1 2018-05-15 Richard Biener <rguenther@suse.de>
3 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4 add by_clobber_p one. Change algorithm to collect all defs
5 representing uses we need to walk and try reducing them to
6 a single one before failing.
7 (dse_dom_walker::dse_optimize_stmt): Adjust.
9 2018-05-13 Mark Wielaard <mark@klomp.org>
11 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
12 (size_of_loc_descr): Likewise.
13 (output_loc_operands): Likewise.
14 (output_loc_operands_raw): Likewise.
15 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
16 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
17 (hash_loc_operands): Likewise.
18 (compare_loc_operands): Likewise.
20 2018-05-14 Mark Wielaard <mark@klomp.org>
22 * dwarf2out.c (count_index_addrs): New function.
23 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
25 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27 PR tree-optimization/83648
28 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
29 return value as malloc candidate.
31 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
34 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
35 as true in call to suggest_attribute.
37 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
39 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
42 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
44 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
46 (aarch64_crypto_aesd_fused): Likewise.
48 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
50 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
51 (movsi_aarch64): Likewise.
52 (load_pairsi): Likewise.
53 (load_pairdi): Likewise.
54 (store_pairsi): Likewise.
55 (store_pairdi): Likewise.
56 (load_pairsf): Likewise.
57 (load_pairdf): Likewise.
58 (store_pairsf): Likewise.
59 (store_pairdf): Likewise.
60 (zero_extend): Likewise.
61 (trunc): Swap alternatives.
62 (fcvt_target): Add '?' to prefer w over r.
64 2018-05-14 Jakub Jelinek <jakub@redhat.com>
67 * config/i386/i386.md: Disallow non-commutative arithmetics in
68 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
69 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
70 in the peephole2 before it.
72 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
74 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
75 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
76 (ix86_handle_option): Handle -mcldemote.
77 * config.gcc: New header.
78 * config/i386/cldemoteintrin.h: New file.
79 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
80 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
82 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
83 OPTION_MASK_ISA_CLDEMOTE.
84 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
85 (ix86_valid_target_attribute_inner_p): Ditto.
86 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
87 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
88 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
89 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
90 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
92 * config/i386/i386.opt: Add -mcldemote.
93 * config/i386/x86intrin.h: New header.
94 * doc/invoke.texi: Add -mcldemote.
96 2018-05-14 Richard Biener <rguenther@suse.de>
98 * doc/match-and-simplify.texi: Adjust :s documentation.
100 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
102 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
103 intended memcpy size.
104 (REORDER_45): Likewise.
106 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
109 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
110 * vec.c (qsort_chk): Use gcc_qsort.
111 * Makefile.in (OBJS-libcommon): Add sort.o.
112 (build/sort.o): New target. Use it...
113 (BUILD_RTL): ... here, and...
114 (build/gencfn-macros): ... here, and...
115 (build/genmatch): ... here.
117 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
118 Chung-Ju Wu <jasonwucj@gmail.com>
120 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
121 * config/nds32/nds32-graywolf.md: New file.
122 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
123 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
125 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
126 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
127 * config/nds32/nds32.md (pipeline_model): Add graywolf.
128 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
129 * config/nds32/pipelines.md: Include n15 settings.
131 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
132 Chung-Ju Wu <jasonwucj@gmail.com>
134 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
135 * config/nds32/nds32-n13.md: New file.
136 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
137 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
139 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
140 * config/nds32/nds32.md (pipeline_model): Add n13.
141 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
142 * config/nds32/pipelines.md: Include n13 settings.
144 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
145 Chung-Ju Wu <jasonwucj@gmail.com>
147 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
148 * config/nds32/nds32-n10.md: New file.
149 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
150 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
152 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
153 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
154 * config/nds32/nds32.md (pipeline_model): Add n10.
155 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
156 * config/nds32/pipelines.md: Include n10 settings.
158 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
159 Kito Cheng <kito.cheng@gmail.com>
160 Chung-Ju Wu <jasonwucj@gmail.com>
162 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
163 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
164 Add enum values for DSP extension instructions.
165 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
167 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
168 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
170 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
171 * config/nds32/nds32-dspext.md: New file for DSP implementation.
172 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
173 * config/nds32/nds32-intrinsic.md: Likewise.
174 * config/nds32/nds32_intrinsic.h: Likewise.
175 * config/nds32/nds32-md-auxiliary.c: Likewise.
176 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
177 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
178 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
179 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
180 * config/nds32/nds32-protos.h: New declarations for DSP extension.
181 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
182 TYPE_DMAC in switch statement.
183 * config/nds32/nds32.c: New checking and implementation for DSP
184 extension instructions.
185 * config/nds32/nds32.h: Likewise.
186 * config/nds32/nds32.md: Likewise.
187 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
188 * config/nds32/predicates.md: Implement new predicates for DSP
191 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
193 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
194 Reformat alternatives and attributes so it is easier to identify
195 which constraints/attributes go with which instruction.
196 (mov<mode>_hardfloat32, FMOVE64): Likewise.
197 (mov<mode>_softfloat32, FMOVE64): Likewise.
198 (mov<mode>_hardfloat64, FMOVE64): Likewise.
199 (mov<mode>_softfloat64, FMOVE64): Likewise.
201 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
203 * doc/extend.texi (PowerPC Built-in Functions): Rename this
205 (Basic PowerPC Built-in Functions): The new name of the
206 subsection previously known as "PowerPC Built-in Functions".
207 (Basic PowerPC Built-in Functions Available on all Configurations):
209 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
210 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
211 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
212 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
214 2018-05-11 Martin Jambor <mjambor@suse.cz>
217 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
220 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
223 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
225 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
227 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
228 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
229 (ix86_handle_option): Handle -mwaitpkg.
230 * config.gcc: New header.
231 * config/i386/cpuid.h (bit_WAITPKG): New bit.
232 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
233 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
235 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
236 OPTION_MASK_ISA_WAITPKG.
237 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
238 (ix86_option_override_internal): Add PTA_WAITPKG.
239 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
240 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
242 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
243 __builtin_ia32_umwait and __builtin_ia32_tpause.
244 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
245 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
246 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
247 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
248 UNSPECV_TPAUSE): New.
249 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
250 * config/i386/i386.opt: Add -mwaitpkg.
251 * config/i386/waitpkgintrin.h: New file.
252 * config/i386/x86intrin.h: New header.
253 * doc/invoke.texi: Add -mwaitpkg.
255 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
258 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
260 (cortex-m0): Use armv6s-m isa.
261 (cortex-m0plus): Likewise.
262 (cortex-m1): Likewise.
263 (cortex-m0.small-multiply): Likewise.
264 (cortex-m0plus.small-multiply): Likewise.
265 (cortex-m1.small-multiply): Likewise.
267 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
268 Jakub Jelinek <jakub@redhat.com>
270 PR tree-optimization/85692
271 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
272 source permute as well.
274 2018-05-11 Martin Liska <mliska@suse.cz>
277 * doc/extend.texi: Document LLVM style format for no_sanitize
280 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
282 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
283 mode_supports_vsx_dform_quad to mode_supports_dq_form.
284 (mode_supports_vsx_dform_quad): Likewise.
285 (mode_supports_vmx_dform): Move these functions to be next to the
286 other mode_supports functions.
287 (mode_supports_dq_form): Likewise.
288 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
289 mode_supports_dq_form.
290 (reg_offset_addressing_ok_p): Likewise.
291 (offsettable_ok_by_alignment): Likewise.
292 (rs6000_legitimate_offset_address_p): Likewise.
293 (legitimate_lo_sum_address_p): Likewise.
294 (rs6000_legitimize_address): Likewise.
295 (rs6000_legitimize_reload_address): Likewise.
296 (rs6000_secondary_reload_inner): Likewise.
297 (rs6000_preferred_reload_class): Likewise.
298 (rs6000_output_move_128bit): Likewise.
300 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
302 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
303 Generate SImode target register for null target.
304 <case IX86_BUILTIN_XGETBV>: Ditto.
305 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
306 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
308 2018-05-10 Carl Love <cel@us.ibm.com>
310 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
311 dcbtt and dcbtstt if operands[2] is 0.
313 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
316 * config/i386/sse.md (usadv64qi): New expander.
318 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
320 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
321 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
322 -maltivec=be support.
323 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
324 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
325 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
326 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
327 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
328 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
329 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
330 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
331 altivec_vsumsws): Adjust.
332 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
333 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
334 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
336 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
337 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
338 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
339 (altivec_lve<VI_char>x): Delete expand.
340 (*altivec_lve<VI_char>x_internal): Rename to...
341 (altivec_lve<VI_char>x): ... this.
342 (altivec_lvxl_<mode>): Delete expand.
343 (*altivec_lvxl_<mode>_internal): Rename to ...
344 (altivec_lvxl_<mode>): ... this.
345 (altivec_stvxl_<mode>): Delete expand.
346 (*altivec_stvxl_<mode>_internal): Rename to ...
347 (altivec_stvxl_<mode>): ... this.
348 (altivec_stve<VI_char>x): Delete expand.
349 (*altivec_stve<VI_char>x_internal): Rename to ...
350 (altivec_stve<VI_char>x): ... this.
351 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
352 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
353 reduc_plus_scal_<mode>): Adjust.
354 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
356 (rs6000_cpu_cpp_builtins): Adjust.
357 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
358 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
359 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
360 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
361 -maltivec=be support.
362 (rs6000_split_vec_extract_var): Adjust.
363 (rs6000_split_v4si_init): Adjust.
364 (swap_selector_for_mode): Delete.
365 (altivec_expand_lvx_be, altivec_expand_stvx_be,
366 altivec_expand_stvex_be): Delete.
367 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
368 -maltivec=be support.
369 (rs6000_gimple_fold_builtin): Ditto.
370 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
372 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
373 (TARGET_DIRECT_MOVE_64BIT): Adjust.
374 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
375 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
376 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
377 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
378 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
379 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
380 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
381 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
382 anonymous split): Adjust.
383 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
384 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
386 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
388 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
389 when --with-gxx-include-dir is also specified.
390 * configure: Regenerate.
392 2018-05-09 Jim Wilson <jimw@sifive.com>
395 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
396 * config/riscv/t-withmultilib: New.
397 * config/riscv/withmultilib.h: New.
398 * doc/install.texi: Document RISC-V --with-multilib-list support.
400 2018-05-09 Richard Biener <rguenther@suse.de>
402 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
404 (vect_bb_vectorization_profitable_p): Adjust. Compute
405 actual scalar cost using the cost vector and the add_stmt_cost
408 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
410 PR rtl-optimization/85645
411 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
412 in the REG_CFA_REGISTER note for LR, don't leave it empty.
414 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
416 PR rtl-optimization/85645
417 * shrink-wrap.c (spread_components): Return a boolean saying if
418 anything was changed.
419 (try_shrink_wrapping_separate): Iterate spread_components until
420 nothing changes anymore.
422 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
424 PR rtl-optimization/85645
425 * regrename.c (build_def_use): Also kill the chains that include the
426 destination of a REG_CFA_REGISTER note.
428 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
430 PR rtl-optimization/85645
431 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
432 insn that has a REG_CFA_REGISTER note.
434 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
436 * cfgexpand.c (expand_clobber): New function.
437 (expand_gimple_stmt_1): Use it.
438 * tree-vect-stmts.c (vect_clobber_variable): New function,
440 (vectorizable_simd_clone_call): ...here.
441 (vectorizable_store): Emit a clobber either side of an
442 IFN_STORE_LANES sequence.
443 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
445 2018-05-09 Tom de Vries <tom@codesourcery.com>
448 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
449 (define_insn "trap_if_false"): Add exit after trap.
451 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
453 PR rtl-optimization/85638
454 * bb-reorder.c: Include common/common-target.h.
455 (create_forwarder_block): New function extracted from...
456 (fix_up_crossing_landing_pad): ...here. Rename into...
457 (dw2_fix_up_crossing_landing_pad): ...this.
458 (sjlj_fix_up_crossing_landing_pad): New function.
459 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
460 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
461 from both partitions and exit the loop after one iteration.
463 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
466 * doc/extend.texi (PowerPC Built-in Functions): Rename this
468 (Basic PowerPC Built-in Functions): The new name of the
469 subsection previously known as "PowerPC Built-in Functions".
470 (Basic PowerPC Built-in Functions Available on all Configurations):
472 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
474 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
476 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
478 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
481 2018-05-08 Jim Wilson <jimw@sifive.com>
483 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
484 (LD_EMUL_SUFFIX): New.
487 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
489 * doc/extend.texi (PowerPC Built-in Functions): Rename this
491 (Basic PowerPC Built-in Functions): The new name of the
492 subsection previously known as "PowerPC Built-in Functions".
493 (Basic PowerPC Built-in Functions Available on all Configurations):
495 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
497 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
499 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
501 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
504 2018-05-08 Jakub Jelinek <jakub@redhat.com>
507 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
508 after cmpelim optimization.
510 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
512 * config.gcc: Support "goldmont".
513 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
514 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
516 * config/i386/i386.c (m_GOLDMONT): Define.
517 (processor_target_table): Add "goldmont".
518 (PTA_GOLDMONT): Define.
519 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
520 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
521 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
522 (fold_builtin_cpu): Add "goldmont".
523 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
524 (ix86_option_override_internal): Add "goldmont".
525 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
526 (processor_type): Add PROCESSOR_GOLDMONT.
527 * config/i386/i386.md: Add CPU "glm".
528 * config/i386/glm.md: New file.
529 * config/i386/x86-tune.def: Add m_GOLDMONT.
530 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
532 2018-05-08 Jakub Jelinek <jakub@redhat.com>
535 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
537 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
538 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
539 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
542 * config/i386/i386.c (ix86_fold_builtin): Handle
543 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
546 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
547 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
549 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
552 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
553 (check_arch): Likewise.
554 (check_fpu): Return the result rather than printing it.
555 (end arch): Fix operator precedence.
557 (END): Print the result from check_fpu.
559 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
560 Alan Hayward <alan.hayward@arm.com>
561 David Sherwood <david.sherwood@arm.com>
563 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
564 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
565 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
566 (*fcmuo<mode>_and): New patterns.
568 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
570 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
571 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
572 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
573 (cmp_op, sve_imm_con): New code attributes.
574 (SVE_COND_INT_CMP, imm_con): Delete.
575 (cmp_op): Remove above unspecs from int attribute.
576 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
578 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
579 comparison-specific unspecs.
580 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
581 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
582 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
583 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
584 (*vec_fcm<cmp_op><mode>): Rename to...
585 (*fcm<cmp_op><mode>): ...this and adjust likewise.
586 (*vec_fcmuo<mode>): Rename to...
587 (*fcmuo<mode>): ...this and adjust likewise.
588 (*pred_fcm<cmp_op><mode>): New pattern.
589 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
590 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
592 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
594 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
595 (aarch64_emit_sve_predicated_cond): New function.
596 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
597 (aarch64_emit_unspec_cond_or): Replace with...
598 (aarch64_emit_sve_or_conds): ...this new function. Use
599 aarch64_emit_sve_ptrue_op for the individual comparisons and
600 aarch64_emit_binop to OR them together.
601 (aarch64_emit_inverted_unspec_cond): Replace with...
602 (aarch64_emit_sve_inverted_cond): ...this new function. Use
603 aarch64_emit_sve_ptrue_op for the comparison and
604 aarch64_emit_unop to invert the result.
605 (aarch64_expand_sve_vec_cmp_float): Update after the above
606 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
608 2018-05-07 Nathan Sidwell <nathan@acm.org>
610 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
611 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
612 (Backwards Compatibility): Likewise.
614 2018-05-07 Luis Machado <luis.machado@linaro.org>
618 2018-05-07 Luis Machado <luis.machado@linaro.org>
620 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
621 <prefetch_dynamic_strides>: New const bool field.
622 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
623 prefetch_dynamic_strides.
624 (exynosm1_prefetch_tune): Likewise.
625 (thunderxt88_prefetch_tune): Likewise.
626 (thunderx_prefetch_tune): Likewise.
627 (thunderx2t99_prefetch_tune): Likewise.
628 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
629 (aarch64_override_options_internal): Update to set
630 PARAM_PREFETCH_DYNAMIC_STRIDES.
631 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
632 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
633 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
634 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
635 prefetch-dynamic-strides setting.
637 2018-05-07 Luis Machado <luis.machado@linaro.org>
639 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
640 <minimum_stride>: New const int field.
641 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
642 minimum_stride field.
643 (exynosm1_prefetch_tune): Likewise.
644 (thunderxt88_prefetch_tune): Likewise.
645 (thunderx_prefetch_tune): Likewise.
646 (thunderx2t99_prefetch_tune): Likewise.
647 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
648 (aarch64_override_options_internal): Update to set
649 PARAM_PREFETCH_MINIMUM_STRIDE.
650 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
651 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
652 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
653 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
654 stride is constant and is below the minimum stride threshold.
656 2018-05-07 Luis Machado <luis.machado@linaro.org>
658 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
661 2018-05-07 Luis Machado <luis.machado@linaro.org>
663 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
664 <prefetch_dynamic_strides>: New const bool field.
665 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
666 prefetch_dynamic_strides.
667 (exynosm1_prefetch_tune): Likewise.
668 (thunderxt88_prefetch_tune): Likewise.
669 (thunderx_prefetch_tune): Likewise.
670 (thunderx2t99_prefetch_tune): Likewise.
671 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
672 (aarch64_override_options_internal): Update to set
673 PARAM_PREFETCH_DYNAMIC_STRIDES.
674 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
675 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
676 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
677 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
678 prefetch-dynamic-strides setting.
680 2018-05-07 Luis Machado <luis.machado@linaro.org>
682 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
683 <minimum_stride>: New const int field.
684 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
685 minimum_stride field.
686 (exynosm1_prefetch_tune): Likewise.
687 (thunderxt88_prefetch_tune): Likewise.
688 (thunderx_prefetch_tune): Likewise.
689 (thunderx2t99_prefetch_tune): Likewise.
690 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
691 (aarch64_override_options_internal): Update to set
692 PARAM_PREFETCH_MINIMUM_STRIDE.
693 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
694 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
695 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
696 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
697 stride is constant and is below the minimum stride threshold.
699 2018-05-06 Jakub Jelinek <jakub@redhat.com>
702 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
703 the type is addressable. Don't force op into register if it has
706 2018-05-05 Roland McGrath <mcgrathr@google.com>
709 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
710 any section for which we don't know a specific type it should have,
711 regardless of name. Previously this was done only for the exact
712 names ".init_array", ".fini_array", and ".preinit_array".
713 (default_elf_asm_named_section): Add comment about
714 relationship with default_section_type_flags and SECTION_NOTYPE.
715 (get_section): Don't consider it a type conflict if one side has
716 SECTION_NOTYPE and the other doesn't, as long as neither has the
717 SECTION_BSS et al used in the default_section_type_flags logic.
719 2018-05-05 Tom de Vries <tom@codesourcery.com>
722 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
723 (workaround_barsyncs): New function.
724 (nvptx_reorg): Use workaround_barsyncs.
725 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
726 (define_expand "nvptx_membar_cta"): New define_expand.
727 (define_insn "*nvptx_membar_cta"): New insn.
729 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
731 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
732 To improve optimization opportunities.
733 * builtin-types.def: The new needed builtin types for the above.
735 2018-05-04 Richard Biener <rguenther@suse.de>
737 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
738 * gimple-ssa-store-merging.c
739 (imm_store_chain_info::output_merged_store): Remove redundant create,
740 release split_store vector contents on failure.
741 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
742 scalar stmt vector on cache hit.
744 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
746 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
748 * config.gcc (powerpc-xilinx-eabi*): Remove.
749 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
751 (fusion_addis_mem_combo_load): Ditto.
752 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
754 (rs6000_cpu_cpp_builtins): Ditto.
755 * config/rs6000/rs6000-linux.c
756 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
757 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
758 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
760 (rs6000_setup_reg_addr_masks): Ditto.
761 (rs6000_init_hard_regno_mode_ok): Ditto.
762 (rs6000_option_override_internal): Ditto.
763 (legitimate_lo_sum_address_p): Ditto.
764 (rs6000_legitimize_address): Ditto.
765 (rs6000_legitimize_reload_address): Ditto.
766 (rs6000_legitimate_address_p): Ditto.
767 (abi_v4_pass_in_fpr): Ditto.
768 (setup_incoming_varargs): Ditto.
769 (rs6000_gimplify_va_arg): Ditto.
770 (rs6000_split_multireg_move): Ditto.
771 (rs6000_savres_strategy): Ditto.
772 (rs6000_emit_prologue_components): Ditto.
773 (rs6000_emit_epilogue_components): Ditto.
774 (rs6000_emit_prologue): Ditto.
775 (rs6000_emit_epilogue): Ditto.
776 (rs6000_elf_file_end): Ditto.
777 (rs6000_function_value): Ditto.
778 (rs6000_libcall_value): Ditto.
779 * config/rs6000/rs6000.h: Ditto.
780 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
781 (TARGET_MINMAX): ... this. New.
782 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
783 * config/rs6000/rs6000.md: Remove Xilinx FP support.
784 (*movsi_internal1_single): Delete.
785 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
786 mfpu=, mxilinx-fpu): Delete.
787 * config/rs6000/singlefp.h: Delete.
788 * config/rs6000/sysv4.h: Remove Xilinx FP support.
789 * config/rs6000/t-rs6000: Ditto.
790 * config/rs6000/t-xilinx: Delete.
791 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
792 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
793 (VStype_simple): Delete.
794 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
795 * config/rs6000/xfpu.h: Delete.
796 * config/rs6000/xfpu.md: Delete.
797 * config/rs6000/xilinx.h: Delete.
798 * config/rs6000/xilinx.opt: Delete.
799 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
800 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
802 2018-05-04 Tom de Vries <tom@codesourcery.com>
805 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
808 2018-05-04 Richard Biener <rguenther@suse.de>
811 * tree-complex.c (update_complex_assignment): We are always in SSA form.
812 (expand_complex_div_wide): Likewise.
813 (expand_complex_operations_1): Likewise.
814 (expand_complex_libcall): Preserve EH info of the original stmt.
815 (tree_lower_complex): Handle removed blocks.
816 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
817 on complex multiplication and division libcall builtins.
819 2018-05-04 Richard Biener <rguenther@suse.de>
822 * fold-const.c (negate_expr_p): Restrict negation of operand
823 zero of a division to when we know that can happen without
825 (fold_negate_expr_1): Likewise.
827 2018-05-04 Jakub Jelinek <jakub@redhat.com>
830 * real.h (real_nextafter): Declare.
831 * real.c (real_nextafter): New function.
832 * fold-const-call.c (fold_const_nextafter): New function.
833 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
835 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
836 even when arg1_mode is different from arg0_mode.
838 2018-05-03 Nathan Sidwell <nathan@acm.org>
840 * doc/extend.texi (Deprecated Features): Remove
842 (Backwards Compatibility): Likewise.
843 * doc/invoke.texi (C++ Language Options): Likewise.
844 (C++ Dialect Options): Likewise.
846 2018-05-03 Jakub Jelinek <jakub@redhat.com>
849 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
850 _mm512_mask_mullox_epi64): New intrinsics.
852 2018-05-03 Tom de Vries <tom@codesourcery.com>
855 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
856 dump files): Add offload-tree.
858 2018-05-03 Richard Biener <rguenther@suse.de>
860 PR tree-optimization/85615
861 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
862 to loops not nested in BBs loop father to avoid creating multi-entry
865 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
867 PR tree-optimization/70291
868 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
869 arguments. Change return type to tree. Emit libcall as a new
870 statement rather than replacing existing one when inplace_p is true.
871 (expand_complex_multiplication_components): New function.
872 (expand_complex_multiplication): Expand floating-point complex
873 multiplication using the above.
874 (expand_complex_division): Rename inner_type parameter to type.
875 Update expand_complex_libcall call-site.
876 (expand_complex_operations_1): Update expand_complex_multiplication
877 and expand_complex_division call-sites.
879 2018-05-02 Jakub Jelinek <jakub@redhat.com>
882 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
883 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
884 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
885 the highest significant bit of the shift count mask is clear. In
886 check whether and[sq]i3 is needed verify that all significant bits
887 of the shift count other than the highest are set.
889 2018-05-02 Tom de Vries <tom@codesourcery.com>
892 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
893 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
894 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
895 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
896 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
897 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
898 __builtin_goacc_parlevel_size.
900 2018-05-02 Richard Biener <rguenther@suse.de>
902 PR tree-optimization/85597
903 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
904 do not use split vect_get_vec_defs call but call vect_get_slp_defs
907 2018-05-02 Tom de Vries <tom@codesourcery.com>
910 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
911 dump files): Add ltrans-tree.
913 2018-05-02 Tom de Vries <tom@codesourcery.com>
916 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
917 dump files): Add wpa-ipa.
919 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
921 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
922 powerpc*-*-linux*paired* target.
923 * config/rs6000/750cl.h: Delete.
924 * config/rs6000/paired.h: Delete.
925 * config/rs6000/paired.md: Delete.
926 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
928 * config/rs6000/rs6000-builtin.def: Remove paired float support.
929 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
930 comment. Remove paired float support.
931 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
932 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
934 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
935 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
937 * config/rs6000/rs6000.c: Remove paired float support.
938 (paired_expand_vector_init, paired_expand_vector_move,
939 paired_emit_vector_compare, paired_emit_vector_cond_expr,
940 (paired_expand_lv_builtin, paired_expand_stv_builtin,
941 paired_expand_builtin, paired_expand_predicate_builtin,
942 paired_init_builtins): Delete.
943 * config/rs6000/rs6000.h: Remove paired float support.
944 * config/rs6000/rs6000.md: Remove paired float support.
945 (move_from_CR_ov_bit): Delete.
946 * config/rs6000/rs6000.opt (mpaired): Delete.
947 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
948 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
950 2018-05-02 Richard Biener <rguenther@suse.de>
953 * gimplify.c (gimplify_save_expr): When in SSA form allow
954 SAVE_EXPRs to compute to SSA vars.
956 2018-05-02 Jakub Jelinek <jakub@redhat.com>
959 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
960 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
961 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
962 clobber operands[2], instead use a new pseudo. Formatting fixes.
964 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
966 PR tree-optimization/85586
967 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
968 exit early for statements in the same group if the accesses are
971 2018-05-02 Tom de Vries <tom@codesourcery.com>
974 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
977 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
979 PR tree-optimization/85143
980 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
982 2018-05-01 Tom de Vries <tom@codesourcery.com>
985 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
986 not found" error message.
988 2018-05-01 Tom de Vries <tom@codesourcery.com>
991 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
992 * vec.c (test_ordered_remove_if): New function.
993 (vec_c_tests): Call test_ordered_remove_if.
994 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
995 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
996 * tree-vect-patterns.c (vect_pattern_recog_1): Use
997 VEC_ORDERED_REMOVE_IF.
999 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1001 PR tree-optimization/82665
1002 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1003 pointer subtraction where arguments come from a memchr call.
1005 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1007 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1008 --push-state --as-needed and --pop-state instead of --as-needed and
1009 --no-as-needed if ld supports it.
1010 * configure: Regenerated.
1013 * doc/install.texi2html: Replace _002d with - and _002a with * in
1014 generated html files using sed.
1016 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1019 * gcc-rich-location.c (blank_line_before_p): New function.
1020 (use_new_line): New function.
1021 (gcc_rich_location::add_fixit_insert_formatted): New function.
1022 * gcc-rich-location.h
1023 (gcc_rich_location::add_fixit_insert_formatted): New function.
1025 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1027 * selftest.c (assert_streq): Rename "expected" and "actual" to
1028 "val1" and "val2". Extend NULL-handling to cover both inputs
1029 symmetrically, while still requiring both to be non-NULL for a pass.
1030 * selftest.h (assert_streq): Rename "expected" and "actual" to
1032 (ASSERT_EQ): Likewise.
1033 (ASSERT_EQ_AT): Likewise.
1034 (ASSERT_KNOWN_EQ): Likewise.
1035 (ASSERT_KNOWN_EQ_AT): Likewise.
1036 (ASSERT_NE): Likewise.
1037 (ASSERT_MAYBE_NE): Likewise.
1038 (ASSERT_MAYBE_NE_AT): Likewise.
1039 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1040 the assertion to pass.
1041 (ASSERT_STREQ_AT): Likewise.
1043 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1045 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1046 interaction with -pie.
1048 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1050 * selftest.h: Fix alphabetization of per-source-file selftest
1053 2018-04-30 Jason Merrill <jason@redhat.com>
1055 PR c++/61982 - dead stores to destroyed objects.
1056 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1059 2018-04-30 Jason Merrill <jason@redhat.com>
1061 * tree.c (build_clobber): New.
1062 * tree.h: Declare it.
1063 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1065 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1067 * diagnostic-show-locus.c (layout::layout): Update for
1068 location_get_source_line returning a char_span.
1069 (struct char_span): Move to input.h.
1070 (struct correction): Update for fields in char_span becoming
1072 (struct source_line): Update for location_get_source_line
1073 returning a char_span.
1074 (layout::print_line): Likewise.
1075 * edit-context.c (edited_file::print_content): Likewise.
1076 (edited_file::print_diff_hunk): Likewise.
1077 (edited_file::print_run_of_changed_lines): Likewise.
1078 (edited_file::get_num_lines): Likewise.
1079 (edited_line::edited_line): Likewise.
1080 * final.c (asm_show_source): Likewise.
1081 * input.c (location_get_source_line): Convert return type
1082 from const char * to char_span, losing the final "line_len"
1084 (dump_location_info): Update for the above.
1085 (get_substring_ranges_for_loc): Likewise. Use a char_span
1086 when handling the literal within the line.
1087 (test_reading_source_line): Update for location_get_source_line
1088 returning a char_span.
1089 * input.h (class char_span): Move here from
1090 diagnostic-show-locus.c, converting from a struct to a class.
1091 Make data members private.
1092 (char_span::operator bool): New.
1093 (char_span::length): New.
1094 (char_span::get_buffer): New.
1095 (char_span::operator[]): New.
1096 (char_span::subspan): Make const.
1097 (char_span::xstrdup): New.
1098 (location_get_source_line): Convert return type from const char *
1099 to char_span, losing the final "line_size" param.
1101 2018-04-30 Jan Hubicka <jh@suse.cz>
1103 * lto-wrapper.c (ltrans_priorities): New static var.
1104 (cmp_priority): New.
1105 (run_gcc): Read priorities and if doing parallel build order
1106 the Makefile by them.
1108 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1110 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1112 2018-04-30 Richard Biener <rguenther@suse.de>
1114 * tree-cfg.c (verify_address): Remove base argument, add
1115 flag whether to check TREE_ADDRESSABLE and do that.
1116 (verify_expr): Remove.
1117 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1118 (verify_gimple_assign_single): Likewise.
1119 (verify_gimple_switch): Likewise.
1120 (verify_expr_location_1): Dereference tp once. Add (disabled)
1121 piece from verify_expr.
1122 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1124 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1126 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1128 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1130 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1131 (small_data_pattern): Likewise.
1132 (arc_rewrite_small_data): Likewise.
1133 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1134 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1135 (get_symbol_alignment): New function.
1136 (legitimate_small_data_address_p): Likewise.
1137 (legitimate_scaled_address): Update, call
1138 legitimate_small_data_address_p.
1139 (output_sdata): New static variable.
1140 (arc_print_operand): Update how we handle small data operands.
1141 (arc_print_operand_address): Likewise.
1142 (arc_legitimate_address_p): Update, use
1143 legitimate_small_data_address_p.
1144 (arc_rewrite_small_data_p): Remove.
1145 (arc_rewrite_small_data_1): Likewise.
1146 (arc_rewrite_small_data): Likewise.
1147 (small_data_pattern): Likewise.
1148 (compact_sda_memory_operand): Update to use
1149 legitimate_small_data_address_p and get_symbol_alignment.
1150 (prepare_move_operands): Don't rewite sdata pattern.
1151 (prepare_extend_operands): Remove.
1152 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1154 (zero_extendqisi2): Likewise.
1155 (zero_extendhisi2): Likewise.
1156 (extendqihi2): Likewise.
1157 (extendqisi2): Likewise.
1158 (extendhisi2): Likewise.
1162 * config/arc/constraints.md (Usd): Change it to memory constraint.
1164 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1166 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1167 as source of std instructions.
1168 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1169 allow 6-bit constants as source for store instructions.
1170 (movdi_insn): Update instruction pattern to allow 6-bit constants
1171 as source for store instructions.
1173 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1175 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1177 2018-04-30 Nathan Sidwell <nathan@acm.org>
1178 Sandra Loosemore <sandra@codesourcery.com>
1180 * dumpfile.c (dump_open): Allow '-' for stdout.
1181 * doc/invoke.texi (Developer Options): Document dump filename
1182 determination early. Document stdin/stdout selection.
1184 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1186 Microblaze Target: PIC data text relative
1188 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1189 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1191 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1192 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1193 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1194 New addressing mode for data-text relative position indepenedent code.
1195 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1196 'ADDRESS_SYMBOLIC_TXT_REL'.
1197 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1198 (microblaze_legitimate_pic_operand): Exclude function calls from
1199 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1200 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1202 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1203 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1204 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1205 for 'address + offset'.
1206 (microblaze_expand_prologue): Add new function prologue call for
1208 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1209 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1210 table in case of TARGET_PIC_DATA_TEXT_REL.
1211 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1212 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1213 Add new macros 'UNSPEC_TEXT',
1214 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1215 + exclude function calls from 'UNSPEC_PLT' in case of data text
1217 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1218 new target hook for generating address diff vector tables in case of
1220 * doc/tm.texi : Regenerate.
1221 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1222 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1223 of addr diff vector generation.
1224 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1225 target hook definition.
1226 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1227 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1228 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1229 option for data text relative.
1231 2018-04-30 Richard Biener <rguenther@suse.de>
1233 * tree-chrec.h (evolution_function_is_constant_p): Remove
1235 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1237 2018-04-30 Richard Biener <rguenther@suse.de>
1240 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1242 2018-04-30 Richard Biener <rguenther@suse.de>
1244 PR tree-optimization/28364
1245 PR tree-optimization/85275
1246 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1247 copying first exit test.
1249 2018-04-28 Mark Wielaard <mark@klomp.org>
1251 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1253 (dwarf_AT): Handle DW_AT_addr_base.
1254 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1256 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1259 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1260 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1261 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1262 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1264 2018-04-28 Richard Biener <rguenther@suse.de>
1266 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1267 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1268 to reflect use. Only add interesting stmts.
1270 2018-04-27 Martin Jambor <mjambor@suse.cz>
1273 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1274 the jump function allows for passing through aggregate values.
1276 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1278 * input.h (in_system_header_at): Convert from macro to inline
1280 (from_macro_expansion_at): Likewise.
1281 (from_macro_definition_at): Likewise.
1283 2018-04-27 Jeff Law <law@redhat.com>
1285 * config.gcc: Mark tile* targets as deprecated/obsolete.
1287 2018-04-27 Richard Biener <rguenther@suse.de>
1289 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1292 2018-04-27 Richard Biener <rguenther@suse.de>
1294 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1296 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1298 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1299 with Yd constraint. Set "preferred_for_speed" attribute from
1300 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1302 (*movdi_internal): Ditto.
1303 (movti_interunit splitters): Remove
1304 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1305 (movdi_interunit splitters): Ditto.
1306 * config/i386/constraints.md (Ye): Remove.
1307 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1309 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1312 * config/aarch64/constraints.md (Usg): Limit to 31.
1315 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1317 PR tree-optimization/85529
1318 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1319 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1320 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1321 zero extension or masking of the MSB bit.
1322 (optimize_range_tests): Add FIRST_BB argument, pass it through
1323 to optimize_range_tests_var_bound.
1324 (maybe_optimize_range_tests, reassociate_bb): Adjust
1325 optimize_range_tests callers.
1327 2018-04-26 Richard Biener <rguenther@suse.de>
1328 Jakub Jelinek <jakub@redhat.com>
1330 * cgraph.h (symbol_table): Just declare debug method here.
1331 * symtab.c (symbol_table::debug): Define.
1333 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1335 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1337 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1339 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1340 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1341 (*movdi_internal): Substitute Yi and Yj constraint with x
1342 and Ym and Yn constraint with y constraint. Update "isa"
1343 attribute and set "preferred_for_speed" attribute from
1344 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1345 (*movsi_internal): Ditto.
1346 (*movdf_internal): Ditto.
1347 (*movsf_internal): Ditto.
1348 (*zero_extendsidi2): Ditto.
1349 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1350 (sse2_loadld): Ditto.
1351 (*vec_extract<ssevecmodelower>_0): Ditto.
1352 (*vec_extractv4si_0_zext_sse4): Ditto.
1353 (vec_concatv2di): Ditto.
1354 (*vec_dup<mode>): Ditto.
1355 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1356 * config/i386/constraints.md (Yi): Remove.
1361 2018-04-26 Nathan Sidwell <nathan@acm.org>
1363 * dumpfile.c (dump_open): New.
1364 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1365 (dump_finish): Detect stdio/stderr by value not name.
1367 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1369 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1371 2018-04-26 Tom de Vries <tom@codesourcery.com>
1374 * config/nvptx/nvptx.c (verify_neutering_jumps)
1375 (verify_neutering_labels): New function
1376 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1378 2018-04-26 Tom de Vries <tom@codesourcery.com>
1381 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1382 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1385 2018-04-26 Richard Biener <rguenther@suse.de>
1386 Tom de Vries <tom@codesourcery.com>
1389 * lto-streamer-out.c (output_function): Fixup loops if required to match
1390 discovery done in the reader.
1392 2018-04-26 Richard Biener <rguenther@suse.de>
1394 PR tree-optimization/85116
1395 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1396 have a loop exit from the single latch predecessor. Remove
1397 case of header with just condition.
1398 (ch_base::copy_headers): Exclude infinite loops from any
1400 (pass_ch::execute): Record exits.
1402 2018-04-26 Richard Biener <rguenther@suse.de>
1404 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1405 prologue cost vector and pass it to vect_get_load_cost.
1406 (vect_get_peeling_costs_all_drs): Likewise.
1407 (vect_peeling_hash_get_lowest_cost): Likewise.
1408 (vect_enhance_data_refs_alignment): Likewise.
1410 2018-04-26 Richard Biener <rguenther@suse.de>
1413 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1414 checking of integer<->pointer conversions.
1415 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1416 sign-/zero-extending pointer types.
1417 (expand_omp_for_static_chunk): Likewise.
1419 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1420 Jean Lee <xiaoyur347@gmail.com>
1422 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1423 (TARGET_ASAN_SHADOW_OFFSET): Define.
1424 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1425 true for -fsanitize=address.
1427 2018-04-25 Mark Wielaard <mark@klomp.org>
1429 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1432 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1434 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1435 than "alu", remove explicit "memory" and "imm_disp" attributes.
1436 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1439 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1440 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1441 gen_lowpart_no_emit.
1443 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1446 * config/i386/i386.c (ix86_expand_builtin): Change memory
1447 operand to XI, extend p0 to Pmode.
1448 * config/i386/i386.md: Change unspec volatile and operand
1449 1 mode to XI, change operand 0 mode to P.
1451 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1453 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1454 GET_MODE_MASK before any checking.
1455 (nds32_can_use_bset_p): Likewise.
1456 (nds32_can_use_btgl_p): Likewise.
1458 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1460 * config/nds32/nds32-doubleword.md: New define_split pattern for
1461 illegal register number.
1463 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1465 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1467 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1469 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1471 2018-04-25 Richard Biener <rguenther@suse.de>
1473 * lto-streamer.h (LTO_major_version): Bump to 8.
1475 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1477 * BASE-VER: Set to 9.0.0.
1479 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1481 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1482 in __abskf2 and __powikf2.
1484 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1487 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1488 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1489 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1490 Use the above on operand 2. Reindent.
1491 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1493 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1496 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1498 * config/i386/i386.opt (mcet): Removed.
1499 * doc/install.texi: Remove -mcet documentation.
1500 * doc/invoke.texi: Likewise.
1502 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1505 * doc/install.texi: Remove -mcet from bootstrap-cet.
1507 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1510 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1511 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1515 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1516 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1517 containing a CONST_VECTOR.
1519 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1521 * doc/install.texi: Update newlib dependency for nvptx.
1523 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1526 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1527 instead of INTVAL when shifting x left.
1529 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1531 PR tree-optimization/85478
1532 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1533 vect_grouped_store_supported for single element vectors.
1535 2018-04-24 Richard Biener <rguenther@suse.de>
1538 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1539 load cost increase to the case of non-constant step.
1541 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1544 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1545 destination if any_malformed_asm.
1547 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1550 * expr.c (store_field): In the bitfield case, if the value comes from
1551 a function call and is returned in registers by means of a PARALLEL,
1552 do not change the mode of the temporary unless BLKmode and VOIDmode.
1554 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1556 PR rtl-optimization/85423
1557 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1558 dependencies to debug insns when the previous insn is non-debug.
1560 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1562 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1563 enums into a single definition.
1564 (fls): Fix predicates and printing.
1567 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1569 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1570 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1571 and short u6 immediate.
1572 (check_if_valid_sleep_operand): Remove.
1573 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1575 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1577 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1578 flag_always_save_lp condition.
1579 * config/nds32/nds32.opt (malways-save-lp): New option.
1581 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1583 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1584 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1585 * config/nds32/nds32.h
1586 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1587 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1589 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1591 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1592 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1594 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1595 Chung-Ju Wu <jasonwucj@gmail.com>
1597 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1599 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1600 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1602 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1604 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1606 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1608 * config/nds32/nds32-protos.h (nds32_data_alignment,
1609 nds32_local_alignment): Declare.
1610 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1611 nds32_local_alignment): New functions.
1612 (TARGET_CONSTANT_ALIGNMENT): Define.
1613 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1615 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1617 * config/nds32/nds32.c
1618 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1619 (TARGET_MODES_TIEABLE_P): Likewise.
1621 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1623 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1626 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1627 Chung-Ju Wu <jasonwucj@gmail.com>
1629 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1630 for unaligned access.
1631 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1633 * config/nds32/nds32-intrinsic.md: Likewise.
1634 * config/nds32/nds32_intrinsic.h: Likewise.
1635 * config/nds32/nds32.h (nds32_builtins): Likewise.
1636 * config/nds32/nds32.opt (munaligned-access): New option.
1637 * config/nds32/nds32.c (nds32_asm_file_start): Display
1638 flag_unaligned_access status.
1640 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1642 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1643 -mno-relax is present.
1644 * config/riscv/linux.h (LINK_SPEC): Ditto.
1646 2018-04-20 Martin Sebor <msebor@redhat.com>
1649 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1651 (gimple_fold_builtin_stxcpy_chk): Same.
1652 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1654 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1657 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1658 __powikf2 when long double is IEEE 128-bit.
1660 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1662 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1663 step to make sure stack always aligned.
1665 2018-04-20 Carl Love <cel@us.ibm.com>
1668 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1669 size check for arg0.
1671 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1672 Tom de Vries <tom@codesourcery.com>
1675 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1676 Emit insns for calls too.
1677 (nvptx_find_par): Always look for worker-level predecessor insn.
1678 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1680 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1681 (nvptx_process_pars): Propagate frames for calls.
1683 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1686 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1688 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1689 (ix86_handle_option): Don't handle OPT_mibt.
1690 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1692 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1694 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1695 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1696 (ix86_target_macros): Define __CET__ with flag_cf_protection
1697 for -fcf-protection.
1698 * config/i386/i386.c (isa2_opts): Remove -mibt.
1699 * config/i386/i386.h (TARGET_IBT): Removed.
1700 (TARGET_IBT_P): Likewise.
1701 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1702 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1703 * config/i386/i386.opt (mcet): Update help message.
1706 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1707 -mcet as an alias for -mshstk.
1709 2018-04-20 Richard Biener <rguenther@suse.de>
1712 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1713 complexity by forcing a single use of the multiply operand.
1715 2018-04-20 Martin Jambor <mjambor@suse.cz>
1718 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1719 recursion dependency to only apply to non-clones.
1721 2018-04-20 Martin Jambor <mjambor@suse.cz>
1724 * ipa-cp.c (create_specialized_node): Check that clones of
1725 self-recursive edges exist during IPA-CP.
1727 2018-04-19 Toon Moene <toon@moene.org>
1729 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1732 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1734 PR tree-optimization/85467
1735 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1736 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1737 VECTOR_CST element to type.
1739 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1742 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1743 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1744 (builtin_longjmp): Likewise.
1745 (save_stack_nonlocal): New pattern.
1746 (restore_stack_nonlocal): Likewise.
1748 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1751 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1752 Replace ASM_OUTPUT_LABEL with fprintf.
1754 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1757 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1758 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1759 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1760 define __IBT__ and __SHSTK__ for -fcf-protection.
1761 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1763 (ix86_trampoline_init): Likewise.
1764 (x86_output_mi_thunk): Likewise.
1765 (ix86_notrack_prefixed_insn_p): Likewise.
1766 (ix86_option_override_internal): Don't disallow -fcf-protection.
1767 * config/i386/i386.md (rdssp<mode>): Also enable for
1769 (incssp<mode>): Likewise.
1770 (nop_endbr): Likewise.
1771 * config/i386/i386.opt (mcet): Change help message to built-in
1775 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1776 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1777 enable CET built-in functions.
1779 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1781 * common/config/i386/i386-common.c
1782 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1783 OPTION_MASK_ISA_MOVDIRI_UNSET,
1784 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1785 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1786 * config.gcc (movdirintrin.h): New header.
1787 * config/i386/cpuid.h (bit_MOVDIRI,
1788 bit_MOVDIR64B): New bits.
1789 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1791 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1792 (VOID, PVOID, PCVOID)): New function types.
1793 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1794 __builtin_ia32_directstoreu_u64,
1795 __builtin_ia32_movdir64b): New builtins.
1796 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1797 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1799 (ix86_valid_target_attribute_inner_p): Ditto.
1800 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1801 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1802 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1803 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1804 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1805 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1806 (movdiri<mode>, movdir64b_<mode>): New.
1807 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1808 * config/i386/immintrin.h: Include movdirintrin.h.
1809 * config/i386/movdirintrin.h: New file.
1810 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1812 2018-04-19 Richard Biener <rguenther@suse.de>
1815 * cfg.c (clear_bb_flags): When loop state says we have
1816 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1818 2018-04-19 Richard Biener <rguenther@suse.de>
1820 PR tree-optimization/84737
1821 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1822 copying restrict info.
1823 (vect_setup_realignment): Use it.
1824 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1825 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1826 the first DR to all generated stores.
1827 (vectorizable_load): Likewise for loads.
1829 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1831 PR tree-optimization/85446
1832 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1833 the integral and pointer types to have the same precision.
1835 * doc/install.texi: Document --disable-cet being the default and
1838 2018-04-18 Martin Liska <mliska@suse.cz>
1840 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1843 2018-04-18 Martin Liska <mliska@suse.cz>
1846 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1849 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1850 arguments if they are comparable.
1852 2018-04-18 Martin Liska <mliska@suse.cz>
1855 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1858 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1861 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1864 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1865 ENDBR after calling __morestack.
1867 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1870 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1871 by using gcc_base_ver to generate a gcc_driver_version, and use
1872 it when generating GCC_DRIVER_NAME.
1873 * configure: Regenerate.
1875 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1878 * config.gcc: Obsolete powerpc*-*-*spe*.
1880 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1883 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1884 (stabstr_D): Change type of unum from unsigned int to
1885 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1888 2018-04-17 Jim Wilson <jimw@sifive.com>
1891 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1892 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1893 Set arg_pointer_offset after using pretend_args_size.
1895 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1897 PR rtl-optimization/85431
1898 * dse.c (record_store): Ignore zero width stores.
1901 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1902 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1903 __builtin_stack_restore rather than after it.
1904 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1905 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1906 argument instead of virtual_dynamic_stack_rtx.
1908 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1910 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1912 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1913 Add note to error message to explain internal mapping of overloaded
1914 built-in function name to non-overloaded built-in function name.
1915 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1918 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1921 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1922 where the inputs overlap with the output.
1924 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1926 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1927 (=v, v) alternative and explicit "memory" attribute.
1928 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1929 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1931 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1932 "sselog1" type instead of "sselog".
1933 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1934 "sselog". Remove explicit "memory" attribute.
1935 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1936 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1938 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1939 "sselog1" type instead of "sselog". Remove explicit "memory"
1941 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1942 use "sselog1" type instead of "sselog". Remove explicit "memory"
1944 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1945 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1947 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1948 "sselog1" type instead of "sselog". Remove explicit "memory"
1950 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1951 use "sselog1" type instead of "sselog". Remove explicit "memory"
1955 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1958 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1961 2018-04-17 Martin Jambor <mjambor@suse.cz>
1964 * ipa-cp.c (create_specialized_node): Call
1965 expand_all_artificial_thunks if necessary.
1967 2018-04-17 Martin Liska <mliska@suse.cz>
1970 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1971 in message, remote space in between '_G' and '('.
1973 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1976 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1977 avx512f_vmcmp<mode>3<round_saeonly_name>,
1978 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1979 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1980 avx512f_rndscale<mode><round_saeonly_name>,
1981 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1982 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1983 Use %<iptr>2 instead of %2 for -masm=intel.
1984 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1985 avx512f_vcvttss2usi<round_saeonly_name>,
1986 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1988 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1989 avx512f_vcvttsd2usi<round_saeonly_name>,
1990 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1991 Use %q1 instead of %1 for -masm=intel.
1992 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1993 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1994 of %3 for -masm=intel.
1995 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1997 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1999 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2001 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2002 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2004 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2006 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2007 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2008 %g1 and one with %0 and %1.
2009 (avx512er_vmrcp28<mode><round_saeonly_name>,
2010 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2012 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2013 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2014 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2015 of %0 and %{%4%} for -masm=intel.
2016 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2017 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2018 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2019 order of %0 and %{%5%}%{z%} for -masm=intel.
2021 2018-04-17 Jan Hubicka <jh@suse.cz>
2024 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2026 2018-04-17 Martin Liska <mliska@suse.cz>
2029 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2030 for target_clone error message. Make default implementation
2031 clone to be a local declaration.
2032 (separate_attrs): Add new argument and check for an empty
2034 (expand_target_clones): Handle it.
2035 (ipa_target_clone): Make redirection just for target_clones
2038 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2039 Tom de Vries <tom@codesourcery.com>
2042 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2043 tiled basic blocks without omp continue statements.
2045 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2048 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2049 vec_extract expression as having side effects to make sure it gets
2052 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2055 * config/i386/i386.c (get_builtin_code_for_version): Check
2058 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2061 * gcc/config.gcc: Support "skylake".
2062 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2064 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2065 (processor_target_table): Add "skylake".
2066 (ix86_option_override_internal): Add "skylake".
2067 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2068 PROCESSOR_CANNONLAKE.
2069 (get_builtin_code_for_version): Fix priority for
2070 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2071 PROCESSOR_SKYLAKE-AVX512.
2072 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2073 (processor_type): Add PROCESSOR_SKYLAKE.
2075 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2076 Jason Merrill <jason@redhat.com>
2079 * convert.c (convert_to_integer_1): Use direct recursion for
2080 enumeral types and types with a precision less than the number
2081 of bits in their mode.
2083 2018-04-16 Julia Koval <julia.koval@intel.com>
2086 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2087 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2089 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2092 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2093 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2094 and -mno-direct-move.
2096 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2099 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2100 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2101 Ensure negative shifts result in {0}.
2103 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2105 PR rtl-optimization/79916
2106 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2107 regs (if any) to define how to gnerate SD moves when LRA is in
2110 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2112 PR rtl-optimization/85393
2113 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2114 * except.c (expand_dw2_landing_pad_for_region): Make static.
2115 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2116 a label and unconditional jump to old_bb, rather than
2117 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2120 PR rtl-optimization/85376
2121 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2122 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2123 instead of a specific value.
2125 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2126 Bin Cheng <bin.cheng@arm.com>
2128 PR tree-optimization/82965
2129 PR tree-optimization/83991
2130 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2131 by_profile_only parameter.
2132 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2133 information if the loop was predicted to iterate too many times.
2134 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2136 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2139 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2142 2018-04-13 Martin Liska <mliska@suse.cz>
2143 Jakub Jelinek <jakub@redhat.com>
2146 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2147 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2148 * builtins.c (expand_builtin_memory_copy_args): Use
2149 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2150 handle dest_addr == pc_rtx.
2152 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2155 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2156 asked to not generate direct moves.
2157 (fix_trunc<mode>si2_stfiwx): Similar.
2158 (fix_trunc<mode>si2_internal): Similar.
2160 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2163 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2164 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2165 lookup if dest in some wider mode is known to be const0_rtx and
2166 if so, record permanent equivalence for it to be ZERO_EXTEND of
2167 the narrower mode destination.
2169 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2171 * lto-streamer-out.c (output_function): Revert 259346.
2172 * omp-expand.c (expand_oacc_for): Likewise.
2174 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2176 PR rtl-optimization/85354
2177 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2178 * sel-sched.c (sel_global_init): ... here.
2180 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2183 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2184 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2185 mode for PE-COFF targets.
2186 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2187 (i386_pe_asm_lto_end): Likewise.
2188 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2189 (TARGET_ASM_LTO_END): Likewise.
2190 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2191 (i386_pe_asm_lto_start): New function.
2192 (i386_pe_asm_lto_end): Likewise.
2194 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2195 Richard Biener <rguenther@suse.de>
2198 * lto-streamer-out.c (output_function): Fix CFG loop state before
2200 * omp-expand.c (expand_oacc_for): Handle calls to internal
2201 functions like regular functions.
2203 2018-04-12 Richard Biener <rguenther@suse.de>
2206 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2207 for the early LTO debug to properly generate references to it
2208 during DIE emission. Do not re-use that for the skeleton for
2210 (dwarf2out_early_finish): Likewise.
2212 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2215 * config/i386/sse.md
2216 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2217 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2218 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2219 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2220 and output is a reg, avoid creating invalid lowpart subreg, but
2221 instead split into a 512-bit move. Don't split if not AVX512VL,
2222 input is xmm16+ reg and output is a mem.
2223 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2224 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2225 xmm16+ reg and output is a mem.
2227 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2229 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2230 also for flag_dwarf2_cfi_asm.
2232 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2234 PR rtl-optimization/85342
2235 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2236 a bool scalar var inside of the loop instead. Don't try to update
2237 recog_data.operand after failed apply_change_group.
2239 2018-04-12 Tom de Vries <tom@codesourcery.com>
2242 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2243 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2244 array with flexible array member as array without given dimension.
2245 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2246 argument for undefined param to true.
2248 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2251 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2252 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2253 from PowerPC section.
2254 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2255 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2256 help text that is too long.
2257 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2258 help text that is too long.
2259 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2260 help text that is too long.
2262 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2264 * config/alpha/alpha.md (stack_probe_internal): Rename
2265 from "probe_stack". Update all callers.
2267 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2269 PR rtl-optimization/84566
2270 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2271 sched_macro_fuse_insns.
2273 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2276 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2277 (compute_block_dependences): ... from here.
2279 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2281 PR tree-optimization/85331
2282 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2283 from int to HOST_WIDE_INT.
2285 2018-04-11 Martin Jambor <mjambor@suse.cz>
2288 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2289 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2290 not the same as the source val.
2291 (cgraph_edge_brings_value_p): New parameter.
2292 (gather_edges_for_value): Pass destination value to
2293 cgraph_edge_brings_value_p.
2294 (perhaps_add_new_callers): Likewise.
2295 (get_info_about_necessary_edges): Likewise and exclude values brought
2296 only by self-recursive edges.
2297 (create_specialized_node): Redirect only clones of self-calling edges.
2298 (+self_recursive_pass_through_p): New function.
2299 (find_more_scalar_values_for_callers_subset): Use it.
2300 (find_aggregate_values_for_callers_subset): Likewise.
2301 (known_aggs_to_agg_replacement_list): Removed.
2302 (decide_whether_version_node): Re-calculate known constants for all
2303 remaining context clones.
2305 2018-04-11 Richard Biener <rguenther@suse.de>
2308 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2309 from early DWARF output.
2310 (dwarf2out_early_finish): Output line info unconditionally into
2311 early DWARF and add reference to it.
2313 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2316 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2317 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2318 other than V2DFmode using iptr mode attribute.
2319 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2321 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2323 PR rtl-optimization/84659
2324 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2326 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2329 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2331 (output_loc_list): Pass address of a dummy size variable even in the
2332 locview handling loop.
2333 (index_location_lists): Add comment on why skip_loc_list_entry can't
2336 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2339 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2342 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2345 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2346 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2347 and -mstring-compare-inline-limit.
2349 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2352 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2353 for stack clash protection in a register whenever we need it to be in
2356 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2358 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2359 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2361 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2364 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2366 (mlong-double-): Ditto.
2367 * config/rs6000/sysv4.opt (msdata=): Ditto.
2368 (mtls-size=): Ditto.
2370 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2372 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2373 erroneous entries for
2374 "vector int vec_ldl (int, long int *)", and
2375 "vector unsigned int vec_ldl (int, unsigned long int *)".
2376 Add comments and entries for
2377 "vector bool char vec_ldl (int, bool char *)",
2378 "vector bool short vec_ldl (int, bool short *)",
2379 "vector bool int vec_ldl (int, bool int *)",
2380 "vector bool long long vec_ldl (int, bool long long *)",
2381 "vector pixel vec_ldl (int, pixel *)",
2382 "vector long long vec_ldl (int, long long *)",
2383 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2384 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2385 type tree bool_long_long_type_node and correct definition of
2386 bool_V2DI_type_node to make reference to this new type tree.
2387 (rs6000_mangle_type): Replace erroneous reference to
2388 bool_long_type_node with bool_long_long_type_node.
2389 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2390 comments to emphasize sign distinctions for char and int types and
2391 replace RS6000_BTI_bool_long constant with
2392 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2393 use of RS6000_BTI_pixel.
2394 (bool_long_type_node): Remove this macro definition.
2395 (bool_long_long_type_node): New macro definition
2397 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2399 PR rtl-optimization/85300
2400 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2401 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2402 simplify_unary_operation fails.
2404 2018-04-10 Martin Liska <mliska@suse.cz>
2406 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2407 cgraph_edge and ipa_ref.
2409 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2413 * config/i386/sse.md
2414 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2415 computation of the VEC_MERGE selector from mask.
2416 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2417 Fix decoding of the VEC_MERGE selector into mask.
2419 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2421 PR tree-optimization/85286
2422 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2424 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2426 * final.c (final_1): Set insn_last_address as well as
2427 insn_current_address.
2429 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2432 * explow.c (emit_stack_probe): Call validize_mem on memory location
2433 before passing it to gen_probe_stack. Create address operand and
2434 legitimize it for the probe_stack_address case.
2436 2018-04-09 Jan Hubicka <jh@suse.cz>
2439 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2440 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2441 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2443 (free_lang_data): Rebuild type inheritance graph.
2445 2018-04-09 Martin Sebor <msebor@redhat.com>
2447 * invoke.texi (-finline-small-functions): Mention other optimization
2449 (-findirect-inlining, -fpartial-inlining): Same.
2450 (-finline-functions-called-once): Same.
2451 (-freorder-blocks-and-partition): Same.
2453 2018-04-09 Jan Hubicka <jh@suse.cz>
2456 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2457 jumps; choose last target that matches the criteria (i.e.
2458 no partition changes for non-crossing jumps).
2459 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2460 support for redirecting crossing jumps to non-crossing.
2462 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2464 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2465 also for naked functions.
2467 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2469 * config/arc/arc.md (add_shift): New pattern.
2470 (add_shift2): Likewise.
2471 (sub_shift): Likewise.
2472 (sub_shift_cmp0_noout): Likewise.
2473 (compare_si_ashiftsi): Likewise.
2474 (xbfu_cmp0_noout): New combine pattern.
2475 (xbfu_cmp0"): Likewise.
2476 (movsi_set_cc_insn): Place the predicable variant first.
2477 (commutative_binary_cmp0_noout): Remove clobber.
2478 (commutative_binary_cmp0): New pattern.
2479 (noncommutative_binary_cmp0): Likewise.
2480 (noncommutative_binary_cmp0_noout): Likewise.
2481 (noncommutative_binary_comparison_result_used): Removed.
2482 (rsub_cmp0): New pattern.
2483 (rsub_cmp0_noout): Likewise.
2484 (extzvsi): Changed, keep only meaningful variants.
2485 (SQH, SEZ): New iterators.
2486 (SQH_postfix): New mode attribute.
2487 (SEZ_prefix): New code attribute.
2488 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2489 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2490 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2492 (noncommutative_operator): Check the availability of barrel
2495 2018-04-09 Richard Biener <rguenther@suse.de>
2497 PR tree-optimization/85284
2498 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2499 Only use the niter constraining form of simple_iv when the exit
2502 2018-04-09 Tom de Vries <tom@codesourcery.com>
2505 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2506 (define_expand "*memory_barrier"): New define_expand.
2507 (define_insn "memory_barrier"): New insn.
2509 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2511 PR rtl-optimization/80463
2512 PR rtl-optimization/83972
2513 PR rtl-optimization/83480
2515 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2516 correct producer for the insn.
2517 (tidy_control_flow): Fixup seqnos in case of debug insns.
2519 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2521 PR rtl-optimization/83913
2523 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2524 different sched-times when merging exprs.
2526 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2528 PR rtl-optimization/83962
2530 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2531 tidy_fallthru_edge and tidy_control_flow.
2533 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2535 PR rtl-optimization/83530
2537 * sel-sched.c (force_next_insn): New global variable.
2538 (remove_insn_for_debug): When force_next_insn is true, also leave only
2539 next insn in the ready list.
2540 (sel_sched_region): When the region wasn't scheduled, make another pass
2541 over it with force_next_insn set to 1.
2543 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2545 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2547 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2548 for interrupt control.
2549 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2550 functions for interrupt control.
2551 * config/nds32/nds32-intrinsic.md: Likewise.
2552 * config/nds32/nds32_intrinsic.h: Likewise.
2553 * config/nds32/nds32.h (nds32_builtins): Likewise.
2555 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2557 * config/nds32/nds32.c (nds32_init_machine_status,
2558 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2559 strict_aligned_p field.
2560 (nds32_expand_to_rtl_hook): New function.
2561 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2562 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2564 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2565 Chung-Ju Wu <jasonwucj@gmail.com>
2567 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2568 * config/nds32/nds32-n7.md: New file.
2569 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2570 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2572 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2573 * config/nds32/nds32.md (pipeline_model): Add n7.
2574 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2575 * config/nds32/pipelines.md: Include n7 settings.
2577 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2578 Chung-Ju Wu <jasonwucj@gmail.com>
2580 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2581 * config/nds32/nds32-e8.md: New file.
2582 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2583 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2585 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2586 * config/nds32/nds32.md (pipeline_model): Add e8.
2587 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2588 * config/nds32/pipelines.md: Include e8 settings.
2590 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2591 Chung-Ju Wu <jasonwucj@gmail.com>
2593 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2594 * config/nds32/nds32-n8.md: New file.
2595 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2596 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2598 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2599 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2600 * config/nds32/nds32.md (pipeline_model): Add n8.
2601 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2602 * config/nds32/pipelines.md: Include n8 settings.
2604 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2605 Chung-Ju Wu <jasonwucj@gmail.com>
2607 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2608 * config/nds32/nds32-n9-2r1w.md: New file.
2609 * config/nds32/nds32-n9-3r2w.md: New file.
2610 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2611 nds32_register_ports): New or modify for cpu n9.
2612 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2614 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2615 * config/nds32/nds32-utils.c: New file.
2616 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2617 TARGET_MUL_SLOW): Define.
2618 * config/nds32/nds32.md (pipeline_model): New attribute.
2619 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2620 New options that support cpu n9.
2621 * config/nds32/pipelines.md: Include n9 settings.
2622 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2624 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2626 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2627 information if necessary.
2628 (output_cond_branch_compare_zero): Likewise.
2629 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2630 (nds32_target_alignment): Refine for alignment.
2631 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2632 (FUNCTION_BOUNDARY): Modify.
2633 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2635 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2637 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2639 * config/nds32/constants.md (unspec_volatile_element): Add values for
2640 TLB operation and data prefetch.
2641 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2642 functions for TLB operation and data prefetch.
2643 * config/nds32/nds32-intrinsic.md: Likewise.
2644 * config/nds32/nds32_intrinsic.h: Likewise.
2645 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2646 (nds32_print_operand): Likewise.
2647 * config/nds32/nds32.h (nds32_builtins): Likewise.
2649 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2650 Andrew Pinski <pinsika@gcc.gnu.org>
2653 * match.pd: Use constant_boolean_node of correct type instead of
2654 boolean_true_node or boolean_false_node for simplifying
2655 pointer comparisons to zero.
2657 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2659 PR tree-optimization/80021
2660 * tree.c (verify_type_variant): Make error call in verify_variant_match
2661 translatable and remove final full stop.
2663 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2665 * config/nds32/constants.md (unspec_volatile_element): Add
2666 UNSPEC_VOLATILE_EH_RETURN.
2667 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2668 nds32_output_stack_pop): Support dwarf exception handling process.
2669 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2670 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2671 exception handling process.
2672 (nds32_compute_stack_frame): Likewise.
2673 (nds32_return_addr_rtx): Likewise.
2674 (nds32_initial_elimination_offset): Likewise.
2675 (nds32_expand_prologue): Likewise.
2676 (nds32_expand_epilogue): Likewise.
2677 (nds32_dynamic_chain_address): New function.
2678 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2680 (DYNAMIC_CHAIN_ADDRESS): Define.
2681 (EH_RETURN_DATA_REGNO): Define.
2682 (EH_RETURN_STACKADJ_RTX): Define.
2683 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2684 patterns for dwarf exception handling.
2686 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2688 * config/nds32/nds32.h: Clean up obsolete macros.
2690 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2692 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2693 Add enum values for particular instructions.
2694 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2695 particular intrinsic functions.
2696 * config/nds32/nds32-intrinsic.md: Likewise.
2697 * config/nds32/nds32_intrinsic.h: Likewise.
2698 * config/nds32/nds32.h (nds32_builtins): Likewise.
2699 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2700 (btst, ave): New patterns for particular instructions.
2702 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2704 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2705 Add enum values for atomic load/store and memory sync.
2706 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2708 * config/nds32/nds32-intrinsic.md: Likewise.
2709 * config/nds32/nds32_intrinsic.h: Likewise.
2710 * config/nds32/nds32.h (nds32_builtins): Likewise.
2712 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2714 PR tree-optimization/85257
2715 * fold-const.c (native_encode_vector): If not all elts could fit
2716 and off is -1, return 0 rather than offset.
2717 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2718 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2719 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2720 adjust buffer in native_interpret_expr call.
2722 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2724 * config/nds32/constants.md (unspec_volatile_element): Add cache
2725 control enum values.
2726 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2727 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2728 * config/nds32/nds32.c (nds32_cctl_names): New.
2729 (nds32_print_operand): Handle cache control register names.
2730 * config/nds32/nds32.h (nds32_builtins): New enum values.
2731 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2733 * config/nds32/nds32.md (type): Add mmu.
2734 * config/nds32/pipelines.md (simple_insn): Add mmu.
2736 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2738 * config/nds32/nds32.md (type): Remove call.
2739 * config/nds32/pipelines.md (simple_insn): Likewise.
2741 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2743 * config/nds32/constants.md (unspec_volatile_element): Add
2744 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2745 UNSPEC_VOLATILE_FMFCFG.
2746 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2747 description for fmfcfg and fmfcsr.
2748 (bdesc_1arg): Add fmtcsr.
2749 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2750 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2751 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2752 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2753 unspec_fmfcfg): New patterns.
2754 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2755 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2756 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2757 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2758 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2759 __nds32__fmfcfg): Define.
2761 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2763 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2764 intrinsic register names.
2765 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2766 intrinsic register enum values and macros.
2768 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2770 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2771 for load/store addressing form.
2772 (nds32_print_operand_address): Likewise.
2774 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2777 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2778 based on LABEL_REF. Remove useless assertion.
2779 (pic_address_needs_scratch): Fix formatting.
2780 (sparc_legitimize_pic_address): Minor tweaks.
2781 (sparc_delegitimize_address): Adjust assertion accordingly.
2782 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2783 into symbolic_operand.
2784 (movsi_high_pic_label_ref): Likewise.
2785 (movsi_lo_sum_pic_label_ref): Likewise.
2786 (movdi_pic_label_ref): Likewise.
2787 (movdi_high_pic_label_ref): Likewise.
2788 (movdi_lo_sum_pic_label_ref): Likewise.
2790 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2792 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2793 custom LIB_SPEC setup.
2795 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2796 Kito Cheng <kito.cheng@gmail.com>
2798 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2799 * config/riscv/freebsd.h: New.
2801 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2803 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2804 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2807 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2808 Kito Cheng <kito.cheng@gmail.com>
2810 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2811 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2812 * config/nds32/nds32-protos.h (nds32_output_call,
2813 nds32_output_return): Declare.
2814 * config/nds32/nds32.md: Refine all the call and return patterns.
2816 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2819 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2820 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2822 PR rtl-optimization/84872
2823 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2824 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2827 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2829 * expr.c (copy_blkmode_to_reg): Revert 254862.
2830 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2832 2018-04-06 Richard Biener <rguenther@suse.de>
2835 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2836 after seeing a component reference with an adjacent field. Treat
2837 refs to arrays at struct end of external decls similar to
2838 refs to unconstrained commons.
2840 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2843 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2844 look through SAVE_EXPRs with non-side-effects argument. Adjust
2846 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2849 2018-04-06 Richard Biener <rguenther@suse.de>
2852 * alias.c (find_base_term): New wrapper around find_base_term
2853 unwinding CSELIB_VAL_PTR changes.
2854 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2857 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2859 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2861 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2862 constant definitions.
2863 ("nop"): lr 0,0 -> nopr r0
2864 ("nop_lr0", "nop_lr1"): New insn definitions.
2866 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2868 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2869 NDS32_V3PUSH_AVAILABLE_P macro.
2871 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2872 Chung-Ju Wu <jasonwucj@gmail.com>
2874 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2875 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2876 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2877 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2878 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2879 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2880 * config/nds32/constraints.md: New constraints and checking for hard
2881 float configuration.
2882 * config/nds32/iterators.md: New mode iterator and attribute for hard
2883 float configuration.
2884 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2886 * config/nds32/nds32-fpu.md: New file.
2887 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2888 deal with hard float code generation.
2889 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2891 (abi_type, float_reg_number): New enum type.
2892 * config/nds32/nds32-predicates.c: New predicates for hard float.
2893 * config/nds32/nds32-protos.h: Declare functions for hard float.
2894 * config/nds32/nds32.c: Implementation for hard float configuration.
2895 * config/nds32/nds32.h: Definitions for hard float configuration.
2896 * config/nds32/nds32.md: Include hard float machine description and
2897 modify patterns for hard float configuration.
2898 * config/nds32/nds32.opt: New options for hard float configuration.
2899 * config/nds32/predicates.md: New predicates for hard float
2902 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2904 * common/config/nds32/nds32-common.c
2905 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2907 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2910 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2911 CONSTRUCTOR_ELT (ctor, ...)->value.
2913 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2916 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2918 2018-04-05 Tom de Vries <tom@codesourcery.com>
2921 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2924 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2925 Kito Cheng <kito.cheng@gmail.com>
2927 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2928 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2929 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2930 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2932 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2933 Kito Cheng <kito.cheng@gmail.com>
2935 * config/nds32/constraints.md (Ufe): New memory constraint.
2936 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2937 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2938 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2940 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2941 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2943 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2945 * config/nds32/nds32.md: Use optimize_size in the condition for
2946 alu-shift instructions.
2948 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2950 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2952 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2954 * config/nds32/nds32.md (negsi2): Refine pattern.
2956 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2957 Chung-Ju Wu <jasonwucj@gmail.com>
2959 * config/nds32/iterators.md (shift_rotate): New code iterator.
2960 (shift): New code attribute.
2961 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2962 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2963 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2964 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2965 bit-wise operations.
2966 (andsi3, *andsi3): Ditto.
2967 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2968 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2969 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2970 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2971 nds32_ior_operand, nds32_xor_operand): New predicates.
2973 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2975 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2976 (addsi3, subsi3): ... this.
2978 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2980 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2982 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2984 * config/nds32/nds32.md: Adjust indention.
2986 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2988 * config/nds32/nds32.md (feature): New attribute.
2990 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2992 * config/nds32/nds32.md (subtype): New attribute.
2994 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2997 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2998 expansion to perform a bitwise AND of the argument followed by a
2999 boolean negation of the result.
3001 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3003 PR rtl-optimization/84878
3004 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3005 the basic block. Assert the use reference is not artificial and that
3006 it has an associated insn.
3008 2018-04-04 Michael Matz <matz@suse.de>
3010 * builtins.c (compute_objsize): Pass correct operand
3011 to array_at_struct_end_p.
3013 2018-04-04 Richard Biener <rguenther@suse.de>
3016 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3017 from contexts for DINFO_LEVEL_TERSE and below.
3019 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3021 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3022 resiter_operand condition.
3023 * config/nds32/nds32.md (*move<mode>): Ditto.
3025 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3026 Monk Chiang <sh.chiang04@gmail.com>
3028 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3030 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3032 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3034 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3035 Kito Cheng <kito.cheng@gmail.com>
3037 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3038 nds32_cond_code_str, output_cond_branch,
3039 output_cond_branch_compare_zero, nds32_expand_cbranch,
3040 nds32_expand_cstore, nds32_expand_movcc,
3041 nds32_output_cbranchsi4_equality_zero,
3042 nds32_output_cbranchsi4_equality_reg,
3043 nds32_output_cbranchsi4_equality_reg_or_const_int,
3044 nds32_output_cbranchsi4_greater_less_zero: New functions.
3045 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3046 nds32_expand_cstore, nds32_expand_movcc,
3047 nds32_output_cbranchsi4_equality_zero,
3048 nds32_output_cbranchsi4_equality_reg,
3049 nds32_output_cbranchsi4_equality_reg_or_const_int,
3050 nds32_output_cbranchsi4_greater_less_zero): Declare.
3051 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3052 nds32_rimm11s_operand): New predicates.
3053 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3054 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3057 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3059 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3060 * config/nds32/nds32.md: Ditto.
3061 * config/nds32/pipelines.md: Ditto.
3063 2018-04-04 Richard Biener <rguenther@suse.de>
3065 PR tree-optimization/85168
3066 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3067 propagating abnormals.
3069 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3071 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3073 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3074 Kito Cheng <kito.cheng@gmail.com>
3076 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3077 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3078 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3079 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3080 * config/nds32/nds32.md (sibcall_internal): New.
3081 (sibcall_register): Remove.
3082 (sibcall_immediate): Remove.
3083 (sibcall_value_internal): New.
3084 (sibcall_value_register): Remove.
3085 (sibcall_value_immediate): Remove.
3086 * config/nds32/predicates.md (nds32_general_register_operand): New.
3087 (nds32_call_address_operand): New.
3089 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3091 PR rtl-optimization/85167
3092 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3093 bb_defs if *split_p, instead preinitialize it to NULL.
3095 PR tree-optimization/85156
3096 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3097 evaluating the argument multiple times.
3099 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3101 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3103 (_mm_cvtpd_ps): Likewise.
3104 (_mm_cvttpd_epi32): Likewise.
3105 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3106 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3107 vector, pixel, and bool following altivec.h include.
3109 2018-04-03 Martin Sebor <msebor@redhat.com>
3111 * doc/extend.texi (Common Function Attributes): Clarify.
3112 (const attribute): Likewise.
3113 (pure attribute): Likewise.
3115 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3118 * config/i386/i386.c (ix86_expand_vector_set): Use
3119 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3121 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3123 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3124 instructions when changing rounding bits to preserve precision bits
3125 in the x87 control word.
3127 2018-04-03 Martin Liska <mliska@suse.cz>
3129 PR tree-optimization/82491
3130 * rtl.h (strip_offset_and_add): Replace += suboffset with
3131 poly_uint64 () + suboffset.
3133 2018-03-29 Martin Liska <mliska@suse.cz>
3134 Martin Jambor <mjambor@suse.cz>
3137 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3138 param_type is not an integral or pointer type.
3140 2018-04-03 Richard Biener <rguenther@suse.de>
3142 * sese.h (recompute_all_dominators): Remove.
3144 2018-04-02 Martin Sebor <msebor@redhat.com>
3146 * doc/invoke.texi (-Wrestrict): Fix typos.
3148 2018-04-02 Jim Wilson <jimw@sifive.com>
3150 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3151 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3152 (<optab>di3, <optab>si3_extend): Likewise.
3153 (<optab>si3_mask, <optab>si3_mask_1): New.
3154 (<optab>di3_mask, <optab>di3_mask_1): New.
3155 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3156 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3157 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3159 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3161 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3164 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3166 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3167 (nds32_canonicalize_comparison): New function.
3169 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3170 Kito Cheng <kito.cheng@gmail.com>
3171 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3173 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3174 * config/nds32/constants.md (unspec_volatile_element): Add
3175 UNSPEC_VOLATILE_RELAX_GROUP.
3176 * config/nds32/nds32-relax-opt.c: New file.
3177 * config/nds32/nds32-predicates.c
3178 (nds32_symbol_load_store_p): New function.
3179 * config/nds32/nds32-protos.h
3180 (nds32_symbol_load_store_p): Declare function.
3181 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3182 * config/nds32/nds32.c
3183 (nds32_register_pass): New function to register pass.
3184 (nds32_register_passes): New function to register passes.
3185 * config/nds32/nds32.md (relax_group): New pattern.
3186 * config/nds32/nds32.opt (mrelax-hint): New option.
3187 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3189 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3191 * config/nds32/t-nds32: Modify files dependency.
3193 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3195 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3196 (PROFILE_HOOK): Define its implementation.
3198 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3200 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3201 type and 32-bit size.
3203 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3206 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3207 (V_128_256): New mode iterator.
3208 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3209 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3210 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3212 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3213 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3215 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3218 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3219 NaN inputs correctly.
3221 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3224 * config/rs6000/vsx.md (??r): New mode attribute.
3225 (*vsx_mov<mode>_64bit): Use it.
3226 (*vsx_mov<mode>_32bit): Likewise.
3228 2018-03-30 Martin Sebor <msebor@redhat.com>
3230 PR tree-optimization/84818
3231 * builtins.c (check_access): Use warning_n.
3233 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3236 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3238 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3241 2018-03-30 Julia Koval <julia.koval@intel.com>
3244 * x86-tune.def (movx, partial_reg_dependency): Enable for
3247 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3250 * lra-constraints.c (process_alt_operands): Move setting
3251 this_alternative_matches below.
3253 2018-03-29 Martin Liska <mliska@suse.cz>
3256 * doc/invoke.texi: Document how LTO works with debug info.
3257 Describe auto-load support of binutils. Mention 'x86-64'
3258 as valid option value of -march option.
3260 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3262 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3265 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3266 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3267 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3270 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3273 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3274 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3275 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3276 for RS6000_BTM_POWERPC64.
3277 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3278 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3279 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3284 2018-03-28 Carl Love <cel@us.ibm.com>
3287 2017-09-27 Carl Love <cel@us.ibm.com>
3289 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3290 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3291 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3294 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3296 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3297 instead of __vector bool.
3298 (_mm_max_pu8): Likewise.
3299 (_mm_min_pi16): Likewise.
3301 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3304 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3305 (DIVWEUO): Likewise.
3307 (DIVDEUO): Likewise.
3308 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3309 DIVWEUO and DIVDEUO.
3310 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3311 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3312 (div_extend): Likewise.
3313 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3315 (__builtin_divweuo): Likewise.
3316 (__builtin_divdeo): Likewise.
3317 (__builtin_divdeuo): Likewise.
3319 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3322 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3323 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3325 PR tree-optimization/82004
3326 * gimple-match-head.c (optimize_pow_to_exp): New function.
3327 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3328 Don't fold to exp if optimize_pow_to_exp is false.
3330 2018-03-28 Martin Liska <mliska@suse.cz>
3333 * calls.c (initialize_argument_information): Fix trailing space.
3334 * common.opt: Fix typo and provide better explanation for
3335 -fsanitize-coverage option.
3336 * config/i386/i386.opt: Fix typo.
3338 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3339 Martin Liska <mliska@suse.cz>
3342 * gimplify.c (asan_poison_variable): Don't do the check for
3343 gimplify_omp_ctxp here.
3344 (gimplify_decl_expr): Do it here.
3345 (gimplify_target_expr): Likewise.
3347 2018-03-28 Martin Liska <mliska@suse.cz>
3350 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3351 chkp_type_bounds_count if MPX is not enabled.
3353 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3355 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3357 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3360 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3361 function to create the function decl for complex long double
3362 multiply and divide for -mabi=ieeelongdouble.
3363 (init_float128_ieee): Call it.
3365 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3368 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3369 -fcf-protection=branch -mibt.
3370 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3372 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3375 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3377 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3380 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3381 extern array declarations.
3383 2018-03-27 Richard Biener <rguenther@suse.de>
3386 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3387 explicit single_use checks.
3389 2018-03-27 Richard Biener <rguenther@suse.de>
3391 PR tree-optimization/85082
3392 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3395 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3397 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3398 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3399 Turn on fasynchronous-unwind-tables and funwind-tables.
3401 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3404 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3405 (*bmi_blsr_<mode>_ccz): Ditto.
3407 2018-03-26 Tom de Vries <tom@codesourcery.com>
3409 PR tree-optimization/85063
3410 * omp-general.c (offloading_function_p): New function. Factor out
3412 * omp-offload.c (pass_omp_target_link::gate): ... here.
3413 * omp-general.h (offloading_function_p): Declare.
3414 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3415 with attribute omp declare target for offloading functions.
3417 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3419 PR tree-optimization/84005
3420 * tree-data-ref.h (get_base_for_alignment): Declare.
3421 * tree-data-ref.c (get_base_for_alignment_1): New function.
3422 (get_base_for_alignment): Likewise.
3423 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3424 get_base_for_alignment to find a suitable base object, instead
3425 of always using drb->base_address.
3427 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3430 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3431 known size by default.
3433 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3436 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3437 and non BLKmode operands.
3439 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3442 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3443 Clean up attributes.
3445 2018-03-23 Richard Biener <rguenther@suse.de>
3448 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3449 we are going to emit early debug for LTO.
3451 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3454 * function.c (match_asm_constraints_1): Don't optimize if input
3455 doesn't satisfy general_operand predicate for output's mode.
3458 * alias.c (write_dependence_p): Don't require for x_canonicalized
3459 non-VOIDmode if x has VOIDmode.
3462 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3463 just don't try to optimize it rather than assert it never happens.
3465 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3467 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3468 macro expansions for definition of ST_INTERNAL_<mode> and
3469 LD_INTERNAL_<mode> builtins.
3470 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3472 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3474 (altivec_expand_st_builtin): Likewise.
3475 (altivec_expand_builtin): Remove calls to deleted functions.
3476 (rs6000_address_for_altivec): Delete this function.
3477 * config/rs6000/vector.md: Remove expands for
3478 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3480 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3483 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3484 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3485 re-computing once computed.
3486 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3487 (arm_init_machine_status): Initialize
3488 machine->static_chain_stack_bytes.
3490 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3493 * doc/extend.texi: Add four new prototypes for vec_ld.
3494 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3495 definitions for more logical presentation.
3496 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3497 entries for V1TI variants of __builtin_altivec_ld builtin.
3498 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3499 handling of V1TI variant of LVX icode pattern.
3500 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3501 (rs6000_gimple_fold_builtin): Likewise.
3502 (altivec_init_builtins): Add code to define
3503 __builtin_altivec_lvx_v1ti function.
3505 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3508 * function.c (match_asm_constraints_1): Don't do the optimization
3509 if input isn't a REG, SUBREG, MEM or constant.
3511 2018-03-22 Tom de Vries <tom@codesourcery.com>
3513 PR tree-optimization/84956
3514 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3515 bb_has_abnormal_pred.
3517 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3520 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3521 DECL_INITIAL (decl) to decl at the end.
3522 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3525 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3527 * doc/extend.texi (__builtin_tgmath): Document when complex
3528 integer types are treated as _Complex _Float64.
3530 2018-03-21 Tom de Vries <tom@codesourcery.com>
3532 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3534 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3536 PR tree-optimization/84960
3537 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3538 if it is ENTRY block, move them into single succ of ENTRY in that case.
3540 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3542 PR tree-optimization/84811
3543 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3544 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3545 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3546 (known_subrange_p): Update accordingly. Cast each value involved
3547 in the size comparison, rather than casting the result of the
3550 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3552 PR tree-optimization/84982
3553 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3554 by flipping the least significant bit rather than all bits from
3555 bitpos to bitpos + bitsize - 1.
3557 2018-03-21 Nathan Sidwell <nathan@acm.org>
3559 * doc/extend.texi (Deprecated Features): Remove mention of
3560 long-deleted deprecations.
3562 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3565 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3566 * configure: Regenerate.
3568 2018-03-21 Tom de Vries <tom@codesourcery.com>
3570 PR tree-optimization/83126
3571 * tree-parloops.c (num_phis): New function.
3572 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3574 2018-03-21 Nathan Sidwell <nathan@acm.org>
3576 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3577 mention anon-struct/union members and trailing attributes.
3579 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3581 PR tree-optimization/84969
3582 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3583 builtin memset partitions if they set different rhs values.
3585 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3587 PR rtl-optimization/84989
3588 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3589 VEC_DUPLICATE with scalar result mode.
3591 2018-03-21 Martin Liska <mliska@suse.cz>
3594 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3595 not intended return statement.
3597 2018-03-21 Martin Liska <mliska@suse.cz>
3600 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3601 (chkp_find_bound_slots_1): Limit number of iterations.
3603 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3606 * Minor grammar fixes for x86 options.
3608 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3611 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3612 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3614 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3617 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3618 (altivec_lvx_<mode>_1op): Likewise.
3619 (altivec_stvx_<mode>_2op): Likewise.
3620 (altivec_stvx_<mode>_1op): Likewise.
3621 (altivec_lvx_<VM2:mode>): New define_expand.
3622 (altivec_stvx_<VM2:mode>): Likewise.
3623 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3624 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3625 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3626 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3627 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3628 (rs6000_gen_lvx): Likewise.
3629 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3630 (altivec_expand_stv_builtin): Likewise.
3631 (altivec_expand_builtin): Likewise.
3632 * config/rs6000/vector.md: Likewise.
3634 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3637 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3640 2018-03-20 Richard Biener <rguenther@suse.de>
3643 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3644 sign-conversions as zero, fall back to standard scalar_stmt
3647 2018-03-20 Martin Liska <mliska@suse.cz>
3650 * predict.c (rebuild_frequencies): Handle case when we have
3651 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3653 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3656 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3657 flag_section_anchors.
3658 * varasm.c (use_blocks_for_decl_p): Remove hack for
3659 dw2_force_const_mem.
3662 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3664 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3665 be created, use lowpart_subreg of operands[0] rather than operands[0]
3667 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3668 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3669 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3670 and n constraint instead of aarch64_shift_imm_di and Usd.
3671 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3672 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3674 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3677 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3678 to favor GPR over NEON registers.
3679 (<shift>di3_neon): Likewise.
3681 2018-03-20 Tom de Vries <tom@codesourcery.com>
3684 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3685 (nvptx_process_pars): Emit bar.sync asap and alap.
3687 2018-03-20 Tom de Vries <tom@codesourcery.com>
3690 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3691 seen_label if seen_label is already set.
3693 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3696 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3697 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3698 Use 1U instead of 1. Formatting fixes.
3701 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3702 instead of TREE_TYPE (s1) for the return value.
3704 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3706 PR tree-optimization/84946
3707 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3708 bitsize + bitsize in poly_uint64 rather than poly_int64.
3711 * dwarf2asm.c: Include fold-const.c.
3712 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3713 of decl rather than decl itself.
3715 PR rtl-optimization/84643
3716 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3718 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3721 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3722 calling assemble_variable.
3724 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3727 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3728 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3730 2018-03-19 Jim Wilson <jimw@sifive.com>
3733 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3734 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3735 (riscv_first_stack_step): Likewise.
3736 (riscv_option_override): Use STACK_BOUNDARY instead of
3738 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3740 (BIGGEST_ALIGNMENT): Set to 128.
3741 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3742 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3745 2018-03-19 Richard Biener <rguenther@suse.de>
3747 PR tree-optimization/84933
3748 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3749 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3751 2018-03-19 Richard Biener <rguenther@suse.de>
3753 PR tree-optimization/84859
3754 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3755 (cond_if_else_store_replacement): Perform sinking operation on
3756 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3757 Generalize what a BB with a single eligible store is.
3759 2018-03-19 Richard Biener <rguenther@suse.de>
3761 PR tree-optimization/84929
3762 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3763 chrec_is_positive against non-chrec arg.
3765 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3768 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3770 2018-03-18 Martin Liska <mliska@suse.cz>
3772 PR rtl-optimization/84635
3773 * regrename.c (build_def_use): Use matches_mode only when
3776 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3778 PR tree-optimization/84913
3779 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3780 vectorize chains of COND_EXPRs.
3782 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3784 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3786 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3788 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3790 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3792 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3794 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3795 Kito Cheng <kito.cheng@gmail.com>
3797 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3798 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3799 (nds32_adjust_reg_alloc_order): New function.
3800 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3802 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3804 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3805 nds32_print_operand, nds32_print_operand_address): Use
3806 HOST_WIDE_INT_PRINT_DEC instead.
3808 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3810 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3812 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3815 * config/i386/i386.c (initial_ix86_tune_features,
3816 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3818 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3819 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3820 rather than 1u << ix86_tune. Formatting fix.
3821 (ix86_option_override_internal): Change ix86_arch_mask from
3822 unsigned int to unsigned HOST_WIDE_INT, initialize to
3823 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3824 (ix86_function_specific_restore): Likewise.
3826 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3829 * postreload.c (reload_combine_recognize_pattern): Perform
3830 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3831 truncate_int_for_mode the result for the destination's mode.
3834 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3835 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3837 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3840 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3843 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3844 regno_allocno_class_array and sorted_pseudos.
3845 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3846 insns where regno is used.
3848 2018-03-16 Martin Liska <mliska@suse.cz>
3851 * multiple_target.c (create_dispatcher_calls): Redirect
3852 reference in the symbol table.
3854 2018-03-16 Martin Liska <mliska@suse.cz>
3857 * multiple_target.c (create_dispatcher_calls): Redirect also
3860 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3864 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3865 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3866 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3868 2018-03-16 Julia Koval <julia.koval@intel.com>
3870 * doc/invoke.texi (Skylake Server): Add CLWB.
3871 Cannonlake): Remove CLWB.
3873 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3875 PR tree-optimization/84841
3876 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3878 (FLOAT_ONE_CONST_TYPE): Define.
3879 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3880 (sort_by_operand_rank): Put entries with higher constant_type last
3881 rather than first to match comments.
3883 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3885 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3888 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3891 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3892 check and use address of target always.
3894 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3897 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3898 (indirect_thunk_bnd_needed): Likewise.
3899 (indirect_thunks_used): Likewise.
3900 (indirect_thunks_bnd_used): Likewise.
3901 (indirect_return_needed): New.
3902 (indirect_return_bnd_needed): Likewise.
3903 (output_indirect_thunk_function): Add a bool argument for
3905 (output_indirect_thunk_function): Don't generate alias for
3906 function return thunk.
3907 (ix86_code_end): Call output_indirect_thunk_function to generate
3908 function return thunks.
3909 (ix86_output_function_return): Set indirect_return_bnd_needed
3910 and indirect_return_needed instead of indirect_thunk_bnd_needed
3911 and indirect_thunk_needed.
3913 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3915 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3916 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3917 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3919 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3920 Paul Hua <paul.hua.gm@gmail.com>
3923 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3925 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3927 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3928 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3931 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3934 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3935 instead of GET_MODE_SIZE when comparing Units.
3937 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3940 * varasm.c (hash_section): Return an unchangeble hash value
3941 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3942 Return !aarch64_can_use_per_function_literal_pools_p ().
3944 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3947 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3948 rather than address of cmode as last argument to prepare_cmp_insn.
3950 2018-03-15 Julia Koval <julia.koval@intel.com>
3952 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3953 F_AVX512VNNI, F_AVX512BITALG): New.
3955 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3958 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3959 insn for floating-point loads and stores.
3961 2018-03-14 Carl Love <cel@us.ibm.com>
3963 * config/rs6000/rs6000-c.c: Add macro definitions for
3964 ALTIVEC_BUILTIN_VEC_PERMXOR.
3965 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3966 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3967 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3968 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3970 * config/doc/extend.texi: Add prototypes for vec_permxor.
3972 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3975 * diagnostic-show-locus.c (class layout_point): Convert m_line
3976 from int to linenum_type.
3977 (line_span::comparator): Use linenum "compare" function when
3978 comparing line numbers.
3979 (test_line_span): New function.
3980 (layout_range::contains_point): Convert param "row" from int to
3982 (layout_range::intersects_line_p): Likewise.
3983 (layout::will_show_line_p): Likewise.
3984 (layout::print_source_line): Likewise.
3985 (layout::should_print_annotation_line_p): Likewise.
3986 (layout::print_annotation_line): Likewise.
3987 (layout::print_leading_fixits): Likewise.
3988 (layout::annotation_line_showed_range_p): Likewise.
3989 (struct line_corrections): Likewise for field m_row.
3990 (line_corrections::line_corrections): Likewise for param "row".
3991 (layout::print_trailing_fixits): Likewise.
3992 (layout::get_state_at_point): Likewise.
3993 (layout::get_x_bound_for_row): Likewise.
3994 (layout::print_line): Likewise.
3995 (diagnostic_show_locus): Likewise for locals "last_line" and
3997 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3998 * input.c (selftest::test_linenum_comparisons): New function.
3999 (selftest::input_c_tests): Call it.
4000 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4001 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4002 * selftest.h (ASSERT_GT): New macro.
4003 (ASSERT_GT_AT): New macro.
4004 (ASSERT_LT): New macro.
4005 (ASSERT_LT_AT): New macro.
4007 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4009 PR rtl-optimization/84780
4010 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4012 2018-03-14 Martin Liska <mliska@suse.cz>
4014 * tree.c (record_node_allocation_statistics): Use
4015 get_stats_node_kind.
4016 (get_stats_node_kind): New function extracted from
4017 record_node_allocation_statistics.
4018 (free_node): Use get_stats_node_kind.
4020 2018-03-14 Richard Biener <rguenther@suse.de>
4022 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4023 that the value-set of ANTIC_IN doesn't grow.
4026 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4028 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4029 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4031 2018-03-14 Julia Koval <julia.koval@intel.com>
4033 * config.gcc (icelake-client, icelake-server): New.
4035 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4036 (initial_ix86_arch_features): Ditto.
4037 (PTA_SKYLAKE): Add SGX.
4038 (PTA_ICELAKE): Remove.
4039 (PTA_ICELAKE_CLIENT): New.
4040 (PTA_ICELAKE_SERVER): New.
4041 (ix86_option_override_internal): Split up icelake on icelake client and
4043 (get_builtin_code_for_version): Ditto.
4044 (fold_builtin_cpu): Ditto.
4045 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4046 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4047 * config/i386/i386.h (processor_type): Ditto.
4048 * doc/invoke.texi: Ditto.
4050 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4053 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4054 INTEGER_CST offset, add it together with bitpos / 8 and
4055 sign extend based on POINTER_SIZE.
4059 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4062 * config/i386/constraints.md (Yc): New register constraint.
4063 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4064 Use Yc constraint for alternative 2 of operand 0. Remove
4065 preferred_for_speed attribute.
4067 2018-03-14 Richard Biener <rguenther@suse.de>
4069 PR tree-optimization/84830
4070 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4071 with the old one to avoid oscillations.
4073 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4076 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4078 (assign_by_spills): Return a flag of reload assignment failure.
4079 Do not process the reload assignment failures. Do not spill other
4080 reload pseudos if they has the same reg class. Update n if
4082 (lra_assign): Add a return arg. Set up from the result of
4083 assign_by_spills call.
4084 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4085 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4086 usage_insns if it is not NULL.
4087 (spill_hard_reg_in_range): New function.
4088 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4089 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4090 function prototypes.
4091 (lra_assign): Change prototype.
4092 * lra.c (lra): Add code to deal with fails by splitting hard reg
4095 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4097 * config/riscv/riscv.opt (mrelax): New option.
4098 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4099 "norelax" when riscv_mrelax is disabled.
4100 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4102 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4105 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4106 reassociation for int modes.
4108 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4110 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4111 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4113 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4114 * config/aarch64/aarch64-sve.md
4115 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4116 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4117 (*extend<mode><Vwide>2): Rename to...
4118 (aarch64_sve_extend<mode><Vwide>2): ...this.
4119 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4120 renaming the old pattern to...
4121 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4123 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4124 define_expand, renaming the old pattern to...
4125 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4126 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4127 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4128 account when deciding which SVE instruction the optab should use.
4129 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4131 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4133 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4134 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4135 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4136 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4137 (tlsdesc_small_<mode>): Turn a define_expand and use
4138 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4139 (tlsdesc_small_advsimd_<mode>): ...this.
4140 (tlsdesc_small_sve_<mode>): New pattern.
4142 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4144 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4145 (UNSPEC_UMUL_HIGHPART): New constants.
4146 (MUL_HIGHPART): New int iteraor.
4147 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4148 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4150 (*<su>mul<mode>3_highpart): New define_insn.
4152 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4155 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4158 2018-03-13 Martin Liska <mliska@suse.cz>
4161 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4163 (sem_item_optimizer::~sem_item_optimizer): Release it.
4164 (sem_item_optimizer::merge_classes): Register variable aliases.
4165 (sem_item_optimizer::fixup_pt_set): New function.
4166 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4167 * ipa-icf.h: Declare new variables and functions.
4169 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4172 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4173 integer_pow2p@2 and test integer_pow2p in condition.
4174 (A < 0 ? C : 0): Similarly for @1.
4177 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4178 characters starting at p contain '\0' character, don't look beyond
4182 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4183 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4186 * reg-stack.c (change_stack): Change update_end var from int to
4187 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4188 also call set_block_for_insn on the newly added insns and rescan.
4191 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4192 on the last operand.
4195 * tree.c (stabilize_reference_1): Return save_expr (e) for
4196 STATEMENT_LIST even if it doesn't have side-effects.
4198 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4200 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4202 2018-03-12 Renlin Li <renlin.li@arm.com>
4204 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4205 aarch64_output_scalar_simd_mov_immediate.
4207 2018-03-12 Martin Sebor <msebor@redhat.com>
4209 PR tree-optimization/83456
4210 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4211 for perfectly overlapping calls to memcpy.
4212 (gimple_fold_builtin_memory_chk): Same.
4213 (gimple_fold_builtin_strcpy): Handle no-warning.
4214 (gimple_fold_builtin_stxcpy_chk): Same.
4215 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4217 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4219 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4220 parameter. Use it for SFmode.
4221 (rs6000_function_arg_advance_1): Adjust.
4222 (rs6000_function_arg): Adjust.
4223 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4225 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4227 PR rtl-optimization/84169
4228 PR rtl-optimization/84780
4229 * combine.c (can_combine_p): Check for a 2-insn combination whether
4230 the destination register is used between the two insns, too.
4232 2018-03-12 Richard Biener <rguenther@suse.de>
4234 PR tree-optimization/84803
4235 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4236 for refs DR analysis didn't process.
4238 2018-03-12 Richard Biener <rguenther@suse.de>
4240 PR tree-optimization/84777
4241 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4242 force-vectorize loops ignore whether we are optimizing for size.
4244 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4246 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4247 (TARGET_MD_ASM_ADJUST): Define.
4249 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4250 Kito Cheng <kito.cheng@gmail.com>
4251 Chung-Ju Wu <jasonwucj@gmail.com>
4253 * config/nds32/nds32.c (nds32_compute_stack_frame,
4254 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4255 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4256 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4257 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4258 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4259 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4260 * config/nds32/nds32.md (prologue, epilogue): Use macro
4261 NDS32_V3PUSH_AVAILABLE_P to do checking.
4263 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4266 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4267 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4268 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4269 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4270 addition of most attributes on !orig_type_die or the attribute not
4271 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4273 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4274 Chung-Ju Wu <jasonwucj@gmail.com>
4276 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4278 * config/nds32/nds32.opt (mvh): New option.
4280 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4281 Chung-Ju Wu <jasonwucj@gmail.com>
4283 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4285 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4286 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4289 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4290 Chung-Ju Wu <jasonwucj@gmail.com>
4292 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4294 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4295 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4297 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4298 Kito Cheng <kito.cheng@gmail.com>
4299 Chung-Ju Wu <jasonwucj@gmail.com>
4301 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4302 UNSPEC_FFMISM and UNSPEC_FLMISM.
4303 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4304 for ffb, ffmism and flmism.
4305 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4306 (unspec_ffmism): Ditto.
4307 (unspec_flmism): Ditto.
4308 (nds32_expand_builtin_impl): Check if string extension is available.
4309 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4310 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4312 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4315 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4318 * lra-assigns.c (assign_by_spills): Return a flag of reload
4319 assignment failure. Do not process the reload assignment
4320 failures. Do not spill other reload pseudos if they has the same
4322 (lra_assign): Add a return arg. Set up from the result of
4323 assign_by_spills call.
4324 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4325 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4326 usage_insns if it is not NULL.
4327 (spill_hard_reg_in_range): New function.
4328 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4329 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4330 function prototypes.
4331 (lra_assign): Change prototype.
4332 * lra.c (lra): Add code to deal with fails by splitting hard reg
4335 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4338 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4340 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4343 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4344 (dw_val_node): Add val_symbolic_view.
4345 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4346 (symview_upper_bound): New.
4347 (new_line_info_table): Initialize symviews_since_reset.
4348 (dwarf2out_source_line): Count symviews_since_reset and set
4349 symview_upper_bound.
4350 (dw_val_equal_p): Handle symview.
4351 (add_AT_symview): New.
4352 (print_dw_val): Handle symview.
4353 (attr_checksum, attr_checksum_ordered): Likewise.
4354 (same_dw_val_p, size_of_die): Likewise.
4355 (value_format, output_die): Likewise.
4356 (add_high_low_attributes): Use add_AT_symview for entry_view.
4357 (dwarf2out_finish): Reset symview_upper_bound, clear
4360 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4363 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4364 Add strict argument and use it.
4365 (rs6000_split_multireg_move): Update for new strict argument.
4366 (mem_operand_gpr): Disallow all non-offsettable addresses.
4367 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4369 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4372 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4373 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4374 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4377 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4378 decl, use remap_type if we want to use the type.
4380 2018-03-09 Martin Sebor <msebor@redhat.com>
4382 PR tree-optimization/84526
4383 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4385 (builtin_access::generic_overlap): Be prepared to handle non-array
4388 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4390 PR rtl-optimization/84682
4391 * lra-constraints.c (process_address_1): Check is_address flag
4392 for address constraints.
4393 (process_alt_operands): Likewise.
4394 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4395 preprocess_constraints.
4396 * recog.h (preprocess_constraints): Add oploc parameter.
4398 * recog.c (preprocess_constraints): Test address_operand for
4399 CT_ADDRESS constraints.
4401 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4404 * lra-assigns.c (assign_by_spills): Return a flag of reload
4405 assignment failure. Do not process the reload assignment
4406 failures. Do not spill other reload pseudos if they has the same
4408 (lra_assign): Add a return arg. Set up from the result of
4409 assign_by_spills call.
4410 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4411 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4412 usage_insns if it is not NULL.
4413 (spill_hard_reg_in_range): New function.
4414 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4415 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4416 function prototypes.
4417 (lra_assign): Change prototype.
4418 * lra.c (lra): Add code to deal with fails by splitting hard reg
4421 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4424 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4425 Accept complain bool parameter. Only emit errors if it is true.
4426 (arm_parse_cpu_option_name): Likewise.
4427 (arm_target_thumb_only): Adjust callers of the above.
4428 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4429 prototype to take a default true bool parameter.
4430 (arm_parse_arch_option_name): Likewise.
4432 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4433 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4437 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4438 * configure: Regenerate.
4439 * configure.ac ("linker --version-script option"): New.
4440 ("linker soname option"): New.
4442 2018-03-09 Richard Biener <rguenther@suse.de>
4444 PR tree-optimization/84775
4445 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4446 immediate uses of predicate stmts and mark them modified.
4449 PR tree-optimization/84178
4450 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4452 (version_loop_for_if_conversion): Delay update_ssa call.
4453 (tree_if_conversion): Delay update_ssa until after predicate
4456 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4459 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4460 when the function accesses prior frames.
4462 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4465 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4466 gen_llsym, otherwise call maybe_gen_llsym.
4469 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4470 has ',' character inside of it.
4472 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4475 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4476 as clobbering CC_REGNUM.
4478 2018-03-08 Richard Biener <rguenther@suse.de>
4481 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4482 (follow_copies_to_constant): Do not follow SSA names registered
4485 2018-03-08 Richard Biener <rguenther@suse.de>
4487 PR tree-optimization/84178
4488 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4490 (version_loop_for_if_conversion): Delay update_ssa call.
4491 (tree_if_conversion): Delay update_ssa until after predicate
4494 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4496 PR tree-optimization/84178
4497 * tree-if-conv.c (release_bb_predicate): Remove the
4498 the assertion that the stmts have NULL use_ops.
4499 Discard the statements, asserting that they haven't
4500 yet been added to a BB.
4502 2018-03-08 Richard Biener <rguenther@suse.de>
4504 PR tree-optimization/84746
4505 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4506 (phi_translate): Pass in destination ANTIC_OUT set.
4507 (phi_translate_1): Likewise. For a simplified result lookup
4508 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4509 (phi_translate_set): Adjust.
4510 (do_pre_regular_insertion): Likewise.
4511 (do_pre_partial_partial_insertion): Likewise.
4513 2018-03-08 Martin Liska <mliska@suse.cz>
4515 PR gcov-profile/84735
4516 * doc/gcov.texi: Document usage of profile files.
4517 * gcov-io.h: Document changes in the format.
4519 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4523 * dwarf2out.c (struct dw_line_info_table): Update comments for
4525 (FORCE_RESET_NEXT_VIEW): New.
4526 (FORCE_RESETTING_VIEW_P): New.
4527 (RESETTING_VIEW_P): Check for -1 too.
4528 (ZERO_VIEW_P): Likewise.
4529 (new_line_info_table): Force-reset next view.
4530 (dwarf2out_begin_function): Likewise.
4531 (dwarf2out_source_line): Simplify zero_view_p initialization.
4532 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4533 view directly. Omit view when omitting .loc at line 0.
4535 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4537 PR tree-optimization/84740
4538 * tree-switch-conversion.c (process_switch): Call build_constructors
4539 only if info.phi_count is non-zero.
4541 PR tree-optimization/84739
4542 * tree-tailcall.c (find_tail_calls): Check call arguments against
4543 DECL_ARGUMENTS (current_function_decl) rather than
4544 DECL_ARGUMENTS (func) when checking for tail recursion.
4546 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4548 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4549 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4550 Volker Reichelt's entry and add entries for people that perform
4551 GCC fuzzy testing and report numerous bugs.
4553 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4556 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4557 readonly data in sdata, if that is disabled.
4558 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4559 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4560 -mreadonly-in-sdata option.
4562 2018-03-07 Martin Sebor <msebor@redhat.com>
4564 PR tree-optimization/84468
4565 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4566 basic block when looking for nul assignment.
4568 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4571 * except.h (output_function_exception_table): Adjust prototype.
4572 * except.c (output_function_exception_table): Remove FNNAME parameter
4573 and add SECTION parameter. Ouput one part of the table at a time.
4574 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4575 the first part of the exception table and emit unwind directives.
4576 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4577 (i386_pe_seh_cold_init): Likewise.
4578 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4579 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4580 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4581 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4582 * config/i386/winnt.c: Include except.h.
4583 (struct seh_frame_state): Add reg_offset, after_prologue and
4584 in_cold_section fields.
4585 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4586 (i386_pe_seh_cold_init): New function.
4587 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4588 to seh->in_cold_section.
4589 (seh_emit_push): Record the offset of the push.
4590 (seh_emit_save): Record the offet of the save.
4591 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4592 Test seh->after_prologue to disregard the epilogue.
4593 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4594 (i386_pe_end_cold_function): New function.
4596 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4599 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4600 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4603 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4605 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4606 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4609 * multiple_target.c: Include tree-inline.h and intl.h.
4610 (expand_target_clones): Diagnose and fail if node->definition and
4611 !tree_versionable_function_p (node->decl).
4613 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4615 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4617 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4618 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4619 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4621 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4624 * combine.c (try_combine): Use reg_or_subregno instead of handling
4625 just paradoxical SUBREGs and REGs.
4627 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4629 * config/arc/arc.c (arc_finalize_pic): Remove function.
4630 (arc_must_save_register): We use single base PIC register, remove
4631 checks to save/restore the PIC register.
4632 (arc_expand_prologue): Likewise.
4633 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4635 (arc_verify_short): Likewise.
4636 (arc_attr_type): Likewise.
4637 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4638 (walk_stores): Likewise.
4639 (arc_address_cost): Make it static.
4640 (arc_verify_short): Likewise.
4641 (branch_dest): Likewise.
4642 (arc_attr_type): Likewise.
4643 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4644 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4645 (arc_final_prescan_insn): Remove inserting the nops due to
4646 hardware hazards. It is done in reorg step.
4647 (insn_length_variant_t): Remove.
4648 (insn_length_parameters_t): Likewise.
4649 (arc_insn_length_parameters): Likewise.
4650 (arc_get_insn_variants): Likewise.
4651 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4653 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4656 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4659 PR tree-optimization/84687
4660 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4662 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4664 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4666 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4667 Rename to ppc_speculation_barrier.
4668 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4669 __builtin_ppc_speculation_barrier.
4671 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4674 * combine.c (combine_simplify_rtx): Don't try to simplify if
4675 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4678 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4680 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4681 to 32 bytes when compiling for POWER9.
4683 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4686 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4687 regparm >= 3 with no arg reg available also for calls with
4688 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4691 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4693 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4695 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4698 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4700 2018-03-05 Richard Biener <rguenther@suse.de>
4702 PR tree-optimization/84486
4703 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4704 When inserting a __builtin_assume_aligned call set the LHS
4705 SSA name alignment info accordingly.
4707 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4709 PR tree-optimization/84114
4710 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4711 Avoid reassociation of FLOAT_MODE addition.
4713 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4715 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4716 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4717 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4718 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4719 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4720 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4721 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4723 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4724 __builtin_ia32_wbinvd): New builtins.
4725 (SPECIAL_ARGS2): New.
4726 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4727 (SPECIAL_ARGS2): New.
4728 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4729 (ix86_valid_target_attribute_inner_p): Ditto.
4730 (ix86_init_mmx_sse_builtins): Add special_args2.
4731 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4732 TARGET_WBNOINVD_P): New.
4733 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4734 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4735 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4736 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4737 * config/i386/pconfigintrin.h: New file.
4738 * config/i386/wbnoinvdintrin.h: Ditto.
4739 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4740 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4742 2018-03-05 Richard Biener <rguenther@suse.de>
4744 PR tree-optimization/84670
4745 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4747 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4748 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4749 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4750 doesn't grow if all successors (recursively) were visited at least
4753 2018-03-05 Richard Biener <rguenther@suse.de>
4755 PR tree-optimization/84650
4756 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4757 if executed in the loop pipeline.
4759 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4761 * doc/configfiles.texi (Configuration Files): Move info about
4762 conditionalizing $target-protos.h to...
4763 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4764 differs from $target-protos.h.
4766 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4767 Chung-Ju Wu <jasonwucj@gmail.com>
4769 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4770 * config/nds32/nds32-multiple.md (setmemsi): Define.
4771 * config/nds32/nds32-memory-manipulation.c
4772 (nds32_gen_dup_4_byte_to_word_value): New.
4773 (emit_setmem_word_loop): New.
4774 (emit_setmem_byte_loop): New.
4775 (nds32_expand_setmem_loop): New.
4776 (nds32_expand_setmem_loop_v3m): New.
4777 (nds32_expand_setmem_unroll): New.
4778 (nds32_expand_setmem): New.
4780 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4781 Chung-Ju Wu <jasonwucj@gmail.com>
4783 * config/nds32/nds32-memory-manipulation.c
4784 (nds32_emit_load_store): New.
4785 (nds32_emit_post_inc_load_store): New.
4786 (nds32_emit_mem_move): New.
4787 (nds32_emit_mem_move_block): New.
4788 (nds32_expand_movmemsi_loop_unknown_size): New.
4789 (nds32_expand_movmemsi_loop_known_size): New.
4790 (nds32_expand_movmemsi_loop): New.
4791 (nds32_expand_movmemsi_unroll): New.
4792 (nds32_expand_movmemqi): Rename ...
4793 (nds32_expand_movmemsi): ... to this.
4794 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4795 (movmemsi): ... to this.
4796 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4797 (nds32_expand_movmemsi): ... to this.
4799 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4800 Monk Chiang <sh.chiang04@gmail.com>
4801 Chung-Ju Wu <jasonwucj@gmail.com>
4803 * config/nds32/nds32-protos.h
4804 (nds32_expand_load_multiple): New arguments.
4805 (nds32_expand_store_multiple): Ditto.
4806 (nds32_valid_multiple_load_store): Rename ...
4807 (nds32_valid_multiple_load_store_p): ... to this.
4808 * config/nds32/nds32-memory-manipulation.c
4809 (nds32_expand_load_multiple): Refine implementation.
4810 (nds32_expand_store_multiple): Ditto.
4811 * config/nds32/nds32-multiple.md
4812 (load_multiple): Update nds32_expand_load_multiple interface.
4813 (store_multiple): Update nds32_expand_store_multiple interface.
4814 * config/nds32/nds32-predicates.c
4815 (nds32_valid_multiple_load_store): Rename ...
4816 (nds32_valid_multiple_load_store_p): ... to this and refine
4818 * config/nds32/predicates.md
4819 (nds32_load_multiple_and_update_address_operation): New predicate.
4820 (nds32_store_multiple_and_update_address_operation): New predicate.
4822 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4823 Chung-Ju Wu <jasonwucj@gmail.com>
4825 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4826 (combo): New attribute.
4827 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4829 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4831 * config/nds32/nds32.opt: Change -mcmodel= default value.
4833 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4834 Monk Chiang <sh.chiang04@gmail.com>
4835 Chung-Ju Wu <jasonwucj@gmail.com>
4837 * config/nds32/constants.md (unspec_element): New enum.
4838 * config/nds32/constraints.md (Umw): New constraint.
4839 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4840 * config/nds32/nds32-intrinsic.md: Likewise.
4841 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4842 (nds32_valid_smw_lwm_base_p): New.
4843 (nds32_output_smw_single_word): New.
4844 (nds32_output_lmw_single_word): New.
4845 (nds32_expand_unaligned_load): New.
4846 (nds32_expand_unaligned_store): New.
4847 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4848 (nds32_output_smw_single_word): Declare.
4849 (nds32_output_lmw_single_word): Declare.
4850 (nds32_expand_unaligned_load): Declare.
4851 (nds32_expand_unaligned_store): Declare.
4852 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4853 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4854 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4855 NDS32_BUILTIN_UASTORE_DW.
4856 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4859 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4860 Kito Cheng <kito.cheng@gmail.com>
4861 Chung-Ju Wu <jasonwucj@gmail.com>
4863 * config/nds32/nds32-intrinsic.c
4864 (nds32_expand_builtin_null_ftype_reg): Delete.
4865 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4866 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4867 (nds32_read_argument): New.
4868 (nds32_legitimize_target): Ditto.
4869 (nds32_legitimize_argument): Ditto.
4870 (nds32_check_constant_argument): Ditto.
4871 (nds32_expand_unop_builtin): Ditto.
4872 (nds32_expand_unopimm_builtin): Ditto.
4873 (nds32_expand_binop_builtin): Ditto.
4874 (nds32_builtin_decl_impl): Ditto.
4875 (builtin_description): Ditto.
4876 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4877 (nds32_init_builtins_impl): Ditto.
4878 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4879 (nds32_builtin_decl): New.
4880 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4881 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4883 2018-03-02 Jeff Law <law@redhat.com>
4885 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4886 (redundant_insn, fill_simple_delay_slots): Likewise.
4887 (fill_slots_from_thread): Likewise.
4888 * resource.c (mark_referenced_resources): Likewise.
4889 (mark_set_resources, find_dead_or_set_registers): Likewise.
4891 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4893 * substring-locations.h (format_warning_va): Formatting fix for
4895 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4897 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4898 * substring-locations.c: Include intl.h.
4899 (format_warning_va): Turned into small wrapper around
4900 format_warning_n_va, renamed to ...
4901 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4902 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4904 (format_warning_at_substring_n): New function.
4905 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4906 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4907 format_warning_at_substring with just a shorter name instead of
4908 const function pointer.
4909 (fmtwarn_n): New function.
4910 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4911 appropriate, get rid of all the fmtstr temporaries, move conditionals
4912 with G_() wrapped string literals directly into fmtwarn arguments,
4913 cast dir.len to (int), formatting fixes.
4915 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4917 * doc/invoke.texi: Remove "Cilk Plus" references.
4919 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4920 Richard Biener <rguenther@suse.de>
4923 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4924 for error or warning attributes if CALL_FROM_THUNK_P is set.
4927 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4930 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4931 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4934 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4935 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4937 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4938 instead of -1U in last predictors element's probability member.
4940 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4943 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4944 arguments if they are comparable.
4946 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4948 PR tree-optimization/84634
4949 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4950 masks and masked_loop_p with a single loop_masks, making sure it's
4951 null for bb vectorization.
4953 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4955 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4956 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4957 loop->force_vectorize to check whether there is no alias.
4959 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4962 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4964 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4966 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4967 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4968 instead of a loop around prev_real_insn.
4969 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4973 * config/i386/i386.c (ix86_print_operand): Use conditional
4974 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4977 2018-03-02 Richard Biener <rguenther@suse.de>
4979 PR tree-optimization/84427
4980 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4981 (bitmap_set_subtract_values): Rewrite to handle multiple
4984 (prune_clobbered_mems): Likewise.
4985 (phi_translate): Take edge instead of pred/phiblock.
4986 (phi_translate_1): Likewise.
4987 (phi_translate_set): Likewise. Insert all translated
4988 exprs for a value into the set, keeping possibly multiple
4989 expressions per value.
4990 (compute_antic_aux): Adjust for phi_translate changes.
4991 When intersecting union the expressions and prune those
4992 not in the final value set, keeping possibly multiple
4993 expressions per value. Do not use value-insertion
4994 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4995 all expressions. Add verification that the value-sets
4996 only shrink during iteration.
4997 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4998 (do_pre_regular_insertion): Likewise.
4999 (do_pre_partial_partial_insertion): Likewise.
5001 2018-03-02 Richard Biener <rguenther@suse.de>
5004 * config/darwin.c (saved_debug_info_level): New static global.
5005 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5006 (darwin_asm_lto_end): Restore debug info generation settings.
5008 2018-03-01 Martin Liska <mliska@suse.cz>
5011 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5014 2018-03-01 Richard Biener <rguenther@suse.de>
5017 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5018 type annotation with LTO when debug was disabled at compile-time.
5020 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5022 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5024 (mips_final_postscan_insn): Likewise.
5026 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5028 PR rtl-optimization/84528
5029 * alias.c (init_alias_target): Add commentary.
5030 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5031 a unique base value if the frame pointer is not eliminated
5032 to the stack pointer.
5034 2018-03-01 Tom de Vries <tom@codesourcery.com>
5036 PR rtl-optimization/83327
5037 * lra-int.h (hard_regs_spilled_into): Declare.
5038 * lra.c (hard_regs_spilled_into): Define.
5039 (init_reg_info): Init hard_regs_spilled_into.
5040 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5041 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5042 (process_bb_lives): Handle hard_regs_spilled_into.
5043 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5044 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5046 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5048 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5049 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5050 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5051 * config/rs6000/aix72.h: New file.
5053 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5055 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5056 instead of warning_at with conditional singular and plural messages
5060 * stor-layout.c (update_alignment_for_field): For
5061 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5062 && !DECL_PACKED (field), do the alignment update, just use
5063 only desired_align instead of MAX (type_align, desired_align)
5065 (place_field): Don't do known_align < desired_align handling
5066 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5067 is non-NULL, instead do it after rli->prev_field handling and
5068 only if not within a bitfield word. For DECL_PACKED (field)
5069 use type_align of BITS_PER_UNIT.
5071 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5073 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5074 superfluous parentheses and trailing spaces.
5076 2018-02-28 Richard Biener <rguenther@suse.de>
5078 PR tree-optimization/84584
5079 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5080 SCoPs with fake exit edge.
5082 2018-02-28 Martin Liska <mliska@suse.cz>
5085 * timevar.c (timer::print): Fix format to properly print 100%
5088 2018-02-28 Richard Biener <rguenther@suse.de>
5091 * genmatch.c (capture_info::walk_match): Do not mark
5092 captured expressions without operands as expr_p given
5093 they act more like predicates and should be subject to
5094 "lost tail" side-effect preserving.
5096 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5098 PR rtl-optimization/81611
5099 * auto-inc-dec.c (attempt_change): Move dead note from
5100 mem_insn if it's the next use of regno
5101 (find_address): Take address use of reg holding
5102 non-incremented value. Add parm to limit search to the named
5104 (merge_in_block): Attempt to use a mem insn that is the next
5105 use of the original regno.
5107 2018-02-27 Martin Sebor <msebor@redhat.com>
5110 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5111 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5113 2018-02-27 Martin Sebor <msebor@redhat.com>
5115 PR translation/84207
5116 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5117 n argument to unsigned HOST_WIDE_INT.
5118 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5119 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5120 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5121 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5123 2018-02-27 Richard Biener <rguenther@suse.de>
5125 PR tree-optimization/84512
5126 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5127 Do not use the estimate returned from record_stmt_cost for
5128 the scalar iteration cost but sum properly using add_stmt_cost.
5130 2018-02-27 Richard Biener <rguenther@suse.de>
5132 PR tree-optimization/84466
5133 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5134 Adjust last change to less strictly validate use operands.
5136 2018-02-27 Martin Liska <mliska@suse.cz>
5138 PR gcov-profile/84548
5139 * gcov.c (process_file): Allow partial overlap and consider it
5140 also as group functions.
5141 (output_lines): Properly calculate range of lines for a group.
5143 2018-02-27 Martin Liska <mliska@suse.cz>
5145 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5146 'ggc' suffixes. Change first column width.
5147 (timer::print): Fix formatting of the column.
5149 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5151 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5152 preserve inline entry blocks for the sake of debug inline
5153 entry point markers alone.
5154 (remove_unused_locals): Suggest in comments a better place to
5155 force the preservation of inline entry blocks that are
5156 otherwise unused, but do not preserve them.
5158 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5160 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5162 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5165 * config/i386/constraints.md (Bs): Replace
5166 ix86_indirect_branch_register with
5167 TARGET_INDIRECT_BRANCH_REGISTER.
5169 * config/i386/i386.md (indirect_jump): Likewise.
5170 (tablejump): Likewise.
5171 (*sibcall_memory): Likewise.
5172 (*sibcall_value_memory): Likewise.
5173 Peepholes of indirect call and jump via memory: Likewise.
5174 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5175 (*sibcall_value_GOT_32): Likewise.
5176 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5177 (GOT_memory_operand): Likewise.
5178 (call_insn_operand): Likewise.
5179 (sibcall_insn_operand): Likewise.
5180 (GOT32_symbol_operand): Likewise.
5181 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5183 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5185 PR rtl-optimization/83496
5186 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5187 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5188 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5189 redundant insn, if any.
5190 (relax_delay_slots): Likewise.
5191 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5193 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5195 PR tree-optimization/83965
5196 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5197 that grouped statements are part of a reduction chain. Return
5198 true if the statement is not marked as a reduction itself but
5200 (vect_recog_dot_prod_pattern): Don't check whether the statement
5201 is part of a group here.
5202 (vect_recog_sad_pattern): Likewise.
5203 (vect_recog_widen_sum_pattern): Likewise.
5205 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5208 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5210 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5213 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5215 (ix86_output_indirect_function_return): New prototype.
5216 (ix86_split_simple_return_pop_internal): Likewise.
5217 * config/i386/i386.c (indirect_return_via_cx): New.
5218 (indirect_return_via_cx_bnd): Likewise.
5219 (indirect_thunk_name): Handle return va CX_REG.
5220 (output_indirect_thunk_function): Create alias for
5221 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5222 (ix86_output_indirect_jmp): Remove the bool argument.
5223 (ix86_output_indirect_function_return): New function.
5224 (ix86_split_simple_return_pop_internal): Likewise.
5225 * config/i386/i386.md (*indirect_jump): Don't pass false
5226 to ix86_output_indirect_jmp.
5227 (*tablejump_1): Likewise.
5228 (simple_return_pop_internal): Change it to define_insn_and_split.
5229 Call ix86_split_simple_return_pop_internal to split it for
5231 (simple_return_indirect_internal): Call
5232 ix86_output_indirect_function_return instead of
5233 ix86_output_indirect_jmp.
5235 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5238 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5239 memset and value initialization afterwards.
5241 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5243 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5245 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5248 * common/config/aarch64/aarch64-common.c
5249 (aarch_option_optimization_table[]): Switch
5250 off fomit-frame-pointer
5252 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5253 Chung-Ju Wu <jasonwucj@gmail.com>
5255 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5257 (store_multiple): Ditto.
5259 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5261 * config.gcc: Add --with-cpu support for nds32 target.
5262 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5263 * config/nds32/nds32.opt: Add -mcpu= option.
5265 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5267 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5268 isel=yes): Warn for these deprecated options.
5270 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5272 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5273 ISA_2_5_MASKS_EMBEDDED.
5275 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5277 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5278 p->max as pointers rather than using iterative_hash_expr.
5280 2018-02-23 Carl Love <cel@us.ibm.com>
5282 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5283 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5285 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5286 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5287 P8V_BUILTIN_VEC_VUNSIGNED2.
5289 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5292 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5293 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5295 * lra-constraints.c (curr_insn_transform): Set up
5296 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5298 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5299 * lra-eliminations.c (spill_pseudos): Ditto.
5300 (process_insn_for_elimination): Ditto.
5301 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5302 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5303 LRA_NON_CLOBBERED_ALT.
5305 2018-02-22 Martin Sebor <msebor@redhat.com>
5307 PR tree-optimization/84480
5308 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5309 to maybe_diag_stxncpy_trunc. Call it.
5310 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5311 from gimple_fold_builtin_strcpy. Print inlining stack.
5312 (handle_builtin_stxncpy): Print inlining stack.
5313 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5315 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5318 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5319 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5320 and -fcheck-pointer-bounds are used together.
5321 (indirect_thunk_prefix): New enum.
5322 (indirect_thunk_need_prefix): New function.
5323 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5324 "_nt" instead of "_bnd" for NOTRACK prefix.
5325 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5326 (output_indirect_thunk_function): Likewise.
5328 (ix86_code_end): Update output_indirect_thunk_function calls.
5329 (ix86_output_indirect_branch_via_reg): Replace
5330 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5331 (ix86_output_indirect_branch_via_push): Likewise.
5332 (ix86_output_function_return): Likewise.
5333 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5334 incompatible with -fcf-protection=branch and
5335 -fcheck-pointer-bounds.
5337 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5340 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5341 Change gcc_assert call to output_operand_lossage.
5343 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5345 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5347 2018-02-22 DJ Delorie <dj@redhat.com>
5348 Sebastian Perta <sebastian.perta@renesas.com>
5349 Oleg Endo <olegendo@gcc.gnu.org>
5351 * config/rx/rx.c (rx_rtx_costs): New function.
5352 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5354 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5356 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5358 2018-02-22 Martin Liska <mliska@suse.cz>
5361 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5362 Add "native" as a possible value.
5364 2018-02-22 Martin Liska <mliska@suse.cz>
5367 * config/i386/i386.c (ix86_option_override_internal):
5368 Add "native" as a possible value for -march and -mtune.
5370 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5373 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5374 to all type variants.
5376 PR tree-optimization/84503
5377 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5378 width as info->bitpos + info->bitsize - start.
5379 (merged_store_group::merge_overlapping): Simplify width computation.
5380 (check_no_overlap): New function.
5381 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5382 start + width and last_order of the group, fail if check_no_overlap
5384 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5385 to group if check_no_overlap fails.
5387 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5389 * config/rs6000/altivec.md: Delete contraint arguments to
5390 define_expand, define_split, and define_peephole2, and in
5391 define_insn_and_split if always unused.
5392 * config/rs6000/darwin.md: Ditto.
5393 * config/rs6000/dfp.md: Ditto.
5394 * config/rs6000/rs6000.md: Ditto.
5395 * config/rs6000/sync.md: Ditto.
5396 * config/rs6000/vector.md: Ditto.
5397 * config/rs6000/vsx.md: Ditto.
5399 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5401 * config/rs6000/altivec.md: Write output control strings as braced
5402 blocks instead of double-quoted strings.
5403 * config/rs6000/darwin.md: Ditto.
5404 * config/rs6000/rs6000.md: Ditto.
5405 * config/rs6000/vector.md: Ditto.
5406 * config/rs6000/vsx.md: Ditto.
5408 2018-02-21 Jason Merrill <jason@redhat.com>
5410 PR c++/84314 - ICE with templates and fastcall attribute.
5411 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5413 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5415 * ipa-cp.c (determine_versionability): Fix comment typos.
5417 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5420 * ipa-cp.c (determine_versionability): Do not version functions caling
5423 2018-02-21 Martin Liska <mliska@suse.cz>
5426 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5427 Add "native" as a possible value.
5428 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5429 the macro when native cpu detection is available.
5431 2018-02-21 Martin Liska <mliska@suse.cz>
5434 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5435 Add "native" as a possible value.
5436 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5437 when native cpu detection is available.
5439 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5440 Martin Sebor <msebor@redhat.com>
5442 PR tree-optimization/84478
5443 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5445 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5446 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5447 support which is conservatively correct, for 2 only stay conservative
5448 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5449 argument to the 2 argument get_range_strlen, adjust 6 arg
5450 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5452 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5453 (gimple_fold_builtin_strlen): Pass true as last argument to
5456 2018-02-20 Martin Sebor <msebor@redhat.com>
5459 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5460 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5461 (builtin_memref::builtin_memref): Factor out parts into
5462 set_base_and_offset and call it.
5464 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5467 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5468 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5469 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5470 search at the associated MODE_INT.
5472 2018-02-20 Jeff Law <law@redhat.com>
5475 PR tree-optimization/81592
5477 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5478 for range data rather than using global data.
5479 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5480 range data rather than using global data.
5481 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5482 pass it to children as needed.
5483 (struct directive::fmtresult): Similarly.
5484 (struct directive::set_width): Similarly.
5485 (struct directive::set_precision): Similarly.
5486 (format_integer, format_directive, parse_directive): Similarly.
5487 (format_none): Accept unnamed vr_values parameter.
5488 (format_percent, format_floating, format_character): Similarly.
5489 (format_string, format_plain): Similarly.
5490 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5491 the EVRP range analyzer for range data rather than using global data.
5492 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5493 gimple-ssa-evrp-analyze.h
5494 (class sprintf_dom_walker): Add after_dom_children member function.
5495 Add evrp_range_analyzer member.
5496 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5497 range analyzer as needed.
5498 (sprintf_dom_walker::after_dom_children): New member function.
5499 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5501 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5502 (evrp_range_analyzer::pop_to_marker): Likewise.
5504 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5506 PR tree-optimization/84419
5507 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5508 with the required type if its current type is compatible but
5511 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5514 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5515 after vectorization.
5517 2018-02-20 Martin Liska <mliska@suse.cz>
5520 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5521 possible values if we don't have a hint.
5523 2018-02-20 Martin Liska <mliska@suse.cz>
5527 * final.c (shorten_branches): Build align_tab array with one
5529 * opts.c (finish_options): Add alignment option limit check.
5530 (MAX_CODE_ALIGN): Likewise.
5531 (MAX_CODE_ALIGN_VALUE): Likewise.
5532 * doc/invoke.texi: Document maximum allowed option value for
5533 all -falign-* options.
5535 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5538 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5539 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5540 * var-tracking.c (emit_note_insn_var_location): Remove all references
5541 to NOTE_INSN_CALL_ARG_LOCATION.
5542 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5543 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5544 Use copy_rtx_if_shared.
5545 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5546 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5547 (dwarf2out_var_location): Remove handling of
5548 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5550 * final.c (final_scan_insn): Remove all references to
5551 NOTE_INSN_CALL_ARG_LOCATION.
5552 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5553 before dumping final insns.
5554 * except.c (emit_note_eh_region_end): Remove all references to
5555 NOTE_INSN_CALL_ARG_LOCATION.
5556 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5557 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5558 * config/arc/arc.c (hwloop_optimize): Likewise.
5559 * config/arm/arm.c (create_fix_barrier): Likewise.
5560 * config/s390/s390.c (s390_chunkify_start): Likewise.
5561 * config/sh/sh.c (find_barrier): Likewise.
5562 * config/i386/i386.c (rest_of_insert_endbranch,
5563 ix86_seh_fixup_eh_fallthru): Likewise.
5564 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5565 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5566 * config/frv/frv.c (frv_function_prologue): Likewise.
5567 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5569 (note_outside_basic_block_p): Remove all references to
5570 NOTE_INSN_CALL_ARG_LOCATION.
5571 * gengtype.c (adjust_field_rtx_def): Likewise.
5572 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5574 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5575 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5578 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5581 PR tree-optimization/84452
5582 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5583 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5586 2018-02-19 Martin Liska <mliska@suse.cz>
5589 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5591 2018-02-19 Martin Liska <mliska@suse.cz>
5592 Richard Sandiford <richard.sandiford@linaro.org>
5594 PR tree-optimization/82491
5595 * gimple-fold.c (get_base_constructor): Make earlier bail out
5598 2018-02-19 Carl Love <cel@us.ibm.com>
5600 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5601 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5603 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5604 P8V_BUILTIN_VEC_NEG.
5606 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5608 * config/rl78/rl78.md (movdf): New define expand.
5610 2018-02-19 Martin Liska <mliska@suse.cz>
5613 * doc/invoke.texi: Fix typo.
5614 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5616 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5618 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5619 handle rs6000_single_float and rs6000_double_float specially for
5622 2018-02-16 Jeff Law <law@redhat.com>
5624 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5627 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5630 * ipa-inline.c (inline_small_functions): Fix a typo.
5632 2018-02-16 Nathan Sidwell <nathan@acm.org>
5634 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5636 2018-02-16 Carl Love <cel@us.ibm.com>
5638 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5639 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5640 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5641 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5642 expansion to P8V_BUILTIN_VEC_FLOAT2.
5644 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5646 PR rtl-optimization/70023
5647 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5648 src_regno into account.
5650 2018-02-16 Carl Love <cel@us.ibm.com>
5652 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5653 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5654 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5655 * config/rs6000/rs6000.c: Remove case statements for
5656 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5657 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5658 and P9V_BUILTIN_VEC_VINSERT4B.
5659 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5660 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5661 * config/rs6000/vsx.md:
5662 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5665 2018-02-16 Carl Love <cel@us.ibm.com>
5667 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5669 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5671 * config/rs6000/rs6000-c.c: Add the definitions for
5672 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5673 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5674 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5675 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5676 definition for insert4b and define insn *insert3b_internal.
5677 * doc/extend.texi: Add documentation for vec_extract4b.
5679 2018-02-16 Nathan Sidwell <nathan@acm.org>
5681 * doc/extend.texi (Backwards Compatibility): Mention friend
5682 injection. Note for-scope is deprecated.
5683 * doc/invoke.texi (-ffriend-injection): Deprecate.
5685 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5687 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5688 that moved to I2, also allow destinations that are a paradoxical
5689 subreg (instead of a normal reg).
5691 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5694 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5697 2018-02-16 Richard Biener <rguenther@suse.de>
5699 PR tree-optimization/84037
5700 PR tree-optimization/84016
5702 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5703 Adjust vec_construct for the fact we need additional higher latency
5704 128bit inserts for AVX256 and AVX512 vector builds.
5705 (ix86_add_stmt_cost): Scale vector construction cost for
5708 2018-02-16 Richard Biener <rguenther@suse.de>
5710 PR tree-optimization/84417
5711 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5712 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5713 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5715 2018-02-16 Martin Liska <mliska@suse.cz>
5718 * internal-fn.def (ASAN_CHECK): Set proper flags.
5719 (ASAN_MARK): Likewise.
5721 2018-02-16 Julia Koval <julia.koval@intel.com>
5723 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5724 from PTA_CANNONLAKE.
5726 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5729 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5730 Use ++iter rather than iter++ for std::list iterators.
5731 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5732 defer deleting them until all nodes in the forest are processed. Do
5733 free even leaf nodes. Change to_process into auto_vec.
5736 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5737 * vec.h (vec_default_construct): Use memset instead of placement new
5738 if BROKEN_VALUE_INITIALIZATION is defined.
5739 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5740 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5743 PR rtl-optimization/83723
5744 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5745 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5746 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5748 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5750 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5752 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5754 PR rtl-optimization/81443
5755 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5756 from inner REGs to paradoxical SUBREGs.
5758 2018-02-16 Richard Biener <rguenther@suse.de>
5760 PR tree-optimization/84399
5761 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5762 For operands we can analyze at their definition make sure we can
5763 analyze them at each use as well.
5765 2018-02-16 Richard Biener <rguenther@suse.de>
5767 PR tree-optimization/84190
5768 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5769 volatile accesses if the decl isn't volatile.
5771 2018-02-15 Jason Merrill <jason@redhat.com>
5773 PR c++/84314 - ICE with templates and fastcall attribute.
5774 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5775 TYPE_CANONICAL on an existing type.
5777 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5779 PR tree-optimization/84383
5780 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5781 dstoff nor call operand_equal_p if dstbase is NULL.
5783 PR tree-optimization/84334
5784 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5785 also a CONSTANT_CLASS_P, punt.
5787 2018-02-14 Jim Wilson <jimw@sifive.com>
5789 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5790 first SMALL_OPERAND check. New local min_second_step. Move assert
5791 to where locals are set. Add TARGET_RVC support.
5792 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5794 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5796 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5798 2018-02-14 Martin Sebor <msebor@redhat.com>
5800 PR tree-optimization/83698
5801 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5802 arrays constrain the offset range to their bounds.
5803 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5804 (builtin_access::overlap): Avoid setting the size of overlap if it's
5806 (maybe_diag_overlap): Also consider arrays when deciding what values
5807 of offsets to include in diagnostics.
5809 2018-02-14 Martin Sebor <msebor@redhat.com>
5812 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5813 that correspond to the kind of a declaration.
5815 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5818 * config/pa/pa.md: Load address of PIC label using the linkage table
5819 if the label is nonlocal.
5821 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5823 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5824 warning message if user requests -maltivec=be.
5825 * doc/invoke.texi: Document deprecation of -maltivec=be.
5827 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5830 * config/rs6000/rs6000-c.c: Update definitions for
5831 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5832 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5834 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5837 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5838 add _get_ssp intrinsics. Remove argument from
5839 __builtin_ia32_rdssp[d|q].
5840 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5841 * config/i386/i386-builtin.def: Remove argument from
5842 __builtin_ia32_rdssp[d|q].
5843 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5844 ix86_expand_special_args_builtin for _rdssp[d|q].
5845 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5846 Clear register before usage.
5847 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5848 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5850 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5852 PR tree-optimization/84357
5853 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5854 operand 1 of an ARRAY_REF too.
5856 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5859 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5860 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5862 (set_of_reg): New struct.
5863 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5864 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5865 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5867 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5868 Split into bitclr, bitset, bitinvert patterns if appropriate.
5869 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5870 use rx_fuse_in_memory_bitop.
5871 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5872 to named insn, correct maximum insn length.
5874 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5877 * machmode.def: Define a complex mode for PARTIAL_INT.
5878 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5880 * doc/rtl.texi: Document CSPImode.
5881 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5883 (msp430_hard_regno_nregs_with_padding): Likewise.
5885 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5888 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5890 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5892 PR rtl-optimization/84169
5893 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5894 we generated a parallel as new i3 and we split that to new i2 and i3
5895 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5896 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5897 those to i2, not i1. Partially rewrite this scan code.
5899 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5902 * stor-layout.c (place_field): For variable length fields, adjust
5903 offset_align afterwards not just based on the field's alignment,
5904 but also on the size.
5907 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5908 of exps and logs in the use_exp2 case.
5910 2018-02-13 Jeff Law <law@redhat.com>
5912 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5915 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5918 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5922 * common.opt (gas-loc-support, gas-locview-support): New.
5923 (ginline-points, ginternal-reset-location-views): New.
5924 * doc/invoke.texi: Document them. Use @itemx where intended.
5925 (gvariable-location-views): Adjust.
5926 * target.def (reset_location_view): New.
5927 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5928 (TARGET_RESET_LOCATION_VIEW): New.
5929 * doc/tm.texi: Rebuilt.
5930 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5931 (dwarf2out_default_as_locview_support): New.
5932 (output_asm_line_debug_info): Use option variables.
5933 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5934 (output_loc_list): Likewise.
5935 (add_high_low_attributes): Check option variables.
5936 Don't output entry view attribute in strict mode.
5937 (gen_inlined_subroutine_die): Check option variables.
5938 (dwarf2out_inline_entry): Likewise.
5939 (init_sections_and_labels): Likewise.
5940 (dwarf2out_early_finish): Likewise.
5941 (maybe_reset_location_view): New, from...
5942 (dwarf2out_var_location): ... here. Call it.
5943 * debug.h (dwarf2out_default_as_loc_support): Declare.
5944 (dwarf2out_default_as_locview_support): Declare.
5945 * hooks.c (hook_int_rtx_insn_0): New.
5946 * hooks.h (hook_int_rtx_insn_0): Declare.
5947 * toplev.c (process_options): Take -gas-loc-support and
5948 -gas-locview-support from dwarf2out. Enable
5949 -gvariable-location-views by default only with locview
5950 assembler support. Enable -ginternal-reset-location-views by
5951 default only if the target defines the corresponding hook.
5952 Enable -ginline-points by default if location views are
5953 enabled; force it disabled if statement frontiers are
5955 * tree-inline.c (expand_call_inline): Check option variables.
5956 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5958 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5960 PR tree-optimization/84321
5961 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5962 handling. Also check whether the anti-range contains any values
5963 that satisfy the mask; switch to a VR_RANGE if not.
5965 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5968 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5970 2018-02-13 Martin Jambor <mjambor@suse.cz>
5973 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5974 of call statements, also set location of a load to a temporary.
5976 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5978 * config/rl78/rl78.c (add_vector_labels): New function.
5979 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5980 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5981 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5982 which checks that no arguments are passed.
5983 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5984 * doc/extend.texi: Documentation for the new attribute.
5986 2018-02-13 Andreas Schwab <schwab@suse.de>
5988 * config/riscv/linux.h (CPP_SPEC): Define.
5990 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5993 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5994 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5995 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5996 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5997 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5998 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5999 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6000 temporarily for AES and PCLMUL builtins.
6002 PR tree-optimization/84339
6003 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6004 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6008 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6009 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6010 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6012 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6014 * omp-simd-clone.h: New file.
6015 * omp-simd-clone.c: Include omp-simd-clone.h.
6016 (expand_simd_clones): No longer static.
6017 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6018 cgraph.h and omp-simd-clone.h.
6019 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6020 (vect_recog_widen_shift_pattern): Formatting fix.
6021 (vect_pattern_recog_1): Don't check optab for calls.
6024 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6025 operands[2] into a REG before using gen_lowpart on it.
6027 2018-02-12 Jeff Law <law@redhat.com>
6030 * config/sh/sh.c (find_barrier): Consider a sibling call
6033 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6034 successfully back substituting a reg.
6036 2018-02-12 Richard Biener <rguenther@suse.de>
6038 PR tree-optimization/84037
6039 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6040 parameter, move visited init to caller.
6041 (vect_slp_analyze_operations): Separate cost from validity
6042 check, initialize visited once for all instances.
6043 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6045 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6047 (vect_model_promotion_demotion_cost): Likewise.
6048 (vectorizable_bswap): Guard cost modeling with !slp_node
6049 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6051 (vectorizable_call): Likewise.
6052 (vectorizable_conversion): Likewise.
6053 (vectorizable_assignment): Likewise.
6054 (vectorizable_shift): Likewise.
6055 (vectorizable_operation): Likewise.
6056 (vectorizable_store): Likewise.
6057 (vectorizable_load): Likewise.
6058 (vectorizable_condition): Likewise.
6059 (vectorizable_comparison): Likewise.
6061 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6064 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6065 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6066 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6068 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6071 * params.def (inline-min-speedup): Increase from 8 to 15.
6072 (max-inline-insns-auto): Decrease from 40 to 30.
6073 * ipa-split.c (consider_split): Add some buffer for function to
6074 be considered inlining candidate.
6075 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6078 2018-02-12 Richard Biener <rguenther@suse.de>
6080 PR tree-optimization/84037
6081 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6082 matched stmts if we cannot swap the non-matched ones.
6084 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6086 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6087 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6088 _mm_maskz_scalef_round_ss): New intrinsics.
6089 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6090 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6091 __builtin_ia32_scalefss_round): Remove.
6092 (__builtin_ia32_scalefsd_mask_round,
6093 __builtin_ia32_scalefss_mask_round): New intrinsics.
6094 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6095 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6096 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6097 "<round_constraint>")): Changed to ...
6098 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6099 "<round_scalar_constraint>")): ... this.
6100 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6101 %0, %1, %2<round_op3>}"): Changed to ...
6102 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6103 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6104 %2<round_scalar_mask_op3>}"): ... this.
6105 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6107 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6109 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6110 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6111 (_mm_maskz_sqrt_round_ss): New intrinsics.
6112 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6113 (__builtin_ia32_sqrtsd_mask_round)
6114 (__builtin_ia32_sqrtss_mask_round): New builtins.
6115 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6116 (__builtin_ia32_sqrtss_round): Remove.
6117 (__builtin_ia32_sqrtsd_mask_round)
6118 (__builtin_ia32_sqrtss_mask_round): New builtins.
6119 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6120 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6121 ((match_operand:VF_128 1 "vector_operand"
6122 "xBm,<round_constraint>")): Changed to ...
6123 ((match_operand:VF_128 1 "vector_operand"
6124 "xBm,<round_scalar_constraint>")): ... this.
6125 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6126 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6127 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6128 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6129 %<iptr>1<round_scalar_mask_op3>}): ... this.
6130 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6131 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6133 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6136 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6137 Cast vec_cmpeq result to correct type.
6138 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6139 Cast vec_cmpgt result to correct type.
6141 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6143 * final.c (final_scan_insn_1): Renamed from...
6144 (final_scan_insn): ... this. New wrapper, to recover
6145 seen from the outermost call in recursive ones.
6146 * config/sparc/sparc.c (output_return): Drop seen from call.
6147 (output_sibcall): Likewise.
6148 * config/visium/visium.c (output_branch): Likewise.
6150 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6152 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6155 2018-02-10 Alan Modra <amodra@gmail.com>
6158 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6159 Specify LR as an input.
6161 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6164 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6165 remove_member_access_dummy_vars): New functions.
6166 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6167 lower_omp_1, execute_lower_omp): Use them.
6169 PR rtl-optimization/84308
6170 * shrink-wrap.c (spread_components): Release todo vector.
6172 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6174 PR rtl-optimization/57193
6175 * ira-color.c (struct allocno_color_data): Add member
6176 conflict_allocno_hard_prefs.
6177 (update_conflict_allocno_hard_prefs): New.
6178 (bucket_allocno_compare_func): Add a preference based on
6179 conflict_allocno_hard_prefs.
6180 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6181 (color_allocnos): Remove a dead code. Initiate
6182 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6184 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6187 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6188 constraint from =wa to wa. Avoid a subreg on the output operand,
6189 instead use a pseudo and subreg it in a move.
6190 (p9_xxbrd_<mode>): Changed to ...
6191 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6192 (p9_xxbrd_v2df): New expander.
6193 (p9_xxbrw_<mode>): Changed to ...
6194 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6195 (p9_xxbrw_v4sf): New expander.
6197 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6199 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6201 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6204 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6205 multiply in 32-bit mode.
6206 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6207 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6210 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6212 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6213 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6214 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6215 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6217 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6220 * dwarf2out.c (is_trivial_indirect_ref): New function.
6221 (dwarf2out_late_global_decl): Do not generate a location
6222 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6223 and that are not defined in the current unit.
6225 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6227 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6228 instead of a libcall for UNORDERED.
6230 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6233 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6234 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6236 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6239 * config/s390/s390.c (s390_set_current_function): Invoke
6240 s390_indirect_branch_settings also if fndecl didn't change.
6242 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6244 * config/rs6000/rs6000.md (blockage): Set length to zero.
6246 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6248 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6250 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6253 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6254 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6258 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6259 PARALLEL incoming that failed vt_get_decl_and_offset check.
6262 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6263 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6265 (get_variable_section): For decls in named .bss* sections pass true as
6266 second argument to bss_initializer_p.
6268 2018-02-09 Marek Polacek <polacek@redhat.com>
6269 Jakub Jelinek <jakub@redhat.com>
6272 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6273 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6274 Sync some changes from cxx_fold_indirect_ref.
6276 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6278 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6280 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6281 (BLOCK_INLINE_ENTRY_LABEL): New.
6282 (dwarf2out_var_location): Disregard inline entry markers.
6283 (inline_entry_data): New struct.
6284 (inline_entry_data_hasher): New hashtable type.
6285 (inline_entry_data_hasher::hash): New.
6286 (inline_entry_data_hasher::equal): New.
6287 (inline_entry_data_table): New variable.
6288 (add_high_low_attributes): Add DW_AT_entry_pc and
6289 DW_AT_GNU_entry_view attributes if a pending entry is found
6290 in inline_entry_data_table. Add old entry_pc attribute only
6291 if debug nonbinding markers are disabled.
6292 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6293 markers are enabled.
6294 (block_within_block_p, dwarf2out_inline_entry): New.
6295 (dwarf2out_finish): Check that no entries remained in
6296 inline_entry_data_table.
6297 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6298 (final_scan_insn, notice_source_line): Likewise.
6299 (rest_of_clean_state): Skip inline entry markers.
6300 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6302 * gimple.c (gimple_build_debug_inline_entry): New.
6303 * gimple.h (enum gimple_debug_subcode): Add
6304 GIMPLE_DEBUG_INLINE_ENTRY.
6305 (gimple_build_debug_inline_entry): Declare.
6306 (gimple_debug_inline_entry_p): New.
6307 (gimple_debug_nonbind_marker_p): Adjust.
6308 * insn-notes.def (INLINE_ENTRY): New.
6309 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6310 inline entry marker notes.
6311 (print_insn): Likewise.
6312 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6313 (INSN_DEBUG_MARKER_KIND): Likewise.
6314 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6315 * tree-inline.c (expand_call_inline): Build and insert
6316 debug_inline_entry stmt.
6317 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6318 inline entry blocks early, if nonbind markers are enabled.
6319 (dump_scope_block): Dump fragment info.
6320 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6321 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6322 (gimple_build_debug_inline_entry): New.
6323 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6324 Enable/disable inline entry points too.
6325 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6326 (DEBUG_INSN): Describe inline entry markers.
6328 * common.opt (gvariable-location-views): New.
6329 (gvariable-location-views=incompat5): New.
6330 * config.in: Rebuilt.
6331 * configure: Rebuilt.
6332 * configure.ac: Test assembler for view support.
6333 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6334 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6335 * dwarf2out.c (var_loc_view): New typedef.
6336 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6337 (dwarf2out_locviews_in_attribute): New.
6338 (dwarf2out_locviews_in_loclist): New.
6339 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6340 (enum dw_line_info_opcode): Add LI_adv_address.
6341 (struct dw_line_info_table): Add view.
6342 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6343 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6344 (zero_view_p): New variable.
6345 (ZERO_VIEW_P): New macro.
6346 (output_asm_line_debug_info): New.
6347 (struct var_loc_node): Add view.
6348 (add_AT_view_list, AT_loc_list): New.
6349 (add_var_loc_to_decl): Add view param. Test it against last.
6350 (new_loc_list): Add view params. Record them.
6351 (AT_loc_list_ptr): Handle loc and view lists.
6352 (view_list_to_loc_list_val_node): New.
6353 (print_dw_val): Handle dw_val_class_view_list.
6354 (size_of_die): Likewise.
6355 (value_format): Likewise.
6356 (loc_list_has_views): New.
6357 (gen_llsym): Set vl_symbol too.
6358 (maybe_gen_llsym, skip_loc_list_entry): New.
6359 (dwarf2out_maybe_output_loclist_view_pair): New.
6360 (output_loc_list): Output view list or entries too.
6361 (output_view_list_offset): New.
6362 (output_die): Handle dw_val_class_view_list.
6363 (output_dwarf_version): New.
6364 (output_compilation_unit_header): Use it.
6365 (output_skeleton_debug_sections): Likewise.
6366 (output_rnglists, output_line_info): Likewise.
6367 (output_pubnames, output_aranges): Update version comments.
6368 (output_one_line_info_table): Output view numbers in asm comments.
6369 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6370 Call maybe_gen_llsym.
6371 (loc_list_from_tree_1): Adjust.
6372 (add_AT_location_description): Create view list attribute if
6373 needed, check it's absent otherwise.
6374 (convert_cfa_to_fb_loc_list): Adjust.
6375 (maybe_emit_file): Call output_asm_line_debug_info for test.
6376 (dwarf2out_var_location): Reset views as needed. Precompute
6377 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6378 attribute. Set view.
6379 (new_line_info_table): Reset next view.
6380 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6381 (dwarf2out_source_line): Likewise. Output view resets and labels to
6382 the assembler, or select appropriate line info opcodes.
6383 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6384 (optimize_string_length): Catch it. Adjust.
6385 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6386 dw_val_class_view_list, and remove it if no longer needed.
6387 (hash_loc_list): Hash view numbers.
6388 (loc_list_hasher::equal): Compare them.
6389 (optimize_location_lists): Check whether a view list symbol is
6390 needed, and whether the locview attribute is present, and
6391 whether they match. Remove the locview attribute if no longer
6393 (index_location_lists): Call skip_loc_list_entry for test.
6394 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6395 Use output_dwarf_version.
6396 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6397 (struct dw_val_node): Add val_view_list.
6398 * final.c (SEEN_NEXT_VIEW): New.
6399 (set_next_view_needed): New.
6400 (clear_next_view_needed): New.
6401 (maybe_output_next_view): New.
6402 (final_start_function): Rename to...
6403 (final_start_function_1): ... this. Take pointer to FIRST,
6404 add SEEN parameter. Emit param bindings in the initial view.
6405 (final_start_function): Reintroduce SEEN-less interface.
6406 (final): Rename to...
6407 (final_1): ... this. Take SEEN parameter. Output final pending
6408 next view at the end.
6409 (final): Reintroduce seen-less interface.
6410 (final_scan_insn): Output pending next view before switching
6411 sections or ending a block. Mark the next view as needed when
6412 outputting variable locations. Notify debug backend of section
6413 changes, and of location view changes.
6414 (rest_of_handle_final): Adjust.
6415 * toplev.c (process_options): Autodetect value for debug variable
6416 location views option. Warn on incompat5 without -gdwarf-5.
6417 * doc/invoke.texi (gvariable-location-views): New.
6418 (gvariable-location-views=incompat5): New.
6419 (gno-variable-location-views): New.
6421 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6423 PR tree-optimization/84136
6424 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6425 that the result of find_edge is non-NULL.
6427 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6430 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6431 storing integer register in SImode. Fix cost of 256 and 512
6432 byte aligned SSE register store.
6434 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6436 * config/i386/i386.c (ix86_multiplication_cost): Fix
6437 multiplication cost for TARGET_AVX512DQ.
6439 2018-02-08 Marek Polacek <polacek@redhat.com>
6441 PR tree-optimization/84238
6442 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6445 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6447 PR tree-optimization/84265
6448 * tree-vect-stmts.c (vectorizable_store): Don't treat
6449 VMAT_CONTIGUOUS accesses as grouped.
6450 (vectorizable_load): Likewise.
6452 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6454 PR tree-optimization/81635
6455 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6456 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6457 (test_round_for_mask): New functions.
6458 (wide_int_cc_tests): Call test_round_for_mask.
6459 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6460 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6461 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6462 range returned by get_range_info.
6464 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6467 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6468 * symtab.c: Include builtins.h
6469 (symtab_node::output_to_lto_symbol_table_p): Move here
6470 from lto-streamer-out.c:output_symbol_p.
6471 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6472 (output_symbol_p): Move all logic to symtab.c
6473 (produce_symtab): Update.
6475 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6477 * config/s390/s390-opts.h (enum indirect_branch): Define.
6478 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6479 (s390_indirect_branch_via_thunk)
6480 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6481 (enum s390_indirect_branch_type): Define.
6482 * config/s390/s390.c (struct s390_frame_layout, struct
6483 machine_function): Remove.
6484 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6485 (indirect_branch_table_label_no, indirect_branch_table_name):
6487 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6488 (enum s390_indirect_branch_option): Define.
6489 (s390_return_addr_from_memory): New function.
6490 (s390_handle_string_attribute): New function.
6491 (s390_attribute_table): Add new attribute handler.
6492 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6493 (s390_indirect_branch_via_thunk): New function.
6494 (s390_indirect_branch_via_inline_thunk): New function.
6495 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6496 sibling call optimization for non z10 compiles.
6497 (s390_emit_call): Force indirect branch target to be a single
6498 register. Add r1 clobber for non-z10 compiles.
6499 (s390_emit_epilogue): Emit return jump via return_use expander.
6500 (s390_reorg): Handle JUMP_INSNs as execute targets.
6501 (s390_option_override_internal): Perform validity checks for the
6502 new command line options.
6503 (s390_indirect_branch_attrvalue): New function.
6504 (s390_indirect_branch_settings): New function.
6505 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6506 (s390_output_indirect_thunk_function): New function.
6507 (s390_code_end): Implement target hook.
6508 (s390_case_values_threshold): Implement target hook.
6509 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6511 * config/s390/s390.h (struct s390_frame_layout)
6512 (struct machine_function): Move here from s390.c.
6513 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6514 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6515 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6516 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6517 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6518 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6519 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6520 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6521 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6522 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6523 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6524 (mnemonic attribute): Add values which aren't recognized
6526 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6527 pattern for branch conversion. Fix mnemonic attribute.
6528 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6529 indirect branch via thunk if requested.
6530 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6531 ("*indirect_jump"): Disable for branch conversion using out of
6533 ("indirect_jump_via_thunk<mode>_z10")
6534 ("indirect_jump_via_thunk<mode>")
6535 ("indirect_jump_via_inlinethunk<mode>_z10")
6536 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6537 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6538 ("casesi_jump_via_inlinethunk<mode>_z10")
6539 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6540 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6541 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6542 ("*indirect2_jump"): Disable for branch conversion.
6543 ("casesi_jump"): Turn into expander and expand patterns for branch
6545 ("return_use"): New expander.
6546 ("*return"): Emit return via thunk and rename it to ...
6547 ("*return<mode>"): ... this one.
6548 * config/s390/s390.opt: Add new options and and enum for the
6551 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6553 * lra-constraints.c (match_reload): Unconditionally use
6554 gen_lowpart_SUBREG, rather than selecting between that
6555 and equivalent gen_rtx_SUBREG code.
6557 2018-02-08 Richard Biener <rguenther@suse.de>
6559 PR tree-optimization/84233
6560 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6561 changed flag instead of boguously re-using phi_inserted.
6563 2018-02-08 Martin Jambor <mjambor@suse.cz>
6565 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6566 static local variables.
6568 2018-02-08 Richard Biener <rguenther@suse.de>
6570 PR tree-optimization/84278
6571 * tree-vect-stmts.c (vectorizable_store): When looking for
6572 smaller vector types to perform grouped strided loads/stores
6573 make sure the mode is supported by the target.
6574 (vectorizable_load): Likewise.
6576 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6578 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6579 Increase LDP/STP opportunities by adding adjacent callee-saves.
6581 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6583 PR rtl-optimization/84068
6584 PR rtl-optimization/83459
6585 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6587 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6589 PR tree-optimization/84224
6590 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6591 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6594 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6597 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6598 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6599 count, remove one USE.
6601 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6603 * doc/install.texi (Configuration): Document the
6604 --with-long-double-format={ibm,ieee} PowerPC configuration
6608 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6609 Convert from define_expand to be define_insn_and_split. Rework
6610 float/double/_Float128 conversions to QI/HI/SImode to work with
6611 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6612 conversions to QI/HImode types did a store and then a load to
6613 truncate the value. For conversions to VSX registers, don't split
6614 the insn, instead emit the code directly. Use the code iterator
6615 any_fix to combine signed and unsigned conversions.
6616 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6617 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6618 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6619 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6620 (fix_<mode>di2_hw): Likewise.
6621 (fixuns_<mode>di2_hw): Likewise.
6622 (fix_<mode>si2_hw): Likewise.
6623 (fixuns_<mode>si2_hw): Likewise.
6624 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6625 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6626 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6627 fix<uns>_trunc<SFDF:mode>si2_p8.
6628 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6630 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6631 (fix<uns>_<mode>_mem): Likewise.
6632 (fctiw<u>z_<mode>_mem): Likewise.
6633 (fix<uns>_<mode>_mem): Likewise.
6634 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6635 the register allocator from doing a direct move to the GPRs to do
6636 a store, and instead use the ISA 3.0 store byte/half-word from
6637 vector register instruction. For IEEE 128-bit floating point,
6638 also optimize stores of 32-bit ints.
6639 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6641 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6643 * genextract.c (push_pathstr_operand): New function to support
6645 (walk_rtx): Call push_pathstr_operand.
6646 (print_path): Support [a-zA-Z].
6648 2018-02-07 Richard Biener <rguenther@suse.de>
6650 PR tree-optimization/84037
6651 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6652 (cse_and_gimplify_to_preheader): Declare.
6653 (vect_get_place_in_interleaving_chain): Likewise.
6654 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6656 (_loop_vec_info::~_loop_vec_info): Delete it.
6657 (cse_and_gimplify_to_preheader): New function.
6658 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6659 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6660 (vectorizable_load): Likewise. For grouped stores always base
6661 the IV on the first element.
6662 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6663 condition before gimplifying.
6665 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6667 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6668 *DIV_EXPR and *MOD_EXPR.
6670 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6673 * config/i386/i386.c (ix86_option_override_internal): Mask out
6674 the CF_SET bit when checking -fcf-protection.
6676 2018-02-07 Tom de Vries <tom@codesourcery.com>
6679 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6682 2018-02-07 Richard Biener <rguenther@suse.de>
6684 PR tree-optimization/84204
6685 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6688 PR tree-optimization/84205
6689 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6690 special-case isl_ast_op_zdiv_r.
6692 PR tree-optimization/84223
6693 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6694 Only add conditions from within the region.
6695 (gather_bbs::after_dom_children): Adjust.
6697 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6700 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6701 * config/avr/avr.md: Only post-reload split REG-REG moves if
6702 either register is GENERAL_REG_P.
6704 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6706 PR tree-optimization/84235
6707 * tree-ssa-scopedtables.c
6708 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6709 if the subtraction is performed in floating point type where NaNs are
6710 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6711 build 1. Formatting fix.
6713 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6716 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6717 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6718 and skip it regardless of bb boundaries. Use CALL_P macro,
6719 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6720 unnecessarily, formatting fix.
6722 2018-02-06 Michael Collison <michael.collison@arm.com>
6724 * config/arm/thumb2.md:
6725 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6726 (*thumb_mov_notscc): Ditto.
6728 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6731 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6732 unsigned_fix, not "s".
6734 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6736 * configure.ac (gcc_fn_eh_frame_ro): New function.
6737 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6738 correct .eh_frame permissions.
6739 * configure: Regenerate.
6741 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6743 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6746 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6748 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6749 Display warning message for -mno-speculate-indirect-jumps.
6751 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6753 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6755 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6757 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6759 PR tree-optimization/84225
6760 * tree-eh.c (find_trapping_overflow): Only call
6761 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6763 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6766 * config/i386/i386.c: Reimplement the check of possible options
6767 -mibt/-mshstk conbination. Change error messages.
6768 * doc/invoke.texi: Fix a typo: remove extra '='.
6770 2018-02-06 Marek Polacek <polacek@redhat.com>
6772 PR tree-optimization/84228
6773 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6775 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6778 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6779 emitted arch directives.
6780 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6781 __ARM_FEATURE_COPROC before changing architectures.
6783 2018-02-06 Richard Biener <rguenther@suse.de>
6785 * config/i386/i386.c (print_reg): Fix typo.
6786 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6788 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6790 * configure: Regenerate.
6792 2018-02-05 Martin Sebor <msebor@redhat.com>
6794 PR tree-optimization/83369
6795 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6798 2018-02-05 Martin Liska <mliska@suse.cz>
6800 * doc/invoke.texi: Cherry-pick upstream r323995.
6802 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6804 * ira.c (ira_init_register_move_cost): Adjust comment.
6806 2018-02-05 Martin Liska <mliska@suse.cz>
6808 PR gcov-profile/84137
6809 * doc/gcov.texi: Fix typo in documentation.
6811 2018-02-05 Martin Liska <mliska@suse.cz>
6813 PR gcov-profile/83879
6814 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6815 using dlopen functionality.
6817 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6819 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6820 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6821 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6822 _mm_maskz_range_round_ss): New intrinsics.
6823 (__builtin_ia32_rangesd128_round)
6824 (__builtin_ia32_rangess128_round): Remove.
6825 (__builtin_ia32_rangesd128_mask_round,
6826 __builtin_ia32_rangess128_mask_round): New builtins.
6827 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6828 __builtin_ia32_rangess128_round): Remove.
6829 (__builtin_ia32_rangesd128_mask_round,
6830 __builtin_ia32_rangess128_mask_round): New builtins.
6831 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6832 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6833 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6834 "<round_saeonly_constraint>")): Changed to ...
6835 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6836 "<round_saeonly_scalar_constraint>")): ... this.
6837 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6838 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6839 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6840 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6841 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6843 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6845 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6847 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6848 Remove all values except native, 8540 and 8548.
6850 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6852 * config/i386/i386.c (ix86_output_function_return): Pass
6853 INVALID_REGNUM, instead of -1, as invalid register number to
6854 indirect_thunk_name and output_indirect_thunk.
6856 2018-02-02 Julia Koval <julia.koval@intel.com>
6858 * config.gcc: Add -march=icelake.
6859 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6860 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6861 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6862 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6863 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6864 (processor_target_table): Add icelake.
6865 (ix86_option_override_internal): Handle new PTAs.
6866 (get_builtin_code_for_version): Handle icelake.
6867 (M_INTEL_COREI7_ICELAKE): New.
6868 (fold_builtin_cpu): Handle icelake.
6869 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6870 * doc/invoke.texi: Add -march=icelake.
6872 2018-02-02 Julia Koval <julia.koval@intel.com>
6874 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6875 to wide_int_bitmask.
6876 * wide-int-bitmask.h: New.
6878 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6881 * config/i386/i386.md: Replace Pmode with word_mode in
6882 builtin_setjmp_setup and builtin_longjmp to support x32.
6884 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6888 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6890 (rs6000_supported_cpu_names): New static variable.
6891 (linux_cpu_translation_table): Likewise.
6892 (elf_platform) <cpu>: Define new static variable and use it.
6893 Translate kernel AT_PLATFORM name to canonical name if needed.
6894 Error if platform name is unknown.
6896 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6899 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6901 2018-02-01 Jeff Law <law@redhat.com>
6904 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6905 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6906 the scratch if RELEASE_VIA_POP is false.
6907 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6908 If we have to save a temporary register, decrement SIZE appropriately.
6909 Pass new arguments to release_scratch_register_on_entry.
6910 (ix86_adjust_stack_and_probe): Likewise.
6911 (ix86_emit_probe_stack_range): Pass new arguments to
6912 release_scratch_register_on_entry.
6914 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6916 PR rtl-optimization/84157
6917 * combine.c (change_zero_ext): Use REG_P predicate in
6918 front of HARD_REGISTER_P predicate.
6920 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6922 * config/avr/avr.c (avr_option_override): Move disabling of
6923 -fdelete-null-pointer-checks to...
6924 * common/config/avr/avr-common.c (avr_option_optimization_table):
6927 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6929 PR tree-optimization/81635
6930 * tree-data-ref.c (split_constant_offset_1): For types that
6931 wrap on overflow, try to use range info to prove that wrapping
6934 2018-02-01 Renlin Li <renlin.li@arm.com>
6937 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6939 (aarch64_register_move_cost): Likewise.
6940 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6942 (REG_CLASS_NAMES): Likewise.
6943 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6944 TAILCALL_ADDR_REGS. Remove IP registers.
6945 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6947 2018-02-01 Richard Biener <rguenther@suse.de>
6949 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6950 for specifying RPO order and allow NULL for that.
6951 * domwalk.c (dom_walker::dom_walker): Likewise.
6952 (dom_walker::walk): Handle NULL RPO order.
6953 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6955 (rewrite_update_dom_walker): Likewise.
6956 (mark_def_dom_walker): Likewise.
6958 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6960 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6961 (aarch64_maybe_expand_sve_subreg_move): Declare.
6962 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6963 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6965 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6966 that are semantically a reverse operation.
6967 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6968 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6969 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6971 (aarch64_can_change_mode_class): For big-endian, forbid changes
6972 between two SVE modes if they have different element sizes.
6974 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6976 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6977 the TImode handling for big-endian targets.
6979 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6981 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6982 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6984 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6985 Remove BSWAP handing for big-endian targets and use the form of
6986 LD1RQ appropariate for the mode.
6988 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6990 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6991 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6994 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6997 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6998 check for operands that need to go through aarch64_sve_reload_be.
7000 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7002 PR tree-optimization/81661
7003 PR tree-optimization/84117
7004 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7005 * tree-eh.c: Include gimplify.h.
7006 (find_trapping_overflow, replace_trapping_overflow,
7007 rewrite_to_non_trapping_overflow): New functions.
7008 * tree-vect-loop.c: Include tree-eh.h.
7009 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7010 * tree-data-ref.c: Include tree-eh.h.
7011 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7013 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7015 PR rtl-optimization/84123
7016 * combine.c (change_zero_ext): Check if hard register satisfies
7017 can_change_dest_mode before calling gen_lowpart_SUBREG.
7019 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7022 * ira.c (ira_init_register_move_cost): Remove assert.
7024 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7026 PR rtl-optimization/84071
7027 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7028 * doc/tm.texi: Regenerate.
7030 2018-01-31 Richard Biener <rguenther@suse.de>
7032 PR tree-optimization/84132
7033 * tree-data-ref.c (analyze_miv_subscript): Properly
7034 check whether evolution_function_is_affine_multivariate_p
7035 before calling gcd_of_steps_may_divide_p.
7037 2018-01-31 Julia Koval <julia.koval@intel.com>
7040 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7041 * config/i386/i386.md (rdpid_rex64) New.
7042 (rdpid): Make 32bit only.
7044 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7047 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7048 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7050 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7053 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7055 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7057 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7059 PR rtl-optimization/84071
7060 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7061 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7063 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7065 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7066 (arc_attribute_table): Add 'aux' attribute.
7067 (arc_in_small_data_p): Consider aux like variables.
7068 (arc_is_aux_reg_p): New function.
7069 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7070 (arc_get_aux_arg): New function.
7071 (prepare_move_operands): Handle aux-register access.
7072 (arc_handle_aux_attribute): New function.
7073 * doc/extend.texi (ARC Variable attributes): Add subsection.
7075 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7077 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7078 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7079 (arc_attribute_table): Add 'uncached' attribute.
7080 (arc_print_operand): Print '.di' flag for uncached memory
7082 (arc_in_small_data_p): Do not consider for small data the uncached
7084 (arc_is_uncached_mem_p): New function.
7085 * config/arc/predicates.md (compact_store_memory_operand): Check
7086 for uncached memory accesses.
7087 (nonvol_nonimm_operand): Likewise.
7088 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7090 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7093 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7094 falign-loops=): Add Optimization flag.
7096 2018-01-30 Jeff Law <law@redhat.com>
7099 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7100 INT_REGISTERS_SAVED. Check it prior to calling
7101 get_scratch_register_on_entry.
7102 (ix86_adjust_stack_and_probe): Similarly.
7103 (ix86_emit_probe_stack_range): Similarly.
7104 (ix86_expand_prologue): Corresponding changes.
7106 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7109 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7110 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7112 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7115 * lra-constraints.c (curr_insn_transform): Process AND in the
7118 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7120 PR rtl-optimization/83986
7121 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7122 dependence against last_pending_memory_flush in addition to
7125 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7127 PR tree-optimization/81611
7128 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7131 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7134 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7137 2018-01-30 Richard Biener <rguenther@suse.de>
7138 Jakub Jelinek <jakub@redhat.com>
7140 PR tree-optimization/84111
7141 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7142 inner loops added during recursion, as they don't have up-to-date
7145 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7148 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7149 (can_inline_edge_by_limits_p): ... here.
7150 (can_early_inline_edge_p, check_callers,
7151 update_caller_keys, update_callee_keys, recursive_inlining,
7152 add_new_edges_to_heap, speculation_useful_p,
7153 inline_small_functions,
7154 inline_small_functions, flatten_function,
7155 inline_to_all_callers_1): Update.
7157 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7159 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7162 2018-01-30 Richard Biener <rguenther@suse.de>
7164 PR tree-optimization/83008
7165 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7166 invariant and constant vector uses in stmts when they need
7169 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7172 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7173 * configure: Regenerate.
7175 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7177 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7179 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7180 Use gen_rtx_REG rather than gen_lowpart.
7182 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7184 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7185 rather than 0 when creating partial subregs.
7187 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7189 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7192 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7195 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7196 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7197 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7198 flags. This restores the settings used before the 2017-07-24.
7199 Turning off pre increment/decrement/modify allows IVOPTS to
7200 optimize DF/SF loops where the index is an int.
7202 2018-01-29 Richard Biener <rguenther@suse.de>
7203 Kelvin Nilsen <kelvin@gcc.gnu.org>
7206 * tree-vect-stmts.c (vectorizable_call): Don't call
7207 targetm.vectorize_builtin_md_vectorized_function if callee is
7210 2018-01-22 Carl Love <cel@us.ibm.com>
7212 * doc/extend.tex: Fix typo in second arg in
7213 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7215 2018-01-29 Richard Biener <rguenther@suse.de>
7217 PR tree-optimization/84086
7218 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7219 (flush_ssaname_freelist): When SSA names were released reset
7220 the SCEV hash table.
7222 2018-01-29 Richard Biener <rguenther@suse.de>
7224 PR tree-optimization/84057
7225 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7226 removed paths when removing edges.
7228 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7230 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7231 -mfunction-return=@var{choice}.
7233 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7236 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7238 (layout::print_source_line): Likewise.
7239 (test_get_line_width_without_trailing_whitespace): Add test cases.
7241 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7244 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7247 2018-01-26 Jim Wilson <jimw@sifive.com>
7249 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7251 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7254 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7256 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7257 and CMP + SUB-immediate -> SUBS.
7259 2018-01-26 Martin Sebor <msebor@redhat.com>
7261 PR tree-optimization/83896
7262 * tree-ssa-strlen.c (get_string_len): Rename...
7263 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7264 Avoid assuming length is constant.
7265 (handle_char_store): Use HOST_WIDE_INT for string length.
7267 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7270 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7271 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7273 2018-01-26 Richard Biener <rguenther@suse.de>
7275 PR rtl-optimization/84003
7276 * dse.c (record_store): Only record redundant stores when
7277 the earlier store aliases at least all accesses the later one does.
7279 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7281 PR rtl-optimization/83985
7282 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7283 REG_CFA_RESTORE insns.
7284 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7285 REG_CFA_RESTORE insns here.
7288 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7289 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7291 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7293 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7294 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7295 (arc_init): Likewise.
7296 (arc_override_options): Likewise.
7297 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7299 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7301 * config/arc/arc.h (TARGET_DBNZ): Define.
7302 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7303 properly set the tune attribute.
7304 (dbnz): Use TARGET_DBNZ guard.
7305 * config/arc/arc.opt (mtune): Add core3 option.
7307 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7309 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7310 recognize new pic like addresses.
7311 (arc_delegitimize_address): Clean up.
7313 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7315 * config/arc/arc-arches.def: Option mrf16 valid for all
7317 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7318 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7319 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7320 * config/arc/arc-tables.opt: Regenerate.
7321 * config/arc/arc.c (arc_conditional_register_usage): Handle
7322 reduced register file case.
7323 (arc_file_start): Set must have build attributes.
7324 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7326 * config/arc/arc.opt (mrf16): Add new option.
7327 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7328 * config/arc/genmultilib.awk: Handle new mrf16 option.
7329 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7330 * config/arc/t-multilib: Regenerate.
7331 * doc/invoke.texi (ARC Options): Document mrf16 option.
7333 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7335 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7336 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7337 (arc_attribute_table): Add 'secure_call' attribute.
7338 (arc_print_operand): Print secure call operand.
7339 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7341 (arc_is_secure_call_p): New function. * config/arc/arc.md
7342 (call_i): Add support for sjli instruction.
7343 (call_value_i): Likewise.
7344 * config/arc/constraints.md (Csc): New constraint.
7346 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7347 John Eric Martin <John.Martin@emmicro-us.com>
7349 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7350 * config/arc/arc.c (_arc_jli_section): New struct.
7351 (arc_jli_section): New type.
7352 (rc_jli_sections): New static variable.
7353 (arc_handle_jli_attribute): New function.
7354 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7355 (arc_file_end): New function.
7356 (TARGET_ASM_FILE_END): Define.
7357 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7358 (arc_add_jli_section): New function.
7359 (jli_call_scan): Likewise.
7360 (arc_reorg): Call jli_call_scan.
7361 (arc_output_addsi): Remove 'S' from printing asm operand.
7362 (arc_is_jli_call_p): New function.
7363 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7365 (movhi_insn): Likewise.
7366 (movsi_insn): Likewise.
7367 (movsi_set_cc_insn): Likewise.
7368 (loadqi_update): Likewise.
7369 (load_zeroextendqisi_update): Likewise.
7370 (load_signextendqisi_update): Likewise.
7371 (loadhi_update): Likewise.
7372 (load_zeroextendhisi_update): Likewise.
7373 (load_signextendhisi_update): Likewise.
7374 (loadsi_update): Likewise.
7375 (loadsf_update): Likewise.
7376 (movsicc_insn): Likewise.
7377 (bset_insn): Likewise.
7378 (bxor_insn): Likewise.
7379 (bclr_insn): Likewise.
7380 (bmsk_insn): Likewise.
7381 (bicsi3_insn): Likewise.
7382 (cmpsi_cc_c_insn): Likewise.
7383 (movsi_ne): Likewise.
7384 (movsi_cond_exec): Likewise.
7385 (clrsbsi2): Likewise.
7396 (call_i): Remove 'S' asm letter, add jli instruction.
7397 (call_value_i): Likewise.
7398 * config/arc/arc.op (mjli-always): New option.
7399 * config/arc/constraints.md (Cji): New constraint.
7400 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7402 (subsf3_fpx): Likewise.
7403 (mulsf3_fpx): Likewise.
7404 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7406 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7408 * doc/invoke.texi (ARC): Document mjli-always option.
7410 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7412 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7413 avoid addition with 0 and use incw and decw where possible.
7415 2018-01-26 Richard Biener <rguenther@suse.de>
7417 PR tree-optimization/81082
7418 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7419 association if it requires casting to unsigned.
7420 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7421 from fold_plusminus_mult_expr to catch important cases late when
7422 range info is available.
7424 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7426 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7427 * configure.ac (hidden_linkonce): New test.
7428 * configure: Regenerate.
7429 * config.in: Regenerate.
7431 2018-01-26 Julia Koval <julia.koval@intel.com>
7433 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7434 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7435 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7436 _mm_mask_bitshuffle_epi64_mask): Fix type.
7437 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7438 USI_FTYPE_V4DI_V4DI_USI): Remove.
7439 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7440 __builtin_ia32_vpshufbitqmb256_mask,
7441 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7442 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7443 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7445 2018-01-26 Alan Modra <amodra@gmail.com>
7448 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7449 UNSPEC_VBPERMQ. Sort other unspecs.
7451 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7453 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7455 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7458 * predict.c (drop_profile): Do not push/pop cfun; update also
7460 (handle_missing_profiles): Fix logic looking for zero profiles.
7462 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7465 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7466 on functions with #pragma omp declare simd or functions with simd
7468 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7469 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7470 Remove trailing \n from warning_at calls.
7472 2018-01-25 Tom de Vries <tom@codesourcery.com>
7475 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7476 for neutered workers.
7478 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7481 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7482 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7484 2018-01-24 Jeff Law <law@redhat.com>
7487 * i386.c (get_probe_interval): Move to earlier point.
7488 (ix86_compute_frame_layout): If -fstack-clash-protection and
7489 the frame is larger than the probe interval, then use pushes
7490 to save registers rather than reg->mem moves.
7491 (ix86_expand_prologue): Remove conditional for int_registers_saved
7494 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7497 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7498 min/max for never referenced object.
7500 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7503 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7505 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7506 attributes from DECL_ATTRIBUTES (decl) without affecting
7507 DECL_ATTRIBUTES (current_function_decl).
7508 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7509 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7511 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7513 PR tree-optimization/83979
7514 * fold-const.c (fold_comparison): Use constant_boolean_node
7515 instead of boolean_{true,false}_node.
7517 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7519 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7522 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7524 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7525 Simplify the clause that sets the length attribute.
7526 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7527 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7528 clause that sets the length attribute.
7529 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7531 2018-01-24 Tom de Vries <tom@codesourcery.com>
7534 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7535 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7536 Add strict parameter.
7537 (prevent_branch_around_nothing): Insert dummy insn between branch to
7538 label and label with no ptx insn inbetween.
7539 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7541 2018-01-24 Tom de Vries <tom@codesourcery.com>
7544 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7545 for neutered threads in warp.
7546 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7548 2018-01-24 Richard Biener <rguenther@suse.de>
7550 PR tree-optimization/83176
7551 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7554 2018-01-24 Richard Biener <rguenther@suse.de>
7556 PR tree-optimization/82819
7557 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7558 code generating pluses that are no-ops in the target precision.
7560 2018-01-24 Richard Biener <rguenther@suse.de>
7563 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7565 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7567 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7568 to merge probabilities.
7569 * predict.c (probably_never_executed): Also mark as cold functions
7570 with global 0 profile and guessed local profile.
7571 * profile-count.c (profile_probability::combine_with_count): New
7573 * profile-count.h (profile_probability::operator*,
7574 profile_probability::operator*=, profile_probability::operator/,
7575 profile_probability::operator/=): Reduce precision to adjusted
7576 and set value to guessed on contradictory divisions.
7577 (profile_probability::combine_with_freq): Remove.
7578 (profile_probability::combine_wiht_count): Declare.
7579 (profile_count::force_nonzero):: Set to adjusted.
7580 (profile_count::probability_in):: Set quality to adjusted.
7581 * tree-ssa-tail-merge.c (replace_block_by): Use
7584 2018-01-23 Andrew Waterman <andrew@sifive.com>
7585 Jim Wilson <jimw@sifive.com>
7587 * config/riscv/riscv.c (riscv_stack_boundary): New.
7588 (riscv_option_override): Set riscv_stack_boundary. Handle
7589 riscv_preferred_stack_boundary_arg.
7590 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7591 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7592 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7593 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7594 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7595 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7597 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7600 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7601 of struct ix86_frame.
7602 (ix86_expand_epilogue): Likewise. Add a local variable for
7603 the reg_save_offset field in struct ix86_frame.
7605 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7607 PR tree-optimization/82604
7608 * tree-loop-distribution.c (enum partition_kind): New enum item
7609 PKIND_PARTIAL_MEMSET.
7610 (partition_builtin_p): Support above new enum item.
7611 (generate_code_for_partition): Ditto.
7612 (compute_access_range): Differentiate cases that equality can be
7613 proven at all loops, the innermost loops or no loops.
7614 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7615 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7616 (finalize_partitions, distribute_loop): Don't fuse partition of
7617 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7618 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7621 2018-01-23 Martin Liska <mliska@suse.cz>
7623 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7624 order to ignore the predictor.
7625 (PRED_POLYMORPHIC_CALL): Likewise.
7626 (PRED_RECURSIVE_CALL): Likewise.
7628 2018-01-23 Martin Liska <mliska@suse.cz>
7630 * tree-profile.c (tree_profiling): Print function header to
7631 aware reader which function we are working on.
7632 * value-prof.c (gimple_find_values_to_profile): Do not print
7633 not interesting value histograms.
7635 2018-01-23 Martin Liska <mliska@suse.cz>
7637 * profile-count.h (enum profile_quality): Add
7638 profile_uninitialized as the first value. Do not number values
7639 as they are zero based.
7640 (profile_count::verify): Update sanity check.
7641 (profile_probability::verify): Likewise.
7643 2018-01-23 Nathan Sidwell <nathan@acm.org>
7645 * doc/invoke.texi (ffor-scope): Deprecate.
7647 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7649 PR tree-optimization/83510
7650 * domwalk.c (set_all_edges_as_executable): New function.
7651 (dom_walker::dom_walker): Convert bool param
7652 "skip_unreachable_blocks" to enum reachability. Move setup of
7653 edge flags to set_all_edges_as_executable and only do it when
7654 reachability is REACHABLE_BLOCKS.
7655 * domwalk.h (enum dom_walker::reachability): New enum.
7656 (dom_walker::dom_walker): Convert bool param
7657 "skip_unreachable_blocks" to enum reachability.
7658 (set_all_edges_as_executable): New decl.
7659 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7660 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7662 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7663 but converting true to REACHABLE_BLOCKS.
7664 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7666 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7667 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7668 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7670 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7671 if check_all_array_refs will be called.
7673 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7675 * tree.c (selftest::test_location_wrappers): Add more test
7678 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7680 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7681 (selftest::test_bit_in_range): Likewise.
7683 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7686 * doc/sourcebuild.texi (vect_float): Say that the selector
7687 only describes the situation when -funsafe-math-optimizations is on.
7688 (vect_float_strict): Document.
7690 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7692 PR tree-optimization/83965
7693 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7694 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7695 instead of checking only for a reduction.
7696 (vect_recog_widen_sum_pattern): Likewise.
7698 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7700 * predict.c (probably_never_executed): Only use precise profile info.
7701 (compute_function_frequency): Skip after inlining hack since we now
7702 have quality checking.
7704 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7706 * profile-count.h (profile_probability::very_unlikely,
7707 profile_probability::unlikely, profile_probability::even): Set
7708 precision to guessed.
7710 2018-01-23 Richard Biener <rguenther@suse.de>
7712 PR tree-optimization/83963
7713 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7714 Properly terminate dominator walk when crossing the exit edge not
7715 when visiting its source block.
7717 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7720 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7721 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7723 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7725 PR tree-optimization/83957
7726 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7727 semicolon after for body surrounded by braces.
7729 PR tree-optimization/83081
7730 * profile-count.h (profile_probability::split): New method.
7731 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7732 Use profile_probability::split.
7733 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7734 when splitting a single conditional jump into 2.
7736 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7738 PR tree-optimization/69452
7739 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7742 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7744 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7745 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7746 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7748 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7750 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7751 * config/rl78/rl78.md (movdi): New define_expand.
7752 * config/rl78/rl78.c (rl78_split_movdi): New function.
7754 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7757 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7759 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7760 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7761 128-bit to produce an UNSPEC move to get the double word with the
7762 signbit and then a shift directly to do signbit.
7763 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7764 implementation with a new version that just does either a direct
7765 move or a regular move. Move memory interface to separate insns.
7766 Move insns so they are next to the expander.
7767 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7768 with signbit move. Split big and little endian case.
7769 (signbit<mode>2_dm_mem_le): Likewise.
7770 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7771 (signbit<mode>2_dm2): Likewise.
7773 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7775 * config/rl78/rl78.md (anddi3): New define_expand.
7777 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7779 * config/rl78/rl78.md (umindi3): New define_expand.
7781 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7783 * config/rl78/rl78.md (smindi3): New define_expand.
7785 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7787 * config/rl78/rl78.md (smaxdi3): New define_expand.
7789 2018-01-22 Carl Love <cel@us.ibm.com>
7791 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7792 LVX_V1TI): Add macro expansion.
7793 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7794 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7795 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7796 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7797 Change check to determine if the instruction is a byte reversing
7798 entry. Fix typo in comment.
7799 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7800 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7801 Add def_builtin calls for new builtins.
7802 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7803 Add define_insn expansion.
7805 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7807 * config/rl78/rl78.md (umaxdi3): New define_expand.
7809 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7811 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7812 for non-QImode registers.
7814 2018-01-22 Richard Biener <rguenther@suse.de>
7816 PR tree-optimization/83963
7817 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7818 including the loop exit block.
7819 (scop_detection::merge_sese): Likewise.
7820 (scop_detection::add_scop): Do it here instead.
7822 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7824 * doc/sourcebuild.texi (arm_softfloat): Document.
7826 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7829 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7830 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7831 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7833 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7834 David Edelsohn <dje.gcc@gmail.com>
7837 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7838 Change "crset eq" to "crset 2".
7839 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7840 (*call_indirect_aix<mode>_nospec): Likewise.
7841 (*call_value_indirect_aix<mode>_nospec): Likewise.
7842 (*call_indirect_elfv2<mode>_nospec): Likewise.
7843 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7844 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7845 change assembly output from . to $.
7846 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7847 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7848 (*tablejump<mode>_internal1_nospec): Likewise.
7850 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7853 * config/sh/sh_optimize_sett_clrt.cc:
7854 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7856 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7858 PR tree-optimization/83940
7859 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7860 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7861 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7862 use it to pass back the definition type.
7863 (vect_check_store_rhs): Likewise rhs_dt_out.
7864 (vect_build_gather_load_calls): Add a mask_dt argument and use
7865 it instead of a call to vect_is_simple_use.
7866 (vectorizable_store): Update calls to vect_check_load_store_mask
7867 and vect_check_store_rhs. Use the dt returned by the latter instead
7868 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7869 instead of calls to vect_is_simple_use. Pass the scalar rather
7870 than the vector operand to vect_is_simple_use when handling
7871 second and subsequent copies of an rhs value.
7872 (vectorizable_load): Update calls to vect_check_load_store_mask
7873 and vect_build_gather_load_calls. Use the cached mask_dt and
7874 gs_info.offset_dt instead of calls to vect_is_simple_use.
7876 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7879 * tree-emutls.c: Include gimplify.h.
7880 (lower_emutls_2): New function.
7881 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7882 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7883 it before further processing.
7886 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7887 UINTVAL (trueop1) instead of INTVAL (op1).
7889 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7893 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7894 INCOMING_FRAME_SP_OFFSET if not defined.
7895 (scan_trace): Add ENTRY argument. If true and
7896 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7897 emit a note to adjust the CFA offset.
7898 (create_cfi_notes): Adjust scan_trace callers.
7899 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7900 INCOMING_FRAME_SP_OFFSET in the CIE.
7901 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7902 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7904 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7905 * doc/tm.texi: Regenerated.
7907 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7909 PR rtl-optimization/83147
7910 * lra-constraints.c (remove_inheritance_pseudos): Use
7911 lra_substitute_pseudo_within_insn.
7913 2018-01-19 Tom de Vries <tom@codesourcery.com>
7914 Cesar Philippidis <cesar@codesourcery.com>
7917 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7919 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7922 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7923 spaces for function labels.
7925 2018-01-19 Martin Liska <mliska@suse.cz>
7927 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7928 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7929 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7930 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7931 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7932 (PRED_CONST_RETURN): Change from 69 to 65.
7933 (PRED_NULL_RETURN): Change from 91 to 71.
7934 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7935 (PRED_LOOP_GUARD): Change from 66 to 73.
7937 2018-01-19 Martin Liska <mliska@suse.cz>
7939 * predict.c (predict_insn_def): Add new assert.
7940 (struct branch_predictor): Change type to signed integer.
7941 (test_prediction_value_range): Amend test to cover
7943 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7944 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7945 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7946 (PRED_LOOP_IV_COMPARE): Likewise.
7947 * predict.h (PROB_UNINITIALIZED): Define new constant.
7949 2018-01-19 Martin Liska <mliska@suse.cz>
7951 * predict.c (dump_prediction): Add new format for
7952 analyze_brprob.py script which is enabled with -details
7954 * profile-count.h (precise_p): New function.
7956 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7958 PR tree-optimization/83922
7959 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7960 there are no statements that need masking.
7961 (vect_active_double_reduction_p): New function.
7962 (vect_analyze_loop_operations): Use it when handling phis that
7963 are not in the loop header.
7965 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7967 PR tree-optimization/83914
7968 * tree-vect-loop.c (vectorizable_induction): Don't convert
7969 init_expr or apply the peeling adjustment for inductions
7970 that are nested within the vectorized loop.
7972 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7974 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7977 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7981 * function.h (gimplify_parameters): Add gimple_seq * argument.
7982 * function.c: Include gimple.h and options.h.
7983 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7984 for the added local temporaries if needed.
7985 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7986 if there are any parameter cleanups, wrap whole body into a
7987 try/finally with the cleanups.
7989 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7992 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7993 Use GET_MODE_CLASS for scalar floating point.
7995 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7999 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8000 Fix call of call_cgraph_insertion_hooks.
8002 2018-01-18 Martin Sebor <msebor@redhat.com>
8004 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8006 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8009 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8012 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8015 * common.opt: (-ffile-prefix-map): New option.
8016 * opts.c (common_handle_option): Defer it.
8017 * opts-global.c (handle_common_deferred_options): Handle it.
8018 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8019 * file-prefix-map.h: New file.
8020 (remap_debug_filename, add_debug_prefix_map): ...here.
8021 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8022 * final.c (debug_prefix_map, add_debug_prefix_map
8023 remap_debug_filename): Move to...
8024 * file-prefix-map.c: New file.
8025 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8026 generalize, get rid of alloca(), use strrchr() instead of strchr().
8027 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8028 Implement in terms of add_prefix_map().
8029 (remap_macro_filename, remap_debug_filename): Implement in term of
8031 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8032 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8033 * dbxout.c: Include file-prefix-map.h.
8034 * varasm.c: Likewise.
8035 * vmsdbgout.c: Likewise.
8036 * xcoffout.c: Likewise.
8037 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8038 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8039 * doc/invoke.texi (-ffile-prefix-map): Document.
8040 (-fdebug-prefix-map): Update description.
8042 2018-01-18 Martin Liska <mliska@suse.cz>
8044 * config/i386/i386.c (indirect_thunk_name): Document that also
8046 (output_indirect_thunk): Document why both instructions
8047 (pause and lfence) are generated.
8049 2018-01-18 Richard Biener <rguenther@suse.de>
8051 PR tree-optimization/83887
8052 * graphite-scop-detection.c
8053 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8054 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8055 (scop_detection::merge_sese): Re-implement with a flood-fill
8056 algorithm that properly finds a SESE region if it exists.
8058 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8061 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8062 pointer_diff optimizations use view_convert instead of convert.
8064 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8066 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8067 Generate different code for -mno-speculate-indirect-jumps.
8068 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8069 (*call_indirect_aix<mode>): Disable for
8070 -mno-speculate-indirect-jumps.
8071 (*call_indirect_aix<mode>_nospec): New define_insn.
8072 (*call_value_indirect_aix<mode>): Disable for
8073 -mno-speculate-indirect-jumps.
8074 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8075 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8076 -mno-speculate-indirect-jumps.
8077 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8079 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8081 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8082 long double type, set the flags for noting the default long double
8083 type, even if we don't pass or return a long double type.
8085 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8088 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8091 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8093 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8094 support for merge[hl].
8095 (fold_mergehl_helper): New helper function.
8096 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8097 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8098 (altivec_vmrglw_direct): Add xxmrglw insn.
8100 2018-01-17 Andrew Waterman <andrew@sifive.com>
8102 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8103 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8105 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8108 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8109 call the lto_location_cache before reading the
8110 DECL_SOURCE_LOCATION of the types.
8112 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8113 Richard Sandiford <richard.sandiford@linaro.org>
8115 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8116 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8117 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8118 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8119 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8121 * config/aarch64/constraints.md (aarch64_movti_operand):
8123 * config/aarch64/predicates.md (Uti): Add new constraint.
8125 2018-01-17 Carl Love <cel@us.ibm.com>
8127 * config/rs6000/vsx.md (define_expand xl_len_r,
8128 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8129 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8131 (define_expand, define_insn): Move the shift left from the
8132 define_insn to the define_expand for lxvl and stxvl instructions.
8133 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8134 and XL_LEN_R definitions to PURE.
8136 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8138 * config/i386/i386.c (indirect_thunk_name): Declare regno
8139 as unsigned int. Compare regno with INVALID_REGNUM.
8140 (output_indirect_thunk): Ditto.
8141 (output_indirect_thunk_function): Ditto.
8142 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8143 in the call to output_indirect_thunk_function.
8145 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8148 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8149 rather than the size of inner_type to determine the stack slot size
8150 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8152 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8155 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8158 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8160 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8161 endian Linux systems to optionally enable multilibs for selecting
8162 the long double type if the user configured an explicit type.
8163 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8164 have no long double multilibs if not defined.
8165 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8166 warn if the user used -mabi={ieee,ibm}longdouble and we built
8167 multilibs for long double.
8168 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8169 appropriate multilib option.
8170 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8172 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8173 for building long double multilibs.
8174 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8176 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8178 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8181 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8183 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8186 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8189 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8192 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8194 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8195 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8197 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8199 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8200 different rtl trees depending on TARGET_64BIT.
8201 (rs6000_gen_lvx): Likewise.
8203 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8205 * config/visium/visium.md (nop): Tweak comment.
8206 (hazard_nop): Likewise.
8208 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8210 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8211 -mspeculate-indirect-jumps.
8212 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8213 for -mno-speculate-indirect-jumps.
8214 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8215 (*call_value_indirect_elfv2<mode>): Disable for
8216 -mno-speculate-indirect-jumps.
8217 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8218 (indirect_jump): Emit different RTL for
8219 -mno-speculate-indirect-jumps.
8220 (*indirect_jump<mode>): Disable for
8221 -mno-speculate-indirect-jumps.
8222 (*indirect_jump<mode>_nospec): New define_insn.
8223 (tablejump): Emit different RTL for
8224 -mno-speculate-indirect-jumps.
8225 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8226 (tablejumpsi_nospec): New define_expand.
8227 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8228 (tablejumpdi_nospec): New define_expand.
8229 (*tablejump<mode>_internal1): Disable for
8230 -mno-speculate-indirect-jumps.
8231 (*tablejump<mode>_internal1_nospec): New define_insn.
8232 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8235 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8237 * caller-save.c (insert_save): Drop unnecessary parameter. All
8240 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8241 Richard Biener <rguenth@suse.de>
8244 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8245 return early, inline manually is_gimple_sizepos. Make sure if we
8246 call gimplify_expr we don't end up with a gimple constant.
8247 * tree.c (variably_modified_type_p): Don't return true for
8248 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8249 * gimplify.h (is_gimple_sizepos): Remove.
8251 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8253 PR tree-optimization/83857
8254 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8255 vectorizable_live_operation for pure SLP statements.
8256 (vectorizable_live_operation): Handle PHIs.
8258 2018-01-16 Richard Biener <rguenther@suse.de>
8260 PR tree-optimization/83867
8261 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8262 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8264 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8267 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8268 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8269 If off is not INTEGER_CST, issue a may not be aligned warning
8270 rather than isn't aligned. Use isn%'t rather than isn't.
8271 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8273 <case MULT_EXPR>: Improve the case when bottom and one of the
8274 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8275 operand, in that case check if the other operand is multiple of
8276 bottom divided by the INTEGER_CST operand.
8278 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8281 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8282 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8283 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8284 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8285 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8286 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8287 (pa_function_arg_size): New function.
8289 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8291 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8292 in a separate statement.
8294 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8296 PR tree-optimization/83847
8297 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8298 group gathers and scatters.
8300 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8302 PR rtl-optimization/86620
8303 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8305 PR rtl-optimization/83213
8306 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8307 to last if both are JUMP_INSNs.
8309 PR tree-optimization/83843
8310 * gimple-ssa-store-merging.c
8311 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8312 store_immediate_info for bswap/nop orig_stores.
8314 2018-01-15 Andrew Waterman <andrew@sifive.com>
8316 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8318 <UDIV>: Increase cost if !TARGET_DIV.
8320 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8322 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8323 (define_attr "cr_logical_3op"): New.
8324 (cceq_ior_compare): Adjust.
8325 (cceq_ior_compare_complement): Adjust.
8326 (*cceq_rev_compare): Adjust.
8327 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8328 (is_cracked_insn): Adjust.
8329 (insn_must_be_first_in_group): Adjust.
8330 * config/rs6000/40x.md: Adjust.
8331 * config/rs6000/440.md: Adjust.
8332 * config/rs6000/476.md: Adjust.
8333 * config/rs6000/601.md: Adjust.
8334 * config/rs6000/603.md: Adjust.
8335 * config/rs6000/6xx.md: Adjust.
8336 * config/rs6000/7450.md: Adjust.
8337 * config/rs6000/7xx.md: Adjust.
8338 * config/rs6000/8540.md: Adjust.
8339 * config/rs6000/cell.md: Adjust.
8340 * config/rs6000/e300c2c3.md: Adjust.
8341 * config/rs6000/e500mc.md: Adjust.
8342 * config/rs6000/e500mc64.md: Adjust.
8343 * config/rs6000/e5500.md: Adjust.
8344 * config/rs6000/e6500.md: Adjust.
8345 * config/rs6000/mpc.md: Adjust.
8346 * config/rs6000/power4.md: Adjust.
8347 * config/rs6000/power5.md: Adjust.
8348 * config/rs6000/power6.md: Adjust.
8349 * config/rs6000/power7.md: Adjust.
8350 * config/rs6000/power8.md: Adjust.
8351 * config/rs6000/power9.md: Adjust.
8352 * config/rs6000/rs64.md: Adjust.
8353 * config/rs6000/titan.md: Adjust.
8355 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8357 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8358 ix86_indirect_branch_register logic.
8360 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8362 * config/i386/constraints.md (Bs): Update
8363 ix86_indirect_branch_register check. Don't check
8364 ix86_indirect_branch_register with GOT_memory_operand.
8366 * config/i386/predicates.md (GOT_memory_operand): Don't check
8367 ix86_indirect_branch_register here.
8368 (GOT32_symbol_operand): Likewise.
8370 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8372 * config/i386/predicates.md (constant_call_address_operand):
8373 Rewrite ix86_indirect_branch_register logic.
8374 (sibcall_insn_operand): Likewise.
8376 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8378 * config/i386/constraints.md (Bs): Replace
8379 ix86_indirect_branch_thunk_register with
8380 ix86_indirect_branch_register.
8382 * config/i386/i386.md (indirect_jump): Likewise.
8383 (tablejump): Likewise.
8384 (*sibcall_memory): Likewise.
8385 (*sibcall_value_memory): Likewise.
8386 Peepholes of indirect call and jump via memory: Likewise.
8387 * config/i386/i386.opt: Likewise.
8388 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8389 (GOT_memory_operand): Likewise.
8390 (call_insn_operand): Likewise.
8391 (sibcall_insn_operand): Likewise.
8392 (GOT32_symbol_operand): Likewise.
8394 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8397 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8398 type rather than type addr's type points to.
8399 (expand_omp_atomic_mutex): Likewise.
8400 (expand_omp_atomic): Likewise.
8402 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8405 * config/i386/i386.c (output_indirect_thunk_function): Use
8406 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8407 for __x86_return_thunk.
8409 2018-01-15 Richard Biener <rguenther@suse.de>
8412 * expmed.c (extract_bit_field_1): Fix typo.
8414 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8417 * config/arm/iterators.md (VF): New mode iterator.
8418 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8419 Remove integer-related logic from pattern.
8420 (neon_vabd<mode>_3): Likewise.
8422 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8425 * common.opt (fstrict-overflow): No longer an alias.
8426 (fwrapv-pointer): New option.
8427 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8428 also for pointer types based on flag_wrapv_pointer.
8429 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8430 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8431 opts->x_flag_wrapv got set.
8432 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8433 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8434 POINTER_TYPE_OVERFLOW_UNDEFINED.
8435 * match.pd: Likewise in address comparison pattern.
8436 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8438 2018-01-15 Richard Biener <rguenther@suse.de>
8441 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8442 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8443 Reset type names to their identifier if their TYPE_DECL doesn't
8444 have linkage (and thus is used for ODR and devirt).
8445 (save_debug_info_for_decl): Remove.
8446 (save_debug_info_for_type): Likewise.
8447 (add_tree_to_fld_list): Adjust.
8448 * tree-pretty-print.c (dump_generic_node): Make dumping of
8449 type names more robust.
8451 2018-01-15 Richard Biener <rguenther@suse.de>
8453 * BASE-VER: Bump to 8.0.1.
8455 2018-01-14 Martin Sebor <msebor@redhat.com>
8458 * builtins.c (check_access): Avoid warning when the no-warning bit
8461 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8463 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8464 * ira-color (allocno_hard_regs_compare): Likewise.
8466 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8469 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8470 Use .pushsection/.popsection.
8472 2018-01-14 Martin Sebor <msebor@redhat.com>
8475 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8477 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8479 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8480 entry from extra_headers.
8481 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8482 extra_headers, make the list bitwise identical to the i?86-*-* one.
8484 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8486 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8487 -mcmodel=large with -mindirect-branch=thunk,
8488 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8489 -mfunction-return=thunk-extern.
8490 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8491 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8492 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8494 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8496 * config/i386/i386.c (print_reg): Print the name of the full
8497 integer register without '%'.
8498 (ix86_print_operand): Handle 'V'.
8499 * doc/extend.texi: Document 'V' modifier.
8501 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8503 * config/i386/constraints.md (Bs): Disallow memory operand for
8504 -mindirect-branch-register.
8506 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8507 (GOT_memory_operand): Likewise.
8508 (call_insn_operand): Likewise.
8509 (sibcall_insn_operand): Likewise.
8510 (GOT32_symbol_operand): Likewise.
8511 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8512 for -mindirect-branch-register.
8513 (tablejump): Likewise.
8514 (*sibcall_memory): Likewise.
8515 (*sibcall_value_memory): Likewise.
8516 Disallow peepholes of indirect call and jump via memory for
8517 -mindirect-branch-register.
8518 (*call_pop): Replace m with Bw.
8519 (*call_value_pop): Likewise.
8520 (*sibcall_pop_memory): Replace m with Bs.
8521 * config/i386/i386.opt (mindirect-branch-register): New option.
8522 * doc/invoke.texi: Document -mindirect-branch-register option.
8524 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8526 * config/i386/i386-protos.h (ix86_output_function_return): New.
8527 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8528 set function_return_type.
8529 (indirect_thunk_name): Add ret_p to indicate thunk for function
8531 (output_indirect_thunk_function): Pass false to
8532 indirect_thunk_name.
8533 (ix86_output_indirect_branch_via_reg): Likewise.
8534 (ix86_output_indirect_branch_via_push): Likewise.
8535 (output_indirect_thunk_function): Create alias for function
8536 return thunk if regno < 0.
8537 (ix86_output_function_return): New function.
8538 (ix86_handle_fndecl_attribute): Handle function_return.
8539 (ix86_attribute_table): Add function_return.
8540 * config/i386/i386.h (machine_function): Add
8541 function_return_type.
8542 * config/i386/i386.md (simple_return_internal): Use
8543 ix86_output_function_return.
8544 (simple_return_internal_long): Likewise.
8545 * config/i386/i386.opt (mfunction-return=): New option.
8546 (indirect_branch): Mention -mfunction-return=.
8547 * doc/extend.texi: Document function_return function attribute.
8548 * doc/invoke.texi: Document -mfunction-return= option.
8550 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8552 * config/i386/i386-opts.h (indirect_branch): New.
8553 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8554 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8555 with local indirect jump when converting indirect call and jump.
8556 (ix86_set_indirect_branch_type): New.
8557 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8558 (indirectlabelno): New.
8559 (indirect_thunk_needed): Likewise.
8560 (indirect_thunk_bnd_needed): Likewise.
8561 (indirect_thunks_used): Likewise.
8562 (indirect_thunks_bnd_used): Likewise.
8563 (INDIRECT_LABEL): Likewise.
8564 (indirect_thunk_name): Likewise.
8565 (output_indirect_thunk): Likewise.
8566 (output_indirect_thunk_function): Likewise.
8567 (ix86_output_indirect_branch_via_reg): Likewise.
8568 (ix86_output_indirect_branch_via_push): Likewise.
8569 (ix86_output_indirect_branch): Likewise.
8570 (ix86_output_indirect_jmp): Likewise.
8571 (ix86_code_end): Call output_indirect_thunk_function if needed.
8572 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8574 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8575 (ix86_attribute_table): Add indirect_branch.
8576 * config/i386/i386.h (machine_function): Add indirect_branch_type
8577 and has_local_indirect_jump.
8578 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8580 (tablejump): Likewise.
8581 (*indirect_jump): Use ix86_output_indirect_jmp.
8582 (*tablejump_1): Likewise.
8583 (simple_return_indirect_internal): Likewise.
8584 * config/i386/i386.opt (mindirect-branch=): New option.
8585 (indirect_branch): New.
8588 (thunk-inline): Likewise.
8589 (thunk-extern): Likewise.
8590 * doc/extend.texi: Document indirect_branch function attribute.
8591 * doc/invoke.texi: Document -mindirect-branch= option.
8593 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8596 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8598 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8600 * ipa-inline.c (want_inline_small_function_p): Return false if
8601 inlining has already failed with CIF_FINAL_ERROR.
8602 (update_caller_keys): Call want_inline_small_function_p before
8604 (update_callee_keys): Likewise.
8606 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8608 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8610 (rs6000_quadword_masked_address_p): Likewise.
8611 (quad_aligned_load_p): Likewise.
8612 (quad_aligned_store_p): Likewise.
8613 (const_load_sequence_p): Add comment to describe the outer-most loop.
8614 (mimic_memory_attributes_and_flags): New function.
8615 (rs6000_gen_stvx): Likewise.
8616 (replace_swapped_aligned_store): Likewise.
8617 (rs6000_gen_lvx): Likewise.
8618 (replace_swapped_aligned_load): Likewise.
8619 (replace_swapped_load_constant): Capitalize argument name in
8620 comment describing this function.
8621 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8622 and stores that access quad-word aligned addresses and replace
8623 with stvx or lvx instructions when appropriate.
8624 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8625 New function prototype.
8626 (rs6000_quadword_masked_address_p): Likewise.
8627 (rs6000_gen_lvx): Likewise.
8628 (rs6000_gen_stvx): Likewise.
8629 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8630 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8631 when memory address is aligned.
8632 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8633 this split to select lvx instruction when memory address is aligned.
8634 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8635 instruction when memory address is aligned.
8636 (*vsx_le_perm_load_v16qi): Likewise.
8637 (four unnamed splitters): Modify to select the stvx instruction
8638 when memory is aligned.
8640 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8642 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8643 which appears in the queue multiple times.
8645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8646 Alan Hayward <alan.hayward@arm.com>
8647 David Sherwood <david.sherwood@arm.com>
8649 * tree-vectorizer.h (vec_lower_bound): New structure.
8650 (_loop_vec_info): Add check_nonzero and lower_bounds.
8651 (LOOP_VINFO_CHECK_NONZERO): New macro.
8652 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8653 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8654 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8655 fields. Make seg_len the distance travelled, not including the
8657 (dr_direction_indicator): Declare.
8658 (dr_zero_step_indicator): Likewise.
8659 (dr_known_forward_stride_p): Likewise.
8660 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8662 (runtime_alias_check_p): Allow runtime alias checks with
8664 (operator ==): Compare access_size and align.
8665 (prune_runtime_alias_test_list): Rework for new distinction between
8666 the access_size and seg_len.
8667 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8669 (get_segment_min_max): New function.
8670 (create_intersect_range_checks): Use it.
8671 (dr_step_indicator): New function.
8672 (dr_direction_indicator): Likewise.
8673 (dr_zero_step_indicator): Likewise.
8674 (dr_known_forward_stride_p): Likewise.
8675 * tree-loop-distribution.c (data_ref_segment_size): Return
8676 DR_STEP * (niters - 1).
8677 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8679 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8680 (vect_preserves_scalar_order_p): New function, split out from...
8681 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8682 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8683 (vect_vfa_access_size): New function.
8684 (vect_vfa_align): Likewise.
8685 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8686 (dump_lower_bound): New function.
8687 (vect_check_lower_bound): Likewise.
8688 (vect_small_gap_p): Likewise.
8689 (vectorizable_with_step_bound_p): Likewise.
8690 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8691 depencies if the vectorization factor is 1. Convert the checks
8692 for nonzero steps into checks on the bounds of DR_STEP. Try using
8693 a bunds check for variable steps if the minimum required step is
8694 relatively small. Update calls to the dr_with_seg_len
8695 constructor and to vect_compile_time_alias.
8696 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8698 (vect_loop_versioning): Call it.
8699 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8701 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8704 Alan Hayward <alan.hayward@arm.com>
8705 David Sherwood <david.sherwood@arm.com>
8707 * doc/sourcebuild.texi (vect_scatter_store): Document.
8708 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8710 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8712 * genopinit.c (main): Add supports_vec_scatter_store and
8713 supports_vec_scatter_store_cached to target_optabs.
8714 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8715 IFN_MASK_SCATTER_STORE.
8716 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8718 * internal-fn.h (internal_store_fn_p): Declare.
8719 (internal_fn_stored_value_index): Likewise.
8720 * internal-fn.c (scatter_store_direct): New macro.
8721 (expand_scatter_store_optab_fn): New function.
8722 (direct_scatter_store_optab_supported_p): New macro.
8723 (internal_store_fn_p): New function.
8724 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8725 IFN_MASK_SCATTER_STORE.
8726 (internal_fn_mask_index): Likewise.
8727 (internal_fn_stored_value_index): New function.
8728 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8730 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8731 * optabs-query.c (supports_vec_scatter_store_p): New function.
8732 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8733 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8734 true for scatter stores.
8735 (vect_gather_scatter_fn_p): Handle scatter stores too.
8736 (vect_check_gather_scatter): Consider using scatter stores if
8737 supports_vec_scatter_store_p.
8738 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8740 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8741 internal_fn_stored_value_index.
8742 (check_load_store_masking): Handle scatter stores too.
8743 (vect_get_store_rhs): Make public.
8744 (vectorizable_call): Use internal_store_fn_p.
8745 (vectorizable_store): Handle scatter store internal functions.
8746 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8747 when deciding whether the end of the group has been reached.
8748 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8749 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8750 (mask_scatter_store<mode>): New insns.
8752 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8753 Alan Hayward <alan.hayward@arm.com>
8754 David Sherwood <david.sherwood@arm.com>
8756 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8757 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8758 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8760 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8761 Use vect_truncate_gather_scatter_offset if we can't treat the
8762 operation as a normal gather load or scatter store.
8763 (get_group_load_store_type): Take the gather_scatter_info
8764 as argument. Try using a gather load or scatter store for
8765 single-element groups.
8766 (get_load_store_type): Update calls to get_group_load_store_type
8767 and vect_use_strided_gather_scatters_p.
8769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8770 Alan Hayward <alan.hayward@arm.com>
8771 David Sherwood <david.sherwood@arm.com>
8773 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8774 optional tree argument.
8775 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8777 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8778 but continue to use the current value as a fallback.
8779 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8780 to compare the updates.
8781 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8782 (get_load_store_type): Use it when handling a strided access.
8783 (vect_get_strided_load_store_ops): New function.
8784 (vect_get_data_ptr_increment): Likewise.
8785 (vectorizable_load): Handle strided gather loads. Always pass
8786 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8788 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8789 Alan Hayward <alan.hayward@arm.com>
8790 David Sherwood <david.sherwood@arm.com>
8792 * doc/md.texi (gather_load@var{m}): Document.
8793 (mask_gather_load@var{m}): Likewise.
8794 * genopinit.c (main): Add supports_vec_gather_load and
8795 supports_vec_gather_load_cached to target_optabs.
8796 * optabs-tree.c (init_tree_optimization_optabs): Use
8797 ggc_cleared_alloc to allocate target_optabs.
8798 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8799 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8801 * internal-fn.h (internal_load_fn_p): Declare.
8802 (internal_gather_scatter_fn_p): Likewise.
8803 (internal_fn_mask_index): Likewise.
8804 (internal_gather_scatter_fn_supported_p): Likewise.
8805 * internal-fn.c (gather_load_direct): New macro.
8806 (expand_gather_load_optab_fn): New function.
8807 (direct_gather_load_optab_supported_p): New macro.
8808 (direct_internal_fn_optab): New function.
8809 (internal_load_fn_p): Likewise.
8810 (internal_gather_scatter_fn_p): Likewise.
8811 (internal_fn_mask_index): Likewise.
8812 (internal_gather_scatter_fn_supported_p): Likewise.
8813 * optabs-query.c (supports_at_least_one_mode_p): New function.
8814 (supports_vec_gather_load_p): Likewise.
8815 * optabs-query.h (supports_vec_gather_load_p): Declare.
8816 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8817 and memory_type field.
8818 (NUM_PATTERNS): Bump to 15.
8819 * tree-vect-data-refs.c: Include internal-fn.h.
8820 (vect_gather_scatter_fn_p): New function.
8821 (vect_describe_gather_scatter_call): Likewise.
8822 (vect_check_gather_scatter): Try using internal functions for
8823 gather loads. Recognize existing calls to a gather load function.
8824 (vect_analyze_data_refs): Consider using gather loads if
8825 supports_vec_gather_load_p.
8826 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8827 (vect_get_gather_scatter_offset_type): Likewise.
8828 (vect_convert_mask_for_vectype): Likewise.
8829 (vect_add_conversion_to_patterm): Likewise.
8830 (vect_try_gather_scatter_pattern): Likewise.
8831 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8832 (vect_vect_recog_func_ptrs): Add it.
8833 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8834 internal_fn_mask_index and internal_gather_scatter_fn_p.
8835 (check_load_store_masking): Take the gather_scatter_info as an
8836 argument and handle gather loads.
8837 (vect_get_gather_scatter_ops): New function.
8838 (vectorizable_call): Check internal_load_fn_p.
8839 (vectorizable_load): Likewise. Handle gather load internal
8841 (vectorizable_store): Update call to check_load_store_masking.
8842 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8843 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8844 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8845 (aarch64_gather_scale_operand_d): New predicates.
8846 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8847 (mask_gather_load<mode>): New insns.
8849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8850 Alan Hayward <alan.hayward@arm.com>
8851 David Sherwood <david.sherwood@arm.com>
8853 * optabs.def (fold_left_plus_optab): New optab.
8854 * doc/md.texi (fold_left_plus_@var{m}): Document.
8855 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8856 * internal-fn.c (fold_left_direct): Define.
8857 (expand_fold_left_optab_fn): Likewise.
8858 (direct_fold_left_optab_supported_p): Likewise.
8859 * fold-const-call.c (fold_const_fold_left): New function.
8860 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8861 * tree-parloops.c (valid_reduction_p): New function.
8862 (gather_scalar_reductions): Use it.
8863 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8864 (vect_finish_replace_stmt): Declare.
8865 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8866 (needs_fold_left_reduction_p): New function, split out from...
8867 (vect_is_simple_reduction): ...here. Accept reductions that
8868 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8869 (vect_force_simple_reduction): Also store the reduction type in
8870 the assignment's STMT_VINFO_REDUC_TYPE.
8871 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8872 (merge_with_identity): New function.
8873 (vect_expand_fold_left): Likewise.
8874 (vectorize_fold_left_reduction): Likewise.
8875 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8876 scalar phi in place for it. Check for target support and reject
8877 cases that would reassociate the operation. Defer the transform
8878 phase to vectorize_fold_left_reduction.
8879 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8880 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8881 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8885 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8886 call to ifc_temp_var.
8888 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8889 Alan Hayward <alan.hayward@arm.com>
8890 David Sherwood <david.sherwood@arm.com>
8892 * target.def (legitimize_address_displacement): Take the original
8893 offset as a poly_int.
8894 * targhooks.h (default_legitimize_address_displacement): Update
8896 * targhooks.c (default_legitimize_address_displacement): Likewise.
8897 * doc/tm.texi: Regenerate.
8898 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8899 as an argument, moving assert of ad->disp == ad->disp_term to...
8900 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8901 Try calling targetm.legitimize_address_displacement before expanding
8902 the address rather than afterwards, and adjust for the new interface.
8903 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8904 Match the new hook interface. Handle SVE addresses.
8905 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8908 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8910 * Makefile.in (OBJS): Add early-remat.o.
8911 * target.def (select_early_remat_modes): New hook.
8912 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8913 * doc/tm.texi: Regenerate.
8914 * targhooks.h (default_select_early_remat_modes): Declare.
8915 * targhooks.c (default_select_early_remat_modes): New function.
8916 * timevar.def (TV_EARLY_REMAT): New timevar.
8917 * passes.def (pass_early_remat): New pass.
8918 * tree-pass.h (make_pass_early_remat): Declare.
8919 * early-remat.c: New file.
8920 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8922 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8925 Alan Hayward <alan.hayward@arm.com>
8926 David Sherwood <david.sherwood@arm.com>
8928 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8929 vfm1 with a bound_epilog parameter.
8930 (vect_do_peeling): Update calls accordingly, and move the prologue
8931 call earlier in the function. Treat the base bound_epilog as 0 for
8932 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8933 this base when peeling for gaps.
8934 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8935 with fully-masked loops.
8936 (vect_estimate_min_profitable_iters): Handle the single peeled
8937 iteration in that case.
8939 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8940 Alan Hayward <alan.hayward@arm.com>
8941 David Sherwood <david.sherwood@arm.com>
8943 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8944 single-element interleaving even if the size is not a power of 2.
8945 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8946 accesses for single-element interleaving if the group size is
8949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8950 Alan Hayward <alan.hayward@arm.com>
8951 David Sherwood <david.sherwood@arm.com>
8953 * doc/md.texi (fold_extract_last_@var{m}): Document.
8954 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8955 * optabs.def (fold_extract_last_optab): New optab.
8956 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8957 * internal-fn.c (fold_extract_direct): New macro.
8958 (expand_fold_extract_optab_fn): Likewise.
8959 (direct_fold_extract_optab_supported_p): Likewise.
8960 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8961 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8962 EXTRACT_LAST_REDUCTION.
8963 (get_initial_def_for_reduction): Do not create an initial vector
8964 for EXTRACT_LAST_REDUCTION reductions.
8965 (vectorizable_reduction): Leave the scalar phi in place for
8966 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8967 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8968 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8969 transform phase to vectorizable_condition.
8970 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8972 (vect_finish_stmt_generation): ...here.
8973 (vect_finish_replace_stmt): New function.
8974 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8975 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8977 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8980 Alan Hayward <alan.hayward@arm.com>
8981 David Sherwood <david.sherwood@arm.com>
8983 * doc/md.texi (extract_last_@var{m}): Document.
8984 * optabs.def (extract_last_optab): New optab.
8985 * internal-fn.def (EXTRACT_LAST): New internal function.
8986 * internal-fn.c (cond_unary_direct): New macro.
8987 (expand_cond_unary_optab_fn): Likewise.
8988 (direct_cond_unary_optab_supported_p): Likewise.
8989 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8990 loops using EXTRACT_LAST.
8991 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8992 (extract_last_<mode>): ...this optab.
8993 (vec_extract<mode><Vel>): Update accordingly.
8995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8996 Alan Hayward <alan.hayward@arm.com>
8997 David Sherwood <david.sherwood@arm.com>
8999 * target.def (empty_mask_is_expensive): New hook.
9000 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9001 * doc/tm.texi: Regenerate.
9002 * targhooks.h (default_empty_mask_is_expensive): Declare.
9003 * targhooks.c (default_empty_mask_is_expensive): New function.
9004 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9005 if the target says that empty masks are expensive.
9006 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9008 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9011 Alan Hayward <alan.hayward@arm.com>
9012 David Sherwood <david.sherwood@arm.com>
9014 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9015 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9016 (vect_use_loop_mask_for_alignment_p): New function.
9017 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9018 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9019 niters_skip argument. Make sure that the first niters_skip elements
9020 of the first iteration are inactive.
9021 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9022 Update call to vect_set_loop_masks_directly.
9023 (get_misalign_in_elems): New function, split out from...
9024 (vect_gen_prolog_loop_niters): ...here.
9025 (vect_update_init_of_dr): Take a code argument that specifies whether
9026 the adjustment should be added or subtracted.
9027 (vect_update_init_of_drs): Likewise.
9028 (vect_prepare_for_masked_peels): New function.
9029 (vect_do_peeling): Skip prologue peeling if we're using a mask
9030 instead. Update call to vect_update_inits_of_drs.
9031 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9033 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9034 alignment. Do not include the number of peeled iterations in
9035 the minimum threshold in that case.
9036 (vectorizable_induction): Adjust the start value down by
9037 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9038 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9039 Take the number of skipped iterations into account when calculating
9041 * tree-vect-stmts.c (vect_gen_while_not): New function.
9043 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9044 Alan Hayward <alan.hayward@arm.com>
9045 David Sherwood <david.sherwood@arm.com>
9047 * doc/sourcebuild.texi (vect_fully_masked): Document.
9048 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9050 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9052 (vect_analyze_loop_2): ...here. Don't check the vectorization
9053 factor against the number of loop iterations if the loop is
9056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9057 Alan Hayward <alan.hayward@arm.com>
9058 David Sherwood <david.sherwood@arm.com>
9060 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9061 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9062 (dump_groups): Update accordingly.
9063 (iv_use::mem_type): New member variable.
9064 (address_p): New function.
9065 (record_use): Add a mem_type argument and initialize the new
9067 (record_group_use): Add a mem_type argument. Use address_p.
9068 Remove obsolete null checks of base_object. Update call to record_use.
9069 (find_interesting_uses_op): Update call to record_group_use.
9070 (find_interesting_uses_cond): Likewise.
9071 (find_interesting_uses_address): Likewise.
9072 (get_mem_type_for_internal_fn): New function.
9073 (find_address_like_use): Likewise.
9074 (find_interesting_uses_stmt): Try find_address_like_use before
9075 calling find_interesting_uses_op.
9076 (addr_offset_valid_p): Use the iv mem_type field as the type
9077 of the addressed memory.
9078 (add_autoinc_candidates): Likewise.
9079 (get_address_cost): Likewise.
9080 (split_small_address_groups_p): Use address_p.
9081 (split_address_groups): Likewise.
9082 (add_iv_candidate_for_use): Likewise.
9083 (autoinc_possible_for_pair): Likewise.
9084 (rewrite_groups): Likewise.
9085 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9086 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9087 (get_alias_ptr_type_for_ptr_address): New function.
9088 (rewrite_use_address): Rewrite address uses in calls that were
9089 identified by find_address_like_use.
9091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9092 Alan Hayward <alan.hayward@arm.com>
9093 David Sherwood <david.sherwood@arm.com>
9095 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9097 * gimple-expr.h (is_gimple_addressable: Likewise.
9098 * gimple-expr.c (is_gimple_address): Likewise.
9099 * internal-fn.c (expand_call_mem_ref): New function.
9100 (expand_mask_load_optab_fn): Use it.
9101 (expand_mask_store_optab_fn): Likewise.
9103 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9104 Alan Hayward <alan.hayward@arm.com>
9105 David Sherwood <david.sherwood@arm.com>
9107 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9108 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9109 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9110 (cond_umax@var{mode}): Document.
9111 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9112 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9113 (cond_umin_optab, cond_umax_optab): New optabs.
9114 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9115 (COND_IOR, COND_XOR): New internal functions.
9116 * internal-fn.h (get_conditional_internal_fn): Declare.
9117 * internal-fn.c (cond_binary_direct): New macro.
9118 (expand_cond_binary_optab_fn): Likewise.
9119 (direct_cond_binary_optab_supported_p): Likewise.
9120 (get_conditional_internal_fn): New function.
9121 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9122 Cope with reduction statements that are vectorized as calls rather
9124 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9125 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9126 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9127 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9128 (UNSPEC_COND_EOR): New unspecs.
9129 (optab): Add mappings for them.
9130 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9131 (sve_int_op, sve_fp_op): New int attributes.
9133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9134 Alan Hayward <alan.hayward@arm.com>
9135 David Sherwood <david.sherwood@arm.com>
9137 * optabs.def (while_ult_optab): New optab.
9138 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9139 * internal-fn.def (WHILE_ULT): New internal function.
9140 * internal-fn.h (direct_internal_fn_supported_p): New override
9141 that takes two types as argument.
9142 * internal-fn.c (while_direct): New macro.
9143 (expand_while_optab_fn): New function.
9144 (convert_optab_supported_p): Likewise.
9145 (direct_while_optab_supported_p): New macro.
9146 * wide-int.h (wi::udiv_ceil): New function.
9147 * tree-vectorizer.h (rgroup_masks): New structure.
9148 (vec_loop_masks): New typedef.
9149 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9151 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9152 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9153 (vect_max_vf): New function.
9154 (slpeel_make_loop_iterate_ntimes): Delete.
9155 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9156 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9157 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9158 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9159 internal-fn.h, stor-layout.h and optabs-query.h.
9160 (vect_set_loop_mask): New function.
9161 (add_preheader_seq): Likewise.
9162 (add_header_seq): Likewise.
9163 (interleave_supported_p): Likewise.
9164 (vect_maybe_permute_loop_masks): Likewise.
9165 (vect_set_loop_masks_directly): Likewise.
9166 (vect_set_loop_condition_masked): Likewise.
9167 (vect_set_loop_condition_unmasked): New function, split out from
9168 slpeel_make_loop_iterate_ntimes.
9169 (slpeel_make_loop_iterate_ntimes): Rename to..
9170 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9171 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9172 (vect_do_peeling): Update call accordingly.
9173 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9175 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9176 mask_compare_type, can_fully_mask_p and fully_masked_p.
9177 (release_vec_loop_masks): New function.
9178 (_loop_vec_info): Use it to free the loop masks.
9179 (can_produce_all_loop_masks_p): New function.
9180 (vect_get_max_nscalars_per_iter): Likewise.
9181 (vect_verify_full_masking): Likewise.
9182 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9183 retries, and free the mask rgroups before retrying. Check loop-wide
9184 reasons for disallowing fully-masked loops. Make the final decision
9185 about whether use a fully-masked loop or not.
9186 (vect_estimate_min_profitable_iters): Do not assume that peeling
9187 for the number of iterations will be needed for fully-masked loops.
9188 (vectorizable_reduction): Disable fully-masked loops.
9189 (vectorizable_live_operation): Likewise.
9190 (vect_halve_mask_nunits): New function.
9191 (vect_double_mask_nunits): Likewise.
9192 (vect_record_loop_mask): Likewise.
9193 (vect_get_loop_mask): Likewise.
9194 (vect_transform_loop): Handle the case in which the final loop
9195 iteration might handle a partial vector. Call vect_set_loop_condition
9196 instead of slpeel_make_loop_iterate_ntimes.
9197 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9198 (check_load_store_masking): New function.
9199 (prepare_load_store_mask): Likewise.
9200 (vectorizable_store): Handle fully-masked loops.
9201 (vectorizable_load): Likewise.
9202 (supportable_widening_operation): Use vect_halve_mask_nunits for
9204 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9205 (vect_gen_while): New function.
9206 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9207 (aarch64_uqdec<mode>): New insn.
9209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9210 Alan Hayward <alan.hayward@arm.com>
9211 David Sherwood <david.sherwood@arm.com>
9213 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9214 (reduc_xor_scal_optab): New optabs.
9215 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9216 (reduc_xor_scal_@var{m}): Document.
9217 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9218 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9220 * fold-const-call.c (fold_const_call): Handle them.
9221 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9222 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9223 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9224 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9225 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9226 (UNSPEC_XORV): New unspecs.
9227 (optab): Add entries for them.
9228 (BITWISEV): New int iterator.
9229 (bit_reduc_op): New int attributes.
9231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9232 Alan Hayward <alan.hayward@arm.com>
9233 David Sherwood <david.sherwood@arm.com>
9235 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9236 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9237 * optabs.def (vec_shl_insert_optab): New optab.
9238 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9239 (duplicate_and_interleave): Likewise.
9240 * tree-vect-loop.c: Include internal-fn.h.
9241 (neutral_op_for_slp_reduction): New function, split out from
9242 get_initial_defs_for_reduction.
9243 (get_initial_def_for_reduction): Handle option 2 for variable-length
9244 vectors by loading the neutral value into a vector and then shifting
9245 the initial value into element 0.
9246 (get_initial_defs_for_reduction): Replace the code argument with
9247 the neutral value calculated by neutral_op_for_slp_reduction.
9248 Use gimple_build_vector for constant-length vectors.
9249 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9250 but the first group_size elements have a neutral value.
9251 Use duplicate_and_interleave otherwise.
9252 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9253 Update call to get_initial_defs_for_reduction. Handle SLP
9254 reductions for variable-length vectors by creating one vector
9255 result for each scalar result, with the elements associated
9256 with other scalar results stubbed out with the neutral value.
9257 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9258 Require IFN_VEC_SHL_INSERT for double reductions on
9259 variable-length vectors, or SLP reductions that have
9260 a neutral value. Require can_duplicate_and_interleave_p
9261 support for variable-length unchained SLP reductions if there
9262 is no neutral value, such as for MIN/MAX reductions. Also require
9263 the number of vector elements to be a multiple of the number of
9264 SLP statements when doing variable-length unchained SLP reductions.
9265 Update call to vect_create_epilog_for_reduction.
9266 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9267 and remove initial values.
9268 (duplicate_and_interleave): Make public.
9269 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9270 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9273 Alan Hayward <alan.hayward@arm.com>
9274 David Sherwood <david.sherwood@arm.com>
9276 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9277 (can_duplicate_and_interleave_p): New function.
9278 (vect_get_and_check_slp_defs): Take the vector of statements
9279 rather than just the current one. Remove excess parentheses.
9280 Restriction rejectinon of vect_constant_def and vect_external_def
9281 for variable-length vectors to boolean types, or types for which
9282 can_duplicate_and_interleave_p is false.
9283 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9284 (duplicate_and_interleave): New function.
9285 (vect_get_constant_vectors): Use gimple_build_vector for
9286 constant-length vectors and suitable variable-length constant
9287 vectors. Use duplicate_and_interleave for other variable-length
9288 vectors. Don't defer the update when inserting new statements.
9290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9291 Alan Hayward <alan.hayward@arm.com>
9292 David Sherwood <david.sherwood@arm.com>
9294 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9295 min_profitable_iters doesn't go negative.
9297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9298 Alan Hayward <alan.hayward@arm.com>
9299 David Sherwood <david.sherwood@arm.com>
9301 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9302 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9303 * optabs.def (vec_mask_load_lanes_optab): New optab.
9304 (vec_mask_store_lanes_optab): Likewise.
9305 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9306 (MASK_STORE_LANES): Likewise.
9307 * internal-fn.c (mask_load_lanes_direct): New macro.
9308 (mask_store_lanes_direct): Likewise.
9309 (expand_mask_load_optab_fn): Handle masked operations.
9310 (expand_mask_load_lanes_optab_fn): New macro.
9311 (expand_mask_store_optab_fn): Handle masked operations.
9312 (expand_mask_store_lanes_optab_fn): New macro.
9313 (direct_mask_load_lanes_optab_supported_p): Likewise.
9314 (direct_mask_store_lanes_optab_supported_p): Likewise.
9315 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9317 (vect_load_lanes_supported): Likewise.
9318 * tree-vect-data-refs.c (strip_conversion): New function.
9319 (can_group_stmts_p): Likewise.
9320 (vect_analyze_data_ref_accesses): Use it instead of checking
9321 for a pair of assignments.
9322 (vect_store_lanes_supported): Take a masked_p parameter.
9323 (vect_load_lanes_supported): Likewise.
9324 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9325 vect_store_lanes_supported and vect_load_lanes_supported.
9326 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9327 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9328 parameter. Don't allow gaps for masked accesses.
9329 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9330 and vect_load_lanes_supported.
9331 (get_load_store_type): Take a masked_p parameter and update
9332 call to get_group_load_store_type.
9333 (vectorizable_store): Update call to get_load_store_type.
9334 Handle IFN_MASK_STORE_LANES.
9335 (vectorizable_load): Update call to get_load_store_type.
9336 Handle IFN_MASK_LOAD_LANES.
9338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9339 Alan Hayward <alan.hayward@arm.com>
9340 David Sherwood <david.sherwood@arm.com>
9342 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9344 * config/aarch64/aarch64-protos.h
9345 (aarch64_sve_struct_memory_operand_p): Declare.
9346 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9347 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9348 (VPRED, vpred): Handle SVE structure modes.
9349 * config/aarch64/constraints.md (Utx): New constraint.
9350 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9351 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9352 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9353 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9354 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9355 structure modes. Split into pieces after RA.
9356 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9357 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9359 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9360 SVE structure modes.
9361 (aarch64_classify_address): Likewise.
9362 (sizetochar): Move earlier in file.
9363 (aarch64_print_operand): Handle SVE register lists.
9364 (aarch64_array_mode): New function.
9365 (aarch64_sve_struct_memory_operand_p): Likewise.
9366 (TARGET_ARRAY_MODE): Redefine.
9368 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9369 Alan Hayward <alan.hayward@arm.com>
9370 David Sherwood <david.sherwood@arm.com>
9372 * target.def (array_mode): New target hook.
9373 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9374 * doc/tm.texi: Regenerate.
9375 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9376 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9377 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9379 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9382 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9383 Alan Hayward <alan.hayward@arm.com>
9384 David Sherwood <david.sherwood@arm.com>
9386 * fold-const.c (fold_binary_loc): Check the argument types
9387 rather than the result type when testing for a vector operation.
9389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9391 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9392 * doc/tm.texi: Regenerate.
9394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9395 Alan Hayward <alan.hayward@arm.com>
9396 David Sherwood <david.sherwood@arm.com>
9398 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9399 (sve): Document new AArch64 extension.
9400 * doc/md.texi (w): Extend the description of the AArch64
9401 constraint to include SVE vectors.
9402 (Upl, Upa): Document new AArch64 predicate constraints.
9403 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9405 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9406 (msve-vector-bits=): New option.
9407 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9408 SVE when these are disabled.
9409 (sve): New extension.
9410 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9411 modes. Adjust their number of units based on aarch64_sve_vg.
9412 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9413 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9414 aarch64_addr_query_type.
9415 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9416 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9417 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9418 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9419 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9420 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9421 (aarch64_simd_imm_zero_p): Delete.
9422 (aarch64_check_zero_based_sve_index_immediate): Declare.
9423 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9424 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9425 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9426 (aarch64_sve_float_mul_immediate_p): Likewise.
9427 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9429 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9430 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9431 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9432 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9433 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9434 (aarch64_regmode_natural_size): Likewise.
9435 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9436 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9438 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9439 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9440 for VG and the SVE predicate registers.
9441 (V_ALIASES): Add a "z"-prefixed alias.
9442 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9443 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9444 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9445 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9446 (REG_CLASS_NAMES): Add entries for them.
9447 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9448 and the predicate registers.
9449 (aarch64_sve_vg): Declare.
9450 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9451 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9452 (REGMODE_NATURAL_SIZE): Define.
9453 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9455 * config/aarch64/aarch64.c: Include cfgrtl.h.
9456 (simd_immediate_info): Add a constructor for series vectors,
9457 and an associated step field.
9458 (aarch64_sve_vg): New variable.
9459 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9460 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9461 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9462 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9463 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9464 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9465 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9466 (aarch64_get_mask_mode): New functions.
9467 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9468 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9469 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9470 predicate modes and predicate registers. Explicitly restrict
9471 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9472 to store a vector mode if it is recognized by
9473 aarch64_classify_vector_mode.
9474 (aarch64_regmode_natural_size): New function.
9475 (aarch64_hard_regno_caller_save_mode): Return the original mode
9477 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9478 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9479 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9480 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9482 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9483 does not overlap dest if the function is frame-related. Handle
9485 (aarch64_split_add_offset): New function.
9486 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9487 them aarch64_add_offset.
9488 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9489 and update call to aarch64_sub_sp.
9490 (aarch64_add_cfa_expression): New function.
9491 (aarch64_expand_prologue): Pass extra temporary registers to the
9492 functions above. Handle the case in which we need to emit new
9493 DW_CFA_expressions for registers that were originally saved
9494 relative to the stack pointer, but now have to be expressed
9495 relative to the frame pointer.
9496 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9498 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9499 IP0 and IP1 values for SVE frames.
9500 (aarch64_expand_vec_series): New function.
9501 (aarch64_expand_sve_widened_duplicate): Likewise.
9502 (aarch64_expand_sve_const_vector): Likewise.
9503 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9504 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9505 into the register, rather than emitting a SET directly.
9506 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9507 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9508 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9509 (offset_9bit_signed_scaled_p): New functions.
9510 (aarch64_replicate_bitmask_imm): New function.
9511 (aarch64_bitmask_imm): Use it.
9512 (aarch64_cannot_force_const_mem): Reject expressions involving
9513 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9514 (aarch64_classify_index): Handle SVE indices, by requiring
9515 a plain register index with a scale that matches the element size.
9516 (aarch64_classify_address): Handle SVE addresses. Assert that
9517 the mode of the address is VOIDmode or an integer mode.
9518 Update call to aarch64_classify_symbol.
9519 (aarch64_classify_symbolic_expression): Update call to
9520 aarch64_classify_symbol.
9521 (aarch64_const_vec_all_in_range_p): New function.
9522 (aarch64_print_vector_float_operand): Likewise.
9523 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9524 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9525 and the FP immediates 1.0 and 0.5.
9526 (aarch64_print_address_internal): Handle SVE addresses.
9527 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9528 (aarch64_regno_regclass): Handle predicate registers.
9529 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9531 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9532 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9533 (aarch64_convert_sve_vector_bits): New function.
9534 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9535 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9537 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9538 Handle SVE vector and predicate modes. Accept VL-based constants
9539 that need only one temporary register, and VL offsets that require
9540 no temporary registers.
9541 (aarch64_conditional_register_usage): Mark the predicate registers
9542 as fixed if SVE isn't available.
9543 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9544 Return true for SVE vector and predicate modes.
9545 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9546 rather than an unsigned int. Handle SVE modes.
9547 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9549 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9551 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9552 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9553 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9554 (aarch64_sve_float_mul_immediate_p): New functions.
9555 (aarch64_sve_valid_immediate): New function.
9556 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9557 Explicitly reject structure modes. Check for INDEX constants.
9558 Handle PTRUE and PFALSE constants.
9559 (aarch64_check_zero_based_sve_index_immediate): New function.
9560 (aarch64_simd_imm_zero_p): Delete.
9561 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9562 vector modes. Accept constants in the range of CNT[BHWD].
9563 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9564 ask for an Advanced SIMD mode.
9565 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9566 (aarch64_simd_vector_alignment): Handle SVE predicates.
9567 (aarch64_vectorize_preferred_vector_alignment): New function.
9568 (aarch64_simd_vector_alignment_reachable): Use it instead of
9570 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9571 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9573 (MAX_VECT_LEN): Delete.
9574 (expand_vec_perm_d): Add a vec_flags field.
9575 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9576 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9577 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9579 (aarch64_evpc_rev): Rename to...
9580 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9581 (aarch64_evpc_rev_global): New function.
9582 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9583 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9585 (aarch64_evpc_sve_tbl): New function.
9586 (aarch64_expand_vec_perm_const_1): Update after rename of
9587 aarch64_evpc_rev. Handle SVE permutes too, trying
9588 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9589 than aarch64_evpc_tbl.
9590 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9591 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9592 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9593 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9594 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9595 (aarch64_expand_sve_vcond): New functions.
9596 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9597 of aarch64_vector_mode_p.
9598 (aarch64_dwarf_poly_indeterminate_value): New function.
9599 (aarch64_compute_pressure_classes): Likewise.
9600 (aarch64_can_change_mode_class): Likewise.
9601 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9602 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9603 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9604 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9605 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9606 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9607 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9608 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9610 (Dn, Dl, Dr): Accept const as well as const_vector.
9611 (Dz): Likewise. Compare against CONST0_RTX.
9612 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9613 of "vector" where appropriate.
9614 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9615 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9616 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9617 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9618 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9619 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9620 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9621 (v_int_equiv): Extend to SVE modes.
9622 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9624 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9625 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9626 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9627 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9628 (SVE_COND_FP_CMP): New int iterators.
9629 (perm_hilo): Handle the new unpack unspecs.
9630 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9632 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9633 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9634 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9635 (aarch64_equality_operator, aarch64_constant_vector_operand)
9636 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9637 (aarch64_sve_nonimmediate_operand): Likewise.
9638 (aarch64_sve_general_operand): Likewise.
9639 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9640 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9641 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9642 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9643 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9644 (aarch64_sve_float_arith_immediate): Likewise.
9645 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9646 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9647 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9648 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9649 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9650 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9651 (aarch64_sve_float_arith_operand): Likewise.
9652 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9653 (aarch64_sve_float_mul_operand): Likewise.
9654 (aarch64_sve_vec_perm_operand): Likewise.
9655 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9656 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9657 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9658 as well as const_vector.
9659 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9660 in file. Use CONST0_RTX and CONSTM1_RTX.
9661 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9662 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9663 Use aarch64_simd_imm_zero.
9664 * config/aarch64/aarch64-sve.md: New file.
9665 * config/aarch64/aarch64.md: Include it.
9666 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9667 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9668 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9669 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9670 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9671 (sve): New attribute.
9672 (enabled): Disable instructions with the sve attribute unless
9674 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9675 aarch64_expand_mov_immediate.
9676 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9677 CNT[BHSD] immediates.
9678 (movti): Split CONST_POLY_INT moves into two halves.
9679 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9680 Split additions that need a temporary here if the destination
9681 is the stack pointer.
9682 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9683 (*add<mode>3_poly_1): New instruction.
9684 (set_clobber_cc): New expander.
9686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9688 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9689 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9690 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9691 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9692 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9693 Change innermode from fixed_mode_size to machine_mode.
9694 (simplify_subreg): Update call accordingly. Handle a constant-sized
9695 subreg of a variable-length CONST_VECTOR.
9697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9698 Alan Hayward <alan.hayward@arm.com>
9699 David Sherwood <david.sherwood@arm.com>
9701 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9702 (add_offset_to_base): New function, split out from...
9703 (create_mem_ref): ...here. When handling a scale other than 1,
9704 check first whether the address is valid without the offset.
9705 Add it into the base if so, leaving the index and scale as-is.
9707 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9710 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9711 fold_for_warn before checking if arg2 is INTEGER_CST.
9713 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9715 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9716 (store_multiple_operation): Delete.
9717 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9718 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9719 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9720 guarded by TARGET_STRING.
9721 (rs6000_output_load_multiple): Delete.
9722 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9723 OPTION_MASK_STRING / TARGET_STRING handling.
9724 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9725 (const rs6000_opt_masks) <"string">: Change mask to 0.
9726 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9727 (MASK_STRING): Delete.
9728 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9730 (load_multiple): Delete.
9737 (store_multiple): Delete.
9744 (movmemsi_8reg): Delete.
9745 (corresponding unnamed define_insn): Delete.
9746 (movmemsi_6reg): Delete.
9747 (corresponding unnamed define_insn): Delete.
9748 (movmemsi_4reg): Delete.
9749 (corresponding unnamed define_insn): Delete.
9750 (movmemsi_2reg): Delete.
9751 (corresponding unnamed define_insn): Delete.
9752 (movmemsi_1reg): Delete.
9753 (corresponding unnamed define_insn): Delete.
9754 * config/rs6000/rs6000.opt (mno-string): New.
9755 (mstring): Replace by deprecation warning stub.
9756 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9758 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9760 * regrename.c (regrename_do_replace): If replacing the same
9761 reg multiple times, try to reuse last created gen_raw_REG.
9764 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9765 main to workaround a bug in GDB.
9767 2018-01-12 Tom de Vries <tom@codesourcery.com>
9770 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9772 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9774 PR rtl-optimization/80481
9775 * ira-color.c (get_cap_member): New function.
9776 (allocnos_conflict_by_live_ranges_p): Use it.
9777 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9778 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9780 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9783 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9784 (*saddl_se_1): Ditto.
9786 (*ssubl_se_1): Ditto.
9788 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9790 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9791 rather than wi::to_widest for DR_INITs.
9792 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9793 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9794 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9796 (vect_analyze_group_access_1): Note that here.
9798 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9800 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9801 polynomial type sizes.
9803 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9805 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9806 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9807 (gimple_add_tmp_var): Likewise.
9809 2018-01-12 Martin Liska <mliska@suse.cz>
9811 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9812 (gimple_alloc_sizes): Likewise.
9813 (dump_gimple_statistics): Use PRIu64 in printf format.
9814 * gimple.h: Change uint64_t to int.
9816 2018-01-12 Martin Liska <mliska@suse.cz>
9818 * tree-core.h: Use uint64_t instead of int.
9819 * tree.c (tree_node_counts): Likewise.
9820 (tree_node_sizes): Likewise.
9821 (dump_tree_statistics): Use PRIu64 in printf format.
9823 2018-01-12 Martin Liska <mliska@suse.cz>
9825 * Makefile.in: As qsort_chk is implemented in vec.c, add
9826 vec.o to linkage of gencfn-macros.
9827 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9828 passing the info to record_node_allocation_statistics.
9829 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9831 * ggc-common.c (struct ggc_usage): Add operator== and use
9832 it in operator< and compare function.
9833 * mem-stats.h (struct mem_usage): Likewise.
9834 * vec.c (struct vec_usage): Remove operator< and compare
9835 function. Can be simply inherited.
9837 2018-01-12 Martin Jambor <mjambor@suse.cz>
9840 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9841 * tree-ssa-math-opts.c: Include domwalk.h.
9842 (convert_mult_to_fma_1): New function.
9843 (fma_transformation_info): New type.
9844 (fma_deferring_state): Likewise.
9845 (cancel_fma_deferring): New function.
9846 (result_of_phi): Likewise.
9847 (last_fma_candidate_feeds_initial_phi): Likewise.
9848 (convert_mult_to_fma): Added deferring logic, split actual
9849 transformation to convert_mult_to_fma_1.
9850 (math_opts_dom_walker): New type.
9851 (math_opts_dom_walker::after_dom_children): New method, body moved
9852 here from pass_optimize_widening_mul::execute, added deferring logic
9854 (pass_optimize_widening_mul::execute): Moved most of code to
9855 math_opts_dom_walker::after_dom_children.
9856 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9857 * config/i386/i386.c (ix86_option_override_internal): Added
9858 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9860 2018-01-12 Richard Biener <rguenther@suse.de>
9863 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9864 inline instance vars.
9866 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9869 * config/rx/rx.c (rx_is_restricted_memory_address):
9872 2018-01-12 Richard Biener <rguenther@suse.de>
9874 PR tree-optimization/80846
9875 * target.def (split_reduction): New target hook.
9876 * targhooks.c (default_split_reduction): New function.
9877 * targhooks.h (default_split_reduction): Declare.
9878 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9879 target requests first reduce vectors by combining low and high
9881 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9882 (get_vectype_for_scalar_type_and_size): Export.
9883 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9884 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9885 * doc/tm.texi: Regenerate.
9886 * config/i386/i386.c (ix86_split_reduction): Implement
9887 TARGET_VECTORIZE_SPLIT_REDUCTION.
9889 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9892 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9893 in PIC mode except for TARGET_VXWORKS_RTP.
9894 * config/sparc/sparc.c: Include cfgrtl.h.
9895 (TARGET_INIT_PIC_REG): Define.
9896 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9897 (sparc_pic_register_p): New predicate.
9898 (sparc_legitimate_address_p): Use it.
9899 (sparc_legitimize_pic_address): Likewise.
9900 (sparc_delegitimize_address): Likewise.
9901 (sparc_mode_dependent_address_p): Likewise.
9902 (gen_load_pcrel_sym): Remove 4th parameter.
9903 (load_got_register): Adjust call to above. Remove obsolete stuff.
9904 (sparc_expand_prologue): Do not call load_got_register here.
9905 (sparc_flat_expand_prologue): Likewise.
9906 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9907 (sparc_use_pseudo_pic_reg): New function.
9908 (sparc_init_pic_reg): Likewise.
9909 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9910 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9912 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9914 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9915 Add item for branch_cost.
9917 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9919 PR rtl-optimization/83565
9920 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9921 not extend the result to a larger mode for rotate operations.
9922 (num_sign_bit_copies1): Likewise.
9924 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9927 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9929 Use values-Xc.o for -pedantic.
9930 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9932 2018-01-12 Martin Liska <mliska@suse.cz>
9935 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9937 (possible_polymorphic_call_targets): Use it.
9938 (ipa_devirt): Likewise.
9940 2018-01-12 Martin Liska <mliska@suse.cz>
9942 * profile-count.h (enum profile_quality): Use 0 as invalid
9943 enum value of profile_quality.
9945 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9947 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9948 -mext-string options.
9950 2018-01-12 Richard Biener <rguenther@suse.de>
9952 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9953 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9954 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9956 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9958 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9960 * configure.ac (--with-long-double-format): Add support for the
9961 configuration option to change the default long double format on
9963 * config.gcc (powerpc*-linux*-*): Likewise.
9964 * configure: Regenerate.
9965 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9966 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9967 used without modification.
9969 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9971 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9972 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9973 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9974 MISC_BUILTIN_SPEC_BARRIER.
9975 (rs6000_init_builtins): Likewise.
9976 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9978 (speculation_barrier): New define_insn.
9979 * doc/extend.texi: Document __builtin_speculation_barrier.
9981 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9984 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9985 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9986 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9988 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9989 integral modes instead of "ss" and "sd".
9990 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9991 vectors with 32-bit and 64-bit elements.
9992 (vecdupssescalarmodesuffix): New mode attribute.
9993 (vec_dup<mode>): Use it.
9995 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9998 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9999 frame if argument is passed on stack.
10001 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10004 * ree.c (combine_reaching_defs): Optimize also
10005 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10006 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10008 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10010 PR middle-end/83189
10011 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10013 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10015 PR middle-end/83718
10016 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10017 after they are computed.
10019 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10021 PR tree-optimization/83695
10022 * gimple-loop-linterchange.cc
10023 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10024 reset cached scev information after interchange.
10025 (pass_linterchange::execute): Remove call to scev_reset_htab.
10027 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10029 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10030 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10031 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10032 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10033 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10034 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10035 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10036 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10037 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10038 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10039 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10040 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10041 (V_lane_reg): Likewise.
10042 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10044 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10045 (vfmal_lane_low<mode>_intrinsic,
10046 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10047 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10048 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10049 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10050 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10051 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10053 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10055 * config/arm/arm-cpus.in (fp16fml): New feature.
10056 (ALL_SIMD): Add fp16fml.
10057 (armv8.2-a): Add fp16fml as an option.
10058 (armv8.3-a): Likewise.
10059 (armv8.4-a): Add fp16fml as part of fp16.
10060 * config/arm/arm.h (TARGET_FP16FML): Define.
10061 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10063 * config/arm/arm-modes.def (V2HF): Define.
10064 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10065 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10066 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10067 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10068 vfmsl_low, vfmsl_high): New set of builtins.
10069 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10070 (vfml_op): New code attribute.
10071 (VFMLHALVES): New int iterator.
10072 (VFML, VFMLSEL): New mode attributes.
10073 (V_reg): Define mapping for V2HF.
10074 (V_hi, V_lo): New mode attributes.
10075 (VF_constraint): Likewise.
10076 (vfml_half, vfml_half_selector): New int attributes.
10077 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10079 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10080 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10082 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10083 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10084 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10085 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10087 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10088 Document new effective target and option set.
10090 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10092 * config/arm/arm-cpus.in (armv8_4): New feature.
10093 (ARMv8_4a): New fgroup.
10094 (armv8.4-a): New arch.
10095 * config/arm/arm-tables.opt: Regenerate.
10096 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10097 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10098 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10099 Add matching rules for -march=armv8.4-a and extensions.
10100 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10102 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10105 * config/rx/rx.md (BW): New mode attribute.
10106 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10108 2018-01-11 Richard Biener <rguenther@suse.de>
10110 PR tree-optimization/83435
10111 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10112 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10113 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10115 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10116 Alan Hayward <alan.hayward@arm.com>
10117 David Sherwood <david.sherwood@arm.com>
10119 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10121 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10122 (aarch64_print_address_internal): Use it to check for a zero offset.
10124 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10125 Alan Hayward <alan.hayward@arm.com>
10126 David Sherwood <david.sherwood@arm.com>
10128 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10129 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10130 Return a poly_int64 rather than a HOST_WIDE_INT.
10131 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10132 rather than a HOST_WIDE_INT.
10133 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10134 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10135 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10136 final_offset from HOST_WIDE_INT to poly_int64.
10137 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10138 to_constant when getting the number of units in an Advanced SIMD
10140 (aarch64_builtin_vectorized_function): Check for a constant number
10142 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10144 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10145 attribute instead of GET_MODE_NUNITS.
10146 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10147 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10148 GET_MODE_SIZE for fixed-size registers.
10149 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10150 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10151 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10152 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10153 (aarch64_print_operand, aarch64_print_address_internal)
10154 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10155 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10156 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10157 Handle polynomial GET_MODE_SIZE.
10158 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10159 wider than SImode without modification.
10160 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10161 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10162 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10163 passing and returning SVE modes.
10164 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10165 rather than GEN_INT.
10166 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10167 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10168 (aarch64_allocate_and_probe_stack_space): Likewise.
10169 (aarch64_layout_frame): Cope with polynomial offsets.
10170 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10171 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10172 polynomial offsets.
10173 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10174 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10175 poly_int64 rather than a HOST_WIDE_INT.
10176 (aarch64_get_separate_components, aarch64_process_components)
10177 (aarch64_expand_prologue, aarch64_expand_epilogue)
10178 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10179 (aarch64_anchor_offset): New function, split out from...
10180 (aarch64_legitimize_address): ...here.
10181 (aarch64_builtin_vectorization_cost): Handle polynomial
10182 TYPE_VECTOR_SUBPARTS.
10183 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10185 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10186 number of elements from the PARALLEL rather than the mode.
10187 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10188 rather than GET_MODE_BITSIZE.
10189 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10190 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10191 (aarch64_expand_vec_perm_const_1): Handle polynomial
10192 d->perm.length () and d->perm elements.
10193 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10194 Apply to_constant to d->perm elements.
10195 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10196 polynomial CONST_VECTOR_NUNITS.
10197 (aarch64_move_pointer): Take amount as a poly_int64 rather
10199 (aarch64_progress_pointer): Avoid temporary variable.
10200 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10201 the mode attribute instead of GET_MODE.
10203 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10204 Alan Hayward <alan.hayward@arm.com>
10205 David Sherwood <david.sherwood@arm.com>
10207 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10208 x exists before using it.
10209 (aarch64_add_constant_internal): Rename to...
10210 (aarch64_add_offset_1): ...this. Replace regnum with separate
10211 src and dest rtxes. Handle the case in which they're different,
10212 including when the offset is zero. Replace scratchreg with an rtx.
10213 Use 2 additions if there is no spare register into which we can
10214 move a 16-bit constant.
10215 (aarch64_add_constant): Delete.
10216 (aarch64_add_offset): Replace reg with separate src and dest
10217 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10218 Use aarch64_add_offset_1.
10219 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10220 an rtx rather than an int. Take the delta as a poly_int64
10221 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10222 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10223 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10224 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10225 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10226 and aarch64_add_sp.
10227 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10228 aarch64_add_constant.
10230 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10232 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10233 Use scalar_float_mode.
10235 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10237 * config/aarch64/aarch64-simd.md
10238 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10239 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10240 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10241 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10242 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10243 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10244 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10245 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10246 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10247 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10249 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10252 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10253 targ_options->x_arm_arch_string is non NULL.
10255 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10257 * config/aarch64/aarch64.h
10258 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10260 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10263 * expmed.c (emit_store_flag_force): Swap if const op0
10264 and change VOIDmode to mode of op0.
10266 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10268 PR rtl-optimization/83761
10269 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10270 than bytes to mode_for_size.
10272 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10274 PR middle-end/83189
10275 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10276 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10279 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10281 PR middle-end/83575
10282 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10283 when in layout mode.
10284 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10285 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10288 2018-01-10 Michael Collison <michael.collison@arm.com>
10290 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10291 * config/aarch64/aarch64-option-extension.def: Add
10292 AARCH64_OPT_EXTENSION of 'fp16fml'.
10293 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10294 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10295 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10296 * config/aarch64/constraints.md (Ui7): New constraint.
10297 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10298 (VFMLA_SEL_W): Ditto.
10301 (VFMLA16_LOW): New int iterator.
10302 (VFMLA16_HIGH): Ditto.
10303 (UNSPEC_FMLAL): New unspec.
10304 (UNSPEC_FMLSL): Ditto.
10305 (UNSPEC_FMLAL2): Ditto.
10306 (UNSPEC_FMLSL2): Ditto.
10307 (f16mac): New code attribute.
10308 * config/aarch64/aarch64-simd-builtins.def
10309 (aarch64_fmlal_lowv2sf): Ditto.
10310 (aarch64_fmlsl_lowv2sf): Ditto.
10311 (aarch64_fmlalq_lowv4sf): Ditto.
10312 (aarch64_fmlslq_lowv4sf): Ditto.
10313 (aarch64_fmlal_highv2sf): Ditto.
10314 (aarch64_fmlsl_highv2sf): Ditto.
10315 (aarch64_fmlalq_highv4sf): Ditto.
10316 (aarch64_fmlslq_highv4sf): Ditto.
10317 (aarch64_fmlal_lane_lowv2sf): Ditto.
10318 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10319 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10320 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10321 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10322 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10323 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10324 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10325 (aarch64_fmlal_lane_highv2sf): Ditto.
10326 (aarch64_fmlsl_lane_highv2sf): Ditto.
10327 (aarch64_fmlal_laneq_highv2sf): Ditto.
10328 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10329 (aarch64_fmlalq_lane_highv4sf): Ditto.
10330 (aarch64_fmlsl_lane_highv4sf): Ditto.
10331 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10332 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10333 * config/aarch64/aarch64-simd.md:
10334 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10335 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10336 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10337 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10338 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10339 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10340 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10341 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10342 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10343 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10344 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10345 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10346 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10347 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10348 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10349 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10350 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10351 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10352 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10353 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10354 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10355 (vfmlsl_low_u32): Ditto.
10356 (vfmlalq_low_u32): Ditto.
10357 (vfmlslq_low_u32): Ditto.
10358 (vfmlal_high_u32): Ditto.
10359 (vfmlsl_high_u32): Ditto.
10360 (vfmlalq_high_u32): Ditto.
10361 (vfmlslq_high_u32): Ditto.
10362 (vfmlal_lane_low_u32): Ditto.
10363 (vfmlsl_lane_low_u32): Ditto.
10364 (vfmlal_laneq_low_u32): Ditto.
10365 (vfmlsl_laneq_low_u32): Ditto.
10366 (vfmlalq_lane_low_u32): Ditto.
10367 (vfmlslq_lane_low_u32): Ditto.
10368 (vfmlalq_laneq_low_u32): Ditto.
10369 (vfmlslq_laneq_low_u32): Ditto.
10370 (vfmlal_lane_high_u32): Ditto.
10371 (vfmlsl_lane_high_u32): Ditto.
10372 (vfmlal_laneq_high_u32): Ditto.
10373 (vfmlsl_laneq_high_u32): Ditto.
10374 (vfmlalq_lane_high_u32): Ditto.
10375 (vfmlslq_lane_high_u32): Ditto.
10376 (vfmlalq_laneq_high_u32): Ditto.
10377 (vfmlslq_laneq_high_u32): Ditto.
10378 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10379 (AARCH64_FL_FOR_ARCH8_4): New.
10380 (AARCH64_ISA_F16FML): New ISA flag.
10381 (TARGET_F16FML): New feature flag for fp16fml.
10382 (doc/invoke.texi): Document new fp16fml option.
10384 2018-01-10 Michael Collison <michael.collison@arm.com>
10386 * config/aarch64/aarch64-builtins.c:
10387 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10388 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10389 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10390 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10391 (AARCH64_ISA_SHA3): New ISA flag.
10392 (TARGET_SHA3): New feature flag for sha3.
10393 * config/aarch64/iterators.md (sha512_op): New int attribute.
10394 (CRYPTO_SHA512): New int iterator.
10395 (UNSPEC_SHA512H): New unspec.
10396 (UNSPEC_SHA512H2): Ditto.
10397 (UNSPEC_SHA512SU0): Ditto.
10398 (UNSPEC_SHA512SU1): Ditto.
10399 * config/aarch64/aarch64-simd-builtins.def
10400 (aarch64_crypto_sha512hqv2di): New builtin.
10401 (aarch64_crypto_sha512h2qv2di): Ditto.
10402 (aarch64_crypto_sha512su0qv2di): Ditto.
10403 (aarch64_crypto_sha512su1qv2di): Ditto.
10404 (aarch64_eor3qv8hi): Ditto.
10405 (aarch64_rax1qv2di): Ditto.
10406 (aarch64_xarqv2di): Ditto.
10407 (aarch64_bcaxqv8hi): Ditto.
10408 * config/aarch64/aarch64-simd.md:
10409 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10410 (aarch64_crypto_sha512su0qv2di): Ditto.
10411 (aarch64_crypto_sha512su1qv2di): Ditto.
10412 (aarch64_eor3qv8hi): Ditto.
10413 (aarch64_rax1qv2di): Ditto.
10414 (aarch64_xarqv2di): Ditto.
10415 (aarch64_bcaxqv8hi): Ditto.
10416 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10417 (vsha512h2q_u64): Ditto.
10418 (vsha512su0q_u64): Ditto.
10419 (vsha512su1q_u64): Ditto.
10420 (veor3q_u16): Ditto.
10421 (vrax1q_u64): Ditto.
10422 (vxarq_u64): Ditto.
10423 (vbcaxq_u16): Ditto.
10424 * config/arm/types.md (crypto_sha512): New type attribute.
10425 (crypto_sha3): Ditto.
10426 (doc/invoke.texi): Document new sha3 option.
10428 2018-01-10 Michael Collison <michael.collison@arm.com>
10430 * config/aarch64/aarch64-builtins.c:
10431 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10432 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10433 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10434 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10435 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10436 (AARCH64_ISA_SM4): New ISA flag.
10437 (TARGET_SM4): New feature flag for sm4.
10438 * config/aarch64/aarch64-simd-builtins.def
10439 (aarch64_sm3ss1qv4si): Ditto.
10440 (aarch64_sm3tt1aq4si): Ditto.
10441 (aarch64_sm3tt1bq4si): Ditto.
10442 (aarch64_sm3tt2aq4si): Ditto.
10443 (aarch64_sm3tt2bq4si): Ditto.
10444 (aarch64_sm3partw1qv4si): Ditto.
10445 (aarch64_sm3partw2qv4si): Ditto.
10446 (aarch64_sm4eqv4si): Ditto.
10447 (aarch64_sm4ekeyqv4si): Ditto.
10448 * config/aarch64/aarch64-simd.md:
10449 (aarch64_sm3ss1qv4si): Ditto.
10450 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10451 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10452 (aarch64_sm4eqv4si): Ditto.
10453 (aarch64_sm4ekeyqv4si): Ditto.
10454 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10455 (sm3part_op): Ditto.
10456 (CRYPTO_SM3TT): Ditto.
10457 (CRYPTO_SM3PART): Ditto.
10458 (UNSPEC_SM3SS1): New unspec.
10459 (UNSPEC_SM3TT1A): Ditto.
10460 (UNSPEC_SM3TT1B): Ditto.
10461 (UNSPEC_SM3TT2A): Ditto.
10462 (UNSPEC_SM3TT2B): Ditto.
10463 (UNSPEC_SM3PARTW1): Ditto.
10464 (UNSPEC_SM3PARTW2): Ditto.
10465 (UNSPEC_SM4E): Ditto.
10466 (UNSPEC_SM4EKEY): Ditto.
10467 * config/aarch64/constraints.md (Ui2): New constraint.
10468 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10469 * config/arm/types.md (crypto_sm3): New type attribute.
10470 (crypto_sm4): Ditto.
10471 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10472 (vsm3tt1aq_u32): Ditto.
10473 (vsm3tt1bq_u32): Ditto.
10474 (vsm3tt2aq_u32): Ditto.
10475 (vsm3tt2bq_u32): Ditto.
10476 (vsm3partw1q_u32): Ditto.
10477 (vsm3partw2q_u32): Ditto.
10478 (vsm4eq_u32): Ditto.
10479 (vsm4ekeyq_u32): Ditto.
10480 (doc/invoke.texi): Document new sm4 option.
10482 2018-01-10 Michael Collison <michael.collison@arm.com>
10484 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10485 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10486 (AARCH64_FL_FOR_ARCH8_4): New.
10487 (AARCH64_FL_V8_4): New flag.
10488 (doc/invoke.texi): Document new armv8.4-a option.
10490 2018-01-10 Michael Collison <michael.collison@arm.com>
10492 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10493 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10494 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10495 * config/aarch64/aarch64-option-extension.def: Add
10496 AARCH64_OPT_EXTENSION of 'sha2'.
10497 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10498 (crypto): Disable sha2 and aes if crypto disabled.
10499 (crypto): Enable aes and sha2 if enabled.
10500 (simd): Disable sha2 and aes if simd disabled.
10501 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10503 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10504 (TARGET_SHA2): New feature flag for sha2.
10505 (TARGET_AES): New feature flag for aes.
10506 * config/aarch64/aarch64-simd.md:
10507 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10508 conditional on TARGET_AES.
10509 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10510 (aarch64_crypto_sha1hsi): Make pattern conditional
10512 (aarch64_crypto_sha1hv4si): Ditto.
10513 (aarch64_be_crypto_sha1hv4si): Ditto.
10514 (aarch64_crypto_sha1su1v4si): Ditto.
10515 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10516 (aarch64_crypto_sha1su0v4si): Ditto.
10517 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10518 (aarch64_crypto_sha256su0v4si): Ditto.
10519 (aarch64_crypto_sha256su1v4si): Ditto.
10520 (doc/invoke.texi): Document new aes and sha2 options.
10522 2018-01-10 Martin Sebor <msebor@redhat.com>
10524 PR tree-optimization/83781
10525 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10528 2018-01-11 Martin Sebor <msebor@gmail.com>
10529 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10531 PR tree-optimization/83501
10532 PR tree-optimization/81703
10534 * tree-ssa-strlen.c (get_string_cst): Rename...
10535 (get_string_len): ...to this. Handle global constants.
10536 (handle_char_store): Adjust.
10538 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10539 Jim Wilson <jimw@sifive.com>
10541 * config/riscv/riscv-protos.h (riscv_output_return): New.
10542 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10543 (riscv_attribute_table, riscv_output_return),
10544 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10545 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10546 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10548 (riscv_expand_prologue): Add early return for naked function.
10549 (riscv_expand_epilogue): Likewise.
10550 (riscv_function_ok_for_sibcall): Return false for naked function.
10551 (riscv_set_current_function): New.
10552 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10553 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10554 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10555 * doc/extend.texi (RISC-V Function Attributes): New.
10557 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10559 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10560 check for 128-bit long double before checking TCmode.
10561 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10562 128-bit long doubles before checking TFmode or TCmode.
10563 (FLOAT128_IBM_P): Likewise.
10565 2018-01-10 Martin Sebor <msebor@redhat.com>
10567 PR tree-optimization/83671
10568 * builtins.c (c_strlen): Unconditionally return zero for the empty
10570 Use -Warray-bounds for warnings.
10571 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10572 for non-constant array indices with COMPONENT_REF, arrays of
10573 arrays, and pointers to arrays.
10574 (gimple_fold_builtin_strlen): Determine and set length range for
10575 non-constant character arrays.
10577 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10579 PR middle-end/81897
10580 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10583 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10585 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10587 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10590 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10591 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10592 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10593 indexed_or_indirect_operand predicate.
10594 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10595 (*vsx_le_perm_load_v8hi): Likewise.
10596 (*vsx_le_perm_load_v16qi): Likewise.
10597 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10598 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10599 (*vsx_le_perm_store_v8hi): Likewise.
10600 (*vsx_le_perm_store_v16qi): Likewise.
10601 (eight unnamed splitters): Likewise.
10603 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10605 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10606 * config/rs6000/emmintrin.h: Likewise.
10607 * config/rs6000/mmintrin.h: Likewise.
10608 * config/rs6000/xmmintrin.h: Likewise.
10610 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10613 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10615 * tree.c (tree_nop_conversion): Return true for location wrapper
10617 (maybe_wrap_with_location): New function.
10618 (selftest::check_strip_nops): New function.
10619 (selftest::test_location_wrappers): New function.
10620 (selftest::tree_c_tests): Call it.
10621 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10622 (maybe_wrap_with_location): New decl.
10623 (EXPR_LOCATION_WRAPPER_P): New macro.
10624 (location_wrapper_p): New inline function.
10625 (tree_strip_any_location_wrapper): New inline function.
10627 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10630 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10631 stack_realign_offset for the largest alignment of stack slot
10633 (ix86_find_max_used_stack_alignment): New function.
10634 (ix86_finalize_stack_frame_flags): Use it. Set
10635 max_used_stack_alignment if we don't realign stack.
10636 * config/i386/i386.h (machine_function): Add
10637 max_used_stack_alignment.
10639 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10641 * config/arm/arm.opt (-mbranch-cost): New option.
10642 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10645 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10648 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10649 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10651 2018-01-10 Richard Biener <rguenther@suse.de>
10654 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10655 early out so it also covers the case where we have a non-NULL
10658 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10660 PR tree-optimization/83753
10661 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10662 for non-strided grouped accesses if the number of elements is 1.
10664 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10667 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10668 * i386.h (TARGET_USE_GATHER): Define.
10669 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10671 2018-01-10 Martin Liska <mliska@suse.cz>
10674 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10675 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10677 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10678 CLEANUP_NO_PARTITIONING is not set.
10680 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10682 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10683 for vectors, as a partial revert of r254296.
10684 * rtl.h (const_vec_p): Delete.
10685 (const_vec_duplicate_p): Don't test for vector CONSTs.
10686 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10687 * expmed.c (make_tree): Likewise.
10690 * common.md (E, F): Use CONSTANT_P instead of checking for
10692 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10693 checking for CONST_VECTOR.
10695 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10697 PR middle-end/83575
10698 * predict.c (force_edge_cold): Handle in more sane way edges
10699 with no prediction.
10701 2018-01-09 Carl Love <cel@us.ibm.com>
10703 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10705 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10706 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10707 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10708 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10709 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10710 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10711 * config/rs6000/rs6000-protos.h: Add extern defition for
10712 rs6000_generate_float2_double_code.
10713 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10715 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10716 (float2_v2df): Add define_expand.
10718 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10721 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10722 op_mode in the force_to_mode call.
10724 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10726 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10727 instead of checking each element individually.
10728 (aarch64_evpc_uzp): Likewise.
10729 (aarch64_evpc_zip): Likewise.
10730 (aarch64_evpc_ext): Likewise.
10731 (aarch64_evpc_rev): Likewise.
10732 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10733 instead of checking each element individually. Return true without
10735 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10736 whether all selected elements come from the same input, instead of
10737 checking each element individually. Remove calls to gen_rtx_REG,
10738 start_sequence and end_sequence and instead assert that no rtl is
10741 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10743 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10744 order of HIGH and CONST checks.
10746 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10748 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10749 if the destination isn't an SSA_NAME.
10751 2018-01-09 Richard Biener <rguenther@suse.de>
10753 PR tree-optimization/83668
10754 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10756 (canonicalize_loop_form): ... here, renamed from ...
10757 (canonicalize_loop_closed_ssa_form): ... this and amended to
10758 swap successor edges for loop exit blocks to make us use
10759 the RPO order we need for initial schedule generation.
10761 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10763 PR tree-optimization/64811
10764 * match.pd: When optimizing comparisons with Inf, avoid
10765 introducing or losing exceptions from comparisons with NaN.
10767 2018-01-09 Martin Liska <mliska@suse.cz>
10770 * asan.c (shadow_mem_size): Add gcc_assert.
10772 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10774 Don't save registers in main().
10777 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10778 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10779 * config/avr/avr.c (avr_set_current_function): Don't error if
10780 naked, OS_task or OS_main are specified at the same time.
10781 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10783 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10785 * common/config/avr/avr-common.c (avr_option_optimization_table):
10786 Switch on -mmain-is-OS_task for optimizing compilations.
10788 2018-01-09 Richard Biener <rguenther@suse.de>
10790 PR tree-optimization/83572
10791 * graphite.c: Include cfganal.h.
10792 (graphite_transform_loops): Connect infinite loops to exit
10793 and remove fake edges at the end.
10795 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10797 * ipa-inline.c (edge_badness): Revert accidental checkin.
10799 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10802 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10803 symbols; not inline clones.
10805 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10808 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10809 hard registers. Formatting fixes.
10811 PR preprocessor/83722
10812 * gcc.c (try_generate_repro): Pass
10813 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10814 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10817 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10818 Kito Cheng <kito.cheng@gmail.com>
10820 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10821 (riscv_leaf_function_p): Delete.
10822 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10824 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10826 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10828 (do_ifelse): New function.
10829 (do_isel): New function.
10830 (do_sub3): New function.
10831 (do_add3): New function.
10832 (do_load_mask_compare): New function.
10833 (do_overlap_load_compare): New function.
10834 (expand_compare_loop): New function.
10835 (expand_block_compare): Call expand_compare_loop() when appropriate.
10836 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10837 option description.
10838 (-mblock-compare-inline-loop-limit): New option.
10840 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10843 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10844 Reverse order of second and third operands in first alternative.
10845 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10846 of first and second elements in UNSPEC_VPERMR vector.
10847 (altivec_expand_vec_perm_le): Likewise.
10849 2018-01-08 Jeff Law <law@redhat.com>
10851 PR rtl-optimizatin/81308
10852 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10853 (process_switch): If group_case_labels makes a change, then set
10855 (pass_convert_switch::execute): If a switch is converted, then
10856 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10858 PR rtl-optimization/81308
10859 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10862 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10864 PR target/83663 - Revert r255946
10865 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10866 generation for cases where splatting a value is not useful.
10867 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10868 across a vec_duplicate and a paradoxical subreg forming a vector
10869 mode to a vec_concat.
10871 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10873 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10874 -march=armv8.3-a variants.
10875 * config/arm/t-multilib: Likewise.
10876 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10878 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10880 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10882 (cceq_ior_compare_complement): Give it a name so I can use it, and
10883 change boolean_or_operator predicate to boolean_operator so it can
10884 be used to generate a crand.
10885 (eqne): New code iterator.
10886 (bd/bd_neg): New code_attrs.
10887 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10888 a single define_insn.
10889 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10890 decrement (bdnzt/bdnzf/bdzt/bdzf).
10891 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10892 with the new names of the branch decrement patterns, and added the
10893 names of the branch decrement conditional patterns.
10895 2018-01-08 Richard Biener <rguenther@suse.de>
10897 PR tree-optimization/83563
10898 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10901 2018-01-08 Richard Biener <rguenther@suse.de>
10903 PR middle-end/83713
10904 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10906 2018-01-08 Richard Biener <rguenther@suse.de>
10908 PR tree-optimization/83685
10909 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10910 references to abnormals.
10912 2018-01-08 Richard Biener <rguenther@suse.de>
10915 * dwarf2out.c (output_indirect_strings): Handle empty
10916 skeleton_debug_str_hash.
10917 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10919 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10921 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10922 (emit_store_direct): Likewise.
10923 (arc_trampoline_adjust_address): Likewise.
10924 (arc_asm_trampoline_template): New function.
10925 (arc_initialize_trampoline): Use asm_trampoline_template.
10926 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10927 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10928 * config/arc/arc.md (flush_icache): Delete pattern.
10930 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10932 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10933 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10936 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10939 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10940 by not USED_FOR_TARGET.
10941 (make_pass_resolve_sw_modes): Likewise.
10943 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10945 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10948 2018-01-08 Richard Biener <rguenther@suse.de>
10950 PR middle-end/83580
10951 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10953 2018-01-08 Richard Biener <rguenther@suse.de>
10955 PR middle-end/83517
10956 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10958 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10960 PR middle-end/81897
10961 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10962 basic blocks with a small number of successors.
10963 (convert_control_dep_chain_into_preds): Improve handling of
10965 (dump_predicates): Split apart into...
10966 (dump_pred_chain): ...here...
10967 (dump_pred_info): ...and here.
10968 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10969 (can_chain_union_be_invalidated_p): Improve check for invalidation
10971 (uninit_uses_cannot_happen): Avoid unnecessary if
10972 convert_control_dep_chain_into_preds yielded nothing.
10974 2018-01-06 Martin Sebor <msebor@redhat.com>
10976 PR tree-optimization/83640
10977 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10978 subtracting negative offset from size.
10979 (builtin_access::overlap): Adjust offset bounds of the access to fall
10980 within the size of the object if possible.
10982 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10984 PR rtl-optimization/83699
10985 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10986 extract_bit_field_as_subreg to cases in which the extracted
10987 value is also a vector.
10989 * lra-constraints.c (process_alt_operands): Test for the equivalence
10990 substitutions when detecting a possible reload cycle.
10992 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10995 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10996 by default if flag_selective_schedling{,2}. Formatting fixes.
10998 PR rtl-optimization/83682
10999 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11000 if it has non-VECTOR_MODE element mode.
11001 (vec_duplicate_p): Likewise.
11003 PR middle-end/83694
11004 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11005 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11007 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11010 * config/i386/i386-builtin.def
11011 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11012 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11013 Require also OPTION_MASK_ISA_AVX512F in addition to
11014 OPTION_MASK_ISA_GFNI.
11015 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11016 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11017 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11018 to OPTION_MASK_ISA_GFNI.
11019 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11020 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11021 OPTION_MASK_ISA_AVX512BW.
11022 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11023 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11024 addition to OPTION_MASK_ISA_GFNI.
11025 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11026 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11027 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11028 to OPTION_MASK_ISA_GFNI.
11029 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11030 a requirement for all ISAs rather than any of them with a few
11032 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11034 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11035 bitmasks to be enabled with 3 exceptions, instead of requiring any
11036 enabled ISA with lots of exceptions.
11037 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11038 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11039 Change avx512bw in isa attribute to avx512f.
11040 * config/i386/sgxintrin.h: Add license boilerplate.
11041 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11042 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11043 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11044 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11046 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11047 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11048 temporarily sse2 rather than sse if not enabled already.
11051 * config/i386/sse.md (VI248_VLBW): Rename to ...
11052 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11053 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11054 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11055 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11056 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11057 mode iterator instead of VI248_VLBW.
11059 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11061 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11062 (record_modified): Skip clobbers; add debug output.
11063 (param_change_prob): Use sreal frequencies.
11065 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11067 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11068 punt for user-aligned variables.
11070 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11072 * tree-chrec.c (chrec_contains_symbols): Return true for
11075 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11078 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11079 of (x|y) == x for BICS pattern.
11081 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11083 PR tree-optimization/83605
11084 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11085 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11088 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11090 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11091 * config/epiphany/rtems.h: New file.
11093 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11094 Uros Bizjak <ubizjak@gmail.com>
11097 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11098 QIreg_operand instead of register_operand predicate.
11099 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11100 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11101 comments instead of -fmitigate[-_]rop.
11103 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11106 * cgraphunit.c (symbol_table::compile): Switch to text_section
11107 before calling assembly_start debug hook.
11108 * run-rtl-passes.c (run_rtl_passes): Likewise.
11111 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11113 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11114 range_int_cst_p rather than !symbolic_range_p before calling
11115 extract_range_from_multiplicative_op_1.
11117 2018-01-04 Jeff Law <law@redhat.com>
11119 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11120 redundant test in assertion.
11122 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11124 * doc/rtl.texi: Document machine_mode wrapper classes.
11126 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11128 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11129 using tree_to_uhwi.
11131 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11133 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11134 the VEC_PERM_EXPR fold to fail.
11136 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11139 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11140 to switched_sections.
11142 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11145 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11146 test for d.testing.
11148 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11151 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11152 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11154 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11157 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11158 is BLKmode and bitpos not zero or mode change is needed.
11160 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11163 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11166 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11169 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11170 instead of MULT rtx. Update all corresponding splitters.
11171 (*saddl_se): Ditto.
11172 (*ssub<modesuffix>): Ditto.
11173 (*ssubl_se): Ditto.
11174 (*cmp_sadd_di): Update split patterns.
11175 (*cmp_sadd_si): Ditto.
11176 (*cmp_sadd_sidi): Ditto.
11177 (*cmp_ssub_di): Ditto.
11178 (*cmp_ssub_si): Ditto.
11179 (*cmp_ssub_sidi): Ditto.
11180 * config/alpha/predicates.md (const23_operand): New predicate.
11181 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11182 Look for ASHIFT, not MULT inner operand.
11183 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11185 2018-01-04 Martin Liska <mliska@suse.cz>
11187 PR gcov-profile/83669
11188 * gcov.c (output_intermediate_file): Add version to intermediate
11190 * doc/gcov.texi: Document new field 'version' in intermediate
11191 file format. Fix location of '-k' option of gcov command.
11193 2018-01-04 Martin Liska <mliska@suse.cz>
11196 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11198 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11200 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11202 2018-01-03 Martin Sebor <msebor@redhat.com>
11204 PR tree-optimization/83655
11205 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11206 checking calls with invalid arguments.
11208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11210 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11211 (vectorizable_mask_load_store): Delete.
11212 (vectorizable_call): Return false for masked loads and stores.
11213 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11214 instead of gimple_assign_rhs1.
11215 (vectorizable_load): Handle IFN_MASK_LOAD.
11216 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11220 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11222 (vectorizable_mask_load_store): ...here.
11223 (vectorizable_load): ...and here.
11225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11227 * tree-vect-stmts.c (vect_build_all_ones_mask)
11228 (vect_build_zero_merge_argument): New functions, split out from...
11229 (vectorizable_load): ...here.
11231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11233 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11235 (vectorizable_mask_load_store): ...here.
11236 (vectorizable_store): ...and here.
11238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11240 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11242 (vectorizable_mask_load_store): ...here.
11244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11246 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11247 (vect_model_store_cost): Take a vec_load_store_type instead of a
11249 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11250 (vect_model_store_cost): Take a vec_load_store_type instead of a
11252 (vectorizable_mask_load_store): Update accordingly.
11253 (vectorizable_store): Likewise.
11254 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11258 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11259 IFN_MASK_LOAD calls here rather than...
11260 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11263 Alan Hayward <alan.hayward@arm.com>
11264 David Sherwood <david.sherwood@arm.com>
11266 * expmed.c (extract_bit_field_1): For vector extracts,
11267 fall back to extract_bit_field_as_subreg if vec_extract
11270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11271 Alan Hayward <alan.hayward@arm.com>
11272 David Sherwood <david.sherwood@arm.com>
11274 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11275 they are variable or constant sized.
11276 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11277 slots for constant-sized data.
11279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11280 Alan Hayward <alan.hayward@arm.com>
11281 David Sherwood <david.sherwood@arm.com>
11283 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11284 handling COND_EXPRs with boolean comparisons, try to find a better
11285 basis for the mask type than the boolean itself.
11287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11289 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11290 is calculated and how it can be overridden.
11291 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11292 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11294 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11298 Alan Hayward <alan.hayward@arm.com>
11299 David Sherwood <david.sherwood@arm.com>
11301 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11302 Remove the mode argument.
11303 (aarch64_simd_valid_immediate): Remove the mode and inverse
11305 * config/aarch64/iterators.md (bitsize): New iterator.
11306 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11307 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11308 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11309 aarch64_simd_valid_immediate.
11310 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11311 (aarch64_reg_or_bic_imm): Likewise.
11312 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11313 with an insn_type enum and msl with a modifier_type enum.
11314 Replace element_width with a scalar_mode. Change the shift
11315 to unsigned int. Add constructors for scalar_float_mode and
11316 scalar_int_mode elements.
11317 (aarch64_vect_float_const_representable_p): Delete.
11318 (aarch64_can_const_movi_rtx_p)
11319 (aarch64_simd_scalar_immediate_valid_for_move)
11320 (aarch64_simd_make_constant): Update call to
11321 aarch64_simd_valid_immediate.
11322 (aarch64_advsimd_valid_immediate_hs): New function.
11323 (aarch64_advsimd_valid_immediate): Likewise.
11324 (aarch64_simd_valid_immediate): Remove mode and inverse
11325 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11326 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11327 and aarch64_float_const_representable_p on the result.
11328 (aarch64_output_simd_mov_immediate): Remove mode argument.
11329 Update call to aarch64_simd_valid_immediate and use of
11330 simd_immediate_info.
11331 (aarch64_output_scalar_simd_mov_immediate): Update call
11334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11335 Alan Hayward <alan.hayward@arm.com>
11336 David Sherwood <david.sherwood@arm.com>
11338 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11339 (mode_nunits): Likewise CONST_MODE_NUNITS.
11340 * machmode.def (ADJUST_NUNITS): Document.
11341 * genmodes.c (mode_data::need_nunits_adj): New field.
11342 (blank_mode): Update accordingly.
11343 (adj_nunits): New variable.
11344 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11346 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11347 listed in adj_nunits.
11348 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11349 listed in adj_nunits. Don't emit case statements for such modes.
11350 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11351 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11352 nothing if adj_nunits is nonnull.
11353 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11354 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11355 (emit_mode_fbit): Update use of print_maybe_const_decl.
11356 (emit_move_size): Likewise. Treat the array as non-const
11358 (emit_mode_adjustments): Handle adj_nunits.
11360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11362 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11363 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11364 (VECTOR_MODES): Use it.
11365 (make_vector_modes): Take the prefix as an argument.
11367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11368 Alan Hayward <alan.hayward@arm.com>
11369 David Sherwood <david.sherwood@arm.com>
11371 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11372 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11373 for MODE_VECTOR_BOOL.
11374 * machmode.def (VECTOR_BOOL_MODE): Document.
11375 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11376 (make_vector_bool_mode): New function.
11377 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11379 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11380 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11382 * stor-layout.c (int_mode_for_mode): Likewise.
11383 * tree.c (build_vector_type_for_mode): Likewise.
11384 * varasm.c (output_constant_pool_2): Likewise.
11385 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11386 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11387 for MODE_VECTOR_BOOL.
11388 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11389 of mode class checks.
11390 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11391 instead of a list of mode class checks.
11392 (expand_vector_scalar_condition): Likewise.
11393 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11396 Alan Hayward <alan.hayward@arm.com>
11397 David Sherwood <david.sherwood@arm.com>
11399 * machmode.h (mode_size): Change from unsigned short to
11401 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11402 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11403 or if measurement_type is not polynomial.
11404 (fixed_size_mode::includes_p): Check for constant-sized modes.
11405 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11406 return a poly_uint16 rather than an unsigned short.
11407 (emit_mode_size): Change the type of mode_size from unsigned short
11408 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11409 (emit_mode_adjustments): Cope with polynomial vector sizes.
11410 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11412 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11414 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11415 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11416 * caller-save.c (setup_save_areas): Likewise.
11417 (replace_reg_with_saved_mem): Likewise.
11418 * calls.c (emit_library_call_value_1): Likewise.
11419 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11420 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11421 (gen_lowpart_for_combine): Likewise.
11422 * convert.c (convert_to_integer_1): Likewise.
11423 * cse.c (equiv_constant, cse_insn): Likewise.
11424 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11425 (cselib_subst_to_values): Likewise.
11426 * dce.c (word_dce_process_block): Likewise.
11427 * df-problems.c (df_word_lr_mark_ref): Likewise.
11428 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11429 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11430 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11431 (rtl_for_decl_location): Likewise.
11432 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11433 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11434 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11435 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11436 (expand_expr_real_1): Likewise.
11437 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11438 (pad_below): Likewise.
11439 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11440 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11441 * ira.c (get_subreg_tracking_sizes): Likewise.
11442 * ira-build.c (ira_create_allocno_objects): Likewise.
11443 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11444 (ira_sort_regnos_for_alter_reg): Likewise.
11445 * ira-costs.c (record_operand_costs): Likewise.
11446 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11447 (resolve_simple_move): Likewise.
11448 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11449 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11450 (lra_constraints): Likewise.
11451 (CONST_POOL_OK_P): Reject variable-sized modes.
11452 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11453 (add_pseudo_to_slot, lra_spill): Likewise.
11454 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11455 * optabs-query.c (get_best_extraction_insn): Likewise.
11456 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11457 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11458 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11459 * recog.c (offsettable_address_addr_space_p): Likewise.
11460 * regcprop.c (maybe_mode_change): Likewise.
11461 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11462 * regrename.c (build_def_use): Likewise.
11463 * regstat.c (dump_reg_info): Likewise.
11464 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11465 (find_reloads, find_reloads_subreg_address): Likewise.
11466 * reload1.c (eliminate_regs_1): Likewise.
11467 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11468 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11469 (simplify_binary_operation_1, simplify_subreg): Likewise.
11470 * targhooks.c (default_function_arg_padding): Likewise.
11471 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11472 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11473 (verify_gimple_assign_ternary): Likewise.
11474 * tree-inline.c (estimate_move_cost): Likewise.
11475 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11476 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11477 (get_address_cost_ainc): Likewise.
11478 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11479 (vect_supportable_dr_alignment): Likewise.
11480 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11481 (vectorizable_reduction): Likewise.
11482 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11483 (vectorizable_operation, vectorizable_load): Likewise.
11484 * tree.c (build_same_sized_truth_vector_type): Likewise.
11485 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11486 * var-tracking.c (emit_note_insn_var_location): Likewise.
11487 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11488 (ADDR_VEC_ALIGN): Likewise.
11490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11491 Alan Hayward <alan.hayward@arm.com>
11492 David Sherwood <david.sherwood@arm.com>
11494 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11496 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11497 or if measurement_type is polynomial.
11498 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11499 * combine.c (make_extraction): Likewise.
11500 * dse.c (find_shift_sequence): Likewise.
11501 * dwarf2out.c (mem_loc_descriptor): Likewise.
11502 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11503 (extract_bit_field, extract_low_bits): Likewise.
11504 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11505 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11506 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11507 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11508 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11509 * reload.c (find_reloads): Likewise.
11510 * reload1.c (alter_reg): Likewise.
11511 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11512 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11513 * tree-if-conv.c (predicate_mem_writes): Likewise.
11514 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11515 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11516 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11517 * valtrack.c (dead_debug_insert_temp): Likewise.
11518 * varasm.c (mergeable_constant_section): Likewise.
11519 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11522 Alan Hayward <alan.hayward@arm.com>
11523 David Sherwood <david.sherwood@arm.com>
11525 * expr.c (expand_assignment): Cope with polynomial mode sizes
11526 when assigning to a CONCAT.
11528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11529 Alan Hayward <alan.hayward@arm.com>
11530 David Sherwood <david.sherwood@arm.com>
11532 * machmode.h (mode_precision): Change from unsigned short to
11534 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11536 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11537 or if measurement_type is not polynomial.
11538 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11539 in which the mode is already known to be a scalar_int_mode.
11540 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11541 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11543 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11544 for GET_MODE_PRECISION.
11545 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11546 for GET_MODE_PRECISION.
11547 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11549 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11550 (expand_field_assignment, make_extraction): Likewise.
11551 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11552 (get_last_value): Likewise.
11553 * convert.c (convert_to_integer_1): Likewise.
11554 * cse.c (cse_insn): Likewise.
11555 * expr.c (expand_expr_real_1): Likewise.
11556 * lra-constraints.c (simplify_operand_subreg): Likewise.
11557 * optabs-query.c (can_atomic_load_p): Likewise.
11558 * optabs.c (expand_atomic_load): Likewise.
11559 (expand_atomic_store): Likewise.
11560 * ree.c (combine_reaching_defs): Likewise.
11561 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11562 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11563 * tree.h (type_has_mode_precision_p): Likewise.
11564 * ubsan.c (instrument_si_overflow): Likewise.
11566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11567 Alan Hayward <alan.hayward@arm.com>
11568 David Sherwood <david.sherwood@arm.com>
11570 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11571 polynomial numbers of units.
11572 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11573 (valid_vector_subparts_p): New function.
11574 (build_vector_type): Remove temporary shim and take the number
11575 of units as a poly_uint64 rather than an int.
11576 (build_opaque_vector_type): Take the number of units as a
11577 poly_uint64 rather than an int.
11578 * tree.c (build_vector_from_ctor): Handle polynomial
11579 TYPE_VECTOR_SUBPARTS.
11580 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11581 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11582 (build_vector_from_val): If the number of units is variable,
11583 use build_vec_duplicate_cst for constant operands and
11584 VEC_DUPLICATE_EXPR otherwise.
11585 (make_vector_type): Remove temporary is_constant ().
11586 (build_vector_type, build_opaque_vector_type): Take the number of
11587 units as a poly_uint64 rather than an int.
11588 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11590 * cfgexpand.c (expand_debug_expr): Likewise.
11591 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11592 (store_constructor, expand_expr_real_1): Likewise.
11593 (const_scalar_mask_from_tree): Likewise.
11594 * fold-const-call.c (fold_const_reduction): Likewise.
11595 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11596 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11597 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11598 (fold_relational_const): Likewise.
11599 (native_interpret_vector): Likewise. Change the size from an
11600 int to an unsigned int.
11601 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11602 TYPE_VECTOR_SUBPARTS.
11603 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11604 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11605 duplicating a non-constant operand into a variable-length vector.
11606 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11607 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11608 * ipa-icf.c (sem_variable::equals): Likewise.
11609 * match.pd: Likewise.
11610 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11611 * print-tree.c (print_node): Likewise.
11612 * stor-layout.c (layout_type): Likewise.
11613 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11614 * tree-cfg.c (verify_gimple_comparison): Likewise.
11615 (verify_gimple_assign_binary): Likewise.
11616 (verify_gimple_assign_ternary): Likewise.
11617 (verify_gimple_assign_single): Likewise.
11618 * tree-pretty-print.c (dump_generic_node): Likewise.
11619 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11620 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11621 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11622 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11623 (vect_shift_permute_load_chain): Likewise.
11624 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11625 (expand_vector_condition, optimize_vector_constructor): Likewise.
11626 (lower_vec_perm, get_compute_type): Likewise.
11627 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11628 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11629 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11630 (vect_recog_mask_conversion_pattern): Likewise.
11631 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11632 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11633 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11634 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11635 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11636 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11637 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11638 (supportable_widening_operation): Likewise.
11639 (supportable_narrowing_operation): Likewise.
11640 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11642 * varasm.c (output_constant): Likewise.
11644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11645 Alan Hayward <alan.hayward@arm.com>
11646 David Sherwood <david.sherwood@arm.com>
11648 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11649 so that both the length == 3 and length != 3 cases set up their
11650 own permute vectors. Add comments explaining why we know the
11651 number of elements is constant.
11652 (vect_permute_load_chain): Likewise.
11654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11655 Alan Hayward <alan.hayward@arm.com>
11656 David Sherwood <david.sherwood@arm.com>
11658 * machmode.h (mode_nunits): Change from unsigned char to
11660 (ONLY_FIXED_SIZE_MODES): New macro.
11661 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11662 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11663 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11665 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11666 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11667 or if measurement_type is not polynomial.
11668 * genmodes.c (ZERO_COEFFS): New macro.
11669 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11671 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11672 Use ZERO_COEFFS when emitting initializers.
11673 * data-streamer.h (bp_pack_poly_value): New function.
11674 (bp_unpack_poly_value): Likewise.
11675 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11676 for GET_MODE_NUNITS.
11677 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11678 for GET_MODE_NUNITS.
11679 * tree.c (make_vector_type): Remove temporary shim and make
11680 the real function take the number of units as a poly_uint64
11681 rather than an int.
11682 (build_vector_type_for_mode): Handle polynomial nunits.
11683 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11684 * emit-rtl.c (const_vec_series_p_1): Likewise.
11685 (gen_rtx_CONST_VECTOR): Likewise.
11686 * fold-const.c (test_vec_duplicate_folding): Likewise.
11687 * genrecog.c (validate_pattern): Likewise.
11688 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11689 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11690 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11691 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11692 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11693 * rtlanal.c (subreg_get_info): Likewise.
11694 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11695 (vect_grouped_load_supported): Likewise.
11696 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11697 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11698 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11699 (simplify_const_unary_operation, simplify_binary_operation_1)
11700 (simplify_const_binary_operation, simplify_ternary_operation)
11701 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11702 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11703 instead of CONST_VECTOR_NUNITS.
11704 * varasm.c (output_constant_pool_2): Likewise.
11705 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11706 explicit-encoded elements in the XVEC for variable-length vectors.
11708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11710 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11713 Alan Hayward <alan.hayward@arm.com>
11714 David Sherwood <david.sherwood@arm.com>
11716 * coretypes.h (fixed_size_mode): Declare.
11717 (fixed_size_mode_pod): New typedef.
11718 * builtins.h (target_builtins::x_apply_args_mode)
11719 (target_builtins::x_apply_result_mode): Change type to
11720 fixed_size_mode_pod.
11721 * builtins.c (apply_args_size, apply_result_size, result_vector)
11722 (expand_builtin_apply_args_1, expand_builtin_apply)
11723 (expand_builtin_return): Update accordingly.
11725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11727 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11728 * cselib.c (cselib_hash_rtx): Likewise.
11729 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11730 CONST_VECTOR encoding.
11732 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11733 Jeff Law <law@redhat.com>
11736 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11737 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11738 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11739 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11742 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11743 explicitly probe *sp in a noreturn function if there were any callee
11744 register saves or frame pointer is needed.
11746 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11749 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11750 BLKmode for ternary, binary or unary expressions.
11753 * var-tracking.c (delete_vta_debug_insn): New inline function.
11754 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11755 insns from get_insns () to NULL instead of each bb separately.
11756 Use delete_vta_debug_insn. No longer static.
11757 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11758 delete_vta_debug_insns callers.
11759 * rtl.h (delete_vta_debug_insns): Declare.
11760 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11761 instead of variable_tracking_main.
11763 2018-01-03 Martin Sebor <msebor@redhat.com>
11765 PR tree-optimization/83603
11766 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11767 arguments past the endof the argument list in functions declared
11768 without a prototype.
11769 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11770 Avoid checking when arguments are null.
11772 2018-01-03 Martin Sebor <msebor@redhat.com>
11775 * doc/extend.texi (attribute const): Fix a typo.
11776 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11777 issuing -Wsuggest-attribute for void functions.
11779 2018-01-03 Martin Sebor <msebor@redhat.com>
11781 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11782 offset_int::from instead of wide_int::to_shwi.
11783 (maybe_diag_overlap): Remove assertion.
11784 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11785 * gimple-ssa-sprintf.c (format_directive): Same.
11786 (parse_directive): Same.
11787 (sprintf_dom_walker::compute_format_length): Same.
11788 (try_substitute_return_value): Same.
11790 2018-01-03 Jeff Law <law@redhat.com>
11792 PR middle-end/83654
11793 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11794 non-constant residual for zero at runtime and avoid probing in
11795 that case. Reorganize code for trailing problem to mirror handling
11798 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11800 PR tree-optimization/83501
11801 * tree-ssa-strlen.c (get_string_cst): New.
11802 (handle_char_store): Call get_string_cst.
11804 2018-01-03 Martin Liska <mliska@suse.cz>
11806 PR tree-optimization/83593
11807 * tree-ssa-strlen.c: Include tree-cfg.h.
11808 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11809 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11810 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11812 (strlen_dom_walker::before_dom_children): Call
11813 gimple_purge_dead_eh_edges. Dump tranformation with details
11815 (strlen_dom_walker::before_dom_children): Update call by adding
11816 new argument cleanup_eh.
11817 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11819 2018-01-03 Martin Liska <mliska@suse.cz>
11822 * cif-code.def (VARIADIC_THUNK): New enum value.
11823 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11826 2018-01-03 Jan Beulich <jbeulich@suse.com>
11828 * sse.md (mov<mode>_internal): Tighten condition for when to use
11829 vmovdqu<ssescalarsize> for TI and OI modes.
11831 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11833 Update copyright years.
11835 2018-01-03 Martin Liska <mliska@suse.cz>
11838 * ipa-visibility.c (function_and_variable_visibility): Skip
11839 functions with noipa attribure.
11841 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11843 * gcc.c (process_command): Update copyright notice dates.
11844 * gcov-dump.c (print_version): Ditto.
11845 * gcov.c (print_version): Ditto.
11846 * gcov-tool.c (print_version): Ditto.
11847 * gengtype.c (create_file): Ditto.
11848 * doc/cpp.texi: Bump @copying's copyright year.
11849 * doc/cppinternals.texi: Ditto.
11850 * doc/gcc.texi: Ditto.
11851 * doc/gccint.texi: Ditto.
11852 * doc/gcov.texi: Ditto.
11853 * doc/install.texi: Ditto.
11854 * doc/invoke.texi: Ditto.
11856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11858 * vector-builder.h (vector_builder::m_full_nelts): Change from
11859 unsigned int to poly_uint64.
11860 (vector_builder::full_nelts): Update prototype accordingly.
11861 (vector_builder::new_vector): Likewise.
11862 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11863 (vector_builder::operator ==): Likewise.
11864 (vector_builder::finalize): Likewise.
11865 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11866 Take the number of elements as a poly_uint64 rather than an
11868 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11869 from unsigned int to poly_uint64.
11870 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11871 (vec_perm_indices::new_vector): Likewise.
11872 (vec_perm_indices::length): Likewise.
11873 (vec_perm_indices::nelts_per_input): Likewise.
11874 (vec_perm_indices::input_nelts): Likewise.
11875 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11876 number of elements per input as a poly_uint64 rather than an
11877 unsigned int. Use the original encoding for variable-length
11878 vectors, rather than clamping each individual element.
11879 For the second and subsequent elements in each pattern,
11880 clamp the step and base before clamping their sum.
11881 (vec_perm_indices::series_p): Handle polynomial element counts.
11882 (vec_perm_indices::all_in_range_p): Likewise.
11883 (vec_perm_indices_to_tree): Likewise.
11884 (vec_perm_indices_to_rtx): Likewise.
11885 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11886 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11887 (tree_vector_builder::new_binary_operation): Handle polynomial
11888 element counts. Return false if we need to know the number
11889 of elements at compile time.
11890 * fold-const.c (fold_vec_perm): Punt if the number of elements
11891 isn't known at compile time.
11893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11895 * vec-perm-indices.h (vec_perm_builder): Change element type
11896 from HOST_WIDE_INT to poly_int64.
11897 (vec_perm_indices::element_type): Update accordingly.
11898 (vec_perm_indices::clamp): Handle polynomial element_types.
11899 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11900 (vec_perm_indices::all_in_range_p): Likewise.
11901 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11903 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11904 polynomial vec_perm_indices element types.
11905 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11906 * fold-const.c (fold_vec_perm): Likewise.
11907 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11908 * tree-vect-generic.c (lower_vec_perm): Likewise.
11909 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11910 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11911 element type to HOST_WIDE_INT.
11913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11914 Alan Hayward <alan.hayward@arm.com>
11915 David Sherwood <david.sherwood@arm.com>
11917 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11918 rather than an int. Use plus_constant.
11919 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11920 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11923 Alan Hayward <alan.hayward@arm.com>
11924 David Sherwood <david.sherwood@arm.com>
11926 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11927 a HOST_WIDE_INT to a poly_int64.
11929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11930 Alan Hayward <alan.hayward@arm.com>
11931 David Sherwood <david.sherwood@arm.com>
11933 * calls.c (load_register_parameters): Cope with polynomial
11934 mode sizes. Require a constant size for BLKmode parameters
11935 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11936 forces a parameter to be padded at the lsb end in order to
11937 fill a complete number of words, require the parameter size
11938 to be ordered wrt UNITS_PER_WORD.
11940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11941 Alan Hayward <alan.hayward@arm.com>
11942 David Sherwood <david.sherwood@arm.com>
11944 * reload1.c (spill_stack_slot_width): Change element type
11945 from unsigned int to poly_uint64_pod.
11946 (alter_reg): Treat mode sizes as polynomial.
11948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11949 Alan Hayward <alan.hayward@arm.com>
11950 David Sherwood <david.sherwood@arm.com>
11952 * reload.c (complex_word_subreg_p): New function.
11953 (reload_inner_reg_of_subreg, push_reload): Use it.
11955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11956 Alan Hayward <alan.hayward@arm.com>
11957 David Sherwood <david.sherwood@arm.com>
11959 * lra-constraints.c (process_alt_operands): Reject matched
11960 operands whose sizes aren't ordered.
11961 (match_reload): Refer to this check here.
11963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11964 Alan Hayward <alan.hayward@arm.com>
11965 David Sherwood <david.sherwood@arm.com>
11967 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11968 that the mode size is in the set {1, 2, 4, 8, 16}.
11970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11971 Alan Hayward <alan.hayward@arm.com>
11972 David Sherwood <david.sherwood@arm.com>
11974 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11975 Use plus_constant instead of gen_rtx_PLUS.
11977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11978 Alan Hayward <alan.hayward@arm.com>
11979 David Sherwood <david.sherwood@arm.com>
11981 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11982 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11983 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11984 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11985 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11986 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11987 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11988 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11989 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11990 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11992 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11993 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11994 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11995 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11996 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11997 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11998 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11999 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12000 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12001 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12003 * expr.c (emit_move_resolve_push): Treat the input and result
12004 of PUSH_ROUNDING as a poly_int64.
12005 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12006 (emit_push_insn): Likewise.
12007 * lra-eliminations.c (mark_not_eliminable): Likewise.
12008 * recog.c (push_operand): Likewise.
12009 * reload1.c (elimination_effects): Likewise.
12010 * rtlanal.c (nonzero_bits1): Likewise.
12011 * calls.c (store_one_arg): Likewise. Require the padding to be
12012 known at compile time.
12014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12015 Alan Hayward <alan.hayward@arm.com>
12016 David Sherwood <david.sherwood@arm.com>
12018 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12019 Use plus_constant instead of gen_rtx_PLUS.
12021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12022 Alan Hayward <alan.hayward@arm.com>
12023 David Sherwood <david.sherwood@arm.com>
12025 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12026 rather than an int.
12028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12029 Alan Hayward <alan.hayward@arm.com>
12030 David Sherwood <david.sherwood@arm.com>
12032 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12033 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12034 via stack temporaries. Treat the mode size as polynomial too.
12036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12037 Alan Hayward <alan.hayward@arm.com>
12038 David Sherwood <david.sherwood@arm.com>
12040 * expr.c (expand_expr_real_2): When handling conversions involving
12041 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12042 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12043 as a poly_uint64 too.
12045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12046 Alan Hayward <alan.hayward@arm.com>
12047 David Sherwood <david.sherwood@arm.com>
12049 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12052 Alan Hayward <alan.hayward@arm.com>
12053 David Sherwood <david.sherwood@arm.com>
12055 * combine.c (can_change_dest_mode): Handle polynomial
12056 REGMODE_NATURAL_SIZE.
12057 * expmed.c (store_bit_field_1): Likewise.
12058 * expr.c (store_constructor): Likewise.
12059 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12060 and polynomial REGMODE_NATURAL_SIZE.
12061 (gen_lowpart_common): Likewise.
12062 * reginfo.c (record_subregs_of_mode): Likewise.
12063 * rtlanal.c (read_modify_subreg_p): Likewise.
12065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12066 Alan Hayward <alan.hayward@arm.com>
12067 David Sherwood <david.sherwood@arm.com>
12069 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12070 numbers of elements.
12072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12073 Alan Hayward <alan.hayward@arm.com>
12074 David Sherwood <david.sherwood@arm.com>
12076 * match.pd: Cope with polynomial numbers of vector elements.
12078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12079 Alan Hayward <alan.hayward@arm.com>
12080 David Sherwood <david.sherwood@arm.com>
12082 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12083 in a POINTER_PLUS_EXPR.
12085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12086 Alan Hayward <alan.hayward@arm.com>
12087 David Sherwood <david.sherwood@arm.com>
12089 * omp-simd-clone.c (simd_clone_subparts): New function.
12090 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12091 (ipa_simd_modify_function_body): Likewise.
12093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12094 Alan Hayward <alan.hayward@arm.com>
12095 David Sherwood <david.sherwood@arm.com>
12097 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12098 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12099 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12100 (expand_vector_condition, vector_element): Likewise.
12101 (subparts_gt): New function.
12102 (get_compute_type): Use subparts_gt.
12103 (count_type_subparts): Delete.
12104 (expand_vector_operations_1): Use subparts_gt instead of
12105 count_type_subparts.
12107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12108 Alan Hayward <alan.hayward@arm.com>
12109 David Sherwood <david.sherwood@arm.com>
12111 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12112 (vect_compile_time_alias): ...this new function. Do the calculation
12113 on poly_ints rather than trees.
12114 (vect_prune_runtime_alias_test_list): Update call accordingly.
12116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12117 Alan Hayward <alan.hayward@arm.com>
12118 David Sherwood <david.sherwood@arm.com>
12120 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12122 (vect_schedule_slp_instance): Likewise.
12124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12125 Alan Hayward <alan.hayward@arm.com>
12126 David Sherwood <david.sherwood@arm.com>
12128 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12129 constant and extern definitions for variable-length vectors.
12130 (vect_get_constant_vectors): Note that the number of units
12131 is known to be constant.
12133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12134 Alan Hayward <alan.hayward@arm.com>
12135 David Sherwood <david.sherwood@arm.com>
12137 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12138 of units as polynomial. Choose between WIDE and NARROW based
12141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12142 Alan Hayward <alan.hayward@arm.com>
12143 David Sherwood <david.sherwood@arm.com>
12145 * tree-vect-stmts.c (simd_clone_subparts): New function.
12146 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12149 Alan Hayward <alan.hayward@arm.com>
12150 David Sherwood <david.sherwood@arm.com>
12152 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12153 vectors as polynomial. Use build_index_vector for
12154 IFN_GOMP_SIMD_LANE.
12156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12157 Alan Hayward <alan.hayward@arm.com>
12158 David Sherwood <david.sherwood@arm.com>
12160 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12161 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12162 for variable-length vectors.
12163 (vectorizable_mask_load_store): Treat the number of units as
12164 polynomial, asserting that it is constant if the condition has
12165 already been enforced.
12166 (vectorizable_store, vectorizable_load): Likewise.
12168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12169 Alan Hayward <alan.hayward@arm.com>
12170 David Sherwood <david.sherwood@arm.com>
12172 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12173 of units as polynomial. Punt if we can't tell at compile time
12174 which vector contains the final result.
12176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12177 Alan Hayward <alan.hayward@arm.com>
12178 David Sherwood <david.sherwood@arm.com>
12180 * tree-vect-loop.c (vectorizable_induction): Treat the number
12181 of units as polynomial. Punt on SLP inductions. Use an integer
12182 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12183 cast of such a series for variable-length floating-point
12186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12187 Alan Hayward <alan.hayward@arm.com>
12188 David Sherwood <david.sherwood@arm.com>
12190 * tree.h (build_index_vector): Declare.
12191 * tree.c (build_index_vector): New function.
12192 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12193 of units as polynomial, forcibly converting it to a constant if
12194 vectorizable_reduction has already enforced the condition.
12195 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12196 to create a {1,2,3,...} vector.
12197 (vectorizable_reduction): Treat the number of units as polynomial.
12198 Choose vectype_in based on the largest scalar element size rather
12199 than the smallest number of units. Enforce the restrictions
12202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12203 Alan Hayward <alan.hayward@arm.com>
12204 David Sherwood <david.sherwood@arm.com>
12206 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12207 number of units as polynomial.
12209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12210 Alan Hayward <alan.hayward@arm.com>
12211 David Sherwood <david.sherwood@arm.com>
12213 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12214 * target.def (autovectorize_vector_sizes): Return the vector sizes
12215 by pointer, using vector_sizes rather than a bitmask.
12216 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12217 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12218 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12220 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12221 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12222 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12223 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12224 * omp-general.c (omp_max_vf): Likewise.
12225 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12226 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12227 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12228 * tree-vect-slp.c (vect_slp_bb): Likewise.
12229 * doc/tm.texi: Regenerate.
12230 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12232 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12233 the vector size as a poly_uint64 rather than an unsigned int.
12234 (current_vector_size): Change from an unsigned int to a poly_uint64.
12235 (get_vectype_for_scalar_type): Update accordingly.
12236 * tree.h (build_truth_vector_type): Take the size and number of
12237 units as a poly_uint64 rather than an unsigned int.
12238 (build_vector_type): Add a temporary overload that takes
12239 the number of units as a poly_uint64 rather than an unsigned int.
12240 * tree.c (make_vector_type): Likewise.
12241 (build_truth_vector_type): Take the number of units as a poly_uint64
12242 rather than an unsigned int.
12244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12245 Alan Hayward <alan.hayward@arm.com>
12246 David Sherwood <david.sherwood@arm.com>
12248 * target.def (get_mask_mode): Take the number of units and length
12249 as poly_uint64s rather than unsigned ints.
12250 * targhooks.h (default_get_mask_mode): Update accordingly.
12251 * targhooks.c (default_get_mask_mode): Likewise.
12252 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12253 * doc/tm.texi: Regenerate.
12255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12256 Alan Hayward <alan.hayward@arm.com>
12257 David Sherwood <david.sherwood@arm.com>
12259 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12260 * omp-general.c (omp_max_vf): Likewise.
12261 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12262 (expand_omp_simd): Handle polynomial safelen.
12263 * omp-low.c (omplow_simd_context): Add a default constructor.
12264 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12265 (lower_rec_simd_input_clauses): Update accordingly.
12266 (lower_rec_input_clauses): Likewise.
12268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12269 Alan Hayward <alan.hayward@arm.com>
12270 David Sherwood <david.sherwood@arm.com>
12272 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12273 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12274 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12275 (vect_analyze_slp_cost): Likewise.
12276 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12277 (vect_model_load_cost): Likewise.
12279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12280 Alan Hayward <alan.hayward@arm.com>
12281 David Sherwood <david.sherwood@arm.com>
12283 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12284 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12285 from an unsigned int * to a poly_uint64_pod *.
12286 (calculate_unrolling_factor): New function.
12287 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12290 Alan Hayward <alan.hayward@arm.com>
12291 David Sherwood <david.sherwood@arm.com>
12293 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12294 from an unsigned int to a poly_uint64.
12295 (_loop_vec_info::slp_unrolling_factor): Likewise.
12296 (_loop_vec_info::vectorization_factor): Change from an int
12298 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12299 (vect_get_num_vectors): New function.
12300 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12301 (vect_get_num_copies): Use vect_get_num_vectors.
12302 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12303 to an unsigned int *.
12304 (vect_analyze_data_refs): Change min_vf from an int * to a
12306 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12307 than an unsigned HOST_WIDE_INT.
12308 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12309 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12310 to an unsigned int *.
12311 (vect_analyze_data_ref_dependences): Likewise.
12312 (vect_compute_data_ref_alignment): Handle polynomial vf.
12313 (vect_enhance_data_refs_alignment): Likewise.
12314 (vect_prune_runtime_alias_test_list): Likewise.
12315 (vect_shift_permute_load_chain): Likewise.
12316 (vect_supportable_dr_alignment): Likewise.
12317 (dependence_distance_ge_vf): Take the vectorization factor as a
12318 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12319 (vect_analyze_data_refs): Change min_vf from an int * to a
12321 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12322 vfm1 as a poly_uint64 rather than an int. Make the same change
12323 for the returned bound_scalar.
12324 (vect_gen_vector_loop_niters): Handle polynomial vf.
12325 (vect_do_peeling): Likewise. Update call to
12326 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12327 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12329 * tree-vect-loop.c (vect_determine_vectorization_factor)
12330 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12331 (vect_get_known_peeling_cost): Likewise.
12332 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12333 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12334 (vect_transform_loop): Likewise. Use the lowest possible VF when
12335 updating the upper bounds of the loop.
12336 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12337 rather than an int.
12338 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12339 polynomial unroll factors.
12340 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12341 (vect_make_slp_decision): Likewise.
12342 (vect_supported_load_permutation_p): Likewise, and polynomial
12344 (vect_analyze_slp_cost): Handle polynomial vf.
12345 (vect_slp_analyze_node_operations): Likewise.
12346 (vect_slp_analyze_bb_1): Likewise.
12347 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12348 than an unsigned HOST_WIDE_INT.
12349 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12350 (vectorizable_load): Handle polynomial vf.
12351 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12353 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12356 Alan Hayward <alan.hayward@arm.com>
12357 David Sherwood <david.sherwood@arm.com>
12359 * match.pd: Handle bit operations involving three constants
12360 and try to fold one pair.
12362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12364 * tree-vect-loop-manip.c: Include gimple-fold.h.
12365 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12366 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12367 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12368 Add a path that uses a step of VF instead of 1, but disable it
12370 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12371 and niters_no_overflow parameters. Update calls to
12372 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12373 Create a new SSA name if the latter choses to use a ste other
12374 than zero, and return it via niters_vector_mult_vf_var.
12375 * tree-vect-loop.c (vect_transform_loop): Update calls to
12376 vect_do_peeling, vect_gen_vector_loop_niters and
12377 slpeel_make_loop_iterate_ntimes.
12378 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12379 (vect_gen_vector_loop_niters): Update declarations after above changes.
12381 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12383 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12384 128-bit round to integer instructions.
12385 (ceil<mode>2): Likewise.
12386 (btrunc<mode>2): Likewise.
12387 (round<mode>2): Likewise.
12389 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12391 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12392 unaligned VSX load/store on P8/P9.
12393 (expand_block_clear): Allow the use of unaligned VSX
12394 load/store on P8/P9.
12396 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12398 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12400 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12401 swap associated with both a load and a store.
12403 2018-01-02 Andrew Waterman <andrew@sifive.com>
12405 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12406 * config/riscv/riscv.md (clear_cache): Use it.
12408 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12410 * web.c: Remove out-of-date comment.
12412 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12414 * expr.c (fixup_args_size_notes): Check that any existing
12415 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12416 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12417 (emit_single_push_insn): ...here.
12419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12421 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12422 (const_vector_encoded_nelts): New function.
12423 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12424 (const_vector_int_elt, const_vector_elt): Declare.
12425 * emit-rtl.c (const_vector_int_elt_1): New function.
12426 (const_vector_elt): Likewise.
12427 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12428 of CONST_VECTOR_ELT.
12430 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12432 * expr.c: Include rtx-vector-builder.h.
12433 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12434 directly on the tree encoding.
12435 (const_vector_from_tree): Likewise.
12436 * optabs.c: Include rtx-vector-builder.h.
12437 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12438 sequence of "u" values.
12439 * vec-perm-indices.c: Include rtx-vector-builder.h.
12440 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12441 directly on the vec_perm_indices encoding.
12443 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12445 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12446 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12447 * rtx-vector-builder.h: New file.
12448 * rtx-vector-builder.c: Likewise.
12449 * rtl.h (rtx_def::u2): Add a const_vector field.
12450 (CONST_VECTOR_NPATTERNS): New macro.
12451 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12452 (CONST_VECTOR_DUPLICATE_P): Likewise.
12453 (CONST_VECTOR_STEPPED_P): Likewise.
12454 (CONST_VECTOR_ENCODED_ELT): Likewise.
12455 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12456 (unwrap_const_vec_duplicate): Likewise.
12457 (const_vec_series_p): Check for a non-duplicated vector encoding.
12458 Say that the function only returns true for integer vectors.
12459 * emit-rtl.c: Include rtx-vector-builder.h.
12460 (gen_const_vec_duplicate_1): Delete.
12461 (gen_const_vector): Call gen_const_vec_duplicate instead of
12462 gen_const_vec_duplicate_1.
12463 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12464 (gen_const_vec_duplicate): Use rtx_vector_builder.
12465 (gen_const_vec_series): Likewise.
12466 (gen_rtx_CONST_VECTOR): Likewise.
12467 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12468 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12469 Build a new vector rather than modifying a CONST_VECTOR in-place.
12470 (handle_special_swappables): Update call accordingly.
12471 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12472 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12473 Build a new vector rather than modifying a CONST_VECTOR in-place.
12474 (handle_special_swappables): Update call accordingly.
12476 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12478 * simplify-rtx.c (simplify_const_binary_operation): Use
12479 CONST_VECTOR_ELT instead of XVECEXP.
12481 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12483 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12484 the selector elements to be different from the data elements
12485 if the selector is a VECTOR_CST.
12486 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12487 ssizetype for the selector.
12489 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12491 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12492 before testing each element individually.
12493 * tree-vect-generic.c (lower_vec_perm): Likewise.
12495 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12497 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12498 * selftest-run-tests.c (selftest::run_tests): Call it.
12499 * vector-builder.h (vector_builder::operator ==): New function.
12500 (vector_builder::operator !=): Likewise.
12501 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12502 (vec_perm_indices::all_from_input_p): New function.
12503 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12504 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12505 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12506 instead of reading the VECTOR_CST directly. Detect whether both
12507 vector inputs are the same before constructing the vec_perm_indices,
12508 and update the number of inputs argument accordingly. Use the
12509 utility functions added above. Only construct sel2 if we need to.
12511 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12513 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12514 the broadcast of the low byte.
12515 (expand_mult_highpart): Use an explicit encoding for the permutes.
12516 * optabs-query.c (can_mult_highpart_p): Likewise.
12517 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12518 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12519 (vectorizable_bswap): Likewise.
12520 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12521 explicit encoding for the power-of-2 permutes.
12522 (vect_permute_store_chain): Likewise.
12523 (vect_grouped_load_supported): Likewise.
12524 (vect_permute_load_chain): Likewise.
12526 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12528 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12529 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12530 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12531 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12532 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12533 (vect_gen_perm_mask_any): Likewise.
12535 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12537 * int-vector-builder.h: New file.
12538 * vec-perm-indices.h: Include int-vector-builder.h.
12539 (vec_perm_indices): Redefine as an int_vector_builder.
12540 (auto_vec_perm_indices): Delete.
12541 (vec_perm_builder): Redefine as a stand-alone class.
12542 (vec_perm_indices::vec_perm_indices): New function.
12543 (vec_perm_indices::clamp): Likewise.
12544 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12545 (vec_perm_indices::new_vector): New function.
12546 (vec_perm_indices::new_expanded_vector): Update for new
12547 vec_perm_indices class.
12548 (vec_perm_indices::rotate_inputs): New function.
12549 (vec_perm_indices::all_in_range_p): Operate directly on the
12550 encoded form, without computing elided elements.
12551 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12552 encoding. Update for new vec_perm_indices class.
12553 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12554 the given vec_perm_builder.
12555 (expand_vec_perm_var): Update vec_perm_builder constructor.
12556 (expand_mult_highpart): Use vec_perm_builder instead of
12557 auto_vec_perm_indices.
12558 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12559 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12560 or double series encoding as appropriate.
12561 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12562 vec_perm_indices instead of auto_vec_perm_indices.
12563 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12564 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12565 (vect_permute_store_chain): Likewise.
12566 (vect_grouped_load_supported): Likewise.
12567 (vect_permute_load_chain): Likewise.
12568 (vect_shift_permute_load_chain): Likewise.
12569 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12570 (vect_transform_slp_perm_load): Likewise.
12571 (vect_schedule_slp_instance): Likewise.
12572 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12573 (vectorizable_mask_load_store): Likewise.
12574 (vectorizable_bswap): Likewise.
12575 (vectorizable_store): Likewise.
12576 (vectorizable_load): Likewise.
12577 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12578 vec_perm_indices instead of auto_vec_perm_indices. Use
12579 tree_to_vec_perm_builder to read the vector from a tree.
12580 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12581 vec_perm_builder instead of a vec_perm_indices.
12582 (have_whole_vector_shift): Use vec_perm_builder and
12583 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12584 truncation to calc_vec_perm_mask_for_shift.
12585 (vect_create_epilog_for_reduction): Likewise.
12586 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12587 from auto_vec_perm_indices to vec_perm_indices.
12588 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12589 instead of changing individual elements.
12590 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12591 the vector in d.perm.
12592 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12593 from auto_vec_perm_indices to vec_perm_indices.
12594 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12595 instead of changing individual elements.
12596 (arm_vectorize_vec_perm_const): Use new_vector to install
12597 the vector in d.perm.
12598 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12599 Update vec_perm_builder constructor.
12600 (rs6000_expand_interleave): Likewise.
12601 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12602 (rs6000_expand_interleave): Likewise.
12604 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12606 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12607 to qimode could truncate the indices.
12608 * optabs.c (expand_vec_perm_var): Likewise.
12610 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12612 * Makefile.in (OBJS): Add vec-perm-indices.o.
12613 * vec-perm-indices.h: New file.
12614 * vec-perm-indices.c: Likewise.
12615 * target.h (vec_perm_indices): Replace with a forward class
12617 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12618 * optabs.h: Include vec-perm-indices.h.
12619 (expand_vec_perm): Delete.
12620 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12621 (expand_vec_perm_const): Declare.
12622 * target.def (vec_perm_const_ok): Replace with...
12623 (vec_perm_const): ...this new hook.
12624 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12625 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12626 * doc/tm.texi: Regenerate.
12627 * optabs.def (vec_perm_const): Delete.
12628 * doc/md.texi (vec_perm_const): Likewise.
12629 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12630 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12631 expand_vec_perm for constant permutation vectors. Assert that
12632 the mode of variable permutation vectors is the integer equivalent
12633 of the mode that is being permuted.
12634 * optabs-query.h (selector_fits_mode_p): Declare.
12635 * optabs-query.c: Include vec-perm-indices.h.
12636 (selector_fits_mode_p): New function.
12637 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12638 is defined, instead of checking whether the vec_perm_const_optab
12639 exists. Use targetm.vectorize.vec_perm_const instead of
12640 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12641 fit in the vector mode before using a variable permute.
12642 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12643 vec_perm_indices instead of an rtx.
12644 (expand_vec_perm): Replace with...
12645 (expand_vec_perm_const): ...this new function. Take the selector
12646 as a vec_perm_indices rather than an rtx. Also take the mode of
12647 the selector. Update call to shift_amt_for_vec_perm_mask.
12648 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12649 Use vec_perm_indices::new_expanded_vector to expand the original
12650 selector into bytes. Check whether the indices fit in the vector
12651 mode before using a variable permute.
12652 (expand_vec_perm_var): Make global.
12653 (expand_mult_highpart): Use expand_vec_perm_const.
12654 * fold-const.c: Includes vec-perm-indices.h.
12655 * tree-ssa-forwprop.c: Likewise.
12656 * tree-vect-data-refs.c: Likewise.
12657 * tree-vect-generic.c: Likewise.
12658 * tree-vect-loop.c: Likewise.
12659 * tree-vect-slp.c: Likewise.
12660 * tree-vect-stmts.c: Likewise.
12661 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12663 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12664 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12665 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12666 (aarch64_vectorize_vec_perm_const): ...this new function.
12667 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12668 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12669 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12670 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12671 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12672 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12673 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12675 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12676 check for NEON modes.
12677 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12678 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12679 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12680 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12682 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12683 the old VEC_PERM_CONST conditions.
12684 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12685 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12686 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12687 (ia64_vectorize_vec_perm_const_ok): Merge into...
12688 (ia64_vectorize_vec_perm_const): ...this new function.
12689 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12690 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12691 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12692 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12693 * config/mips/mips.c (mips_expand_vec_perm_const)
12694 (mips_vectorize_vec_perm_const_ok): Merge into...
12695 (mips_vectorize_vec_perm_const): ...this new function.
12696 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12697 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12698 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12699 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12700 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12701 (rs6000_expand_vec_perm_const): Delete.
12702 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12704 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12705 (altivec_expand_vec_perm_const_le): Take each operand individually.
12706 Operate on constant selectors rather than rtxes.
12707 (altivec_expand_vec_perm_const): Likewise. Update call to
12708 altivec_expand_vec_perm_const_le.
12709 (rs6000_expand_vec_perm_const): Delete.
12710 (rs6000_vectorize_vec_perm_const_ok): Delete.
12711 (rs6000_vectorize_vec_perm_const): New function.
12712 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12713 an element count and rtx array.
12714 (rs6000_expand_extract_even): Update call accordingly.
12715 (rs6000_expand_interleave): Likewise.
12716 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12717 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12718 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12719 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12720 (rs6000_expand_vec_perm_const): Delete.
12721 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12722 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12723 (altivec_expand_vec_perm_const_le): Take each operand individually.
12724 Operate on constant selectors rather than rtxes.
12725 (altivec_expand_vec_perm_const): Likewise. Update call to
12726 altivec_expand_vec_perm_const_le.
12727 (rs6000_expand_vec_perm_const): Delete.
12728 (rs6000_vectorize_vec_perm_const_ok): Delete.
12729 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12730 reference to the SPE evmerge intructions.
12731 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12732 an element count and rtx array.
12733 (rs6000_expand_extract_even): Update call accordingly.
12734 (rs6000_expand_interleave): Likewise.
12735 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12736 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12738 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12740 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12742 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12743 vector mode and that that mode matches the mode of the data
12745 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12746 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12747 directly using expand_vec_perm_1 when forcing selectors into
12749 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12751 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12753 * optabs-query.h (can_vec_perm_p): Delete.
12754 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12755 * optabs-query.c (can_vec_perm_p): Split into...
12756 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12757 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12758 particular selector is valid.
12759 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12760 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12761 (vect_grouped_load_supported): Likewise.
12762 (vect_shift_permute_load_chain): Likewise.
12763 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12764 (vect_transform_slp_perm_load): Likewise.
12765 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12766 (vectorizable_bswap): Likewise.
12767 (vect_gen_perm_mask_checked): Likewise.
12768 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12769 implementations of variable permutation vectors into account
12770 when deciding which selector to use.
12771 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12772 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12773 with a false third argument.
12774 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12775 to test whether the constant selector is valid and can_vec_perm_var_p
12776 to test whether a variable selector is valid.
12778 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12780 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12781 * optabs-query.c (can_vec_perm_p): Likewise.
12782 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12783 instead of vec_perm_indices.
12784 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12785 (vect_gen_perm_mask_checked): Likewise,
12786 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12787 (vect_gen_perm_mask_checked): Likewise,
12789 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12791 * optabs-query.h (qimode_for_vec_perm): Declare.
12792 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12793 (qimode_for_vec_perm): ...this new function.
12794 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12796 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12798 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12799 does not have a conditional at the top.
12801 2018-01-02 Richard Biener <rguenther@suse.de>
12803 * ipa-inline.c (big_speedup_p): Fix expression.
12805 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12808 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12811 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12815 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12816 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12817 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12818 cond_taken_branch_cost 3->4.
12820 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12822 PR tree-optimization/83581
12823 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12824 TODO_cleanup_cfg if any changes have been made.
12826 PR middle-end/83608
12827 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12828 convert_modes if target mode has the right side, but different mode
12831 PR middle-end/83609
12832 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12833 last argument when extracting from CONCAT. If either from_real or
12834 from_imag is NULL, use expansion through memory. If result is not
12835 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12836 the parts directly to inner mode, if even that fails, use expansion
12839 PR middle-end/83623
12840 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12841 check for bswap in mode rather than HImode and use that in expand_unop
12844 Copyright (C) 2018 Free Software Foundation, Inc.
12846 Copying and distribution of this file, with or without modification,
12847 are permitted in any medium without royalty provided the copyright
12848 notice and this notice are preserved.