get_translation_unit_decl): New function.
39809         (asan_add_global): Extract modules file name from globals
39810         TRANSLATION_UNIT_DECL name.
39812 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39814         PR target/77439
39815         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39816         for long calls with APCS frame and VFP.
39818 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39820         * cfg.c (original_copy_tables_initialized_p): New function.
39821         * cfg.h (original_copy_tables_initialized_p): New decl.
39822         * cfgrtl.c (relink_block_chain): Guard the call to
39823         free_original_copy_tables with a call to
39824         original_copy_tables_initialized_p.
39825         * cgraph.h (symtab_node::native_rtl_p): New decl.
39826         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39827         (symtab_node::needed_p): Don't assert for early assembly output
39828         for __RTL functions.
39829         (cgraph_node::finalize_function): Set "force_output" for __RTL
39830         functions.
39831         (cgraph_node::analyze): Bail out early for __RTL functions.
39832         (analyze_functions): Update assertion to support __RTL functions.
39833         (cgraph_node::expand): Bail out early for __RTL functions.
39834         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39835         __RTL functions.
39836         * function.h (struct function): Update comment for field
39837         "pass_startwith".
39838         * gimple-expr.c: Include "tree-pass.h".
39839         (gimple_has_body_p): Return false for __RTL functions.
39840         * Makefile.in (OBJS): Add run-rtl-passes.o.
39841         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39842         accessor.
39843         (gcc::pass_manager::get_clean_slate): New accessor.
39844         * passes.c: Include "insn-addr.h".
39845         (should_skip_pass_p): Add logging.  Update logic for running
39846         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39847         property-provider override so it is only done for gimple passes.
39848         Don't skip dfinit.
39849         (skip_pass): New function.
39850         (execute_one_pass): Call skip_pass when skipping passes.
39851         * read-md.c (md_reader::read_char): Support filtering
39852         the input to a subset of line numbers.
39853         (md_reader::md_reader): Initialize fields
39854         m_first_line and m_last_line.
39855         (md_reader::read_file_fragment): New function.
39856         * read-md.h (md_reader::read_file_fragment): New decl.
39857         (md_reader::m_first_line): New field.
39858         (md_reader::m_last_line): New field.
39859         * read-rtl-function.c (function_reader::create_function): Only
39860         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39861         curr_properties.  Set DECL_INITIAL to a dummy block.
39862         (read_rtl_function_body_from_file_range): New function.
39863         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39864         New decl.
39865         * run-rtl-passes.c: New file.
39866         * run-rtl-passes.h: New file.
39868 2017-01-24  Jeff Law  <law@redhat.com>
39870         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39871         buffer size.
39873 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39875         PR tree-optimization/79159
39876         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39877         (record_nonwrapping_iv): Improve boundary using above function if no
39878         value range information.
39880 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39881             Martin Jambor  <mjambor@suse.cz>
39883         * brig-builtins.def: New file.
39884         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39885         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39886         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39887         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39888         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39889         * builtin-types.def (BT_INT8): New.
39890         (BT_INT16): Likewise.
39891         (BT_UINT8): Likewise.
39892         (BT_UINT16): Likewise.
39893         (BT_FN_ULONG): Likewise.
39894         (BT_FN_UINT_INT): Likewise.
39895         (BT_FN_UINT_ULONG): Likewise.
39896         (BT_FN_UINT_LONG): Likewise.
39897         (BT_FN_UINT_PTR): Likewise.
39898         (BT_FN_ULONG_PTR): Likewise.
39899         (BT_FN_INT8_FLOAT): Likewise.
39900         (BT_FN_INT16_FLOAT): Likewise.
39901         (BT_FN_UINT32_FLOAT): Likewise.
39902         (BT_FN_UINT16_FLOAT): Likewise.
39903         (BT_FN_UINT8_FLOAT): Likewise.
39904         (BT_FN_UINT64_FLOAT): Likewise.
39905         (BT_FN_UINT16_UINT32): Likewise.
39906         (BT_FN_UINT32_UINT16): Likewise.
39907         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39908         (BT_FN_INT_PTR_INT): Likewise.
39909         (BT_FN_UINT_PTR_UINT): Likewise.
39910         (BT_FN_LONG_PTR_LONG): Likewise.
39911         (BT_FN_ULONG_PTR_ULONG): Likewise.
39912         (BT_FN_VOID_UINT64_UINT64): Likewise.
39913         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39914         (BT_FN_INT8_INT8_INT8): Likewise.
39915         (BT_FN_INT16_INT16_INT16): Likewise.
39916         (BT_FN_INT_INT_INT): Likewise.
39917         (BT_FN_UINT_FLOAT_UINT): Likewise.
39918         (BT_FN_FLOAT_UINT_UINT): Likewise.
39919         (BT_FN_ULONG_UINT_UINT): Likewise.
39920         (BT_FN_ULONG_UINT_PTR): Likewise.
39921         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39922         (BT_FN_UINT_UINT_UINT): Likewise.
39923         (BT_FN_VOID_UINT_PTR): Likewise.
39924         (BT_FN_UINT_UINT_PTR: Likewise.
39925         (BT_FN_UINT32_UINT64_PTR): Likewise.
39926         (BT_FN_INT_INT_UINT_UINT): Likewise.
39927         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39928         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39929         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39930         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39931         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39932         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39933         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39934         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39935         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39936         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39937         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39938         * doc/frontends.texi: List BRIG FE.
39939         * doc/install.texi (Testing): Add BRIG tesring requirements.
39940         * doc/invoke.texi (Overall Options): Mention BRIG.
39941         * doc/standards.texi (Standards): Doucment BRIG HSA version.
39943 2017-01-24  Richard Biener  <rguenther@suse.de>
39945         PR translation/79208
39946         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
39948 2017-01-24  Martin Jambor  <mjambor@suse.cz>
39950         PR bootstrap/79198
39951         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
39952         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
39953         and known_contexts.
39955 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
39957         PR middle-end/79123
39958         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
39959         casts from signed to unsigned really don't have a range.
39961 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
39963         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
39964         GMP_RNDx for compatiblity.
39966 2017-01-24  Martin Liska  <mliska@suse.cz>
39968         PR bootstrap/79132
39969         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
39970         that would prevent us to call alloca with -1 as argument.
39972 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
39974         * dwarf2out.c (output_compilation_unit_header, output_file_names):
39975         Avoid -Wformat-security warning.
39977 2017-01-23  Andrew Pinski  <apinski@cavium.com>
39979         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
39980         cost table.
39982 2017-01-23  Martin Sebor  <msebor@redhat.com>
39984         PR middle-end/78703
39985         * gimple-ssa-sprintf.c (warn_level): New global.
39986         (format_integer): Use it here and throughout the rest of the file.
39987         Use the same switch to compute sign as base.
39988         (maybe_warn): New function.
39989         (format_directive): Factor out warnings into maybe_warn.
39990         Add debugging output.  Use warn_level.
39991         (add_bytes): Use warn_level.
39992         (pass_sprintf_length::compute_format_length): Add debugging output.
39993         (try_substitute_return_value): Same.
39994         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
39996         PR middle-end/78703
39997         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
39998         (struct fmtresult, format_integer, format_floating): Adjust.
39999         (fmtresult::fmtresult): Set max correctly in two argument ctor.
40000         (get_string_length, format_string,format_directive): Same.
40001         (pass_sprintf_length::compute_format_length): Same.
40002         (try_substitute_return_value): Simplify slightly.
40004         PR middle-end/78703
40005         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
40006         (fmtresult::operator+=): Outlined.
40007         (struct fmtresult): Add ctors.
40008         (struct conversion_spec): Rename...
40009         (struct directive): ...to this.  Add and remove data members.
40010         (directive::set_width, directive::set_precision): New functions.
40011         (format_percent): Use fmtresult ctor.
40012         (get_width_and_precision): Remove.
40013         (format_integer): Make naming changes.  Avoid computing width and
40014         precision.
40015         (format_floating): Same.  Adjust indentation.
40016         (format_character, format_none): New functions.
40017         (format_string): Moved character handling to format_character.
40018         (format_directive): Remove arguments, change return type.
40019         (parse_directive): New function.
40020         (pass_sprintf_length::compute_format_length): Move directive
40021         parsing to parse_directive.
40023 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40025         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40026         (assign_assembler_name_if_needed): ... this.
40027         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40028         (assign_assembler_name_if_needed): ... this.
40029         (free_lang_data_in_cgraph): Adjust callers.
40030         * cgraphunit.c (cgraph_node::analyze): Likewise.
40031         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40032         Likewise.
40034 2017-01-23  Richard Biener  <rguenther@suse.de>
40036         PR tree-optimization/79088
40037         PR tree-optimization/79188
40038         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40039         resetting loop bounds after last path deletion.  Reset loop
40040         bounds of the target loop, make code match the comments.
40041         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40042         Make sure loops need no fixups.
40044 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40046         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40047         exponent support with double type for first argument.
40048         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40049         type returned by __builtin_vec_extract_sig,
40050         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40051         functions from "vector int" to "vector unsigned int" or from
40052         "vector long long int" to "vector unsigned long long int".
40053         Changed type returned by __builtin_vec_extract_exp,
40054         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40055         functions from "vector int" to "vector unsigned int" or from
40056         "vector long long int" to "vector unsigned long long int".
40057         Changed return type of __builtin_vec_test_data_class,
40058         __builtin_vec_test_data_class_sp, and
40059         __builtin_vec_test_data_class_dp from "vector int" to
40060         "vector bool int" or from "vector long long int" to "vector bool
40061         long long int" and changed second argument type from "unsigned
40062         int" to "int".  Added new overloaded function forms "vector float
40063         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40064         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40065         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40066         double, vector unsigned long long int)" and "vector double
40067         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40068         long int)".  Changed return type of
40069         __builtin_scalar_test_data_class and
40070         __builtin_scalar_test_data_class_sp and
40071         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40072         int" and changed second argument from "unsigned int" to "int".
40073         Changed type returned by __builtin_scalar_test_neg,
40074         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40075         from "int" to "bool int".  Added new overloaded function form
40076         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40077         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40078         exponent double-precision with floating point first argument.
40079         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40080         documentation of scalar_test_data_class, scalar_test_neg,
40081         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40082         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40083         vec_test_data_class built-in functions to reflect refinements in
40084         their type signatures.
40086 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40088         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40089         size of buf.
40090         (aarch64_elf_asm_destructor): Likewise.
40092 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40094         PR rtl-optimization/78634
40095         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40096         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40097         * ifcvt.c (noce_try_cmove): Add missing cost check.
40099         PR rtl-optimization/71724
40100         * combine.c (if_then_else_cond): Look for situations where it is
40101         beneficial to undo the work of one of the recursive calls.
40103 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40105         PR tree-optimization/70754
40106         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40107         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40108         combined stmt before it if not NULL.
40109         (combine_chains): Process refs reversely and compute dominance point
40110         for root ref.
40112 2017-01-23  Martin Liska  <mliska@suse.cz>
40114         PR tree-optimization/79196
40115         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40116         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40117         instead of memcmp.
40118         (strlen_optimize_stmt): Call the renamed function.
40120 2017-01-23  Michael Matz  <matz@suse.de>
40122         PR tree-optimization/78384
40123         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40125 2017-01-23  Richard Biener  <rguenther@suse.de>
40127         PR tree-optimization/79186
40128         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40129         both incoming edges before moving an assert.
40131 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40133         * ipa-prop.c (load_from_param_1): Removed.
40134         (load_from_unmodified_param): Bits from load_from_param_1 put back
40135         here.
40136         (load_from_param): Removed.
40137         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40138         with stmt.  Reverted back to use of load_from_unmodified_param.
40140 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40142         PR ipa/79108
40143         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40144         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40145         field a pointer to garbage collected vector, mark lattices and
40146         ipcp_orig_node with GTY((skip)).
40147         (ipa_get_param_count): Adjust to descriptors being a pointer.
40148         (ipa_get_param): Likewise.
40149         (ipa_get_type): Likewise.
40150         (ipa_get_param_move_cost): Likewise.
40151         (ipa_set_param_used): Likewise.
40152         (ipa_get_controlled_uses): Likewise.
40153         (ipa_set_controlled_uses): Likewise.
40154         (ipa_is_param_used): Likewise.
40155         (ipa_node_params_t): Move into garbage collector.  New methods insert
40156         and remove.
40157         (ipa_node_params_sum): Annotate wth GTY(()).
40158         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40159         garbage collected.
40160         (ipa_load_from_parm_agg): Adjust declaration.
40161         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40162         * ipa-profile.c (ipa_profile): Likewise.
40163         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40164         (ipa_populate_param_decls): Make descriptors parameter garbage
40165         collected.
40166         (ipa_dump_param): Adjust to descriptors being a pointer.
40167         (ipa_alloc_node_params): Likewise.
40168         (ipa_initialize_node_params): Likewise.
40169         (load_from_param_1): Make descriptors parameter garbage collected.
40170         (load_from_unmodified_param): Likewise.
40171         (load_from_param): Likewise.
40172         (ipa_load_from_parm_agg): Likewise.
40173         (ipa_node_params::~ipa_node_params): Removed.
40174         (ipa_free_all_node_params): Remove call to delete operator.
40175         (ipa_node_params_t::insert): New.
40176         (ipa_node_params_t::remove): Likewise.
40177         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40178         copy known_csts and known_contexts vectors.
40179         (ipa_read_node_info): Adjust to descriptors being a pointer.
40180         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40181         collected.
40182         (ipcp_transform_function): Make descriptors variable garbage
40183         collected.
40185 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40187         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40188         * config/i386/avx512dqintrin.h: Ditto.
40189         * config/i386/avx512fintrin.h: Ditto.
40190         * config/i386/i386.c: Handle new builtins.
40191         * config/i386/i386-builtin.def: Add new builtins.
40192         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40193         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40195 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40196             Martin Liska  <mliska@suse.cz>
40198         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40199         * asan.c (asan_expand_poison_ifn): Support stores and use
40200         appropriate ASAN report function.
40201         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40202         * internal-fn.def (ASAN_POISON_USE): Declare.
40203         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40204         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40205         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40206         ASAN_POISON calls w/o LHS.
40207         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40208         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40209         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40210         * gimplify.c (asan_poison_variables): Add attribute
40211         use_after_scope_memory to variables that really needs to live
40212         in memory.
40213         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40214         having the attribute.
40216 2017-01-23  Martin Liska  <mliska@suse.cz>
40218         * asan.c (create_asan_shadow_var): New function.
40219         (asan_expand_poison_ifn): Likewise.
40220         * asan.h (asan_expand_poison_ifn): New declaration.
40221         * internal-fn.c (expand_ASAN_POISON): Likewise.
40222         * internal-fn.def (ASAN_POISON): New builtin.
40223         * sanopt.c (pass_sanopt::execute): Expand
40224         asan_expand_poison_ifn.
40225         * tree-inline.c (copy_decl_for_dup_finish): Make function
40226         external.
40227         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40228         * tree-ssa.c (is_asan_mark_p): New function.
40229         (execute_update_addresses_taken): Rewrite local variables
40230         (identified just by use-after-scope as addressable) into SSA.
40232 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40234         * doc/install.texi (Specific): opensource.apple.com uses https
40235         now. Remove trailing slash.
40237 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40239         * README.Portability: Remove note on an Irix compatibility issue.
40241 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40243         * gcov.c (INCLUDE_ALGORITHM): Define.
40244         (INCLUDE_VECTOR): Define.
40245         No longer include <vector> and <algorithm> directly.
40247 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40249         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40250         to https.
40251         * doc/invoke.texi (Code Gen Options): Ditto.
40253 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40255         PR lto/78407
40256         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40258 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40260         rtl-optimization/79125
40261         * cprop.c (local_cprop_pass): Handle cases where we make an
40262         unconditional trap.
40264 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40266         PR target/61729
40267         PR target/77850
40268         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40269         read from, for big endian.
40271 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40273         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40274         register pauth builtins for LP64 only.
40276 2017-01-20  Marek Polacek  <polacek@redhat.com>
40278         PR c/79152
40279         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40280         non-case labels.
40282 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40284         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40285         of safelen status.
40286         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40287         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40288         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40290 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40292         PR target/71270
40293         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40294         in big-endian mode when they are not a single duplicated value.
40296 2017-01-20  Richard Biener  <rguenther@suse.de>
40298         * BASE-VER: Bump to 7.0.1.
40300 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40302         * omp-low.c (omplow_simd_context): New struct.  Use it...
40303         (lower_rec_simd_input_clauses): ...here and...
40304         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40305         references to idx, lane, max_vf, is_simt.
40307 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40309         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40310         mcpu=nps400.
40312 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40314         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40315         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40316         gt-hsa-common.h.
40317         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40318         (GTFILES): Rename hsa.c to hsa-common.c.
40319         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40320         * hsa-dump.c: Likewise.
40321         * hsa-gen.c: Likewise.
40322         * hsa-regalloc.c: Likewise.
40323         * ipa-hsa.c: Likewise.
40324         * omp-expand.c: Likewise.
40325         * omp-low.c: Likewise.
40326         * toplev.c: Likewise.
40328 2017-01-20  Marek Polacek  <polacek@redhat.com>
40330         PR c/64279
40331         * doc/invoke.texi: Document -Wduplicated-branches.
40332         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40333         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40334         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40335         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40336         return 0 only when not OEP_LEXICOGRAPHIC.
40337         (fold_build_cleanup_point_expr): Use the expression
40338         location when building CLEANUP_POINT_EXPR.
40339         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40340         * tree.c (add_expr): Handle error_mark_node.
40342 2017-01-20  Martin Liska  <mliska@suse.cz>
40344         PR lto/69188
40345         * tree-profile.c (init_ic_make_global_vars): Do not call
40346         finalize_decl.
40347         (gimple_init_gcov_profiler): Likewise.
40349 2017-01-20  Martin Liska  <mliska@suse.cz>
40351         PR ipa/71190
40352         * cgraph.h (maybe_create_reference): Remove argument and
40353         update comment.
40354         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40355         argument.
40356         * ipa-cp.c (create_specialized_node): Likewise.
40357         * symtab.c (symtab_node::maybe_create_reference): Handle
40358         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40360 2017-01-20  Martin Liska  <mliska@suse.cz>
40362         * read-rtl-function.c (function_reader::create_function): Use
40363         build_decl instread of build_decl_stat.
40365 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40367         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40368         * config/i386/avx512dqintrin.h: Ditto.
40369         * config/i386/avx512fintrin.h: Ditto.
40370         * config/i386/i386-builtin-types.def: Add new types.
40371         * config/i386/i386.c: Handle new types.
40372         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40373         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40374         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40375         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40376         (__builtin_ia32_kshiftridi): New.
40377         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40379 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40381         PR target/78875
40382         PR target/79140
40383         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40384         define to rs6000_init_stack_protect_guard.
40385         (rs6000_init_stack_protect_guard): New function.
40387 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40388             Yunqiang Su  <yunqiang.su@imgtec.com>
40390         * config.gcc (supported_defaults): Add madd4.
40391         (with_madd4): Add validation.
40392         (all_defaults): Add madd4.
40393         * config/mips/mips.opt (mmadd4): New option.
40394         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40395         mmadd4.
40396         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40397         __mips_no_madd4.
40398         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40399         (ISA_HAS_FUSED_MADD4): Likewise.
40400         * doc/invoke.texi (-mmadd4): Document the new option.
40401         * doc/install.texi (--with-madd4): Document the new option.
40403 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40405         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40406         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40407         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40408         (aarch64_init_pauth_hint_builtins): New.
40409         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40410         (aarch64_expand_builtin): Expand new builtins.
40412 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40414         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40415         * combine-stack-adj.c (no_unhandled_cfa): Handle
40416         REG_CFA_TOGGLE_RA_MANGLE.
40417         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40418         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40419         info for return address signing.
40420         (aarch64_expand_epilogue): Likewise.
40422 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40424         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40425         * config/aarch64/aarch64-protos.h
40426         (aarch64_return_address_signing_enabled): New declaration.
40427         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40428         New function.
40429         (aarch64_expand_prologue): Sign return address before it's pushed onto
40430         stack.
40431         (aarch64_expand_epilogue): Authenticate return address fetched from
40432         stack.
40433         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40434         (aarch64_attributes): New function attributes for "sign-return-address".
40435         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40436         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40437         ("*do_return"): Generate combined instructions according to key index.
40438         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40439         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40440         iterators.
40441         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40442         * config/aarch64/aarch64.opt (msign-return-address=): New.
40443         * doc/extend.texi (AArch64 Function Attributes): Documents
40444         "sign-return-address=".
40445         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40447 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40449         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40450         overall option summary.
40452 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40454         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40455         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40456         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40457         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40459 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40461         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40462         -mpower9-minmax by default for -mcpu=power9.
40463         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40464         128-bit floating point.
40466 2017-01-20  Alan Modra  <amodra@gmail.com>
40468         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40469         optimizing for size.
40471 2017-01-20  Alan Modra  <amodra@gmail.com>
40473         PR target/79144
40474         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40475         for strcmp and strncmp from corresponding builtin decl.
40477 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40479         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40480         instead of i386/rtems-64.h.
40481         * config/i386/rtems-64.h: Remove.
40483 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40485         PR target/78478
40486         Revert:
40487         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40489         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40491 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40493         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40494         Change int to HOST_WIDE_INT.
40495         * config/aarch64/aarch64-protos.h
40496         (aarch64_simd_gen_const_vector_dup): Likewise.
40497         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40499 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40501         * langhooks-def.h (lhd_type_for_size): New decl.
40502         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40503         * langhooks.c (lhd_type_for_size): New function, taken from
40504         lto_type_for_size.
40506 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40508         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40509         define_bypass for CR latency.
40510         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40511         (power9-alu2): Add define_bypass for CR latency.
40512         (power9-cmp): New.
40513         (power9-mul): Update insn latency.
40514         (power9-mul-compare): Update insn latency, bypass latency and remove
40515         power9-branch.
40517 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40519         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40520         Delete.
40521         * config/aarch64/aarch64.md
40522         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40523         aarch64_nopcrelative_literal_loads.
40524         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40526 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40528         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40529         TARGET_LOONGSON_3A.
40530         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40532 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40534         PR target/78176
40535         * config.gcc (supported_defaults): Add lxc1-sxc1.
40536         (with_lxc1_sxc1): Add validation.
40537         (all_defaults): Add lxc1-sxc1.
40538         * config/mips/mips.opt (mlxc1-sxc1): New option.
40539         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40540         mlxc1-sxc1.
40541         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40542         __mips_no_lxc1_sxc1.
40543         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40544         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40545         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40547 2017-01-19  Richard Biener  <rguenther@suse.de>
40549         PR tree-optimization/72488
40550         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40551         sure to restore SSA info.
40552         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40554 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40556         PR rtl-optimization/79121
40557         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40558         of the inner type when shifting an extended value.
40560 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40562         PR lto/78407
40563         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40564         interposable aliases.
40566 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40568         PR target/78516
40569         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40570         Use the evmergelohi instruction.
40571         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40572         (mov_sitf_e500_subreg8_2_be): Likewise.
40573         (mov_sitf_e500_subreg12_2_le): Likewise.
40574         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40575         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40576         (mov_sitf_e500_subreg8_2_le): Likewise.
40577         (mov_sitf_e500_subreg12_2_be): Likewise.
40579 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40581         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40582         attribute from vecsimple to vecperm.
40583         (altivec_vbpermq2): Likewise.
40585 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40587         PR target/79040
40588         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40590 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40591         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40592         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40593         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40594         case where N arg is SIZE_MAX.
40595         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40596         (cmpstrsi): Add pattern.
40598 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40600         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40601         __builtin_vec_revb builtins.
40602         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40603         built-in functions to support generation of the ISA 3.0 XXBR<x>
40604         vector byte reverse instructions.
40605         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40606         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40607         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40608         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40609         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40610         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40611         (P9V_BUILTIN_VEC_REVB): Likewise.
40612         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40613         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40614         (p9_xxbrq_v16qi): Likewise.
40615         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40616         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40617         (p9_xxbrh_v8hi): Likewise.
40618         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40619         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40620         vec_revb built-in functions.
40622 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40624         PR rtl-optimization/78952
40625         * config/i386/i386.md (any_extract): New code iterator.
40626         (*insvqi_2): Use any_extract for source operand.
40627         (*insvqi_3): Use any_shiftrt for source operand.
40629 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40631         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40632         New function.
40633         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40635 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40637         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40639 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40641         * config/rs6000/altivec.h (vec_bperm): Change #define.
40642         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40643         (altivec_vbpermq2): New define_insn.
40644         (altivec_vbpermd): Likewise.
40645         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40646         function interface.
40647         (VBPERMD): Likewise.
40648         (VBPERM): New polymorphic function interface.
40649         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40650         Add entries for P9V_BUILTIN_VEC_VBPERM.
40651         * doc/extend.texi: Add interfaces for vec_bperm.
40653 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40655         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40656         first letter of error messages.
40657         (s390_resolve_overloaded_builtin): Likewise.
40658         * config/s390/s390.c (s390_expand_builtin): Likewise.
40659         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40660         (s390_valid_target_attribute_inner_p): Likewise.
40661         * config/s390/s390.md ("tabort"): Likewise.
40663 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40665         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40666         (ISA_AVOID_DIV_HILO): New macro.
40667         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40668         (ISA_HAS_DDIV): Likewise.
40670 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40672         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40674 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40676         * doc/invoke.texi (fabi-version): Spelling fix.
40678 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40680         PR c++/70182
40681         * doc/invoke.texi (fabi-version): Mention mangling fix for
40682         operator names.
40684 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40686         PR c++/77489
40687         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40689 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40691         PR target/78875
40692         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40693         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40694         the new options.
40695         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40696         flexible settings.
40697         (stack_protect_test): Ditto.
40698         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40699         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40700         options.
40701         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40702         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40703         -mstack-protector-guard-offset=.
40704         (RS/6000 and PowerPC Options): Ditto.
40706 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40708         * config/i386/i386.h (MASK_CLASS_P): New define.
40709         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40710         there are no registers from different register sets also when
40711         mask registers are used.  Update function comment.
40712         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40713         to (*k/*r) and (*k/*km) alternatives.
40715 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40717         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40718         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40719         (EH_RETURN_HANDLER_RTX): New define.
40720         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40721         Force frame pointer in EH return functions.
40722         (aarch64_expand_epilogue): Add barrier for eh_return.
40723         (aarch64_final_eh_return_addr): Remove.
40724         (aarch64_eh_return_handler_rtx): New function.
40725         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40726         Remove.
40727         (aarch64_eh_return_handler_rtx): New prototype.
40729 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40731         * config/rs6000/altivec.h (vec_rlmi): New #define.
40732         (vec_vrlnm): Likewise.
40733         (vec_rlnm): Likewise.
40734         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40735         (UNSPEC_VRLNM): Likewise.
40736         (VIlong): New mode iterator.
40737         (altivec_vrl<VI_char>mi): New define_insn.
40738         (altivec_vrl<VI_char>nm): Likewise.
40739         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40740         function entry.
40741         (VRLDNM): Likewise.
40742         (RLNM): New polymorphic function entry.
40743         (VRLWMI): New monomorphic function entry.
40744         (VRLDMI): Likewise.
40745         (RLMI): New polymorphic function entry.
40746         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40747         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40748         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40749         vec_vrlnm.
40751 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40753         PR debug/78839
40754         * dwarf2out.c (field_byte_offset): Restore the
40755         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40756         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40757         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40758         of build2 + fold.
40760 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40762         PR ada/67205
40763         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40765 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40767         PR debug/71669
40768         * dwarf2out.c (add_data_member_location_attribute): For constant
40769         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40770         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40771         DW_AT_byte_size attributes.
40773 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40775         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40776         after forcing to constant memory when the code model is medium.
40778 2017-01-17  Julia Koval  <julia.koval@intel.com>
40780         PR target/76731
40781         * config/i386/avx512fintrin.h
40782         (_mm512_i32gather_ps): Change __addr type to void const*.
40783         (_mm512_mask_i32gather_ps): Ditto.
40784         (_mm512_i32gather_pd): Ditto.
40785         (_mm512_mask_i32gather_pd): Ditto.
40786         (_mm512_i64gather_ps): Ditto.
40787         (_mm512_mask_i64gather_ps): Ditto.
40788         (_mm512_i64gather_pd): Ditto.
40789         (_mm512_mask_i64gather_pd): Ditto.
40790         (_mm512_i32gather_epi32): Ditto.
40791         (_mm512_mask_i32gather_epi32): Ditto.
40792         (_mm512_i32gather_epi64): Ditto.
40793         (_mm512_mask_i32gather_epi64): Ditto.
40794         (_mm512_i64gather_epi32): Ditto.
40795         (_mm512_mask_i64gather_epi32): Ditto.
40796         (_mm512_i64gather_epi64): Ditto.
40797         (_mm512_mask_i64gather_epi64): Ditto.
40798         (_mm512_i32scatter_ps): Change __addr type to void*.
40799         (_mm512_mask_i32scatter_ps): Ditto.
40800         (_mm512_i32scatter_pd): Ditto.
40801         (_mm512_mask_i32scatter_pd): Ditto.
40802         (_mm512_i64scatter_ps): Ditto.
40803         (_mm512_mask_i64scatter_ps): Ditto.
40804         (_mm512_i64scatter_pd): Ditto.
40805         (_mm512_mask_i64scatter_pd): Ditto.
40806         (_mm512_i32scatter_epi32): Ditto.
40807         (_mm512_mask_i32scatter_epi32): Ditto.
40808         (_mm512_i32scatter_epi64): Ditto.
40809         (_mm512_mask_i32scatter_epi64): Ditto.
40810         (_mm512_i64scatter_epi32): Ditto.
40811         (_mm512_mask_i64scatter_epi32): Ditto.
40812         (_mm512_i64scatter_epi64): Ditto.
40813         (_mm512_mask_i64scatter_epi64): Ditto.
40814         * config/i386/avx512pfintrin.h
40815         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40816         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40817         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40818         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40819         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40820         (_mm512_prefetch_i32scatter_ps): Ditto.
40821         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40822         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40823         (_mm512_prefetch_i64scatter_pd): Ditto.
40824         (_mm512_prefetch_i64scatter_ps): Ditto.
40825         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40826         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40827         * config/i386/avx512vlintrin.h
40828         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40829         (_mm_mmask_i32gather_ps): Ditto.
40830         (_mm256_mmask_i32gather_pd): Ditto.
40831         (_mm_mmask_i32gather_pd): Ditto.
40832         (_mm256_mmask_i64gather_ps): Ditto.
40833         (_mm_mmask_i64gather_ps): Ditto.
40834         (_mm256_mmask_i64gather_pd): Ditto.
40835         (_mm_mmask_i64gather_pd): Ditto.
40836         (_mm256_mmask_i32gather_epi32): Ditto.
40837         (_mm_mmask_i32gather_epi32): Ditto.
40838         (_mm256_mmask_i32gather_epi64): Ditto.
40839         (_mm_mmask_i32gather_epi64): Ditto.
40840         (_mm256_mmask_i64gather_epi32): Ditto.
40841         (_mm_mmask_i64gather_epi32): Ditto.
40842         (_mm256_mmask_i64gather_epi64): Ditto.
40843         (_mm_mmask_i64gather_epi64): Ditto.
40844         (_mm256_i32scatter_ps): Change __addr type to void*.
40845         (_mm256_mask_i32scatter_ps): Ditto.
40846         (_mm_i32scatter_ps): Ditto.
40847         (_mm_mask_i32scatter_ps): Ditto.
40848         (_mm256_i32scatter_pd): Ditto.
40849         (_mm256_mask_i32scatter_pd): Ditto.
40850         (_mm_i32scatter_pd): Ditto.
40851         (_mm_mask_i32scatter_pd): Ditto.
40852         (_mm256_i64scatter_ps): Ditto.
40853         (_mm256_mask_i64scatter_ps): Ditto.
40854         (_mm_i64scatter_ps): Ditto.
40855         (_mm_mask_i64scatter_ps): Ditto.
40856         (_mm256_i64scatter_pd): Ditto.
40857         (_mm256_mask_i64scatter_pd): Ditto.
40858         (_mm_i64scatter_pd): Ditto.
40859         (_mm_mask_i64scatter_pd): Ditto.
40860         (_mm256_i32scatter_epi32): Ditto.
40861         (_mm256_mask_i32scatter_epi32): Ditto.
40862         (_mm_i32scatter_epi32): Ditto.
40863         (_mm_mask_i32scatter_epi32): Ditto.
40864         (_mm256_i32scatter_epi64): Ditto.
40865         (_mm256_mask_i32scatter_epi64): Ditto.
40866         (_mm_i32scatter_epi64): Ditto.
40867         (_mm_mask_i32scatter_epi64): Ditto.
40868         (_mm256_i64scatter_epi32): Ditto.
40869         (_mm256_mask_i64scatter_epi32): Ditto.
40870         (_mm_i64scatter_epi32): Ditto.
40871         (_mm_mask_i64scatter_epi32): Ditto.
40872         (_mm256_i64scatter_epi64): Ditto.
40873         (_mm256_mask_i64scatter_epi64): Ditto.
40874         (_mm_i64scatter_epi64): Ditto.
40875         (_mm_mask_i64scatter_epi64): Ditto.
40876         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40877         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40878         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40879         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40880         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40881         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40882         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40883         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40884         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40885         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40886         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40887         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40888         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40889         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40890         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40891         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40892         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40893         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40894         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40895         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40896         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40897         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40898         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40899         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40900         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40901         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40902         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40903         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40904         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40905         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40906         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40907         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40908         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40909         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40910         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40911         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40912         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40913         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40914         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40915         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40916         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40917         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40918         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40919         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40920         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40921         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40922         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40923         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40924         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40925         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40926         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40927         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40928         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40929         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40930         definitions accordingly.
40932 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
40933             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
40935         PR target/79079
40936         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40937         gen_lowpart.
40939 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
40941         PR target/79058
40942         * ira-conflicts.c (ira_build_conflicts): Update total conflict
40943         hard regs for inner regno.
40945 2017-01-17  Martin Liska  <mliska@suse.cz>
40947         PR ipa/71207
40948         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
40949         assumption and add comment.
40951 2017-01-17  Nathan Sidwell  <nathan@acm.org>
40953         * ipa-visibility.c (localize_node): New function, broken out of ...
40954         (function_and_variable_visibility): ... here. Call it.
40956 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40958         PR middle-end/77445
40959         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
40960         correctly set frequency of oudgoing edge.
40961         (duplicate_thread_path): Fix profile updating.
40963 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40965         PR other/79046
40966         * configure.ac: Add GCC_BASE_VER.
40967         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
40968         version from BASE-VER file.
40969         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
40970         (gcc.o): Depend on $(BASEVER).
40971         * common.opt (dumpfullversion): New option.
40972         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
40973         * doc/invoke.texi: Document -dumpfullversion.
40974         * doc/install.texi: Document --with-gcc-major-version-only.
40975         * configure: Regenerated.
40977 2017-01-17  Richard Biener  <rguenther@suse.de>
40979         PR tree-optimization/71433
40980         * tree-vrp.c (register_new_assert_for): Merge same asserts
40981         on all incoming edges.
40982         (process_assert_insertions_for): Handle insertions at the
40983         beginning of BBs.
40985 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
40987         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
40988         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
40990 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
40992         PR target/78633
40993         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
40994         RTL sharing.
40996 2017-01-17  Alan Modra  <amodra@gmail.com>
40998         PR target/79066
40999         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
41000         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
41001         symbolic stack limit when pic.
41003 2017-01-16  Martin Sebor  <msebor@redhat.com>
41005         PR tree-optimization/78608
41006         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
41008 2017-01-16  Jeff Law  <law@redhat.com>
41010         Revert:
41011         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
41012         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
41013         for several include directories that may be relative to sysroot.
41014         * config/i386/x-mingw32 (gplus_includedir): Define.
41015         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
41016         (native_system_includedir): Likewise.
41017         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
41018         override if TARGET_SYSTEM_ROOT is defined.
41019         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
41021         PR tree-optimization/79090
41022         PR tree-optimization/33562
41023         PR tree-optimization/61912
41024         PR tree-optimization/77485
41025         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41026         and computed trims into the dump file.
41028 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41030         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41032 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41034         PR c/79089
41035         * gimplify.c (gimplify_init_constructor): If want_value and
41036         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41037         fix.
41039         PR target/79080
41040         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41041         sequence.  Formatting fixes.
41042         (doloop_optimize): Formatting fixes.
41044         PR driver/49726
41045         * gcc.c (debug_level_greater_than_spec_func): New function.
41046         (static_spec_functions): Add debug-level-gt spec function.
41047         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41048         !g0.
41049         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41050         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41051         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41052         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41053         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41054         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41056 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41058         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41059         QImode fixups to general and mask registers only.
41061 2017-01-16  Carl Love  <cel@us.ibm.com>
41063         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41064         for built-in functions
41065         vector signed char vec_nabs (vector signed char)
41066         vector signed short vec_nabs (vector signed short)
41067         vector signed int vec_nabs (vector signed int)
41068         vector signed long long vec_nabs (vector signed long long)
41069         vector float vec_nabs (vector float)
41070         vector double vec_nabs (vector double)
41071         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41072         and NABS overload.
41073         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41074         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41075         * doc/extend.texi: Update the documentation file for the new built-in
41076         functions.
41078 2017-01-16  Martin Sebor  <msebor@redhat.com>
41080         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41081         message.
41083 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41085         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41086         UNSPEC_VSX__XXSPLTD to require special splat handling.
41088 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41090         PR bootstrap/78616
41091         * system.h: Poison strndup.
41093 2017-01-16  Alan Modra  <amodra@gmail.com>
41095         PR target/79098
41096         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41097         use a switch.
41099 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41101         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41103 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41105         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41106         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41108 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41110         PR target/72749
41111         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41112         fallthrough.
41113         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41114         in the currently scheduled RTL fragment.
41116 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41118         PR rtl-optimization/78751
41119         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41120         give up.
41122 2017-01-14  Jeff Law  <law@redhat.com>
41124         PR tree-optimization/79090
41125         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41126         variable length stores.
41127         (compute_trims): Delete dead assignment to *trim_tail.
41128         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41129         zero length.
41131 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41133         PR rtl-optimization/78626
41134         PR rtl-optimization/78727
41135         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41136         of a block, and split such blocks after everything else is finished.
41138 2017-01-14  Alan Modra  <amodra@gmail.com>
41140         PR target/72749
41141         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41142         target legitimate_combined_insn.
41143         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41144         (rs6000_legitimate_combined_insn): New function.
41145         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41146         all uses.
41147         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41148         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41149         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41151 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41153         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41155 2017-01-13  Jeff Law  <law@redhat.com>
41157         PR tree-optimization/33562
41158         PR tree-optimization/61912
41159         PR tree-optimization/77485
41160         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41161         a statement.
41162         (delete_dead_assignment): Likewise.
41163         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41164         statement to delete_dead_call and delete_dead_assignment.
41166 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41168         PR c/78304
41169         * substring-locations.c (format_warning_va): Strengthen case 1 so
41170         that both endpoints of the substring must be within the format
41171         range for just the substring to be printed.
41173 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41175         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41176         * config/i386/i386.c (ix86_target_string): Add missing options
41177         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41178         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41179         flags_other and ix86_target_other to flags2_other.  Display unknown
41180         isa2 options.
41181         (ix86_valid_target_attribute_inner_p): Add missing options and
41182         reorder options by implied ISAs, as in ix86_target_string.
41184 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41186         * hash-table.h (hash_table::too_empty_p): New function.
41187         (hash_table::expand): Use it.
41188         (hash_table::traverse): Likewise.
41189         (hash_table::empty_slot): Use sizeof (value_type) instead of
41190         sizeof (PTR) to convert bytes to elements.  Shrink the table
41191         if the current size is excessive for the current number of
41192         elements.
41194 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41196         * ira-costs.c (record_reg_classes): Break from the inner loop
41197         early once alt_fail is known to be true.  Update outer loop
41198         handling accordingly.
41200 2017-01-13  Jeff Law  <law@redhat.com>
41202         * tree-ssa-dse.c (decrement_count): New function.
41203         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41204         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41205         when we know the partially dead statement is a mem* function.
41207         PR tree-optimization/61912
41208         PR tree-optimization/77485
41209         * tree-ssa-dse.c: Include expr.h.
41210         (maybe_trim_constructor_store): New function.
41211         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41213         PR tree-optimization/33562
41214         PR tree-optimization/61912
41215         PR tree-optimization/77485
41216         * doc/invoke.texi: Document new dse-max-object-size param.
41217         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41218         * tree-ssa-dse.c: Include params.h.
41219         (dse_store_status): New enum.
41220         (initialize_ao_ref_for_dse): New, partially extracted from
41221         dse_optimize_stmt.
41222         (valid_ao_ref_for_dse, normalize_ref): New.
41223         (setup_live_bytes_from_ref, compute_trims): Likewise.
41224         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41225         (maybe_trim_partially_dead_store): Likewise.
41226         (maybe_trim_complex_store): Likewise.
41227         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41228         Track what bytes live from the original store.  Return tri-state
41229         for dead, partially dead or live.
41230         (dse_dom_walker): Add constructor, destructor and new private members.
41231         (delete_dead_call, delete_dead_assignment): New extracted from
41232         dse_optimize_stmt.
41233         (dse_optimize_stmt): Make a member of dse_dom_walker.
41234         Use initialize_ao_ref_for_dse.
41236         PR tree-optimization/33562
41237         PR tree-optimization/61912
41238         PR tree-optimization/77485
41239         * sbitmap.h (bitmap_count_bits): Prototype.
41240         (bitmap_clear_range, bitmap_set_range): Likewise.
41241         * sbitmap.c (bitmap_clear_range): New function.
41242         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41244 2017-01-13  Martin Liska  <mliska@suse.cz>
41246         PR ipa/79043
41247         * function.c (set_cfun): Add new argument force.
41248         * function.h (set_cfun): Likewise.
41249         * ipa-inline-transform.c (inline_call): Use the function when
41250         strict alising from is dropped for function we inline to.
41252 2017-01-13  Richard Biener  <rguenther@suse.de>
41254         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41255         for dumping GIMPLE INTEGER_CSTs.
41257 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41259         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41260         to 201112L since C++17.
41262 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41264         PR sanitizer/78887
41265         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41266         if -fsanitize=kernel-address is present.
41268 2017-01-13  Richard Biener  <rguenther@suse.de>
41270         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41271         as _Literal ( type ) number in case usual suffixes do not
41272         preserve all information.
41274 2017-01-13  Richard Biener  <rguenther@suse.de>
41276         PR tree-optimization/77283
41277         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41278         and ssa-iterators.h.
41279         (is_feasible_trace): Implement a cost model based on joiner
41280         PHI node uses.
41282 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41284         PR target/79004
41285         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41286         char or short to __float128/_Float128 directly.
41288 2017-01-12  Martin Sebor  <msebor@redhat.com>
41290         to -Wformat-overflow.
41291         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41292         (min_bytes_remaining): Same.
41293         (get_string_length): Same.
41294         (format_string): Same.
41295         (format_directive): Same.
41296         (add_bytes): Same.
41297         (pass_sprintf_length::handle_gimple_call): Same.
41299 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41301         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41302         info.nowrite calls with no lhs that can't throw.  Return bool
41303         whether gsi_remove has been called or not.
41304         (pass_sprintf_length::handle_gimple_call): Return bool whether
41305         try_substitute_return_value called gsi_remove.  Formatting fix.
41306         (pass_sprintf_length::execute): Don't use gsi_remove if
41307         handle_gimple_call returned true.
41309         PR bootstrap/79069
41310         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41311         be removed due to side-effects, don't remove following barrier nor
41312         turn the successor edge into fallthru edge.
41314 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41316         PR target/79044
41317         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41318         element-reversing loads and stores as not swappable.
41320 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41321             Nicolai Stange  <nicstange@gmail.com>
41323         * combine.c (try_combine): Don't ignore result of overlap checking
41324         loop.  Combine overlap & asm check into single loop.
41326 2017-01-12  Richard Biener  <rguenther@suse.de>
41328         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41329         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41331 2017-01-12  Richard Biener  <rguenther@suse.de>
41333         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41334         and TS_TARGET_OPTION directly derive from TS_BASE.
41335         * tree-core.h (tree_optimization_option): Derive from tree_base.
41336         (tree_target_option): Likewise.
41338 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41340         * config/i386/i386.c (memory_address_length): Increase len
41341         only when rip_relative_addr_p returns false.
41343 2017-01-11  Julia Koval  <julia.koval@intel.com>
41345         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41346         (OPTION_MASK_ISA_SGX_SET): New.
41347         (ix86_handle_option): Handle OPT_msgx.
41348         * config.gcc: Added sgxintrin.h.
41349         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41350         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41351         * config/i386/i386.c (ix86_target_string): Add -msgx.
41352         (PTA_SGX): New.
41353         (ix86_option_override_internal): Handle new options.
41354         (ix86_valid_target_attribute_inner_p): Add sgx.
41355         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41356         * config/i386/i386.opt: Add msgx.
41357         * config/i386/sgxintrin.h: New file.
41358         * config/i386/x86intrin.h: Add sgxintrin.h.
41360 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41362         PR c++/71537
41363         * fold-const.c (maybe_nonzero_address): Return 1 for function
41364         local objects.
41365         (tree_single_nonzero_warnv_p): Don't handle function local objects
41366         here.
41368         PR c++/72813
41369         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41370         of c-header.
41372 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41374         PR driver/78877
41375         * opts.c: Include "spellcheck.h"
41376         (struct string_fragment): New struct.
41377         (struct edit_distance_traits<const string_fragment &>): New
41378         struct.
41379         (get_closest_sanitizer_option): New function.
41380         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41382 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41384         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41385         by 12.
41386         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41387         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41388         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41389         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41390         for initial die_offset if dwarf_split_debug_info.
41391         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41392         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41393         fields.
41394         (output_skeleton_debug_sections): Formatting fix.  Use
41395         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41396         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41398 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41400         * config/arm/cortex-a53.md: Add bypasses for
41401         cortex_a53_r2f_cvt.
41402         (cortex_a53_r2f): Only use for transfers.
41403         (cortex_a53_f2r): Likewise.
41404         (cortex_a53_r2f_cvt): Add reservation for conversions.
41405         (cortex_a53_f2r_cvt): Likewise.
41407 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41409         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41410         to all inlined functions, change static to extern.
41412 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41414         PR target/78253
41415         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41416         weak symbol.
41417         (arm_assemble_integer): Likewise.
41419 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41421         * config.gcc: Use new awk script to check CPU, FPU and architecture
41422         parameters for --with-... options.
41423         * config/arm/parsecpu.awk: New file
41424         * config/arm/arm-cpus.in: New file.
41425         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41426         files.
41427         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41428         files.
41429         * config/arm/t-arm: Update dependency rules.
41430         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41431         of processing .def files.
41432         * config/arm/genopt.sh: Deleted.
41433         * config/arm/gentune.sh: Deleted.
41434         * config/arm/arm-cores.def: Deleted.
41435         * config/arm/arm-arches.def: Deleted.
41436         * config/arm/arm-fpus.def: Deleted.
41437         * config/arm/arm-tune.md: Regenerated.
41438         * config/arm/arm-tables.opt: Regenerated.
41439         * config/arm/arm-cpu.h: New generated file.
41440         * config/arm/arm-cpu-data.h: New generated file.
41441         * config/arm/arm-cpu-cdata.h: New generated file.
41443 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41445         PR lto/79042
41446         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41447         bit.
41448         (input_varpool_node): Unpack dynamically_initialized bit.
41450 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41452         PR rtl-optimization/79032
41453         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41454         the alignment of the adjusted memory reference against that of MODE,
41455         instead of the alignment of the original memory reference.
41457 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41459         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41460         test.
41461         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41462         decorated functions.
41464 2017-01-11  Richard Biener  <rguenther@suse.de>
41466         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41467         set range/nonnull info for PHI results.  Do not set it on
41468         stmts marked for removal.
41470 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41472         * expr.c (store_field): In the bitfield case, fetch the return value
41473         from the registers before applying a single big-endian adjustment.
41474         Always do a final load for a BLKmode value not larger than a word.
41476 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41478         PR c++/77949
41479         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41480         that we correctly handle column numbers greater than
41481         LINE_MAP_MAX_COLUMN_NUMBER.
41483 2017-01-10  Martin Sebor  <msebor@redhat.com>
41485         PR middle-end/78245
41486         * gimple-ssa-sprintf.c (get_destination_size): Call
41487         {init,fini}object_sizes.
41488         * tree-object-size.c (addr_object_size): Adjust.
41489         (pass_through_call): Adjust.
41490         (pass_object_sizes::execute): Adjust.
41491         * tree-object-size.h (fini_object_sizes): Declare.
41493 2017-01-10  Martin Sebor  <msebor@redhat.com>
41495         PR tree-optimization/78775
41496         * builtins.c (get_size_range): Move...
41497         * calls.c: ...to here.
41498         (alloc_max_size): Accept zero argument.
41499         (operand_signed_p): Remove.
41500         (maybe_warn_alloc_args_overflow): Call get_size_range.
41501         * calls.h (get_size_range): Declare.
41503 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41505         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41506         from TI's devices.csv file as of September 2016.
41507         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41509 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41511         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41512         * doc/invoke.texi: Likewise.
41513         * doc/md.texi: Likewise.
41514         * doc/objc.texi: Likewise.
41516 2017-01-10  Joshua Conner  <joshconner@google.com>
41518         * config/arm/fuchsia-elf.h: New file.
41519         * config/fuchsia.h: New file.
41520         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41521         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41522         targets.
41523         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41525 2016-01-10  Richard Biener  <rguenther@suse.de>
41527         PR tree-optimization/79034
41528         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41529         Propagate out degenerate PHIs in the joiner.
41531 2017-01-10  Martin Liska  <mliska@suse.cz>
41533         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41534         (sort_congruence_classes_by_decl_uid): Likewise.
41535         (sort_congruence_class_groups_by_decl_uid): Likewise.
41536         (sem_item_optimizer::merge_classes): Sort class, groups in these
41537         classes and members in the groups by DECL_UID of declarations.
41538         This would make merge operations stable.
41540 2017-01-10  Martin Liska  <mliska@suse.cz>
41542         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41543         usage of m_classes_vec.
41544         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41545         (sem_item_optimizer::get_group_by_hash): Likewise.
41546         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41547         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41548         (sem_item_optimizer::verify_classes): Likewise.
41549         (sem_item_optimizer::process_cong_reduction): Likewise.
41550         (sem_item_optimizer::dump_cong_classes): Likewise.
41551         (sem_item_optimizer::merge_classes): Likewise.
41552         * ipa-icf.h (congruence_class_hash): Rename from
41553         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41555 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41557         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41558         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41559         * config.gcc: Add avx512vpopcntdqintrin.h.
41560         * config/i386/avx512vpopcntdqintrin.h: New.
41561         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41562         * config/i386/i386-builtin-types.def: Add new types.
41563         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41564         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41565         __builtin_ia32_vpopcountq_v8di_mask): New.
41566         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41567         __AVX512VPOPCNTDQ__.
41568         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41569         (PTA_AVX512VPOPCNTDQ): Define.
41570         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41571         TARGET_AVX512VPOPCNTDQ_P): Define.
41572         * config/i386/i386.opt: Add mavx512vpopcntdq.
41573         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41574         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41576 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41578         PR middle-end/77484
41579         * predict.def (PRED_CALL): Set to 67.
41581 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41583         * expr.c (store_field): In the bitfield case, if the value comes from
41584         a function call and is of an aggregate type returned in registers, do
41585         not modify the field mode; extract the value in all cases if the mode
41586         is BLKmode and the size is not larger than a word.
41588 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41590         PR target/71017
41591         * config/i386/cpuid.h: Fix undefined behavior.
41593 2017-01-04  Jeff Law  <law@redhat.com>
41595         PR tree-optimization/79007
41596         PR tree-optimization/67955
41597         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41598         conservative for pt.null when flag_non_call_exceptions is on.
41600 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41602         PR translation/79019
41603         PR translation/79020
41604         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41605         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41606         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41607         in descriptions.
41608         * config/avr/avr.opt (maccumulate-args): Likewise.
41609         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41610         * common.opt (freport-bug): Likewise.
41611         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41612         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41613         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41614         translatable string.
41615         * config/i386/i386.c (function_value_32): Likewise.
41616         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41617         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41618         Likewise.
41619         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41620         * common/config/msp430/msp430-common.c (msp430_handle_option):
41621         Likewise.
41622         * symtab.c (symtab_node::verify_base): Likewise.
41623         * opts.c (set_debug_level): Likewise.
41624         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41625         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41626         missing whitespace to translatable strings.
41627         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41628         * config/sh/superh.h: Likewise.
41629         * config/i386/xopintrin.h: Likewise.
41630         * config/i386/znver1.md: Likewise.
41631         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41632         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41633         * double-int.h (struct double_int): Likewise.
41634         * double-int.c (div_and_round_double): Likewise.
41635         * wide-int.cc: Likewise.
41636         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41637         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41638         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41639         (crossjumps_occurred): ... this.
41640         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41641         Adjust all uses.
41643         PR tree-optimization/78899
41644         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41645         returning bool return struct loop *, NULL for failure and the new
41646         loop on success.
41647         (versionable_outer_loop_p): Don't version outer loop if it has
41648         dont_vectorized bit set.
41649         (tree_if_conversion): When versioning outer loop, ensure
41650         tree_if_conversion is performed also on the inner loop of the
41651         non-vectorizable outer loop copy.
41652         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41653         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41654         prevent vectorization of it.
41655         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41656         the outer loop vectorization of the non-scalar version is attempted
41657         before vectorization of the inner loop in scalar version.  If
41658         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41659         vectorization of its inner loop.
41660         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41661         has 2 inner loops, rename also on edges from bb whose single pred
41662         is outer_loop->header.  Fix typo in function comment.
41664 2017-01-09  Martin Sebor  <msebor@redhat.com>
41666         PR bootstrap/79033
41667         * asan.c (asan_emit_stack_protection): Increase local buffer size
41668         to avoid snprintf truncation warning.
41670 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41672         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41673         to reference thunderx2t99 for the tuning structure
41674         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41675         Rename to ...
41676         (thunderx2t99_extra_costs): This.
41677         * config/aarch64/aarch64-tune.md: Regenerate.
41678         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41679         (vulcan_addrcost_table): This.
41680         (vulcan_regmove_cost): Rename to ...
41681         (thunderx2t99_regmove_cost): This.
41682         (vulcan_vector_cost): Rename to ...
41683         (thunderx2t99_vector_cost): this.
41684         (vulcan_branch_cost): Rename to ...
41685         (thunderx2t99_branch_cost): This.
41686         (vulcan_tunings): Rename to ...
41687         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41688         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41690 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41692         PR ipa/78365
41693         PR ipa/78599
41694         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41695         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41696         (propagate_vr_accross_jump_function): Use the above function for all
41697         value range computations for pass-through jump functions and type
41698         converasion from explicit value range values.
41699         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41700         parameters from TYPE_ARG_TYPES.
41701         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41702         (ipa_write_node_info): Stream type of the actual argument.
41703         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41705 2017-01-09  Martin Liska  <mliska@suse.cz>
41707         PR pch/78970
41708         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41709         (lookup_compiler): Do not show error message with have_E.
41711 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41713         PR tree-optimization/78938
41714         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41715         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41716         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41717         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41718         fixes.
41720 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41722         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41723         is const0_rtx.
41725 2017-01-09  Richard Biener  <rguenther@suse.de>
41727         PR tree-optimization/78997
41728         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41729         name condition properly.
41731 2017-01-09  Richard Biener  <rguenther@suse.de>
41733         PR debug/79000
41734         * dwarf2out.c (is_cxx): New overload with context.
41735         (is_naming_typedef_decl): Use it.
41737 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41739         * invoke.texi (Option Summary): Correct spacing in option lists
41740         and add line breaks to fix over-long lines.
41742 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41744         PR middle-end/17660
41746         * extend.texi (Common Variable Attributes): Add xref to GCC
41747         Internals manual to explain mode attribute keywords.
41749 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41751         PR other/16519
41752         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41753         and Preprocessor Options.
41754         (Options for Linking): Document -pthread here....
41755         (RS/6000 and PowerPC Options): ...not here.
41756         (Solaris 2 Options): ...or here.
41757         * doc/cppopts.texi: Document -pthread.
41759 2017-01-08  Martin Sebor  <msebor@redhat.com>
41761         PR middle-end/77708
41762         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41763         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41764         New member functions.
41765         (format_directive): Used them.
41766         (add_bytes): Same.
41767         (pass_sprintf_length::handle_gimple_call): Same.
41768         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41769         to avoid truncation for any argument.
41770         (extract_affine_mul): Same.
41771         * tree.c (get_file_function_name): Same.
41773 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41775         PR middle-end/77484
41776         * predict.def (PRED_INDIR_CALL): Set to 86.
41778 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41780         PR preprocessor/54124
41781         * doc/cppopts.texi: Reformat -d subtable to list the full name
41782         of the options.  Add cross-reference to the docs for the general
41783         compiler -d options.
41784         * doc/invoke.texi (Developer Options): Add cross-reference to the
41785         preprocessor-specific -d option documentation.
41787 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41789         PR preprocessor/13498
41790         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41791         redudant material, and reflect new command-line options.
41792         (System Headers): Likewise.
41794 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41796         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41797         -isystem, and -idirafter.  Copy-edit.
41798         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41799         default for -ftrack-macro-expansion.  Delete obsolete and
41800         badly-formatted implementation details about -fdebug-cpp output.
41801         * doc/cppwarnopts.texi: Copy-edit.
41803 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41805         PR c++/72803
41806         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41807         that the transition from a max line width >= 1<<10 to narrower
41808         lines works correctly.
41810 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41812         * doc/options.texi (PerFunction): New.
41813         * opt-functions.awk (switch_flags): Map both Optimization and
41814         PerFunction to CL_OPTIMIZATION.
41815         * opth-gen.awk: Test for PerFunction flag along with
41816         Optimization.
41817         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41818         it only when the latter is present.  Skip those that don't in
41819         the hash function generator.
41820         * common.opt (fvar-tracking): Mark as PerFunction instead of
41821         Optimization.
41822         (fvar-tracking-assignments): Likewise.
41823         (fvar-tracking-assignments-toggle): Likewise.
41824         (fvar-tracking-uninit): Likewise.
41826 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41828         PR translation/79018
41829         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41830         the and store.
41832 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41834         PR target/57583
41835         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41836         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41837         TARGET_LONG_JUMP_TABLE_OFFSETS.
41838         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41839         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41840         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41841         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41842         * config/m68k/m68k.md (tablejump expander): Likewise.
41843         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41844         TARGET_LONG_JUMP_TABLE_OFFSETS.
41845         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41846         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41848 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41849             David Holsgrove <david.holsgrove@xilinx.com>
41851         * common/config/microblaze/microblaze-common.c
41852         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41853         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41854         New prototype.
41855         * config/microblaze/microblaze.c (microblaze_must_save_register)
41856         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41857         calls_eh_return.
41858         (microblaze_eh_return): New function.
41859         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41860         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41861         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41862         * config/microblaze/microblaze.md (eh_return): New pattern.
41864 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41866         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41867         GCC_DIAGNOSTIC_STRINGIFY): Define.
41869         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41871 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41873         * config/arm/arm.md (<mcrr>): New.
41874         (<mrrc>): New.
41875         * config/arm/arm.c (arm_arch5te): New.
41876         (arm_option_override): Set arm_arch5te.
41877         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41878         and mrrc2.
41879         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41880         (arm_mcrr_qualifiers): ... this. New.
41881         (MRRC_QUALIFIERS): Define to...
41882         (arm_mrrc_qualifiers): ... this. New.
41883         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41884         __arm_mrrc2): New.
41885         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41886         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41887         (MRRCI, mrrc, MRRC): New.
41888         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41889         VUNSPEC_MRRC2): New.
41891 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41893         * config/arm/arm.md (<mcr>): New.
41894         (<mrc>): New.
41895         * config/arm/arm.c (arm_coproc_builtin_available): Add
41896         support for mcr, mrc, mcr2 and mrc2.
41897         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41898         (arm_mcr_qualifiers): ... this. New.
41899         (MRC_QUALIFIERS): Define to ...
41900         (arm_mrc_qualifiers): ... this. New.
41901         (MCR_QUALIFIERS): Define to ...
41902         (arm_mcr_qualifiers): ... this. New.
41903         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41904         __arm_mrc2): New.
41905         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41906         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41907         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41908         VUNSPEC_MRC2): New.
41910 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41912         * config/arm/arm.md (*ldc): New.
41913         (*stc): New.
41914         (<ldc>): New.
41915         (<stc>): New.
41916         * config/arm/arm.c (arm_coproc_builtin_available): Add
41917         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41918         (arm_coproc_ldc_stc_legitimate_address): New.
41919         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41920         'qualifier_const_pointer'.
41921         (LDC_QUALIFIERS): Define to...
41922         (arm_ldc_qualifiers): ... this. New.
41923         (STC_QUALIFIERS): Define to...
41924         (arm_stc_qualifiers): ... this. New.
41925         * config/arm/arm-protos.h
41926         (arm_coproc_ldc_stc_legitimate_address): New.
41927         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41928         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41929         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41930         stc2, stcl, stc2l): New.
41931         * config/arm/constraints.md (Uz): New.
41932         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41933         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41934         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41935         VUNSPEC_STC2L): New.
41937 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41939         * config/arm/arm.md (<cdp>): New.
41940         * config/arm/arm.c (neon_const_bounds): Rename this ...
41941         (arm_const_bounds): ... this.
41942         (arm_coproc_builtin_available): New.
41943         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
41944         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
41945         (CDP_QUALIFIERS): Define to...
41946         (arm_cdp_qualifiers): ... this. New.
41947         (void_UP): Define.
41948         (arm_expand_builtin_args): Add case for 6 arguments.
41949         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
41950         (arm_const_bounds): ... this.
41951         (arm_coproc_builtin_available): New.
41952         * config/arm/arm_acle.h (__arm_cdp): New.
41953         (__arm_cdp2): New.
41954         * config/arm/arm_acle_builtins.def (cdp): New.
41955         (cdp2): New.
41956         * config/arm/iterators.md (CDPI,CDP,cdp): New.
41957         * config/arm/neon.md: Rename all 'neon_const_bounds' to
41958         'arm_const_bounds'.
41959         * config/arm/types.md (coproc): New.
41960         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
41961         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
41962         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
41963         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
41965 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41967         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
41968         (UBINOP_QUALIFIERS): New.
41969         (si_UP): Define.
41970         (acle_builtin_data): New. Change comment.
41971         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
41972         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
41973         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
41974         arm_acle_builtins.def.
41975         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
41976         (arm_init_acle_builtins): New.
41977         (CRC32_BUILTIN): Remove.
41978         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
41979         crc32cb, crc32ch and crc32cw.
41980         (arm_init_crc32_builtins): Remove.
41981         (arm_init_builtins): Use arm_init_acle_builtins rather
41982         than arm_init_crc32_builtins.
41983         (arm_expand_acle_builtin): New.
41984         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
41985         * config/arm/arm_acle_builtins.def: New.
41987 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41989         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
41990         (arm_builtin_datum): ... this.
41991         (arm_init_neon_builtin): Rename to ...
41992         (arm_init_builtin): ... this. Add a new parameters PREFIX
41993         and USE_SIG_IN_NAME.
41994         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
41995         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
41996         'arm_builtin_datum'.
41997         (arm_init_vfp_builtins): Likewise.
41998         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
41999         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
42000         (arm_expand_neon_args): Rename to ...
42001         (arm_expand_builtin_args): ... this. Rename builtin_arg
42002         enum values and differentiate between ARG_BUILTIN_MEMORY
42003         and ARG_BUILTIN_NEON_MEMORY.
42004         (arm_expand_neon_builtin_1): Rename to ...
42005         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
42006         values, arm_expand_builtin_args and add bool parameter NEON.
42007         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
42008         (arm_expand_vfp_builtin): Likewise.
42009         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
42011 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42013         PR middle-end/77484
42014         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
42015         * predict.c (tree_estimate_probability_bb): Reverse direction of
42016         polymorphic call predictor.
42018 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
42020         * passes.c (execute_one_pass): Split out pass-skipping logic into...
42021         (determine_pass_name_match): ...this new function and...
42022         (should_skip_pass_p): ...this new function.
42024 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42026         * ipa-visibility.c (function_and_variable_visibility): Reformat
42027         comments and long lines.  Remove extrneous if.
42028         * symtab.c (symtab_node::make_decl_local): Fix code format.
42029         (symtab_node::set_section_for_node): Fix comment typo.
42031 2017-01-06  Martin Liska  <mliska@suse.cz>
42033         PR bootstrap/79003
42034         * lra-constraints.c: Rename invariant to lra_invariant.
42035         * predict.c (set_even_probabilities): Initialize e to NULL.
42037 2017-01-05  Martin Sebor  <msebor@redhat.com>
42039         PR tree-optimization/78910
42040         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42041         (format_integer): Correct off-by-one error in the handling
42042         of precision with negative numbers in signed conversions..
42044 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42046         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42048 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42050         PR tree-optimization/71016
42051         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42052         factor_out_conditional_conversion.  Formatting fix.
42053         (factor_out_conditional_conversion): Add cond_stmt argument.
42054         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42055         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42056         Formatting fix.
42058 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42060         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42061         read-rtl-function.o, and selftest-rtl.o.
42062         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42063         (selftest::aarch64_test_loading_full_dump): New function.
42064         (selftest::aarch64_run_selftests): New function.
42065         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42066         selftest::aarch64_run_selftests.
42067         * config/i386/i386.c
42068         (selftest::ix86_test_loading_dump_fragment_1): New function.
42069         (selftest::ix86_test_loading_call_insn): New function.
42070         (selftest::ix86_test_loading_full_dump): New function.
42071         (selftest::ix86_test_loading_unspec): New function.
42072         (selftest::ix86_run_selftests): Call the new functions.
42073         * emit-rtl.c (maybe_set_max_label_num): New function.
42074         * emit-rtl.h (maybe_set_max_label_num): New decl.
42075         * function.c (instantiate_decls): Guard call to
42076         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42077         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42078         "static".
42079         * gensupport.c (gen_reader::gen_reader): Pass "false"
42080         for new "compact" param of rtx_reader.
42081         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42082         rather than an empty string for NULL strings.
42083         * read-md.c: Potentially include config.h rather than bconfig.h.
42084         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42085         (have_error): New global, copied from errors.c.
42086         (md_reader::read_name): Rename to...
42087         (md_reader::read_name_1): ...this, adding "out_loc" param,
42088         and converting "missing name or number" to returning false, rather
42089         than failing.
42090         (md_reader::read_name): Reimplement in terms of read_name_1.
42091         (md_reader::read_name_or_nil): New function.
42092         (md_reader::read_string): Handle "(nil)" by returning NULL.
42093         (md_reader::md_reader): Add new param "compact".
42094         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42095         (md_reader::read_file): New method.
42096         * read-md.h (md_reader::md_reader): Add new param "compact".
42097         (md_reader::read_file): New method.
42098         (md_reader::is_compact): New accessor.
42099         (md_reader::read_name): Convert return type from void to file_location.
42100         (md_reader::read_name_or_nil): New decl.
42101         (md_reader::read_name_1): New decl.
42102         (md_reader::m_compact): New field.
42103         (noop_reader::noop_reader): Pass "false" for new "compact" param
42104         of rtx_reader.
42105         (rtx_reader::rtx_reader): Add new "compact" param.
42106         (rtx_reader::read_rtx_operand): Make virtual and convert return
42107         type from void to rtx.
42108         (rtx_reader::read_until): New decl.
42109         (rtx_reader::handle_any_trailing_information): New virtual function.
42110         (rtx_reader::postprocess): New virtual function.
42111         (rtx_reader::finalize_string): New virtual function.
42112         (rtx_reader::m_in_call_function_usage): New field.
42113         (rtx_reader::m_reuse_rtx_by_id): New field.
42114         * read-rtl-function.c: New file.
42115         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42116         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42117         (selftest::verify_three_block_rtl_cfg): New decl.
42118         * read-rtl-function.h: New file.
42119         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42120         For host, include function.h, memmodel.h, and emit-rtl.h.
42121         (one_time_initialization): New function.
42122         (struct compact_insn_name): New struct.
42123         (compact_insn_names): New array.
42124         (find_code): Handle insn codes in compact dumps.
42125         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42126         (bind_subst_iter_and_attr): Likewise.
42127         (add_condition_to_string): Likewise.
42128         (add_condition_to_rtx): Likewise.
42129         (apply_attribute_uses): Likewise.
42130         (add_current_iterators): Likewise.
42131         (apply_iterators): Likewise.
42132         (initialize_iterators): Guard usage of apply_subst_iterator with
42133         #ifdef GENERATOR_FILE.
42134         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42135         (md_reader::read_mapping): Likewise.
42136         (add_define_attr_for_define_subst): Likewise.
42137         (add_define_subst_attr): Likewise.
42138         (read_subst_mapping): Likewise.
42139         (check_code_iterator): Likewise.
42140         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42141         logic to...
42142         (one_time_initialization): New function.
42143         (rtx_reader::read_until): New method.
42144         (read_flags): New function.
42145         (parse_reg_note_name): New function.
42146         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42147         Handle reuse_rtx ids.
42148         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42149         Add parsing support for RTL dumps, mirroring the special-cases in
42150         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42151         values, and calling handle_any_trailing_information.
42152         (rtx_reader::read_rtx_operand): Convert return type from void
42153         to rtx, returning return_rtx.  Handle case 'e'.  Call
42154         finalize_string on XSTR and XTMPL fields.
42155         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42156          "(nil)" values were omitted.  Call the postprocess vfunc on the
42157         return_rtx.
42158         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42159         class ctor.  Initialize m_in_call_function_usage.  Call
42160         one_time_initialization.
42161         * rtl-tests.c (selftest::test_uncond_jump): Call
42162         set_new_first_and_last_insn.
42163         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42164         * selftest-rtl.c: New file.
42165         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42166         (selftest::get_insn_by_uid): New decl.
42167         * selftest-run-tests.c (selftest::run_tests): Call
42168         read_rtl_function_c_tests.
42169         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42170         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42171         dumps.
42173 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42175         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42176         operands in a special way.  Assert that pos+len <= mode precision.
42178 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42180         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42181         3 argument Alias with unlimited for the negative form.
42182         (fno-vect-cost-model): Removed.
42184 2017-01-05  Martin Liska  <mliska@suse.cz>
42186         * hsa-gen.c (gen_hsa_divmod): New function.
42187         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42189 2017-01-05  Martin Liska  <mliska@suse.cz>
42191         PR pch/78970
42192         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42193         header.
42195 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42197         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42198         small constant length operands.
42200 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42202         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42203         between loop iterations.
42205 2017-01-05  Martin Liska  <mliska@suse.cz>
42207         PR sanitizer/78815
42208         * gimplify.c (gimplify_decl_expr): Compare to
42209         asan_poisoned_variables instread of checking flags.
42210         (gimplify_target_expr): Likewise.
42211         (gimplify_expr): Likewise.
42212         (gimplify_function_tree): Conditionally initialize
42213         asan_poisoned_variables.
42215 2017-01-04  Jeff Law  <law@redhat.com>
42217         PR tree-optimizatin/78812
42218         * rtl.h (contains_mem_rtx_p): Prototype.
42219         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42220         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42221         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42222         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42223         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42225 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42227         * input.c (assert_char_at_range): Default-initialize actual_range.
42229 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42231         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42232         to match the caller.
42234 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42236         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42237         insns after final jump in test to emit dummy move.
42239 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42241         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42242         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42244 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42246         * multiple_target.c (create_dispatcher_calls): Init e_next.
42247         * tree-ssa-loop-split.c (split_loop): Init border.
42248         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42249         scalar_type.
42251 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42253         PR target/71977
42254         PR target/70568
42255         PR target/78823
42256         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42257         (altivec_register_operand): Do not return true if the operand
42258         contains a SUBREG mixing SImode and SFmode.
42259         (vsx_register_operand): Likewise.
42260         (vsx_reg_sfsubreg_ok): New predicate.
42261         (vfloat_operand): Do not return true if the operand contains a
42262         SUBREG mixing SImode and SFmode.
42263         (vint_operand): Likewise.
42264         (vlogical_operand): Likewise.
42265         (gpc_reg_operand): Likewise.
42266         (int_reg_operand): Likewise.
42267         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42268         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42269         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42270         SImode and SFmode.
42271         (rs6000_emit_move_si_sf_subreg): New helper function.
42272         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42273         fixup SUBREGs involving SImode and SFmode.
42274         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42275         numbers for the new peephole2 optimization.
42276         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42277         the GLIBC math library that do AND/IOR/XOR operations on single
42278         precision floating point.
42279         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42280         target macros to say whether we need to avoid SUBREGs mixing
42281         SImode and SFmode.
42282         (TARGET_ALLOW_SF_SUBREG): Likewise.
42283         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42284         (UNSPEC_SI_FROM_SF): Likewise.
42285         (iorxor): Change spacing.
42286         (and_ior_xor): New iterator for AND, IOR, and XOR.
42287         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42288         (movdi_from_sf_zero_ext): Likewise.
42289         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42290         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42291         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42292         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42293         (fms<mode>4): Likewise.
42294         (fnma<mode>4): Likewise.
42295         (fnms<mode>4): Likewise.
42296         (nfma<mode>4): Likewise.
42297         (nfms<mode>4): Likewise.
42299 2017-01-04  Marek Polacek  <polacek@redhat.com>
42301         PR c++/64767
42302         * doc/invoke.texi: Document -Wpointer-compare.
42304 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42306         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42307         RejectNegative.
42309         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42310         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42311         2-byte data.
42313 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42315         PR target/78056
42316         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42317         documentation of the powerpc_popcntb_ok attribute.
42318         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42319         code to issue warning messages if a requested CPU configuration is
42320         not supported by the binary (assembler and loader) toolchain.
42321         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42322         made to define a built-in function that has been disabled.
42323         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42324         made to define a built-in function that has been disabled.
42325         (altivec_init_builtins): Add comment explaining why definition
42326         of the DST built-in functions is not preceded by an assertion
42327         check.  Add assertions to prevent ICE if attempts are made to
42328         define an altivec predicate or an abs* built-in function that has
42329         been disabled.
42330         (htm_init_builtins): Add comment explaining why definition of the
42331         htm built-in functions is not preceded by an assertion check.
42333 2017-01-04  Jeff Law  <law@redhat.com>
42335         PR tree-optimizatin/67955
42336         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42337         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42338         the points-to solution does not include pt_null.  Use DECL_PT_UID
42339         unconditionally.
42341 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42343         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42344         Use gen_int_mode instead of gen_lopwart for const_int operands.
42346 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42348         PR tree-optimization/71563
42349         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42350         out of range value - has low bits known to be zero.
42352 2017-01-04  Alan Modra  <amodra@gmail.com>
42354         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42355         * configure: Regenerate.
42356         * config.in: Regenerate.
42358 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42360         PR bootstrap/77569
42361         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42362         a substring of the message, but strcmp with the whole message.  Ifdef
42363         ENABLE_NLS, translate the message first using dgettext.
42365 2017-01-03  Jeff Law  <law@redhat.com>
42367         PR tree-optimizatin/78856
42368         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42369         (mark_threaded_blocks): Remove code to truncate thread paths that
42370         cross multiple loop headers.  Instead invalidate the cached loop
42371         iteration information and handle case of a thread path walking
42372         into an irreducible region.
42374 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42376         PR target/78900
42377         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42378         assertions.  Add support for doing the signbit if the IEEE 128-bit
42379         floating point value is in a GPR.
42380         * config/rs6000/rs6000.md (Fsignbit): Delete.
42381         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42382         Update the length attribute if the value is in a GPR.
42383         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42384         the sign or zero extension instruction, since the value is always 0/1.
42385         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42387         PR target/78953
42388         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42389         extracting SImode to a GPR register so that we can generate a
42390         store, limit the vector to be in a traditional Altivec register
42391         for the vextuwrx instruction.
42393 2017-01-03  Ian Lance Taylor  <iant@google.com>
42395         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42397 2017-01-03  Martin Sebor  <msebor@redhat.com>
42399         PR tree-optimization/78696
42400         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42401         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42402         of %g.
42403         (pass_sprintf_length::compute_format_length): Set width and precision
42404         specified by asrerisk to void_node for vararg functions.
42405         (try_substitute_return_value): Adjust dump output.
42407 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42409         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42411 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42413         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42414         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42415         -mlra/-mno-lra was passed to the compiler.
42417 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42419         PR rtl-optimization/65618
42420         * emit-rtl.c (try_split): Move initialization of "before" and
42421         "after" to just before the call to emit_insn_after_setloc.
42423 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42425         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42427 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42429         * dwarf2out.c (gen_enumeration_type_die): When
42430         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42432 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42434         PR tree-optimization/78965
42435         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42436         Change first argument from const call_info & to call_info &.  For %n
42437         set info.nowrite to false.
42439         PR middle-end/78901
42440         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42441         possibly throwing calls.
42443         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42444         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42445         and fns handling, rather than in a separate case SSA_NAME.
42447 2017-01-02  Jeff Law  <law@redhat.com>
42449         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42450         fixes for first_period and second_period variables.
42452 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42454         PR target/78967
42455         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42456         (*insvqi_1): New insn pattern.
42457         (*insvqi_1_mem_rex64): Ditto.
42458         (*insvqi_2): Ditto.
42459         (*insvqi_3): Rename from *insvqi.
42461         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42463 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42465         * doc/cfg.texi (Edges): Remove reference to Java.
42466         (Maintaining the CFG): Ditto.
42468 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42470         PR middle-end/77674
42471         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42472         transparent aliases.
42474 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42476         PR middle-end/77484
42477         * predict.def (PRED_CALL): Update hitrate.
42478         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42479         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42480         into direct/indirect/polymorphic variants.
42482 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42484         Update copyright years.
42486         * gcc.c (process_command): Update copyright notice dates.
42487         * gcov-dump.c (print_version): Ditto.
42488         * gcov.c (print_version): Ditto.
42489         * gcov-tool.c (print_version): Ditto.
42490         * gengtype.c (create_file): Ditto.
42491         * doc/cpp.texi: Bump @copying's copyright year.
42492         * doc/cppinternals.texi: Ditto.
42493         * doc/gcc.texi: Ditto.
42494         * doc/gccint.texi: Ditto.
42495         * doc/gcov.texi: Ditto.
42496         * doc/install.texi: Ditto.
42497         * doc/invoke.texi: Ditto.
42499 Copyright (C) 2017 Free Software Foundation, Inc.
42501 Copying and distribution of this file, with or without modification,
42502 are permitted in any medium without royalty provided the copyright
42503 notice and this notice are preserved.