1 2018-05-15 Martin Sebor <msebor@redhat.com>
3 PR tree-optimization/85753
4 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5 RECORD_TYPE in addition to ARRAY_TYPE.
7 2018-05-15 Martin Sebor <msebor@redhat.com>
10 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
12 2018-05-15 Richard Biener <rguenther@suse.de>
14 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
15 add by_clobber_p one. Change algorithm to collect all defs
16 representing uses we need to walk and try reducing them to
17 a single one before failing.
18 (dse_dom_walker::dse_optimize_stmt): Adjust.
20 2018-05-13 Mark Wielaard <mark@klomp.org>
22 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
23 (size_of_loc_descr): Likewise.
24 (output_loc_operands): Likewise.
25 (output_loc_operands_raw): Likewise.
26 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
27 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
28 (hash_loc_operands): Likewise.
29 (compare_loc_operands): Likewise.
31 2018-05-14 Mark Wielaard <mark@klomp.org>
33 * dwarf2out.c (count_index_addrs): New function.
34 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
36 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
38 PR tree-optimization/83648
39 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
40 return value as malloc candidate.
42 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
45 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
46 as true in call to suggest_attribute.
48 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
50 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
53 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
55 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
57 (aarch64_crypto_aesd_fused): Likewise.
59 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
61 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
62 (movsi_aarch64): Likewise.
63 (load_pairsi): Likewise.
64 (load_pairdi): Likewise.
65 (store_pairsi): Likewise.
66 (store_pairdi): Likewise.
67 (load_pairsf): Likewise.
68 (load_pairdf): Likewise.
69 (store_pairsf): Likewise.
70 (store_pairdf): Likewise.
71 (zero_extend): Likewise.
72 (trunc): Swap alternatives.
73 (fcvt_target): Add '?' to prefer w over r.
75 2018-05-14 Jakub Jelinek <jakub@redhat.com>
78 * config/i386/i386.md: Disallow non-commutative arithmetics in
79 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
80 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
81 in the peephole2 before it.
83 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
85 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
86 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
87 (ix86_handle_option): Handle -mcldemote.
88 * config.gcc: New header.
89 * config/i386/cldemoteintrin.h: New file.
90 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
91 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
93 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
94 OPTION_MASK_ISA_CLDEMOTE.
95 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
96 (ix86_valid_target_attribute_inner_p): Ditto.
97 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
98 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
99 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
100 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
101 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
103 * config/i386/i386.opt: Add -mcldemote.
104 * config/i386/x86intrin.h: New header.
105 * doc/invoke.texi: Add -mcldemote.
107 2018-05-14 Richard Biener <rguenther@suse.de>
109 * doc/match-and-simplify.texi: Adjust :s documentation.
111 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
113 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
114 intended memcpy size.
115 (REORDER_45): Likewise.
117 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
120 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
121 * vec.c (qsort_chk): Use gcc_qsort.
122 * Makefile.in (OBJS-libcommon): Add sort.o.
123 (build/sort.o): New target. Use it...
124 (BUILD_RTL): ... here, and...
125 (build/gencfn-macros): ... here, and...
126 (build/genmatch): ... here.
128 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
129 Chung-Ju Wu <jasonwucj@gmail.com>
131 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
132 * config/nds32/nds32-graywolf.md: New file.
133 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
134 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
136 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
137 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
138 * config/nds32/nds32.md (pipeline_model): Add graywolf.
139 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
140 * config/nds32/pipelines.md: Include n15 settings.
142 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
143 Chung-Ju Wu <jasonwucj@gmail.com>
145 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
146 * config/nds32/nds32-n13.md: New file.
147 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
148 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
150 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
151 * config/nds32/nds32.md (pipeline_model): Add n13.
152 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
153 * config/nds32/pipelines.md: Include n13 settings.
155 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
156 Chung-Ju Wu <jasonwucj@gmail.com>
158 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
159 * config/nds32/nds32-n10.md: New file.
160 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
161 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
163 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
164 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
165 * config/nds32/nds32.md (pipeline_model): Add n10.
166 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
167 * config/nds32/pipelines.md: Include n10 settings.
169 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
170 Kito Cheng <kito.cheng@gmail.com>
171 Chung-Ju Wu <jasonwucj@gmail.com>
173 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
174 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
175 Add enum values for DSP extension instructions.
176 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
178 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
179 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
181 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
182 * config/nds32/nds32-dspext.md: New file for DSP implementation.
183 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
184 * config/nds32/nds32-intrinsic.md: Likewise.
185 * config/nds32/nds32_intrinsic.h: Likewise.
186 * config/nds32/nds32-md-auxiliary.c: Likewise.
187 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
188 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
189 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
190 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
191 * config/nds32/nds32-protos.h: New declarations for DSP extension.
192 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
193 TYPE_DMAC in switch statement.
194 * config/nds32/nds32.c: New checking and implementation for DSP
195 extension instructions.
196 * config/nds32/nds32.h: Likewise.
197 * config/nds32/nds32.md: Likewise.
198 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
199 * config/nds32/predicates.md: Implement new predicates for DSP
202 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
204 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
205 Reformat alternatives and attributes so it is easier to identify
206 which constraints/attributes go with which instruction.
207 (mov<mode>_hardfloat32, FMOVE64): Likewise.
208 (mov<mode>_softfloat32, FMOVE64): Likewise.
209 (mov<mode>_hardfloat64, FMOVE64): Likewise.
210 (mov<mode>_softfloat64, FMOVE64): Likewise.
212 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
214 * doc/extend.texi (PowerPC Built-in Functions): Rename this
216 (Basic PowerPC Built-in Functions): The new name of the
217 subsection previously known as "PowerPC Built-in Functions".
218 (Basic PowerPC Built-in Functions Available on all Configurations):
220 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
221 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
222 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
223 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
225 2018-05-11 Martin Jambor <mjambor@suse.cz>
228 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
231 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
234 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
236 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
238 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
239 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
240 (ix86_handle_option): Handle -mwaitpkg.
241 * config.gcc: New header.
242 * config/i386/cpuid.h (bit_WAITPKG): New bit.
243 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
244 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
246 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
247 OPTION_MASK_ISA_WAITPKG.
248 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
249 (ix86_option_override_internal): Add PTA_WAITPKG.
250 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
251 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
253 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
254 __builtin_ia32_umwait and __builtin_ia32_tpause.
255 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
256 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
257 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
258 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
259 UNSPECV_TPAUSE): New.
260 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
261 * config/i386/i386.opt: Add -mwaitpkg.
262 * config/i386/waitpkgintrin.h: New file.
263 * config/i386/x86intrin.h: New header.
264 * doc/invoke.texi: Add -mwaitpkg.
266 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
269 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
271 (cortex-m0): Use armv6s-m isa.
272 (cortex-m0plus): Likewise.
273 (cortex-m1): Likewise.
274 (cortex-m0.small-multiply): Likewise.
275 (cortex-m0plus.small-multiply): Likewise.
276 (cortex-m1.small-multiply): Likewise.
278 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
279 Jakub Jelinek <jakub@redhat.com>
281 PR tree-optimization/85692
282 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
283 source permute as well.
285 2018-05-11 Martin Liska <mliska@suse.cz>
288 * doc/extend.texi: Document LLVM style format for no_sanitize
291 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
293 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
294 mode_supports_vsx_dform_quad to mode_supports_dq_form.
295 (mode_supports_vsx_dform_quad): Likewise.
296 (mode_supports_vmx_dform): Move these functions to be next to the
297 other mode_supports functions.
298 (mode_supports_dq_form): Likewise.
299 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
300 mode_supports_dq_form.
301 (reg_offset_addressing_ok_p): Likewise.
302 (offsettable_ok_by_alignment): Likewise.
303 (rs6000_legitimate_offset_address_p): Likewise.
304 (legitimate_lo_sum_address_p): Likewise.
305 (rs6000_legitimize_address): Likewise.
306 (rs6000_legitimize_reload_address): Likewise.
307 (rs6000_secondary_reload_inner): Likewise.
308 (rs6000_preferred_reload_class): Likewise.
309 (rs6000_output_move_128bit): Likewise.
311 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
313 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
314 Generate SImode target register for null target.
315 <case IX86_BUILTIN_XGETBV>: Ditto.
316 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
317 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
319 2018-05-10 Carl Love <cel@us.ibm.com>
321 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
322 dcbtt and dcbtstt if operands[2] is 0.
324 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
327 * config/i386/sse.md (usadv64qi): New expander.
329 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
331 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
332 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
333 -maltivec=be support.
334 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
335 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
336 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
337 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
338 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
339 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
340 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
341 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
342 altivec_vsumsws): Adjust.
343 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
344 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
345 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
347 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
348 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
349 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
350 (altivec_lve<VI_char>x): Delete expand.
351 (*altivec_lve<VI_char>x_internal): Rename to...
352 (altivec_lve<VI_char>x): ... this.
353 (altivec_lvxl_<mode>): Delete expand.
354 (*altivec_lvxl_<mode>_internal): Rename to ...
355 (altivec_lvxl_<mode>): ... this.
356 (altivec_stvxl_<mode>): Delete expand.
357 (*altivec_stvxl_<mode>_internal): Rename to ...
358 (altivec_stvxl_<mode>): ... this.
359 (altivec_stve<VI_char>x): Delete expand.
360 (*altivec_stve<VI_char>x_internal): Rename to ...
361 (altivec_stve<VI_char>x): ... this.
362 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
363 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
364 reduc_plus_scal_<mode>): Adjust.
365 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
367 (rs6000_cpu_cpp_builtins): Adjust.
368 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
369 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
370 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
371 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
372 -maltivec=be support.
373 (rs6000_split_vec_extract_var): Adjust.
374 (rs6000_split_v4si_init): Adjust.
375 (swap_selector_for_mode): Delete.
376 (altivec_expand_lvx_be, altivec_expand_stvx_be,
377 altivec_expand_stvex_be): Delete.
378 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
379 -maltivec=be support.
380 (rs6000_gimple_fold_builtin): Ditto.
381 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
383 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
384 (TARGET_DIRECT_MOVE_64BIT): Adjust.
385 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
386 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
387 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
388 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
389 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
390 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
391 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
392 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
393 anonymous split): Adjust.
394 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
395 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
397 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
399 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
400 when --with-gxx-include-dir is also specified.
401 * configure: Regenerate.
403 2018-05-09 Jim Wilson <jimw@sifive.com>
406 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
407 * config/riscv/t-withmultilib: New.
408 * config/riscv/withmultilib.h: New.
409 * doc/install.texi: Document RISC-V --with-multilib-list support.
411 2018-05-09 Richard Biener <rguenther@suse.de>
413 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
415 (vect_bb_vectorization_profitable_p): Adjust. Compute
416 actual scalar cost using the cost vector and the add_stmt_cost
419 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
421 PR rtl-optimization/85645
422 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
423 in the REG_CFA_REGISTER note for LR, don't leave it empty.
425 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
427 PR rtl-optimization/85645
428 * shrink-wrap.c (spread_components): Return a boolean saying if
429 anything was changed.
430 (try_shrink_wrapping_separate): Iterate spread_components until
431 nothing changes anymore.
433 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
435 PR rtl-optimization/85645
436 * regrename.c (build_def_use): Also kill the chains that include the
437 destination of a REG_CFA_REGISTER note.
439 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
441 PR rtl-optimization/85645
442 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
443 insn that has a REG_CFA_REGISTER note.
445 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
447 * cfgexpand.c (expand_clobber): New function.
448 (expand_gimple_stmt_1): Use it.
449 * tree-vect-stmts.c (vect_clobber_variable): New function,
451 (vectorizable_simd_clone_call): ...here.
452 (vectorizable_store): Emit a clobber either side of an
453 IFN_STORE_LANES sequence.
454 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
456 2018-05-09 Tom de Vries <tom@codesourcery.com>
459 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
460 (define_insn "trap_if_false"): Add exit after trap.
462 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
464 PR rtl-optimization/85638
465 * bb-reorder.c: Include common/common-target.h.
466 (create_forwarder_block): New function extracted from...
467 (fix_up_crossing_landing_pad): ...here. Rename into...
468 (dw2_fix_up_crossing_landing_pad): ...this.
469 (sjlj_fix_up_crossing_landing_pad): New function.
470 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
471 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
472 from both partitions and exit the loop after one iteration.
474 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
477 * doc/extend.texi (PowerPC Built-in Functions): Rename this
479 (Basic PowerPC Built-in Functions): The new name of the
480 subsection previously known as "PowerPC Built-in Functions".
481 (Basic PowerPC Built-in Functions Available on all Configurations):
483 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
485 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
487 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
489 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
492 2018-05-08 Jim Wilson <jimw@sifive.com>
494 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
495 (LD_EMUL_SUFFIX): New.
498 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
500 * doc/extend.texi (PowerPC Built-in Functions): Rename this
502 (Basic PowerPC Built-in Functions): The new name of the
503 subsection previously known as "PowerPC Built-in Functions".
504 (Basic PowerPC Built-in Functions Available on all Configurations):
506 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
508 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
510 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
512 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
515 2018-05-08 Jakub Jelinek <jakub@redhat.com>
518 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
519 after cmpelim optimization.
521 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
523 * config.gcc: Support "goldmont".
524 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
525 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
527 * config/i386/i386.c (m_GOLDMONT): Define.
528 (processor_target_table): Add "goldmont".
529 (PTA_GOLDMONT): Define.
530 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
531 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
532 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
533 (fold_builtin_cpu): Add "goldmont".
534 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
535 (ix86_option_override_internal): Add "goldmont".
536 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
537 (processor_type): Add PROCESSOR_GOLDMONT.
538 * config/i386/i386.md: Add CPU "glm".
539 * config/i386/glm.md: New file.
540 * config/i386/x86-tune.def: Add m_GOLDMONT.
541 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
543 2018-05-08 Jakub Jelinek <jakub@redhat.com>
546 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
548 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
549 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
550 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
553 * config/i386/i386.c (ix86_fold_builtin): Handle
554 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
557 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
558 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
560 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
563 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
564 (check_arch): Likewise.
565 (check_fpu): Return the result rather than printing it.
566 (end arch): Fix operator precedence.
568 (END): Print the result from check_fpu.
570 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
571 Alan Hayward <alan.hayward@arm.com>
572 David Sherwood <david.sherwood@arm.com>
574 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
575 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
576 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
577 (*fcmuo<mode>_and): New patterns.
579 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
581 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
582 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
583 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
584 (cmp_op, sve_imm_con): New code attributes.
585 (SVE_COND_INT_CMP, imm_con): Delete.
586 (cmp_op): Remove above unspecs from int attribute.
587 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
589 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
590 comparison-specific unspecs.
591 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
592 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
593 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
594 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
595 (*vec_fcm<cmp_op><mode>): Rename to...
596 (*fcm<cmp_op><mode>): ...this and adjust likewise.
597 (*vec_fcmuo<mode>): Rename to...
598 (*fcmuo<mode>): ...this and adjust likewise.
599 (*pred_fcm<cmp_op><mode>): New pattern.
600 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
601 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
603 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
605 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
606 (aarch64_emit_sve_predicated_cond): New function.
607 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
608 (aarch64_emit_unspec_cond_or): Replace with...
609 (aarch64_emit_sve_or_conds): ...this new function. Use
610 aarch64_emit_sve_ptrue_op for the individual comparisons and
611 aarch64_emit_binop to OR them together.
612 (aarch64_emit_inverted_unspec_cond): Replace with...
613 (aarch64_emit_sve_inverted_cond): ...this new function. Use
614 aarch64_emit_sve_ptrue_op for the comparison and
615 aarch64_emit_unop to invert the result.
616 (aarch64_expand_sve_vec_cmp_float): Update after the above
617 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
619 2018-05-07 Nathan Sidwell <nathan@acm.org>
621 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
622 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
623 (Backwards Compatibility): Likewise.
625 2018-05-07 Luis Machado <luis.machado@linaro.org>
629 2018-05-07 Luis Machado <luis.machado@linaro.org>
631 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
632 <prefetch_dynamic_strides>: New const bool field.
633 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
634 prefetch_dynamic_strides.
635 (exynosm1_prefetch_tune): Likewise.
636 (thunderxt88_prefetch_tune): Likewise.
637 (thunderx_prefetch_tune): Likewise.
638 (thunderx2t99_prefetch_tune): Likewise.
639 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
640 (aarch64_override_options_internal): Update to set
641 PARAM_PREFETCH_DYNAMIC_STRIDES.
642 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
643 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
644 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
645 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
646 prefetch-dynamic-strides setting.
648 2018-05-07 Luis Machado <luis.machado@linaro.org>
650 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
651 <minimum_stride>: New const int field.
652 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
653 minimum_stride field.
654 (exynosm1_prefetch_tune): Likewise.
655 (thunderxt88_prefetch_tune): Likewise.
656 (thunderx_prefetch_tune): Likewise.
657 (thunderx2t99_prefetch_tune): Likewise.
658 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
659 (aarch64_override_options_internal): Update to set
660 PARAM_PREFETCH_MINIMUM_STRIDE.
661 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
662 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
663 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
664 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
665 stride is constant and is below the minimum stride threshold.
667 2018-05-07 Luis Machado <luis.machado@linaro.org>
669 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
672 2018-05-07 Luis Machado <luis.machado@linaro.org>
674 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
675 <prefetch_dynamic_strides>: New const bool field.
676 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
677 prefetch_dynamic_strides.
678 (exynosm1_prefetch_tune): Likewise.
679 (thunderxt88_prefetch_tune): Likewise.
680 (thunderx_prefetch_tune): Likewise.
681 (thunderx2t99_prefetch_tune): Likewise.
682 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
683 (aarch64_override_options_internal): Update to set
684 PARAM_PREFETCH_DYNAMIC_STRIDES.
685 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
686 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
687 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
688 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
689 prefetch-dynamic-strides setting.
691 2018-05-07 Luis Machado <luis.machado@linaro.org>
693 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
694 <minimum_stride>: New const int field.
695 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
696 minimum_stride field.
697 (exynosm1_prefetch_tune): Likewise.
698 (thunderxt88_prefetch_tune): Likewise.
699 (thunderx_prefetch_tune): Likewise.
700 (thunderx2t99_prefetch_tune): Likewise.
701 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
702 (aarch64_override_options_internal): Update to set
703 PARAM_PREFETCH_MINIMUM_STRIDE.
704 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
705 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
706 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
707 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
708 stride is constant and is below the minimum stride threshold.
710 2018-05-06 Jakub Jelinek <jakub@redhat.com>
713 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
714 the type is addressable. Don't force op into register if it has
717 2018-05-05 Roland McGrath <mcgrathr@google.com>
720 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
721 any section for which we don't know a specific type it should have,
722 regardless of name. Previously this was done only for the exact
723 names ".init_array", ".fini_array", and ".preinit_array".
724 (default_elf_asm_named_section): Add comment about
725 relationship with default_section_type_flags and SECTION_NOTYPE.
726 (get_section): Don't consider it a type conflict if one side has
727 SECTION_NOTYPE and the other doesn't, as long as neither has the
728 SECTION_BSS et al used in the default_section_type_flags logic.
730 2018-05-05 Tom de Vries <tom@codesourcery.com>
733 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
734 (workaround_barsyncs): New function.
735 (nvptx_reorg): Use workaround_barsyncs.
736 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
737 (define_expand "nvptx_membar_cta"): New define_expand.
738 (define_insn "*nvptx_membar_cta"): New insn.
740 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
742 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
743 To improve optimization opportunities.
744 * builtin-types.def: The new needed builtin types for the above.
746 2018-05-04 Richard Biener <rguenther@suse.de>
748 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
749 * gimple-ssa-store-merging.c
750 (imm_store_chain_info::output_merged_store): Remove redundant create,
751 release split_store vector contents on failure.
752 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
753 scalar stmt vector on cache hit.
755 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
757 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
759 * config.gcc (powerpc-xilinx-eabi*): Remove.
760 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
762 (fusion_addis_mem_combo_load): Ditto.
763 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
765 (rs6000_cpu_cpp_builtins): Ditto.
766 * config/rs6000/rs6000-linux.c
767 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
768 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
769 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
771 (rs6000_setup_reg_addr_masks): Ditto.
772 (rs6000_init_hard_regno_mode_ok): Ditto.
773 (rs6000_option_override_internal): Ditto.
774 (legitimate_lo_sum_address_p): Ditto.
775 (rs6000_legitimize_address): Ditto.
776 (rs6000_legitimize_reload_address): Ditto.
777 (rs6000_legitimate_address_p): Ditto.
778 (abi_v4_pass_in_fpr): Ditto.
779 (setup_incoming_varargs): Ditto.
780 (rs6000_gimplify_va_arg): Ditto.
781 (rs6000_split_multireg_move): Ditto.
782 (rs6000_savres_strategy): Ditto.
783 (rs6000_emit_prologue_components): Ditto.
784 (rs6000_emit_epilogue_components): Ditto.
785 (rs6000_emit_prologue): Ditto.
786 (rs6000_emit_epilogue): Ditto.
787 (rs6000_elf_file_end): Ditto.
788 (rs6000_function_value): Ditto.
789 (rs6000_libcall_value): Ditto.
790 * config/rs6000/rs6000.h: Ditto.
791 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
792 (TARGET_MINMAX): ... this. New.
793 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
794 * config/rs6000/rs6000.md: Remove Xilinx FP support.
795 (*movsi_internal1_single): Delete.
796 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
797 mfpu=, mxilinx-fpu): Delete.
798 * config/rs6000/singlefp.h: Delete.
799 * config/rs6000/sysv4.h: Remove Xilinx FP support.
800 * config/rs6000/t-rs6000: Ditto.
801 * config/rs6000/t-xilinx: Delete.
802 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
803 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
804 (VStype_simple): Delete.
805 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
806 * config/rs6000/xfpu.h: Delete.
807 * config/rs6000/xfpu.md: Delete.
808 * config/rs6000/xilinx.h: Delete.
809 * config/rs6000/xilinx.opt: Delete.
810 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
811 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
813 2018-05-04 Tom de Vries <tom@codesourcery.com>
816 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
819 2018-05-04 Richard Biener <rguenther@suse.de>
822 * tree-complex.c (update_complex_assignment): We are always in SSA form.
823 (expand_complex_div_wide): Likewise.
824 (expand_complex_operations_1): Likewise.
825 (expand_complex_libcall): Preserve EH info of the original stmt.
826 (tree_lower_complex): Handle removed blocks.
827 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
828 on complex multiplication and division libcall builtins.
830 2018-05-04 Richard Biener <rguenther@suse.de>
833 * fold-const.c (negate_expr_p): Restrict negation of operand
834 zero of a division to when we know that can happen without
836 (fold_negate_expr_1): Likewise.
838 2018-05-04 Jakub Jelinek <jakub@redhat.com>
841 * real.h (real_nextafter): Declare.
842 * real.c (real_nextafter): New function.
843 * fold-const-call.c (fold_const_nextafter): New function.
844 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
846 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
847 even when arg1_mode is different from arg0_mode.
849 2018-05-03 Nathan Sidwell <nathan@acm.org>
851 * doc/extend.texi (Deprecated Features): Remove
853 (Backwards Compatibility): Likewise.
854 * doc/invoke.texi (C++ Language Options): Likewise.
855 (C++ Dialect Options): Likewise.
857 2018-05-03 Jakub Jelinek <jakub@redhat.com>
860 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
861 _mm512_mask_mullox_epi64): New intrinsics.
863 2018-05-03 Tom de Vries <tom@codesourcery.com>
866 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
867 dump files): Add offload-tree.
869 2018-05-03 Richard Biener <rguenther@suse.de>
871 PR tree-optimization/85615
872 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
873 to loops not nested in BBs loop father to avoid creating multi-entry
876 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
878 PR tree-optimization/70291
879 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
880 arguments. Change return type to tree. Emit libcall as a new
881 statement rather than replacing existing one when inplace_p is true.
882 (expand_complex_multiplication_components): New function.
883 (expand_complex_multiplication): Expand floating-point complex
884 multiplication using the above.
885 (expand_complex_division): Rename inner_type parameter to type.
886 Update expand_complex_libcall call-site.
887 (expand_complex_operations_1): Update expand_complex_multiplication
888 and expand_complex_division call-sites.
890 2018-05-02 Jakub Jelinek <jakub@redhat.com>
893 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
894 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
895 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
896 the highest significant bit of the shift count mask is clear. In
897 check whether and[sq]i3 is needed verify that all significant bits
898 of the shift count other than the highest are set.
900 2018-05-02 Tom de Vries <tom@codesourcery.com>
903 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
904 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
905 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
906 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
907 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
908 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
909 __builtin_goacc_parlevel_size.
911 2018-05-02 Richard Biener <rguenther@suse.de>
913 PR tree-optimization/85597
914 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
915 do not use split vect_get_vec_defs call but call vect_get_slp_defs
918 2018-05-02 Tom de Vries <tom@codesourcery.com>
921 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
922 dump files): Add ltrans-tree.
924 2018-05-02 Tom de Vries <tom@codesourcery.com>
927 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
928 dump files): Add wpa-ipa.
930 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
932 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
933 powerpc*-*-linux*paired* target.
934 * config/rs6000/750cl.h: Delete.
935 * config/rs6000/paired.h: Delete.
936 * config/rs6000/paired.md: Delete.
937 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
939 * config/rs6000/rs6000-builtin.def: Remove paired float support.
940 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
941 comment. Remove paired float support.
942 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
943 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
945 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
946 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
948 * config/rs6000/rs6000.c: Remove paired float support.
949 (paired_expand_vector_init, paired_expand_vector_move,
950 paired_emit_vector_compare, paired_emit_vector_cond_expr,
951 (paired_expand_lv_builtin, paired_expand_stv_builtin,
952 paired_expand_builtin, paired_expand_predicate_builtin,
953 paired_init_builtins): Delete.
954 * config/rs6000/rs6000.h: Remove paired float support.
955 * config/rs6000/rs6000.md: Remove paired float support.
956 (move_from_CR_ov_bit): Delete.
957 * config/rs6000/rs6000.opt (mpaired): Delete.
958 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
959 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
961 2018-05-02 Richard Biener <rguenther@suse.de>
964 * gimplify.c (gimplify_save_expr): When in SSA form allow
965 SAVE_EXPRs to compute to SSA vars.
967 2018-05-02 Jakub Jelinek <jakub@redhat.com>
970 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
971 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
972 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
973 clobber operands[2], instead use a new pseudo. Formatting fixes.
975 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
977 PR tree-optimization/85586
978 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
979 exit early for statements in the same group if the accesses are
982 2018-05-02 Tom de Vries <tom@codesourcery.com>
985 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
988 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
990 PR tree-optimization/85143
991 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
993 2018-05-01 Tom de Vries <tom@codesourcery.com>
996 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
997 not found" error message.
999 2018-05-01 Tom de Vries <tom@codesourcery.com>
1002 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1003 * vec.c (test_ordered_remove_if): New function.
1004 (vec_c_tests): Call test_ordered_remove_if.
1005 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1006 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1007 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1008 VEC_ORDERED_REMOVE_IF.
1010 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1012 PR tree-optimization/82665
1013 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1014 pointer subtraction where arguments come from a memchr call.
1016 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1018 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1019 --push-state --as-needed and --pop-state instead of --as-needed and
1020 --no-as-needed if ld supports it.
1021 * configure: Regenerated.
1024 * doc/install.texi2html: Replace _002d with - and _002a with * in
1025 generated html files using sed.
1027 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1030 * gcc-rich-location.c (blank_line_before_p): New function.
1031 (use_new_line): New function.
1032 (gcc_rich_location::add_fixit_insert_formatted): New function.
1033 * gcc-rich-location.h
1034 (gcc_rich_location::add_fixit_insert_formatted): New function.
1036 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1038 * selftest.c (assert_streq): Rename "expected" and "actual" to
1039 "val1" and "val2". Extend NULL-handling to cover both inputs
1040 symmetrically, while still requiring both to be non-NULL for a pass.
1041 * selftest.h (assert_streq): Rename "expected" and "actual" to
1043 (ASSERT_EQ): Likewise.
1044 (ASSERT_EQ_AT): Likewise.
1045 (ASSERT_KNOWN_EQ): Likewise.
1046 (ASSERT_KNOWN_EQ_AT): Likewise.
1047 (ASSERT_NE): Likewise.
1048 (ASSERT_MAYBE_NE): Likewise.
1049 (ASSERT_MAYBE_NE_AT): Likewise.
1050 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1051 the assertion to pass.
1052 (ASSERT_STREQ_AT): Likewise.
1054 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1056 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1057 interaction with -pie.
1059 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1061 * selftest.h: Fix alphabetization of per-source-file selftest
1064 2018-04-30 Jason Merrill <jason@redhat.com>
1066 PR c++/61982 - dead stores to destroyed objects.
1067 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1070 2018-04-30 Jason Merrill <jason@redhat.com>
1072 * tree.c (build_clobber): New.
1073 * tree.h: Declare it.
1074 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1076 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1078 * diagnostic-show-locus.c (layout::layout): Update for
1079 location_get_source_line returning a char_span.
1080 (struct char_span): Move to input.h.
1081 (struct correction): Update for fields in char_span becoming
1083 (struct source_line): Update for location_get_source_line
1084 returning a char_span.
1085 (layout::print_line): Likewise.
1086 * edit-context.c (edited_file::print_content): Likewise.
1087 (edited_file::print_diff_hunk): Likewise.
1088 (edited_file::print_run_of_changed_lines): Likewise.
1089 (edited_file::get_num_lines): Likewise.
1090 (edited_line::edited_line): Likewise.
1091 * final.c (asm_show_source): Likewise.
1092 * input.c (location_get_source_line): Convert return type
1093 from const char * to char_span, losing the final "line_len"
1095 (dump_location_info): Update for the above.
1096 (get_substring_ranges_for_loc): Likewise. Use a char_span
1097 when handling the literal within the line.
1098 (test_reading_source_line): Update for location_get_source_line
1099 returning a char_span.
1100 * input.h (class char_span): Move here from
1101 diagnostic-show-locus.c, converting from a struct to a class.
1102 Make data members private.
1103 (char_span::operator bool): New.
1104 (char_span::length): New.
1105 (char_span::get_buffer): New.
1106 (char_span::operator[]): New.
1107 (char_span::subspan): Make const.
1108 (char_span::xstrdup): New.
1109 (location_get_source_line): Convert return type from const char *
1110 to char_span, losing the final "line_size" param.
1112 2018-04-30 Jan Hubicka <jh@suse.cz>
1114 * lto-wrapper.c (ltrans_priorities): New static var.
1115 (cmp_priority): New.
1116 (run_gcc): Read priorities and if doing parallel build order
1117 the Makefile by them.
1119 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1121 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1123 2018-04-30 Richard Biener <rguenther@suse.de>
1125 * tree-cfg.c (verify_address): Remove base argument, add
1126 flag whether to check TREE_ADDRESSABLE and do that.
1127 (verify_expr): Remove.
1128 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1129 (verify_gimple_assign_single): Likewise.
1130 (verify_gimple_switch): Likewise.
1131 (verify_expr_location_1): Dereference tp once. Add (disabled)
1132 piece from verify_expr.
1133 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1135 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1137 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1139 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1141 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1142 (small_data_pattern): Likewise.
1143 (arc_rewrite_small_data): Likewise.
1144 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1145 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1146 (get_symbol_alignment): New function.
1147 (legitimate_small_data_address_p): Likewise.
1148 (legitimate_scaled_address): Update, call
1149 legitimate_small_data_address_p.
1150 (output_sdata): New static variable.
1151 (arc_print_operand): Update how we handle small data operands.
1152 (arc_print_operand_address): Likewise.
1153 (arc_legitimate_address_p): Update, use
1154 legitimate_small_data_address_p.
1155 (arc_rewrite_small_data_p): Remove.
1156 (arc_rewrite_small_data_1): Likewise.
1157 (arc_rewrite_small_data): Likewise.
1158 (small_data_pattern): Likewise.
1159 (compact_sda_memory_operand): Update to use
1160 legitimate_small_data_address_p and get_symbol_alignment.
1161 (prepare_move_operands): Don't rewite sdata pattern.
1162 (prepare_extend_operands): Remove.
1163 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1165 (zero_extendqisi2): Likewise.
1166 (zero_extendhisi2): Likewise.
1167 (extendqihi2): Likewise.
1168 (extendqisi2): Likewise.
1169 (extendhisi2): Likewise.
1173 * config/arc/constraints.md (Usd): Change it to memory constraint.
1175 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1177 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1178 as source of std instructions.
1179 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1180 allow 6-bit constants as source for store instructions.
1181 (movdi_insn): Update instruction pattern to allow 6-bit constants
1182 as source for store instructions.
1184 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1186 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1188 2018-04-30 Nathan Sidwell <nathan@acm.org>
1189 Sandra Loosemore <sandra@codesourcery.com>
1191 * dumpfile.c (dump_open): Allow '-' for stdout.
1192 * doc/invoke.texi (Developer Options): Document dump filename
1193 determination early. Document stdin/stdout selection.
1195 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1197 Microblaze Target: PIC data text relative
1199 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1200 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1202 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1203 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1204 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1205 New addressing mode for data-text relative position indepenedent code.
1206 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1207 'ADDRESS_SYMBOLIC_TXT_REL'.
1208 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1209 (microblaze_legitimate_pic_operand): Exclude function calls from
1210 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1211 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1213 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1214 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1215 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1216 for 'address + offset'.
1217 (microblaze_expand_prologue): Add new function prologue call for
1219 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1220 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1221 table in case of TARGET_PIC_DATA_TEXT_REL.
1222 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1223 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1224 Add new macros 'UNSPEC_TEXT',
1225 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1226 + exclude function calls from 'UNSPEC_PLT' in case of data text
1228 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1229 new target hook for generating address diff vector tables in case of
1231 * doc/tm.texi : Regenerate.
1232 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1233 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1234 of addr diff vector generation.
1235 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1236 target hook definition.
1237 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1238 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1239 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1240 option for data text relative.
1242 2018-04-30 Richard Biener <rguenther@suse.de>
1244 * tree-chrec.h (evolution_function_is_constant_p): Remove
1246 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1248 2018-04-30 Richard Biener <rguenther@suse.de>
1251 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1253 2018-04-30 Richard Biener <rguenther@suse.de>
1255 PR tree-optimization/28364
1256 PR tree-optimization/85275
1257 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1258 copying first exit test.
1260 2018-04-28 Mark Wielaard <mark@klomp.org>
1262 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1264 (dwarf_AT): Handle DW_AT_addr_base.
1265 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1267 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1270 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1271 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1272 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1273 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1275 2018-04-28 Richard Biener <rguenther@suse.de>
1277 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1278 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1279 to reflect use. Only add interesting stmts.
1281 2018-04-27 Martin Jambor <mjambor@suse.cz>
1284 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1285 the jump function allows for passing through aggregate values.
1287 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1289 * input.h (in_system_header_at): Convert from macro to inline
1291 (from_macro_expansion_at): Likewise.
1292 (from_macro_definition_at): Likewise.
1294 2018-04-27 Jeff Law <law@redhat.com>
1296 * config.gcc: Mark tile* targets as deprecated/obsolete.
1298 2018-04-27 Richard Biener <rguenther@suse.de>
1300 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1303 2018-04-27 Richard Biener <rguenther@suse.de>
1305 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1307 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1309 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1310 with Yd constraint. Set "preferred_for_speed" attribute from
1311 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1313 (*movdi_internal): Ditto.
1314 (movti_interunit splitters): Remove
1315 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1316 (movdi_interunit splitters): Ditto.
1317 * config/i386/constraints.md (Ye): Remove.
1318 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1320 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1323 * config/aarch64/constraints.md (Usg): Limit to 31.
1326 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1328 PR tree-optimization/85529
1329 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1330 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1331 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1332 zero extension or masking of the MSB bit.
1333 (optimize_range_tests): Add FIRST_BB argument, pass it through
1334 to optimize_range_tests_var_bound.
1335 (maybe_optimize_range_tests, reassociate_bb): Adjust
1336 optimize_range_tests callers.
1338 2018-04-26 Richard Biener <rguenther@suse.de>
1339 Jakub Jelinek <jakub@redhat.com>
1341 * cgraph.h (symbol_table): Just declare debug method here.
1342 * symtab.c (symbol_table::debug): Define.
1344 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1346 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1348 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1350 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1351 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1352 (*movdi_internal): Substitute Yi and Yj constraint with x
1353 and Ym and Yn constraint with y constraint. Update "isa"
1354 attribute and set "preferred_for_speed" attribute from
1355 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1356 (*movsi_internal): Ditto.
1357 (*movdf_internal): Ditto.
1358 (*movsf_internal): Ditto.
1359 (*zero_extendsidi2): Ditto.
1360 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1361 (sse2_loadld): Ditto.
1362 (*vec_extract<ssevecmodelower>_0): Ditto.
1363 (*vec_extractv4si_0_zext_sse4): Ditto.
1364 (vec_concatv2di): Ditto.
1365 (*vec_dup<mode>): Ditto.
1366 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1367 * config/i386/constraints.md (Yi): Remove.
1372 2018-04-26 Nathan Sidwell <nathan@acm.org>
1374 * dumpfile.c (dump_open): New.
1375 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1376 (dump_finish): Detect stdio/stderr by value not name.
1378 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1380 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1382 2018-04-26 Tom de Vries <tom@codesourcery.com>
1385 * config/nvptx/nvptx.c (verify_neutering_jumps)
1386 (verify_neutering_labels): New function
1387 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1389 2018-04-26 Tom de Vries <tom@codesourcery.com>
1392 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1393 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1396 2018-04-26 Richard Biener <rguenther@suse.de>
1397 Tom de Vries <tom@codesourcery.com>
1400 * lto-streamer-out.c (output_function): Fixup loops if required to match
1401 discovery done in the reader.
1403 2018-04-26 Richard Biener <rguenther@suse.de>
1405 PR tree-optimization/85116
1406 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1407 have a loop exit from the single latch predecessor. Remove
1408 case of header with just condition.
1409 (ch_base::copy_headers): Exclude infinite loops from any
1411 (pass_ch::execute): Record exits.
1413 2018-04-26 Richard Biener <rguenther@suse.de>
1415 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1416 prologue cost vector and pass it to vect_get_load_cost.
1417 (vect_get_peeling_costs_all_drs): Likewise.
1418 (vect_peeling_hash_get_lowest_cost): Likewise.
1419 (vect_enhance_data_refs_alignment): Likewise.
1421 2018-04-26 Richard Biener <rguenther@suse.de>
1424 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1425 checking of integer<->pointer conversions.
1426 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1427 sign-/zero-extending pointer types.
1428 (expand_omp_for_static_chunk): Likewise.
1430 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1431 Jean Lee <xiaoyur347@gmail.com>
1433 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1434 (TARGET_ASAN_SHADOW_OFFSET): Define.
1435 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1436 true for -fsanitize=address.
1438 2018-04-25 Mark Wielaard <mark@klomp.org>
1440 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1443 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1445 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1446 than "alu", remove explicit "memory" and "imm_disp" attributes.
1447 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1450 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1451 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1452 gen_lowpart_no_emit.
1454 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1457 * config/i386/i386.c (ix86_expand_builtin): Change memory
1458 operand to XI, extend p0 to Pmode.
1459 * config/i386/i386.md: Change unspec volatile and operand
1460 1 mode to XI, change operand 0 mode to P.
1462 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1464 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1465 GET_MODE_MASK before any checking.
1466 (nds32_can_use_bset_p): Likewise.
1467 (nds32_can_use_btgl_p): Likewise.
1469 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1471 * config/nds32/nds32-doubleword.md: New define_split pattern for
1472 illegal register number.
1474 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1476 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1478 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1480 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1482 2018-04-25 Richard Biener <rguenther@suse.de>
1484 * lto-streamer.h (LTO_major_version): Bump to 8.
1486 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1488 * BASE-VER: Set to 9.0.0.
1490 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1492 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1493 in __abskf2 and __powikf2.
1495 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1498 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1499 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1500 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1501 Use the above on operand 2. Reindent.
1502 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1504 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1507 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1509 * config/i386/i386.opt (mcet): Removed.
1510 * doc/install.texi: Remove -mcet documentation.
1511 * doc/invoke.texi: Likewise.
1513 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1516 * doc/install.texi: Remove -mcet from bootstrap-cet.
1518 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1521 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1522 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1526 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1527 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1528 containing a CONST_VECTOR.
1530 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1532 * doc/install.texi: Update newlib dependency for nvptx.
1534 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1537 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1538 instead of INTVAL when shifting x left.
1540 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1542 PR tree-optimization/85478
1543 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1544 vect_grouped_store_supported for single element vectors.
1546 2018-04-24 Richard Biener <rguenther@suse.de>
1549 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1550 load cost increase to the case of non-constant step.
1552 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1555 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1556 destination if any_malformed_asm.
1558 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1561 * expr.c (store_field): In the bitfield case, if the value comes from
1562 a function call and is returned in registers by means of a PARALLEL,
1563 do not change the mode of the temporary unless BLKmode and VOIDmode.
1565 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1567 PR rtl-optimization/85423
1568 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1569 dependencies to debug insns when the previous insn is non-debug.
1571 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1573 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1574 enums into a single definition.
1575 (fls): Fix predicates and printing.
1578 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1580 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1581 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1582 and short u6 immediate.
1583 (check_if_valid_sleep_operand): Remove.
1584 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1586 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1588 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1589 flag_always_save_lp condition.
1590 * config/nds32/nds32.opt (malways-save-lp): New option.
1592 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1594 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1595 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1596 * config/nds32/nds32.h
1597 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1598 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1600 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1602 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1603 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1605 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1606 Chung-Ju Wu <jasonwucj@gmail.com>
1608 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1610 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1611 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1613 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1615 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1617 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1619 * config/nds32/nds32-protos.h (nds32_data_alignment,
1620 nds32_local_alignment): Declare.
1621 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1622 nds32_local_alignment): New functions.
1623 (TARGET_CONSTANT_ALIGNMENT): Define.
1624 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1626 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1628 * config/nds32/nds32.c
1629 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1630 (TARGET_MODES_TIEABLE_P): Likewise.
1632 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1634 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1637 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1638 Chung-Ju Wu <jasonwucj@gmail.com>
1640 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1641 for unaligned access.
1642 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1644 * config/nds32/nds32-intrinsic.md: Likewise.
1645 * config/nds32/nds32_intrinsic.h: Likewise.
1646 * config/nds32/nds32.h (nds32_builtins): Likewise.
1647 * config/nds32/nds32.opt (munaligned-access): New option.
1648 * config/nds32/nds32.c (nds32_asm_file_start): Display
1649 flag_unaligned_access status.
1651 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1653 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1654 -mno-relax is present.
1655 * config/riscv/linux.h (LINK_SPEC): Ditto.
1657 2018-04-20 Martin Sebor <msebor@redhat.com>
1660 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1662 (gimple_fold_builtin_stxcpy_chk): Same.
1663 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1665 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1668 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1669 __powikf2 when long double is IEEE 128-bit.
1671 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1673 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1674 step to make sure stack always aligned.
1676 2018-04-20 Carl Love <cel@us.ibm.com>
1679 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1680 size check for arg0.
1682 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1683 Tom de Vries <tom@codesourcery.com>
1686 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1687 Emit insns for calls too.
1688 (nvptx_find_par): Always look for worker-level predecessor insn.
1689 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1691 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1692 (nvptx_process_pars): Propagate frames for calls.
1694 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1697 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1699 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1700 (ix86_handle_option): Don't handle OPT_mibt.
1701 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1703 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1705 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1706 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1707 (ix86_target_macros): Define __CET__ with flag_cf_protection
1708 for -fcf-protection.
1709 * config/i386/i386.c (isa2_opts): Remove -mibt.
1710 * config/i386/i386.h (TARGET_IBT): Removed.
1711 (TARGET_IBT_P): Likewise.
1712 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1713 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1714 * config/i386/i386.opt (mcet): Update help message.
1717 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1718 -mcet as an alias for -mshstk.
1720 2018-04-20 Richard Biener <rguenther@suse.de>
1723 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1724 complexity by forcing a single use of the multiply operand.
1726 2018-04-20 Martin Jambor <mjambor@suse.cz>
1729 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1730 recursion dependency to only apply to non-clones.
1732 2018-04-20 Martin Jambor <mjambor@suse.cz>
1735 * ipa-cp.c (create_specialized_node): Check that clones of
1736 self-recursive edges exist during IPA-CP.
1738 2018-04-19 Toon Moene <toon@moene.org>
1740 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1743 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1745 PR tree-optimization/85467
1746 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1747 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1748 VECTOR_CST element to type.
1750 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1753 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1754 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1755 (builtin_longjmp): Likewise.
1756 (save_stack_nonlocal): New pattern.
1757 (restore_stack_nonlocal): Likewise.
1759 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1762 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1763 Replace ASM_OUTPUT_LABEL with fprintf.
1765 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1768 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1769 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1770 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1771 define __IBT__ and __SHSTK__ for -fcf-protection.
1772 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1774 (ix86_trampoline_init): Likewise.
1775 (x86_output_mi_thunk): Likewise.
1776 (ix86_notrack_prefixed_insn_p): Likewise.
1777 (ix86_option_override_internal): Don't disallow -fcf-protection.
1778 * config/i386/i386.md (rdssp<mode>): Also enable for
1780 (incssp<mode>): Likewise.
1781 (nop_endbr): Likewise.
1782 * config/i386/i386.opt (mcet): Change help message to built-in
1786 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1787 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1788 enable CET built-in functions.
1790 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1792 * common/config/i386/i386-common.c
1793 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1794 OPTION_MASK_ISA_MOVDIRI_UNSET,
1795 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1796 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1797 * config.gcc (movdirintrin.h): New header.
1798 * config/i386/cpuid.h (bit_MOVDIRI,
1799 bit_MOVDIR64B): New bits.
1800 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1802 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1803 (VOID, PVOID, PCVOID)): New function types.
1804 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1805 __builtin_ia32_directstoreu_u64,
1806 __builtin_ia32_movdir64b): New builtins.
1807 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1808 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1810 (ix86_valid_target_attribute_inner_p): Ditto.
1811 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1812 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1813 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1814 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1815 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1816 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1817 (movdiri<mode>, movdir64b_<mode>): New.
1818 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1819 * config/i386/immintrin.h: Include movdirintrin.h.
1820 * config/i386/movdirintrin.h: New file.
1821 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1823 2018-04-19 Richard Biener <rguenther@suse.de>
1826 * cfg.c (clear_bb_flags): When loop state says we have
1827 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1829 2018-04-19 Richard Biener <rguenther@suse.de>
1831 PR tree-optimization/84737
1832 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1833 copying restrict info.
1834 (vect_setup_realignment): Use it.
1835 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1836 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1837 the first DR to all generated stores.
1838 (vectorizable_load): Likewise for loads.
1840 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1842 PR tree-optimization/85446
1843 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1844 the integral and pointer types to have the same precision.
1846 * doc/install.texi: Document --disable-cet being the default and
1849 2018-04-18 Martin Liska <mliska@suse.cz>
1851 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1854 2018-04-18 Martin Liska <mliska@suse.cz>
1857 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1860 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1861 arguments if they are comparable.
1863 2018-04-18 Martin Liska <mliska@suse.cz>
1866 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1869 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1872 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1875 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1876 ENDBR after calling __morestack.
1878 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1881 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1882 by using gcc_base_ver to generate a gcc_driver_version, and use
1883 it when generating GCC_DRIVER_NAME.
1884 * configure: Regenerate.
1886 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1889 * config.gcc: Obsolete powerpc*-*-*spe*.
1891 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1894 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1895 (stabstr_D): Change type of unum from unsigned int to
1896 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1899 2018-04-17 Jim Wilson <jimw@sifive.com>
1902 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1903 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1904 Set arg_pointer_offset after using pretend_args_size.
1906 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1908 PR rtl-optimization/85431
1909 * dse.c (record_store): Ignore zero width stores.
1912 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1913 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1914 __builtin_stack_restore rather than after it.
1915 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1916 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1917 argument instead of virtual_dynamic_stack_rtx.
1919 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1921 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1923 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1924 Add note to error message to explain internal mapping of overloaded
1925 built-in function name to non-overloaded built-in function name.
1926 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1929 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1932 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1933 where the inputs overlap with the output.
1935 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1937 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1938 (=v, v) alternative and explicit "memory" attribute.
1939 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1940 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1942 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1943 "sselog1" type instead of "sselog".
1944 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1945 "sselog". Remove explicit "memory" attribute.
1946 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1947 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1949 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1950 "sselog1" type instead of "sselog". Remove explicit "memory"
1952 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1953 use "sselog1" type instead of "sselog". Remove explicit "memory"
1955 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1956 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1958 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1959 "sselog1" type instead of "sselog". Remove explicit "memory"
1961 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1962 use "sselog1" type instead of "sselog". Remove explicit "memory"
1966 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1969 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1972 2018-04-17 Martin Jambor <mjambor@suse.cz>
1975 * ipa-cp.c (create_specialized_node): Call
1976 expand_all_artificial_thunks if necessary.
1978 2018-04-17 Martin Liska <mliska@suse.cz>
1981 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1982 in message, remote space in between '_G' and '('.
1984 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1987 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1988 avx512f_vmcmp<mode>3<round_saeonly_name>,
1989 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1990 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1991 avx512f_rndscale<mode><round_saeonly_name>,
1992 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1993 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1994 Use %<iptr>2 instead of %2 for -masm=intel.
1995 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1996 avx512f_vcvttss2usi<round_saeonly_name>,
1997 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1999 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2000 avx512f_vcvttsd2usi<round_saeonly_name>,
2001 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2002 Use %q1 instead of %1 for -masm=intel.
2003 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2004 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2005 of %3 for -masm=intel.
2006 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2008 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2010 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2012 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2013 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2015 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2017 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2018 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2019 %g1 and one with %0 and %1.
2020 (avx512er_vmrcp28<mode><round_saeonly_name>,
2021 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2023 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2024 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2025 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2026 of %0 and %{%4%} for -masm=intel.
2027 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2028 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2029 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2030 order of %0 and %{%5%}%{z%} for -masm=intel.
2032 2018-04-17 Jan Hubicka <jh@suse.cz>
2035 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2037 2018-04-17 Martin Liska <mliska@suse.cz>
2040 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2041 for target_clone error message. Make default implementation
2042 clone to be a local declaration.
2043 (separate_attrs): Add new argument and check for an empty
2045 (expand_target_clones): Handle it.
2046 (ipa_target_clone): Make redirection just for target_clones
2049 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2050 Tom de Vries <tom@codesourcery.com>
2053 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2054 tiled basic blocks without omp continue statements.
2056 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2059 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2060 vec_extract expression as having side effects to make sure it gets
2063 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2066 * config/i386/i386.c (get_builtin_code_for_version): Check
2069 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2072 * gcc/config.gcc: Support "skylake".
2073 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2075 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2076 (processor_target_table): Add "skylake".
2077 (ix86_option_override_internal): Add "skylake".
2078 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2079 PROCESSOR_CANNONLAKE.
2080 (get_builtin_code_for_version): Fix priority for
2081 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2082 PROCESSOR_SKYLAKE-AVX512.
2083 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2084 (processor_type): Add PROCESSOR_SKYLAKE.
2086 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2087 Jason Merrill <jason@redhat.com>
2090 * convert.c (convert_to_integer_1): Use direct recursion for
2091 enumeral types and types with a precision less than the number
2092 of bits in their mode.
2094 2018-04-16 Julia Koval <julia.koval@intel.com>
2097 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2098 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2100 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2103 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2104 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2105 and -mno-direct-move.
2107 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2110 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2111 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2112 Ensure negative shifts result in {0}.
2114 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2116 PR rtl-optimization/79916
2117 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2118 regs (if any) to define how to gnerate SD moves when LRA is in
2121 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2123 PR rtl-optimization/85393
2124 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2125 * except.c (expand_dw2_landing_pad_for_region): Make static.
2126 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2127 a label and unconditional jump to old_bb, rather than
2128 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2131 PR rtl-optimization/85376
2132 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2133 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2134 instead of a specific value.
2136 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2137 Bin Cheng <bin.cheng@arm.com>
2139 PR tree-optimization/82965
2140 PR tree-optimization/83991
2141 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2142 by_profile_only parameter.
2143 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2144 information if the loop was predicted to iterate too many times.
2145 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2147 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2150 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2153 2018-04-13 Martin Liska <mliska@suse.cz>
2154 Jakub Jelinek <jakub@redhat.com>
2157 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2158 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2159 * builtins.c (expand_builtin_memory_copy_args): Use
2160 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2161 handle dest_addr == pc_rtx.
2163 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2166 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2167 asked to not generate direct moves.
2168 (fix_trunc<mode>si2_stfiwx): Similar.
2169 (fix_trunc<mode>si2_internal): Similar.
2171 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2174 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2175 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2176 lookup if dest in some wider mode is known to be const0_rtx and
2177 if so, record permanent equivalence for it to be ZERO_EXTEND of
2178 the narrower mode destination.
2180 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2182 * lto-streamer-out.c (output_function): Revert 259346.
2183 * omp-expand.c (expand_oacc_for): Likewise.
2185 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2187 PR rtl-optimization/85354
2188 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2189 * sel-sched.c (sel_global_init): ... here.
2191 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2194 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2195 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2196 mode for PE-COFF targets.
2197 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2198 (i386_pe_asm_lto_end): Likewise.
2199 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2200 (TARGET_ASM_LTO_END): Likewise.
2201 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2202 (i386_pe_asm_lto_start): New function.
2203 (i386_pe_asm_lto_end): Likewise.
2205 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2206 Richard Biener <rguenther@suse.de>
2209 * lto-streamer-out.c (output_function): Fix CFG loop state before
2211 * omp-expand.c (expand_oacc_for): Handle calls to internal
2212 functions like regular functions.
2214 2018-04-12 Richard Biener <rguenther@suse.de>
2217 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2218 for the early LTO debug to properly generate references to it
2219 during DIE emission. Do not re-use that for the skeleton for
2221 (dwarf2out_early_finish): Likewise.
2223 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2226 * config/i386/sse.md
2227 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2228 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2229 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2230 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2231 and output is a reg, avoid creating invalid lowpart subreg, but
2232 instead split into a 512-bit move. Don't split if not AVX512VL,
2233 input is xmm16+ reg and output is a mem.
2234 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2235 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2236 xmm16+ reg and output is a mem.
2238 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2240 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2241 also for flag_dwarf2_cfi_asm.
2243 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2245 PR rtl-optimization/85342
2246 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2247 a bool scalar var inside of the loop instead. Don't try to update
2248 recog_data.operand after failed apply_change_group.
2250 2018-04-12 Tom de Vries <tom@codesourcery.com>
2253 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2254 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2255 array with flexible array member as array without given dimension.
2256 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2257 argument for undefined param to true.
2259 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2262 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2263 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2264 from PowerPC section.
2265 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2266 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2267 help text that is too long.
2268 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2269 help text that is too long.
2270 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2271 help text that is too long.
2273 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2275 * config/alpha/alpha.md (stack_probe_internal): Rename
2276 from "probe_stack". Update all callers.
2278 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2280 PR rtl-optimization/84566
2281 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2282 sched_macro_fuse_insns.
2284 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2287 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2288 (compute_block_dependences): ... from here.
2290 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2292 PR tree-optimization/85331
2293 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2294 from int to HOST_WIDE_INT.
2296 2018-04-11 Martin Jambor <mjambor@suse.cz>
2299 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2300 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2301 not the same as the source val.
2302 (cgraph_edge_brings_value_p): New parameter.
2303 (gather_edges_for_value): Pass destination value to
2304 cgraph_edge_brings_value_p.
2305 (perhaps_add_new_callers): Likewise.
2306 (get_info_about_necessary_edges): Likewise and exclude values brought
2307 only by self-recursive edges.
2308 (create_specialized_node): Redirect only clones of self-calling edges.
2309 (+self_recursive_pass_through_p): New function.
2310 (find_more_scalar_values_for_callers_subset): Use it.
2311 (find_aggregate_values_for_callers_subset): Likewise.
2312 (known_aggs_to_agg_replacement_list): Removed.
2313 (decide_whether_version_node): Re-calculate known constants for all
2314 remaining context clones.
2316 2018-04-11 Richard Biener <rguenther@suse.de>
2319 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2320 from early DWARF output.
2321 (dwarf2out_early_finish): Output line info unconditionally into
2322 early DWARF and add reference to it.
2324 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2327 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2328 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2329 other than V2DFmode using iptr mode attribute.
2330 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2332 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2334 PR rtl-optimization/84659
2335 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2337 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2340 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2342 (output_loc_list): Pass address of a dummy size variable even in the
2343 locview handling loop.
2344 (index_location_lists): Add comment on why skip_loc_list_entry can't
2347 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2350 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2353 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2356 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2357 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2358 and -mstring-compare-inline-limit.
2360 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2363 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2364 for stack clash protection in a register whenever we need it to be in
2367 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2369 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2370 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2372 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2375 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2377 (mlong-double-): Ditto.
2378 * config/rs6000/sysv4.opt (msdata=): Ditto.
2379 (mtls-size=): Ditto.
2381 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2383 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2384 erroneous entries for
2385 "vector int vec_ldl (int, long int *)", and
2386 "vector unsigned int vec_ldl (int, unsigned long int *)".
2387 Add comments and entries for
2388 "vector bool char vec_ldl (int, bool char *)",
2389 "vector bool short vec_ldl (int, bool short *)",
2390 "vector bool int vec_ldl (int, bool int *)",
2391 "vector bool long long vec_ldl (int, bool long long *)",
2392 "vector pixel vec_ldl (int, pixel *)",
2393 "vector long long vec_ldl (int, long long *)",
2394 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2395 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2396 type tree bool_long_long_type_node and correct definition of
2397 bool_V2DI_type_node to make reference to this new type tree.
2398 (rs6000_mangle_type): Replace erroneous reference to
2399 bool_long_type_node with bool_long_long_type_node.
2400 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2401 comments to emphasize sign distinctions for char and int types and
2402 replace RS6000_BTI_bool_long constant with
2403 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2404 use of RS6000_BTI_pixel.
2405 (bool_long_type_node): Remove this macro definition.
2406 (bool_long_long_type_node): New macro definition
2408 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2410 PR rtl-optimization/85300
2411 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2412 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2413 simplify_unary_operation fails.
2415 2018-04-10 Martin Liska <mliska@suse.cz>
2417 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2418 cgraph_edge and ipa_ref.
2420 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2424 * config/i386/sse.md
2425 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2426 computation of the VEC_MERGE selector from mask.
2427 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2428 Fix decoding of the VEC_MERGE selector into mask.
2430 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2432 PR tree-optimization/85286
2433 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2435 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2437 * final.c (final_1): Set insn_last_address as well as
2438 insn_current_address.
2440 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2443 * explow.c (emit_stack_probe): Call validize_mem on memory location
2444 before passing it to gen_probe_stack. Create address operand and
2445 legitimize it for the probe_stack_address case.
2447 2018-04-09 Jan Hubicka <jh@suse.cz>
2450 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2451 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2452 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2454 (free_lang_data): Rebuild type inheritance graph.
2456 2018-04-09 Martin Sebor <msebor@redhat.com>
2458 * invoke.texi (-finline-small-functions): Mention other optimization
2460 (-findirect-inlining, -fpartial-inlining): Same.
2461 (-finline-functions-called-once): Same.
2462 (-freorder-blocks-and-partition): Same.
2464 2018-04-09 Jan Hubicka <jh@suse.cz>
2467 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2468 jumps; choose last target that matches the criteria (i.e.
2469 no partition changes for non-crossing jumps).
2470 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2471 support for redirecting crossing jumps to non-crossing.
2473 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2475 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2476 also for naked functions.
2478 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2480 * config/arc/arc.md (add_shift): New pattern.
2481 (add_shift2): Likewise.
2482 (sub_shift): Likewise.
2483 (sub_shift_cmp0_noout): Likewise.
2484 (compare_si_ashiftsi): Likewise.
2485 (xbfu_cmp0_noout): New combine pattern.
2486 (xbfu_cmp0"): Likewise.
2487 (movsi_set_cc_insn): Place the predicable variant first.
2488 (commutative_binary_cmp0_noout): Remove clobber.
2489 (commutative_binary_cmp0): New pattern.
2490 (noncommutative_binary_cmp0): Likewise.
2491 (noncommutative_binary_cmp0_noout): Likewise.
2492 (noncommutative_binary_comparison_result_used): Removed.
2493 (rsub_cmp0): New pattern.
2494 (rsub_cmp0_noout): Likewise.
2495 (extzvsi): Changed, keep only meaningful variants.
2496 (SQH, SEZ): New iterators.
2497 (SQH_postfix): New mode attribute.
2498 (SEZ_prefix): New code attribute.
2499 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2500 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2501 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2503 (noncommutative_operator): Check the availability of barrel
2506 2018-04-09 Richard Biener <rguenther@suse.de>
2508 PR tree-optimization/85284
2509 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2510 Only use the niter constraining form of simple_iv when the exit
2513 2018-04-09 Tom de Vries <tom@codesourcery.com>
2516 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2517 (define_expand "*memory_barrier"): New define_expand.
2518 (define_insn "memory_barrier"): New insn.
2520 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2522 PR rtl-optimization/80463
2523 PR rtl-optimization/83972
2524 PR rtl-optimization/83480
2526 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2527 correct producer for the insn.
2528 (tidy_control_flow): Fixup seqnos in case of debug insns.
2530 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2532 PR rtl-optimization/83913
2534 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2535 different sched-times when merging exprs.
2537 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2539 PR rtl-optimization/83962
2541 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2542 tidy_fallthru_edge and tidy_control_flow.
2544 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2546 PR rtl-optimization/83530
2548 * sel-sched.c (force_next_insn): New global variable.
2549 (remove_insn_for_debug): When force_next_insn is true, also leave only
2550 next insn in the ready list.
2551 (sel_sched_region): When the region wasn't scheduled, make another pass
2552 over it with force_next_insn set to 1.
2554 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2556 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2558 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2559 for interrupt control.
2560 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2561 functions for interrupt control.
2562 * config/nds32/nds32-intrinsic.md: Likewise.
2563 * config/nds32/nds32_intrinsic.h: Likewise.
2564 * config/nds32/nds32.h (nds32_builtins): Likewise.
2566 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2568 * config/nds32/nds32.c (nds32_init_machine_status,
2569 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2570 strict_aligned_p field.
2571 (nds32_expand_to_rtl_hook): New function.
2572 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2573 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2575 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2576 Chung-Ju Wu <jasonwucj@gmail.com>
2578 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2579 * config/nds32/nds32-n7.md: New file.
2580 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2581 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2583 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2584 * config/nds32/nds32.md (pipeline_model): Add n7.
2585 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2586 * config/nds32/pipelines.md: Include n7 settings.
2588 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2589 Chung-Ju Wu <jasonwucj@gmail.com>
2591 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2592 * config/nds32/nds32-e8.md: New file.
2593 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2594 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2596 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2597 * config/nds32/nds32.md (pipeline_model): Add e8.
2598 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2599 * config/nds32/pipelines.md: Include e8 settings.
2601 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2602 Chung-Ju Wu <jasonwucj@gmail.com>
2604 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2605 * config/nds32/nds32-n8.md: New file.
2606 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2607 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2609 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2610 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2611 * config/nds32/nds32.md (pipeline_model): Add n8.
2612 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2613 * config/nds32/pipelines.md: Include n8 settings.
2615 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2616 Chung-Ju Wu <jasonwucj@gmail.com>
2618 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2619 * config/nds32/nds32-n9-2r1w.md: New file.
2620 * config/nds32/nds32-n9-3r2w.md: New file.
2621 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2622 nds32_register_ports): New or modify for cpu n9.
2623 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2625 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2626 * config/nds32/nds32-utils.c: New file.
2627 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2628 TARGET_MUL_SLOW): Define.
2629 * config/nds32/nds32.md (pipeline_model): New attribute.
2630 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2631 New options that support cpu n9.
2632 * config/nds32/pipelines.md: Include n9 settings.
2633 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2635 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2637 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2638 information if necessary.
2639 (output_cond_branch_compare_zero): Likewise.
2640 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2641 (nds32_target_alignment): Refine for alignment.
2642 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2643 (FUNCTION_BOUNDARY): Modify.
2644 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2646 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2648 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2650 * config/nds32/constants.md (unspec_volatile_element): Add values for
2651 TLB operation and data prefetch.
2652 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2653 functions for TLB operation and data prefetch.
2654 * config/nds32/nds32-intrinsic.md: Likewise.
2655 * config/nds32/nds32_intrinsic.h: Likewise.
2656 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2657 (nds32_print_operand): Likewise.
2658 * config/nds32/nds32.h (nds32_builtins): Likewise.
2660 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2661 Andrew Pinski <pinsika@gcc.gnu.org>
2664 * match.pd: Use constant_boolean_node of correct type instead of
2665 boolean_true_node or boolean_false_node for simplifying
2666 pointer comparisons to zero.
2668 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2670 PR tree-optimization/80021
2671 * tree.c (verify_type_variant): Make error call in verify_variant_match
2672 translatable and remove final full stop.
2674 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2676 * config/nds32/constants.md (unspec_volatile_element): Add
2677 UNSPEC_VOLATILE_EH_RETURN.
2678 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2679 nds32_output_stack_pop): Support dwarf exception handling process.
2680 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2681 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2682 exception handling process.
2683 (nds32_compute_stack_frame): Likewise.
2684 (nds32_return_addr_rtx): Likewise.
2685 (nds32_initial_elimination_offset): Likewise.
2686 (nds32_expand_prologue): Likewise.
2687 (nds32_expand_epilogue): Likewise.
2688 (nds32_dynamic_chain_address): New function.
2689 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2691 (DYNAMIC_CHAIN_ADDRESS): Define.
2692 (EH_RETURN_DATA_REGNO): Define.
2693 (EH_RETURN_STACKADJ_RTX): Define.
2694 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2695 patterns for dwarf exception handling.
2697 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2699 * config/nds32/nds32.h: Clean up obsolete macros.
2701 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2703 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2704 Add enum values for particular instructions.
2705 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2706 particular intrinsic functions.
2707 * config/nds32/nds32-intrinsic.md: Likewise.
2708 * config/nds32/nds32_intrinsic.h: Likewise.
2709 * config/nds32/nds32.h (nds32_builtins): Likewise.
2710 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2711 (btst, ave): New patterns for particular instructions.
2713 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2715 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2716 Add enum values for atomic load/store and memory sync.
2717 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2719 * config/nds32/nds32-intrinsic.md: Likewise.
2720 * config/nds32/nds32_intrinsic.h: Likewise.
2721 * config/nds32/nds32.h (nds32_builtins): Likewise.
2723 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2725 PR tree-optimization/85257
2726 * fold-const.c (native_encode_vector): If not all elts could fit
2727 and off is -1, return 0 rather than offset.
2728 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2729 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2730 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2731 adjust buffer in native_interpret_expr call.
2733 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2735 * config/nds32/constants.md (unspec_volatile_element): Add cache
2736 control enum values.
2737 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2738 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2739 * config/nds32/nds32.c (nds32_cctl_names): New.
2740 (nds32_print_operand): Handle cache control register names.
2741 * config/nds32/nds32.h (nds32_builtins): New enum values.
2742 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2744 * config/nds32/nds32.md (type): Add mmu.
2745 * config/nds32/pipelines.md (simple_insn): Add mmu.
2747 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2749 * config/nds32/nds32.md (type): Remove call.
2750 * config/nds32/pipelines.md (simple_insn): Likewise.
2752 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2754 * config/nds32/constants.md (unspec_volatile_element): Add
2755 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2756 UNSPEC_VOLATILE_FMFCFG.
2757 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2758 description for fmfcfg and fmfcsr.
2759 (bdesc_1arg): Add fmtcsr.
2760 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2761 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2762 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2763 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2764 unspec_fmfcfg): New patterns.
2765 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2766 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2767 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2768 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2769 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2770 __nds32__fmfcfg): Define.
2772 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2774 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2775 intrinsic register names.
2776 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2777 intrinsic register enum values and macros.
2779 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2781 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2782 for load/store addressing form.
2783 (nds32_print_operand_address): Likewise.
2785 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2788 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2789 based on LABEL_REF. Remove useless assertion.
2790 (pic_address_needs_scratch): Fix formatting.
2791 (sparc_legitimize_pic_address): Minor tweaks.
2792 (sparc_delegitimize_address): Adjust assertion accordingly.
2793 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2794 into symbolic_operand.
2795 (movsi_high_pic_label_ref): Likewise.
2796 (movsi_lo_sum_pic_label_ref): Likewise.
2797 (movdi_pic_label_ref): Likewise.
2798 (movdi_high_pic_label_ref): Likewise.
2799 (movdi_lo_sum_pic_label_ref): Likewise.
2801 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2803 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2804 custom LIB_SPEC setup.
2806 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2807 Kito Cheng <kito.cheng@gmail.com>
2809 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2810 * config/riscv/freebsd.h: New.
2812 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2814 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2815 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2818 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2819 Kito Cheng <kito.cheng@gmail.com>
2821 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2822 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2823 * config/nds32/nds32-protos.h (nds32_output_call,
2824 nds32_output_return): Declare.
2825 * config/nds32/nds32.md: Refine all the call and return patterns.
2827 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2830 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2831 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2833 PR rtl-optimization/84872
2834 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2835 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2838 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2840 * expr.c (copy_blkmode_to_reg): Revert 254862.
2841 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2843 2018-04-06 Richard Biener <rguenther@suse.de>
2846 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2847 after seeing a component reference with an adjacent field. Treat
2848 refs to arrays at struct end of external decls similar to
2849 refs to unconstrained commons.
2851 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2854 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2855 look through SAVE_EXPRs with non-side-effects argument. Adjust
2857 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2860 2018-04-06 Richard Biener <rguenther@suse.de>
2863 * alias.c (find_base_term): New wrapper around find_base_term
2864 unwinding CSELIB_VAL_PTR changes.
2865 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2868 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2870 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2872 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2873 constant definitions.
2874 ("nop"): lr 0,0 -> nopr r0
2875 ("nop_lr0", "nop_lr1"): New insn definitions.
2877 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2879 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2880 NDS32_V3PUSH_AVAILABLE_P macro.
2882 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2883 Chung-Ju Wu <jasonwucj@gmail.com>
2885 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2886 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2887 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2888 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2889 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2890 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2891 * config/nds32/constraints.md: New constraints and checking for hard
2892 float configuration.
2893 * config/nds32/iterators.md: New mode iterator and attribute for hard
2894 float configuration.
2895 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2897 * config/nds32/nds32-fpu.md: New file.
2898 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2899 deal with hard float code generation.
2900 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2902 (abi_type, float_reg_number): New enum type.
2903 * config/nds32/nds32-predicates.c: New predicates for hard float.
2904 * config/nds32/nds32-protos.h: Declare functions for hard float.
2905 * config/nds32/nds32.c: Implementation for hard float configuration.
2906 * config/nds32/nds32.h: Definitions for hard float configuration.
2907 * config/nds32/nds32.md: Include hard float machine description and
2908 modify patterns for hard float configuration.
2909 * config/nds32/nds32.opt: New options for hard float configuration.
2910 * config/nds32/predicates.md: New predicates for hard float
2913 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2915 * common/config/nds32/nds32-common.c
2916 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2918 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2921 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2922 CONSTRUCTOR_ELT (ctor, ...)->value.
2924 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2927 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2929 2018-04-05 Tom de Vries <tom@codesourcery.com>
2932 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2935 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2936 Kito Cheng <kito.cheng@gmail.com>
2938 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2939 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2940 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2941 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2943 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2944 Kito Cheng <kito.cheng@gmail.com>
2946 * config/nds32/constraints.md (Ufe): New memory constraint.
2947 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2948 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2949 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2951 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2952 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2954 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2956 * config/nds32/nds32.md: Use optimize_size in the condition for
2957 alu-shift instructions.
2959 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2961 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2963 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2965 * config/nds32/nds32.md (negsi2): Refine pattern.
2967 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2968 Chung-Ju Wu <jasonwucj@gmail.com>
2970 * config/nds32/iterators.md (shift_rotate): New code iterator.
2971 (shift): New code attribute.
2972 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2973 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2974 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2975 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2976 bit-wise operations.
2977 (andsi3, *andsi3): Ditto.
2978 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2979 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2980 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2981 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2982 nds32_ior_operand, nds32_xor_operand): New predicates.
2984 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2986 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2987 (addsi3, subsi3): ... this.
2989 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2991 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2993 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2995 * config/nds32/nds32.md: Adjust indention.
2997 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2999 * config/nds32/nds32.md (feature): New attribute.
3001 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3003 * config/nds32/nds32.md (subtype): New attribute.
3005 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3008 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3009 expansion to perform a bitwise AND of the argument followed by a
3010 boolean negation of the result.
3012 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3014 PR rtl-optimization/84878
3015 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3016 the basic block. Assert the use reference is not artificial and that
3017 it has an associated insn.
3019 2018-04-04 Michael Matz <matz@suse.de>
3021 * builtins.c (compute_objsize): Pass correct operand
3022 to array_at_struct_end_p.
3024 2018-04-04 Richard Biener <rguenther@suse.de>
3027 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3028 from contexts for DINFO_LEVEL_TERSE and below.
3030 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3032 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3033 resiter_operand condition.
3034 * config/nds32/nds32.md (*move<mode>): Ditto.
3036 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3037 Monk Chiang <sh.chiang04@gmail.com>
3039 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3041 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3043 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3045 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3046 Kito Cheng <kito.cheng@gmail.com>
3048 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3049 nds32_cond_code_str, output_cond_branch,
3050 output_cond_branch_compare_zero, nds32_expand_cbranch,
3051 nds32_expand_cstore, nds32_expand_movcc,
3052 nds32_output_cbranchsi4_equality_zero,
3053 nds32_output_cbranchsi4_equality_reg,
3054 nds32_output_cbranchsi4_equality_reg_or_const_int,
3055 nds32_output_cbranchsi4_greater_less_zero: New functions.
3056 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3057 nds32_expand_cstore, nds32_expand_movcc,
3058 nds32_output_cbranchsi4_equality_zero,
3059 nds32_output_cbranchsi4_equality_reg,
3060 nds32_output_cbranchsi4_equality_reg_or_const_int,
3061 nds32_output_cbranchsi4_greater_less_zero): Declare.
3062 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3063 nds32_rimm11s_operand): New predicates.
3064 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3065 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3068 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3070 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3071 * config/nds32/nds32.md: Ditto.
3072 * config/nds32/pipelines.md: Ditto.
3074 2018-04-04 Richard Biener <rguenther@suse.de>
3076 PR tree-optimization/85168
3077 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3078 propagating abnormals.
3080 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3082 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3084 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3085 Kito Cheng <kito.cheng@gmail.com>
3087 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3088 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3089 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3090 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3091 * config/nds32/nds32.md (sibcall_internal): New.
3092 (sibcall_register): Remove.
3093 (sibcall_immediate): Remove.
3094 (sibcall_value_internal): New.
3095 (sibcall_value_register): Remove.
3096 (sibcall_value_immediate): Remove.
3097 * config/nds32/predicates.md (nds32_general_register_operand): New.
3098 (nds32_call_address_operand): New.
3100 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3102 PR rtl-optimization/85167
3103 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3104 bb_defs if *split_p, instead preinitialize it to NULL.
3106 PR tree-optimization/85156
3107 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3108 evaluating the argument multiple times.
3110 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3112 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3114 (_mm_cvtpd_ps): Likewise.
3115 (_mm_cvttpd_epi32): Likewise.
3116 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3117 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3118 vector, pixel, and bool following altivec.h include.
3120 2018-04-03 Martin Sebor <msebor@redhat.com>
3122 * doc/extend.texi (Common Function Attributes): Clarify.
3123 (const attribute): Likewise.
3124 (pure attribute): Likewise.
3126 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3129 * config/i386/i386.c (ix86_expand_vector_set): Use
3130 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3132 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3134 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3135 instructions when changing rounding bits to preserve precision bits
3136 in the x87 control word.
3138 2018-04-03 Martin Liska <mliska@suse.cz>
3140 PR tree-optimization/82491
3141 * rtl.h (strip_offset_and_add): Replace += suboffset with
3142 poly_uint64 () + suboffset.
3144 2018-03-29 Martin Liska <mliska@suse.cz>
3145 Martin Jambor <mjambor@suse.cz>
3148 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3149 param_type is not an integral or pointer type.
3151 2018-04-03 Richard Biener <rguenther@suse.de>
3153 * sese.h (recompute_all_dominators): Remove.
3155 2018-04-02 Martin Sebor <msebor@redhat.com>
3157 * doc/invoke.texi (-Wrestrict): Fix typos.
3159 2018-04-02 Jim Wilson <jimw@sifive.com>
3161 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3162 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3163 (<optab>di3, <optab>si3_extend): Likewise.
3164 (<optab>si3_mask, <optab>si3_mask_1): New.
3165 (<optab>di3_mask, <optab>di3_mask_1): New.
3166 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3167 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3168 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3170 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3172 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3175 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3177 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3178 (nds32_canonicalize_comparison): New function.
3180 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3181 Kito Cheng <kito.cheng@gmail.com>
3182 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3184 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3185 * config/nds32/constants.md (unspec_volatile_element): Add
3186 UNSPEC_VOLATILE_RELAX_GROUP.
3187 * config/nds32/nds32-relax-opt.c: New file.
3188 * config/nds32/nds32-predicates.c
3189 (nds32_symbol_load_store_p): New function.
3190 * config/nds32/nds32-protos.h
3191 (nds32_symbol_load_store_p): Declare function.
3192 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3193 * config/nds32/nds32.c
3194 (nds32_register_pass): New function to register pass.
3195 (nds32_register_passes): New function to register passes.
3196 * config/nds32/nds32.md (relax_group): New pattern.
3197 * config/nds32/nds32.opt (mrelax-hint): New option.
3198 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3200 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3202 * config/nds32/t-nds32: Modify files dependency.
3204 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3206 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3207 (PROFILE_HOOK): Define its implementation.
3209 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3211 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3212 type and 32-bit size.
3214 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3217 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3218 (V_128_256): New mode iterator.
3219 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3220 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3221 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3223 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3224 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3226 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3229 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3230 NaN inputs correctly.
3232 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3235 * config/rs6000/vsx.md (??r): New mode attribute.
3236 (*vsx_mov<mode>_64bit): Use it.
3237 (*vsx_mov<mode>_32bit): Likewise.
3239 2018-03-30 Martin Sebor <msebor@redhat.com>
3241 PR tree-optimization/84818
3242 * builtins.c (check_access): Use warning_n.
3244 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3247 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3249 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3252 2018-03-30 Julia Koval <julia.koval@intel.com>
3255 * x86-tune.def (movx, partial_reg_dependency): Enable for
3258 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3261 * lra-constraints.c (process_alt_operands): Move setting
3262 this_alternative_matches below.
3264 2018-03-29 Martin Liska <mliska@suse.cz>
3267 * doc/invoke.texi: Document how LTO works with debug info.
3268 Describe auto-load support of binutils. Mention 'x86-64'
3269 as valid option value of -march option.
3271 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3273 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3276 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3277 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3278 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3281 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3284 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3285 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3286 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3287 for RS6000_BTM_POWERPC64.
3288 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3289 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3290 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3295 2018-03-28 Carl Love <cel@us.ibm.com>
3298 2017-09-27 Carl Love <cel@us.ibm.com>
3300 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3301 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3302 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3305 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3307 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3308 instead of __vector bool.
3309 (_mm_max_pu8): Likewise.
3310 (_mm_min_pi16): Likewise.
3312 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3315 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3316 (DIVWEUO): Likewise.
3318 (DIVDEUO): Likewise.
3319 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3320 DIVWEUO and DIVDEUO.
3321 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3322 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3323 (div_extend): Likewise.
3324 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3326 (__builtin_divweuo): Likewise.
3327 (__builtin_divdeo): Likewise.
3328 (__builtin_divdeuo): Likewise.
3330 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3333 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3334 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3336 PR tree-optimization/82004
3337 * gimple-match-head.c (optimize_pow_to_exp): New function.
3338 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3339 Don't fold to exp if optimize_pow_to_exp is false.
3341 2018-03-28 Martin Liska <mliska@suse.cz>
3344 * calls.c (initialize_argument_information): Fix trailing space.
3345 * common.opt: Fix typo and provide better explanation for
3346 -fsanitize-coverage option.
3347 * config/i386/i386.opt: Fix typo.
3349 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3350 Martin Liska <mliska@suse.cz>
3353 * gimplify.c (asan_poison_variable): Don't do the check for
3354 gimplify_omp_ctxp here.
3355 (gimplify_decl_expr): Do it here.
3356 (gimplify_target_expr): Likewise.
3358 2018-03-28 Martin Liska <mliska@suse.cz>
3361 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3362 chkp_type_bounds_count if MPX is not enabled.
3364 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3366 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3368 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3371 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3372 function to create the function decl for complex long double
3373 multiply and divide for -mabi=ieeelongdouble.
3374 (init_float128_ieee): Call it.
3376 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3379 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3380 -fcf-protection=branch -mibt.
3381 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3383 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3386 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3388 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3391 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3392 extern array declarations.
3394 2018-03-27 Richard Biener <rguenther@suse.de>
3397 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3398 explicit single_use checks.
3400 2018-03-27 Richard Biener <rguenther@suse.de>
3402 PR tree-optimization/85082
3403 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3406 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3408 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3409 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3410 Turn on fasynchronous-unwind-tables and funwind-tables.
3412 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3415 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3416 (*bmi_blsr_<mode>_ccz): Ditto.
3418 2018-03-26 Tom de Vries <tom@codesourcery.com>
3420 PR tree-optimization/85063
3421 * omp-general.c (offloading_function_p): New function. Factor out
3423 * omp-offload.c (pass_omp_target_link::gate): ... here.
3424 * omp-general.h (offloading_function_p): Declare.
3425 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3426 with attribute omp declare target for offloading functions.
3428 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3430 PR tree-optimization/84005
3431 * tree-data-ref.h (get_base_for_alignment): Declare.
3432 * tree-data-ref.c (get_base_for_alignment_1): New function.
3433 (get_base_for_alignment): Likewise.
3434 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3435 get_base_for_alignment to find a suitable base object, instead
3436 of always using drb->base_address.
3438 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3441 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3442 known size by default.
3444 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3447 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3448 and non BLKmode operands.
3450 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3453 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3454 Clean up attributes.
3456 2018-03-23 Richard Biener <rguenther@suse.de>
3459 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3460 we are going to emit early debug for LTO.
3462 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3465 * function.c (match_asm_constraints_1): Don't optimize if input
3466 doesn't satisfy general_operand predicate for output's mode.
3469 * alias.c (write_dependence_p): Don't require for x_canonicalized
3470 non-VOIDmode if x has VOIDmode.
3473 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3474 just don't try to optimize it rather than assert it never happens.
3476 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3478 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3479 macro expansions for definition of ST_INTERNAL_<mode> and
3480 LD_INTERNAL_<mode> builtins.
3481 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3483 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3485 (altivec_expand_st_builtin): Likewise.
3486 (altivec_expand_builtin): Remove calls to deleted functions.
3487 (rs6000_address_for_altivec): Delete this function.
3488 * config/rs6000/vector.md: Remove expands for
3489 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3491 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3494 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3495 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3496 re-computing once computed.
3497 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3498 (arm_init_machine_status): Initialize
3499 machine->static_chain_stack_bytes.
3501 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3504 * doc/extend.texi: Add four new prototypes for vec_ld.
3505 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3506 definitions for more logical presentation.
3507 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3508 entries for V1TI variants of __builtin_altivec_ld builtin.
3509 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3510 handling of V1TI variant of LVX icode pattern.
3511 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3512 (rs6000_gimple_fold_builtin): Likewise.
3513 (altivec_init_builtins): Add code to define
3514 __builtin_altivec_lvx_v1ti function.
3516 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3519 * function.c (match_asm_constraints_1): Don't do the optimization
3520 if input isn't a REG, SUBREG, MEM or constant.
3522 2018-03-22 Tom de Vries <tom@codesourcery.com>
3524 PR tree-optimization/84956
3525 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3526 bb_has_abnormal_pred.
3528 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3531 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3532 DECL_INITIAL (decl) to decl at the end.
3533 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3536 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3538 * doc/extend.texi (__builtin_tgmath): Document when complex
3539 integer types are treated as _Complex _Float64.
3541 2018-03-21 Tom de Vries <tom@codesourcery.com>
3543 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3545 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3547 PR tree-optimization/84960
3548 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3549 if it is ENTRY block, move them into single succ of ENTRY in that case.
3551 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3553 PR tree-optimization/84811
3554 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3555 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3556 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3557 (known_subrange_p): Update accordingly. Cast each value involved
3558 in the size comparison, rather than casting the result of the
3561 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3563 PR tree-optimization/84982
3564 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3565 by flipping the least significant bit rather than all bits from
3566 bitpos to bitpos + bitsize - 1.
3568 2018-03-21 Nathan Sidwell <nathan@acm.org>
3570 * doc/extend.texi (Deprecated Features): Remove mention of
3571 long-deleted deprecations.
3573 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3576 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3577 * configure: Regenerate.
3579 2018-03-21 Tom de Vries <tom@codesourcery.com>
3581 PR tree-optimization/83126
3582 * tree-parloops.c (num_phis): New function.
3583 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3585 2018-03-21 Nathan Sidwell <nathan@acm.org>
3587 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3588 mention anon-struct/union members and trailing attributes.
3590 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3592 PR tree-optimization/84969
3593 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3594 builtin memset partitions if they set different rhs values.
3596 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3598 PR rtl-optimization/84989
3599 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3600 VEC_DUPLICATE with scalar result mode.
3602 2018-03-21 Martin Liska <mliska@suse.cz>
3605 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3606 not intended return statement.
3608 2018-03-21 Martin Liska <mliska@suse.cz>
3611 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3612 (chkp_find_bound_slots_1): Limit number of iterations.
3614 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3617 * Minor grammar fixes for x86 options.
3619 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3622 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3623 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3625 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3628 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3629 (altivec_lvx_<mode>_1op): Likewise.
3630 (altivec_stvx_<mode>_2op): Likewise.
3631 (altivec_stvx_<mode>_1op): Likewise.
3632 (altivec_lvx_<VM2:mode>): New define_expand.
3633 (altivec_stvx_<VM2:mode>): Likewise.
3634 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3635 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3636 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3637 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3638 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3639 (rs6000_gen_lvx): Likewise.
3640 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3641 (altivec_expand_stv_builtin): Likewise.
3642 (altivec_expand_builtin): Likewise.
3643 * config/rs6000/vector.md: Likewise.
3645 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3648 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3651 2018-03-20 Richard Biener <rguenther@suse.de>
3654 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3655 sign-conversions as zero, fall back to standard scalar_stmt
3658 2018-03-20 Martin Liska <mliska@suse.cz>
3661 * predict.c (rebuild_frequencies): Handle case when we have
3662 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3664 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3667 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3668 flag_section_anchors.
3669 * varasm.c (use_blocks_for_decl_p): Remove hack for
3670 dw2_force_const_mem.
3673 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3675 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3676 be created, use lowpart_subreg of operands[0] rather than operands[0]
3678 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3679 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3680 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3681 and n constraint instead of aarch64_shift_imm_di and Usd.
3682 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3683 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3685 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3688 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3689 to favor GPR over NEON registers.
3690 (<shift>di3_neon): Likewise.
3692 2018-03-20 Tom de Vries <tom@codesourcery.com>
3695 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3696 (nvptx_process_pars): Emit bar.sync asap and alap.
3698 2018-03-20 Tom de Vries <tom@codesourcery.com>
3701 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3702 seen_label if seen_label is already set.
3704 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3707 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3708 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3709 Use 1U instead of 1. Formatting fixes.
3712 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3713 instead of TREE_TYPE (s1) for the return value.
3715 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3717 PR tree-optimization/84946
3718 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3719 bitsize + bitsize in poly_uint64 rather than poly_int64.
3722 * dwarf2asm.c: Include fold-const.c.
3723 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3724 of decl rather than decl itself.
3726 PR rtl-optimization/84643
3727 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3729 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3732 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3733 calling assemble_variable.
3735 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3738 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3739 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3741 2018-03-19 Jim Wilson <jimw@sifive.com>
3744 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3745 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3746 (riscv_first_stack_step): Likewise.
3747 (riscv_option_override): Use STACK_BOUNDARY instead of
3749 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3751 (BIGGEST_ALIGNMENT): Set to 128.
3752 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3753 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3756 2018-03-19 Richard Biener <rguenther@suse.de>
3758 PR tree-optimization/84933
3759 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3760 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3762 2018-03-19 Richard Biener <rguenther@suse.de>
3764 PR tree-optimization/84859
3765 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3766 (cond_if_else_store_replacement): Perform sinking operation on
3767 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3768 Generalize what a BB with a single eligible store is.
3770 2018-03-19 Richard Biener <rguenther@suse.de>
3772 PR tree-optimization/84929
3773 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3774 chrec_is_positive against non-chrec arg.
3776 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3779 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3781 2018-03-18 Martin Liska <mliska@suse.cz>
3783 PR rtl-optimization/84635
3784 * regrename.c (build_def_use): Use matches_mode only when
3787 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3789 PR tree-optimization/84913
3790 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3791 vectorize chains of COND_EXPRs.
3793 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3795 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3797 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3799 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3801 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3803 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3805 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3806 Kito Cheng <kito.cheng@gmail.com>
3808 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3809 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3810 (nds32_adjust_reg_alloc_order): New function.
3811 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3813 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3815 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3816 nds32_print_operand, nds32_print_operand_address): Use
3817 HOST_WIDE_INT_PRINT_DEC instead.
3819 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3821 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3823 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3826 * config/i386/i386.c (initial_ix86_tune_features,
3827 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3829 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3830 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3831 rather than 1u << ix86_tune. Formatting fix.
3832 (ix86_option_override_internal): Change ix86_arch_mask from
3833 unsigned int to unsigned HOST_WIDE_INT, initialize to
3834 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3835 (ix86_function_specific_restore): Likewise.
3837 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3840 * postreload.c (reload_combine_recognize_pattern): Perform
3841 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3842 truncate_int_for_mode the result for the destination's mode.
3845 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3846 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3848 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3851 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3854 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3855 regno_allocno_class_array and sorted_pseudos.
3856 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3857 insns where regno is used.
3859 2018-03-16 Martin Liska <mliska@suse.cz>
3862 * multiple_target.c (create_dispatcher_calls): Redirect
3863 reference in the symbol table.
3865 2018-03-16 Martin Liska <mliska@suse.cz>
3868 * multiple_target.c (create_dispatcher_calls): Redirect also
3871 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3875 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3876 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3877 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3879 2018-03-16 Julia Koval <julia.koval@intel.com>
3881 * doc/invoke.texi (Skylake Server): Add CLWB.
3882 Cannonlake): Remove CLWB.
3884 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3886 PR tree-optimization/84841
3887 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3889 (FLOAT_ONE_CONST_TYPE): Define.
3890 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3891 (sort_by_operand_rank): Put entries with higher constant_type last
3892 rather than first to match comments.
3894 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3896 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3899 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3902 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3903 check and use address of target always.
3905 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3908 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3909 (indirect_thunk_bnd_needed): Likewise.
3910 (indirect_thunks_used): Likewise.
3911 (indirect_thunks_bnd_used): Likewise.
3912 (indirect_return_needed): New.
3913 (indirect_return_bnd_needed): Likewise.
3914 (output_indirect_thunk_function): Add a bool argument for
3916 (output_indirect_thunk_function): Don't generate alias for
3917 function return thunk.
3918 (ix86_code_end): Call output_indirect_thunk_function to generate
3919 function return thunks.
3920 (ix86_output_function_return): Set indirect_return_bnd_needed
3921 and indirect_return_needed instead of indirect_thunk_bnd_needed
3922 and indirect_thunk_needed.
3924 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3926 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3927 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3928 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3930 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3931 Paul Hua <paul.hua.gm@gmail.com>
3934 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3936 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3938 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3939 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3942 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3945 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3946 instead of GET_MODE_SIZE when comparing Units.
3948 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3951 * varasm.c (hash_section): Return an unchangeble hash value
3952 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3953 Return !aarch64_can_use_per_function_literal_pools_p ().
3955 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3958 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3959 rather than address of cmode as last argument to prepare_cmp_insn.
3961 2018-03-15 Julia Koval <julia.koval@intel.com>
3963 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3964 F_AVX512VNNI, F_AVX512BITALG): New.
3966 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3969 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3970 insn for floating-point loads and stores.
3972 2018-03-14 Carl Love <cel@us.ibm.com>
3974 * config/rs6000/rs6000-c.c: Add macro definitions for
3975 ALTIVEC_BUILTIN_VEC_PERMXOR.
3976 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3977 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3978 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3979 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3981 * config/doc/extend.texi: Add prototypes for vec_permxor.
3983 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3986 * diagnostic-show-locus.c (class layout_point): Convert m_line
3987 from int to linenum_type.
3988 (line_span::comparator): Use linenum "compare" function when
3989 comparing line numbers.
3990 (test_line_span): New function.
3991 (layout_range::contains_point): Convert param "row" from int to
3993 (layout_range::intersects_line_p): Likewise.
3994 (layout::will_show_line_p): Likewise.
3995 (layout::print_source_line): Likewise.
3996 (layout::should_print_annotation_line_p): Likewise.
3997 (layout::print_annotation_line): Likewise.
3998 (layout::print_leading_fixits): Likewise.
3999 (layout::annotation_line_showed_range_p): Likewise.
4000 (struct line_corrections): Likewise for field m_row.
4001 (line_corrections::line_corrections): Likewise for param "row".
4002 (layout::print_trailing_fixits): Likewise.
4003 (layout::get_state_at_point): Likewise.
4004 (layout::get_x_bound_for_row): Likewise.
4005 (layout::print_line): Likewise.
4006 (diagnostic_show_locus): Likewise for locals "last_line" and
4008 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4009 * input.c (selftest::test_linenum_comparisons): New function.
4010 (selftest::input_c_tests): Call it.
4011 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4012 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4013 * selftest.h (ASSERT_GT): New macro.
4014 (ASSERT_GT_AT): New macro.
4015 (ASSERT_LT): New macro.
4016 (ASSERT_LT_AT): New macro.
4018 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4020 PR rtl-optimization/84780
4021 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4023 2018-03-14 Martin Liska <mliska@suse.cz>
4025 * tree.c (record_node_allocation_statistics): Use
4026 get_stats_node_kind.
4027 (get_stats_node_kind): New function extracted from
4028 record_node_allocation_statistics.
4029 (free_node): Use get_stats_node_kind.
4031 2018-03-14 Richard Biener <rguenther@suse.de>
4033 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4034 that the value-set of ANTIC_IN doesn't grow.
4037 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4039 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4040 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4042 2018-03-14 Julia Koval <julia.koval@intel.com>
4044 * config.gcc (icelake-client, icelake-server): New.
4046 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4047 (initial_ix86_arch_features): Ditto.
4048 (PTA_SKYLAKE): Add SGX.
4049 (PTA_ICELAKE): Remove.
4050 (PTA_ICELAKE_CLIENT): New.
4051 (PTA_ICELAKE_SERVER): New.
4052 (ix86_option_override_internal): Split up icelake on icelake client and
4054 (get_builtin_code_for_version): Ditto.
4055 (fold_builtin_cpu): Ditto.
4056 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4057 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4058 * config/i386/i386.h (processor_type): Ditto.
4059 * doc/invoke.texi: Ditto.
4061 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4064 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4065 INTEGER_CST offset, add it together with bitpos / 8 and
4066 sign extend based on POINTER_SIZE.
4070 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4073 * config/i386/constraints.md (Yc): New register constraint.
4074 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4075 Use Yc constraint for alternative 2 of operand 0. Remove
4076 preferred_for_speed attribute.
4078 2018-03-14 Richard Biener <rguenther@suse.de>
4080 PR tree-optimization/84830
4081 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4082 with the old one to avoid oscillations.
4084 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4087 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4089 (assign_by_spills): Return a flag of reload assignment failure.
4090 Do not process the reload assignment failures. Do not spill other
4091 reload pseudos if they has the same reg class. Update n if
4093 (lra_assign): Add a return arg. Set up from the result of
4094 assign_by_spills call.
4095 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4096 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4097 usage_insns if it is not NULL.
4098 (spill_hard_reg_in_range): New function.
4099 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4100 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4101 function prototypes.
4102 (lra_assign): Change prototype.
4103 * lra.c (lra): Add code to deal with fails by splitting hard reg
4106 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4108 * config/riscv/riscv.opt (mrelax): New option.
4109 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4110 "norelax" when riscv_mrelax is disabled.
4111 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4113 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4116 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4117 reassociation for int modes.
4119 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4121 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4122 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4124 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4125 * config/aarch64/aarch64-sve.md
4126 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4127 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4128 (*extend<mode><Vwide>2): Rename to...
4129 (aarch64_sve_extend<mode><Vwide>2): ...this.
4130 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4131 renaming the old pattern to...
4132 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4134 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4135 define_expand, renaming the old pattern to...
4136 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4137 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4138 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4139 account when deciding which SVE instruction the optab should use.
4140 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4142 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4144 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4145 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4146 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4147 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4148 (tlsdesc_small_<mode>): Turn a define_expand and use
4149 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4150 (tlsdesc_small_advsimd_<mode>): ...this.
4151 (tlsdesc_small_sve_<mode>): New pattern.
4153 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4155 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4156 (UNSPEC_UMUL_HIGHPART): New constants.
4157 (MUL_HIGHPART): New int iteraor.
4158 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4159 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4161 (*<su>mul<mode>3_highpart): New define_insn.
4163 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4166 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4169 2018-03-13 Martin Liska <mliska@suse.cz>
4172 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4174 (sem_item_optimizer::~sem_item_optimizer): Release it.
4175 (sem_item_optimizer::merge_classes): Register variable aliases.
4176 (sem_item_optimizer::fixup_pt_set): New function.
4177 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4178 * ipa-icf.h: Declare new variables and functions.
4180 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4183 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4184 integer_pow2p@2 and test integer_pow2p in condition.
4185 (A < 0 ? C : 0): Similarly for @1.
4188 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4189 characters starting at p contain '\0' character, don't look beyond
4193 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4194 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4197 * reg-stack.c (change_stack): Change update_end var from int to
4198 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4199 also call set_block_for_insn on the newly added insns and rescan.
4202 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4203 on the last operand.
4206 * tree.c (stabilize_reference_1): Return save_expr (e) for
4207 STATEMENT_LIST even if it doesn't have side-effects.
4209 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4211 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4213 2018-03-12 Renlin Li <renlin.li@arm.com>
4215 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4216 aarch64_output_scalar_simd_mov_immediate.
4218 2018-03-12 Martin Sebor <msebor@redhat.com>
4220 PR tree-optimization/83456
4221 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4222 for perfectly overlapping calls to memcpy.
4223 (gimple_fold_builtin_memory_chk): Same.
4224 (gimple_fold_builtin_strcpy): Handle no-warning.
4225 (gimple_fold_builtin_stxcpy_chk): Same.
4226 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4228 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4230 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4231 parameter. Use it for SFmode.
4232 (rs6000_function_arg_advance_1): Adjust.
4233 (rs6000_function_arg): Adjust.
4234 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4236 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4238 PR rtl-optimization/84169
4239 PR rtl-optimization/84780
4240 * combine.c (can_combine_p): Check for a 2-insn combination whether
4241 the destination register is used between the two insns, too.
4243 2018-03-12 Richard Biener <rguenther@suse.de>
4245 PR tree-optimization/84803
4246 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4247 for refs DR analysis didn't process.
4249 2018-03-12 Richard Biener <rguenther@suse.de>
4251 PR tree-optimization/84777
4252 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4253 force-vectorize loops ignore whether we are optimizing for size.
4255 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4257 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4258 (TARGET_MD_ASM_ADJUST): Define.
4260 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4261 Kito Cheng <kito.cheng@gmail.com>
4262 Chung-Ju Wu <jasonwucj@gmail.com>
4264 * config/nds32/nds32.c (nds32_compute_stack_frame,
4265 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4266 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4267 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4268 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4269 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4270 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4271 * config/nds32/nds32.md (prologue, epilogue): Use macro
4272 NDS32_V3PUSH_AVAILABLE_P to do checking.
4274 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4277 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4278 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4279 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4280 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4281 addition of most attributes on !orig_type_die or the attribute not
4282 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4284 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4285 Chung-Ju Wu <jasonwucj@gmail.com>
4287 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4289 * config/nds32/nds32.opt (mvh): New option.
4291 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4292 Chung-Ju Wu <jasonwucj@gmail.com>
4294 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4296 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4297 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4300 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4301 Chung-Ju Wu <jasonwucj@gmail.com>
4303 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4305 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4306 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4308 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4309 Kito Cheng <kito.cheng@gmail.com>
4310 Chung-Ju Wu <jasonwucj@gmail.com>
4312 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4313 UNSPEC_FFMISM and UNSPEC_FLMISM.
4314 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4315 for ffb, ffmism and flmism.
4316 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4317 (unspec_ffmism): Ditto.
4318 (unspec_flmism): Ditto.
4319 (nds32_expand_builtin_impl): Check if string extension is available.
4320 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4321 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4323 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4326 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4329 * lra-assigns.c (assign_by_spills): Return a flag of reload
4330 assignment failure. Do not process the reload assignment
4331 failures. Do not spill other reload pseudos if they has the same
4333 (lra_assign): Add a return arg. Set up from the result of
4334 assign_by_spills call.
4335 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4336 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4337 usage_insns if it is not NULL.
4338 (spill_hard_reg_in_range): New function.
4339 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4340 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4341 function prototypes.
4342 (lra_assign): Change prototype.
4343 * lra.c (lra): Add code to deal with fails by splitting hard reg
4346 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4349 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4351 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4354 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4355 (dw_val_node): Add val_symbolic_view.
4356 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4357 (symview_upper_bound): New.
4358 (new_line_info_table): Initialize symviews_since_reset.
4359 (dwarf2out_source_line): Count symviews_since_reset and set
4360 symview_upper_bound.
4361 (dw_val_equal_p): Handle symview.
4362 (add_AT_symview): New.
4363 (print_dw_val): Handle symview.
4364 (attr_checksum, attr_checksum_ordered): Likewise.
4365 (same_dw_val_p, size_of_die): Likewise.
4366 (value_format, output_die): Likewise.
4367 (add_high_low_attributes): Use add_AT_symview for entry_view.
4368 (dwarf2out_finish): Reset symview_upper_bound, clear
4371 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4374 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4375 Add strict argument and use it.
4376 (rs6000_split_multireg_move): Update for new strict argument.
4377 (mem_operand_gpr): Disallow all non-offsettable addresses.
4378 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4380 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4383 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4384 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4385 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4388 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4389 decl, use remap_type if we want to use the type.
4391 2018-03-09 Martin Sebor <msebor@redhat.com>
4393 PR tree-optimization/84526
4394 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4396 (builtin_access::generic_overlap): Be prepared to handle non-array
4399 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4401 PR rtl-optimization/84682
4402 * lra-constraints.c (process_address_1): Check is_address flag
4403 for address constraints.
4404 (process_alt_operands): Likewise.
4405 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4406 preprocess_constraints.
4407 * recog.h (preprocess_constraints): Add oploc parameter.
4409 * recog.c (preprocess_constraints): Test address_operand for
4410 CT_ADDRESS constraints.
4412 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4415 * lra-assigns.c (assign_by_spills): Return a flag of reload
4416 assignment failure. Do not process the reload assignment
4417 failures. Do not spill other reload pseudos if they has the same
4419 (lra_assign): Add a return arg. Set up from the result of
4420 assign_by_spills call.
4421 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4422 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4423 usage_insns if it is not NULL.
4424 (spill_hard_reg_in_range): New function.
4425 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4426 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4427 function prototypes.
4428 (lra_assign): Change prototype.
4429 * lra.c (lra): Add code to deal with fails by splitting hard reg
4432 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4435 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4436 Accept complain bool parameter. Only emit errors if it is true.
4437 (arm_parse_cpu_option_name): Likewise.
4438 (arm_target_thumb_only): Adjust callers of the above.
4439 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4440 prototype to take a default true bool parameter.
4441 (arm_parse_arch_option_name): Likewise.
4443 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4444 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4448 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4449 * configure: Regenerate.
4450 * configure.ac ("linker --version-script option"): New.
4451 ("linker soname option"): New.
4453 2018-03-09 Richard Biener <rguenther@suse.de>
4455 PR tree-optimization/84775
4456 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4457 immediate uses of predicate stmts and mark them modified.
4460 PR tree-optimization/84178
4461 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4463 (version_loop_for_if_conversion): Delay update_ssa call.
4464 (tree_if_conversion): Delay update_ssa until after predicate
4467 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4470 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4471 when the function accesses prior frames.
4473 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4476 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4477 gen_llsym, otherwise call maybe_gen_llsym.
4480 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4481 has ',' character inside of it.
4483 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4486 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4487 as clobbering CC_REGNUM.
4489 2018-03-08 Richard Biener <rguenther@suse.de>
4492 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4493 (follow_copies_to_constant): Do not follow SSA names registered
4496 2018-03-08 Richard Biener <rguenther@suse.de>
4498 PR tree-optimization/84178
4499 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4501 (version_loop_for_if_conversion): Delay update_ssa call.
4502 (tree_if_conversion): Delay update_ssa until after predicate
4505 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4507 PR tree-optimization/84178
4508 * tree-if-conv.c (release_bb_predicate): Remove the
4509 the assertion that the stmts have NULL use_ops.
4510 Discard the statements, asserting that they haven't
4511 yet been added to a BB.
4513 2018-03-08 Richard Biener <rguenther@suse.de>
4515 PR tree-optimization/84746
4516 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4517 (phi_translate): Pass in destination ANTIC_OUT set.
4518 (phi_translate_1): Likewise. For a simplified result lookup
4519 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4520 (phi_translate_set): Adjust.
4521 (do_pre_regular_insertion): Likewise.
4522 (do_pre_partial_partial_insertion): Likewise.
4524 2018-03-08 Martin Liska <mliska@suse.cz>
4526 PR gcov-profile/84735
4527 * doc/gcov.texi: Document usage of profile files.
4528 * gcov-io.h: Document changes in the format.
4530 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4534 * dwarf2out.c (struct dw_line_info_table): Update comments for
4536 (FORCE_RESET_NEXT_VIEW): New.
4537 (FORCE_RESETTING_VIEW_P): New.
4538 (RESETTING_VIEW_P): Check for -1 too.
4539 (ZERO_VIEW_P): Likewise.
4540 (new_line_info_table): Force-reset next view.
4541 (dwarf2out_begin_function): Likewise.
4542 (dwarf2out_source_line): Simplify zero_view_p initialization.
4543 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4544 view directly. Omit view when omitting .loc at line 0.
4546 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4548 PR tree-optimization/84740
4549 * tree-switch-conversion.c (process_switch): Call build_constructors
4550 only if info.phi_count is non-zero.
4552 PR tree-optimization/84739
4553 * tree-tailcall.c (find_tail_calls): Check call arguments against
4554 DECL_ARGUMENTS (current_function_decl) rather than
4555 DECL_ARGUMENTS (func) when checking for tail recursion.
4557 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4559 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4560 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4561 Volker Reichelt's entry and add entries for people that perform
4562 GCC fuzzy testing and report numerous bugs.
4564 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4567 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4568 readonly data in sdata, if that is disabled.
4569 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4570 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4571 -mreadonly-in-sdata option.
4573 2018-03-07 Martin Sebor <msebor@redhat.com>
4575 PR tree-optimization/84468
4576 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4577 basic block when looking for nul assignment.
4579 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4582 * except.h (output_function_exception_table): Adjust prototype.
4583 * except.c (output_function_exception_table): Remove FNNAME parameter
4584 and add SECTION parameter. Ouput one part of the table at a time.
4585 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4586 the first part of the exception table and emit unwind directives.
4587 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4588 (i386_pe_seh_cold_init): Likewise.
4589 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4590 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4591 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4592 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4593 * config/i386/winnt.c: Include except.h.
4594 (struct seh_frame_state): Add reg_offset, after_prologue and
4595 in_cold_section fields.
4596 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4597 (i386_pe_seh_cold_init): New function.
4598 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4599 to seh->in_cold_section.
4600 (seh_emit_push): Record the offset of the push.
4601 (seh_emit_save): Record the offet of the save.
4602 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4603 Test seh->after_prologue to disregard the epilogue.
4604 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4605 (i386_pe_end_cold_function): New function.
4607 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4610 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4611 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4614 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4616 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4617 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4620 * multiple_target.c: Include tree-inline.h and intl.h.
4621 (expand_target_clones): Diagnose and fail if node->definition and
4622 !tree_versionable_function_p (node->decl).
4624 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4626 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4628 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4629 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4630 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4632 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4635 * combine.c (try_combine): Use reg_or_subregno instead of handling
4636 just paradoxical SUBREGs and REGs.
4638 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4640 * config/arc/arc.c (arc_finalize_pic): Remove function.
4641 (arc_must_save_register): We use single base PIC register, remove
4642 checks to save/restore the PIC register.
4643 (arc_expand_prologue): Likewise.
4644 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4646 (arc_verify_short): Likewise.
4647 (arc_attr_type): Likewise.
4648 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4649 (walk_stores): Likewise.
4650 (arc_address_cost): Make it static.
4651 (arc_verify_short): Likewise.
4652 (branch_dest): Likewise.
4653 (arc_attr_type): Likewise.
4654 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4655 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4656 (arc_final_prescan_insn): Remove inserting the nops due to
4657 hardware hazards. It is done in reorg step.
4658 (insn_length_variant_t): Remove.
4659 (insn_length_parameters_t): Likewise.
4660 (arc_insn_length_parameters): Likewise.
4661 (arc_get_insn_variants): Likewise.
4662 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4664 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4667 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4670 PR tree-optimization/84687
4671 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4673 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4675 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4677 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4678 Rename to ppc_speculation_barrier.
4679 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4680 __builtin_ppc_speculation_barrier.
4682 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4685 * combine.c (combine_simplify_rtx): Don't try to simplify if
4686 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4689 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4691 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4692 to 32 bytes when compiling for POWER9.
4694 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4697 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4698 regparm >= 3 with no arg reg available also for calls with
4699 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4702 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4704 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4706 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4709 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4711 2018-03-05 Richard Biener <rguenther@suse.de>
4713 PR tree-optimization/84486
4714 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4715 When inserting a __builtin_assume_aligned call set the LHS
4716 SSA name alignment info accordingly.
4718 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4720 PR tree-optimization/84114
4721 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4722 Avoid reassociation of FLOAT_MODE addition.
4724 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4726 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4727 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4728 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4729 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4730 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4731 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4732 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4734 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4735 __builtin_ia32_wbinvd): New builtins.
4736 (SPECIAL_ARGS2): New.
4737 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4738 (SPECIAL_ARGS2): New.
4739 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4740 (ix86_valid_target_attribute_inner_p): Ditto.
4741 (ix86_init_mmx_sse_builtins): Add special_args2.
4742 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4743 TARGET_WBNOINVD_P): New.
4744 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4745 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4746 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4747 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4748 * config/i386/pconfigintrin.h: New file.
4749 * config/i386/wbnoinvdintrin.h: Ditto.
4750 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4751 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4753 2018-03-05 Richard Biener <rguenther@suse.de>
4755 PR tree-optimization/84670
4756 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4758 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4759 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4760 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4761 doesn't grow if all successors (recursively) were visited at least
4764 2018-03-05 Richard Biener <rguenther@suse.de>
4766 PR tree-optimization/84650
4767 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4768 if executed in the loop pipeline.
4770 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4772 * doc/configfiles.texi (Configuration Files): Move info about
4773 conditionalizing $target-protos.h to...
4774 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4775 differs from $target-protos.h.
4777 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4778 Chung-Ju Wu <jasonwucj@gmail.com>
4780 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4781 * config/nds32/nds32-multiple.md (setmemsi): Define.
4782 * config/nds32/nds32-memory-manipulation.c
4783 (nds32_gen_dup_4_byte_to_word_value): New.
4784 (emit_setmem_word_loop): New.
4785 (emit_setmem_byte_loop): New.
4786 (nds32_expand_setmem_loop): New.
4787 (nds32_expand_setmem_loop_v3m): New.
4788 (nds32_expand_setmem_unroll): New.
4789 (nds32_expand_setmem): New.
4791 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4792 Chung-Ju Wu <jasonwucj@gmail.com>
4794 * config/nds32/nds32-memory-manipulation.c
4795 (nds32_emit_load_store): New.
4796 (nds32_emit_post_inc_load_store): New.
4797 (nds32_emit_mem_move): New.
4798 (nds32_emit_mem_move_block): New.
4799 (nds32_expand_movmemsi_loop_unknown_size): New.
4800 (nds32_expand_movmemsi_loop_known_size): New.
4801 (nds32_expand_movmemsi_loop): New.
4802 (nds32_expand_movmemsi_unroll): New.
4803 (nds32_expand_movmemqi): Rename ...
4804 (nds32_expand_movmemsi): ... to this.
4805 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4806 (movmemsi): ... to this.
4807 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4808 (nds32_expand_movmemsi): ... to this.
4810 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4811 Monk Chiang <sh.chiang04@gmail.com>
4812 Chung-Ju Wu <jasonwucj@gmail.com>
4814 * config/nds32/nds32-protos.h
4815 (nds32_expand_load_multiple): New arguments.
4816 (nds32_expand_store_multiple): Ditto.
4817 (nds32_valid_multiple_load_store): Rename ...
4818 (nds32_valid_multiple_load_store_p): ... to this.
4819 * config/nds32/nds32-memory-manipulation.c
4820 (nds32_expand_load_multiple): Refine implementation.
4821 (nds32_expand_store_multiple): Ditto.
4822 * config/nds32/nds32-multiple.md
4823 (load_multiple): Update nds32_expand_load_multiple interface.
4824 (store_multiple): Update nds32_expand_store_multiple interface.
4825 * config/nds32/nds32-predicates.c
4826 (nds32_valid_multiple_load_store): Rename ...
4827 (nds32_valid_multiple_load_store_p): ... to this and refine
4829 * config/nds32/predicates.md
4830 (nds32_load_multiple_and_update_address_operation): New predicate.
4831 (nds32_store_multiple_and_update_address_operation): New predicate.
4833 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4834 Chung-Ju Wu <jasonwucj@gmail.com>
4836 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4837 (combo): New attribute.
4838 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4840 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4842 * config/nds32/nds32.opt: Change -mcmodel= default value.
4844 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4845 Monk Chiang <sh.chiang04@gmail.com>
4846 Chung-Ju Wu <jasonwucj@gmail.com>
4848 * config/nds32/constants.md (unspec_element): New enum.
4849 * config/nds32/constraints.md (Umw): New constraint.
4850 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4851 * config/nds32/nds32-intrinsic.md: Likewise.
4852 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4853 (nds32_valid_smw_lwm_base_p): New.
4854 (nds32_output_smw_single_word): New.
4855 (nds32_output_lmw_single_word): New.
4856 (nds32_expand_unaligned_load): New.
4857 (nds32_expand_unaligned_store): New.
4858 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4859 (nds32_output_smw_single_word): Declare.
4860 (nds32_output_lmw_single_word): Declare.
4861 (nds32_expand_unaligned_load): Declare.
4862 (nds32_expand_unaligned_store): Declare.
4863 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4864 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4865 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4866 NDS32_BUILTIN_UASTORE_DW.
4867 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4870 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4871 Kito Cheng <kito.cheng@gmail.com>
4872 Chung-Ju Wu <jasonwucj@gmail.com>
4874 * config/nds32/nds32-intrinsic.c
4875 (nds32_expand_builtin_null_ftype_reg): Delete.
4876 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4877 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4878 (nds32_read_argument): New.
4879 (nds32_legitimize_target): Ditto.
4880 (nds32_legitimize_argument): Ditto.
4881 (nds32_check_constant_argument): Ditto.
4882 (nds32_expand_unop_builtin): Ditto.
4883 (nds32_expand_unopimm_builtin): Ditto.
4884 (nds32_expand_binop_builtin): Ditto.
4885 (nds32_builtin_decl_impl): Ditto.
4886 (builtin_description): Ditto.
4887 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4888 (nds32_init_builtins_impl): Ditto.
4889 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4890 (nds32_builtin_decl): New.
4891 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4892 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4894 2018-03-02 Jeff Law <law@redhat.com>
4896 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4897 (redundant_insn, fill_simple_delay_slots): Likewise.
4898 (fill_slots_from_thread): Likewise.
4899 * resource.c (mark_referenced_resources): Likewise.
4900 (mark_set_resources, find_dead_or_set_registers): Likewise.
4902 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4904 * substring-locations.h (format_warning_va): Formatting fix for
4906 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4908 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4909 * substring-locations.c: Include intl.h.
4910 (format_warning_va): Turned into small wrapper around
4911 format_warning_n_va, renamed to ...
4912 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4913 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4915 (format_warning_at_substring_n): New function.
4916 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4917 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4918 format_warning_at_substring with just a shorter name instead of
4919 const function pointer.
4920 (fmtwarn_n): New function.
4921 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4922 appropriate, get rid of all the fmtstr temporaries, move conditionals
4923 with G_() wrapped string literals directly into fmtwarn arguments,
4924 cast dir.len to (int), formatting fixes.
4926 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4928 * doc/invoke.texi: Remove "Cilk Plus" references.
4930 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4931 Richard Biener <rguenther@suse.de>
4934 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4935 for error or warning attributes if CALL_FROM_THUNK_P is set.
4938 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4941 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4942 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4945 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4946 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4948 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4949 instead of -1U in last predictors element's probability member.
4951 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4954 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4955 arguments if they are comparable.
4957 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4959 PR tree-optimization/84634
4960 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4961 masks and masked_loop_p with a single loop_masks, making sure it's
4962 null for bb vectorization.
4964 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4966 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4967 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4968 loop->force_vectorize to check whether there is no alias.
4970 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4973 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4975 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4977 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4978 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4979 instead of a loop around prev_real_insn.
4980 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4984 * config/i386/i386.c (ix86_print_operand): Use conditional
4985 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4988 2018-03-02 Richard Biener <rguenther@suse.de>
4990 PR tree-optimization/84427
4991 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4992 (bitmap_set_subtract_values): Rewrite to handle multiple
4995 (prune_clobbered_mems): Likewise.
4996 (phi_translate): Take edge instead of pred/phiblock.
4997 (phi_translate_1): Likewise.
4998 (phi_translate_set): Likewise. Insert all translated
4999 exprs for a value into the set, keeping possibly multiple
5000 expressions per value.
5001 (compute_antic_aux): Adjust for phi_translate changes.
5002 When intersecting union the expressions and prune those
5003 not in the final value set, keeping possibly multiple
5004 expressions per value. Do not use value-insertion
5005 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5006 all expressions. Add verification that the value-sets
5007 only shrink during iteration.
5008 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5009 (do_pre_regular_insertion): Likewise.
5010 (do_pre_partial_partial_insertion): Likewise.
5012 2018-03-02 Richard Biener <rguenther@suse.de>
5015 * config/darwin.c (saved_debug_info_level): New static global.
5016 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5017 (darwin_asm_lto_end): Restore debug info generation settings.
5019 2018-03-01 Martin Liska <mliska@suse.cz>
5022 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5025 2018-03-01 Richard Biener <rguenther@suse.de>
5028 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5029 type annotation with LTO when debug was disabled at compile-time.
5031 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5033 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5035 (mips_final_postscan_insn): Likewise.
5037 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5039 PR rtl-optimization/84528
5040 * alias.c (init_alias_target): Add commentary.
5041 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5042 a unique base value if the frame pointer is not eliminated
5043 to the stack pointer.
5045 2018-03-01 Tom de Vries <tom@codesourcery.com>
5047 PR rtl-optimization/83327
5048 * lra-int.h (hard_regs_spilled_into): Declare.
5049 * lra.c (hard_regs_spilled_into): Define.
5050 (init_reg_info): Init hard_regs_spilled_into.
5051 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5052 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5053 (process_bb_lives): Handle hard_regs_spilled_into.
5054 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5055 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5057 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5059 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5060 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5061 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5062 * config/rs6000/aix72.h: New file.
5064 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5066 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5067 instead of warning_at with conditional singular and plural messages
5071 * stor-layout.c (update_alignment_for_field): For
5072 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5073 && !DECL_PACKED (field), do the alignment update, just use
5074 only desired_align instead of MAX (type_align, desired_align)
5076 (place_field): Don't do known_align < desired_align handling
5077 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5078 is non-NULL, instead do it after rli->prev_field handling and
5079 only if not within a bitfield word. For DECL_PACKED (field)
5080 use type_align of BITS_PER_UNIT.
5082 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5084 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5085 superfluous parentheses and trailing spaces.
5087 2018-02-28 Richard Biener <rguenther@suse.de>
5089 PR tree-optimization/84584
5090 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5091 SCoPs with fake exit edge.
5093 2018-02-28 Martin Liska <mliska@suse.cz>
5096 * timevar.c (timer::print): Fix format to properly print 100%
5099 2018-02-28 Richard Biener <rguenther@suse.de>
5102 * genmatch.c (capture_info::walk_match): Do not mark
5103 captured expressions without operands as expr_p given
5104 they act more like predicates and should be subject to
5105 "lost tail" side-effect preserving.
5107 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5109 PR rtl-optimization/81611
5110 * auto-inc-dec.c (attempt_change): Move dead note from
5111 mem_insn if it's the next use of regno
5112 (find_address): Take address use of reg holding
5113 non-incremented value. Add parm to limit search to the named
5115 (merge_in_block): Attempt to use a mem insn that is the next
5116 use of the original regno.
5118 2018-02-27 Martin Sebor <msebor@redhat.com>
5121 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5122 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5124 2018-02-27 Martin Sebor <msebor@redhat.com>
5126 PR translation/84207
5127 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5128 n argument to unsigned HOST_WIDE_INT.
5129 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5130 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5131 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5132 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5134 2018-02-27 Richard Biener <rguenther@suse.de>
5136 PR tree-optimization/84512
5137 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5138 Do not use the estimate returned from record_stmt_cost for
5139 the scalar iteration cost but sum properly using add_stmt_cost.
5141 2018-02-27 Richard Biener <rguenther@suse.de>
5143 PR tree-optimization/84466
5144 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5145 Adjust last change to less strictly validate use operands.
5147 2018-02-27 Martin Liska <mliska@suse.cz>
5149 PR gcov-profile/84548
5150 * gcov.c (process_file): Allow partial overlap and consider it
5151 also as group functions.
5152 (output_lines): Properly calculate range of lines for a group.
5154 2018-02-27 Martin Liska <mliska@suse.cz>
5156 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5157 'ggc' suffixes. Change first column width.
5158 (timer::print): Fix formatting of the column.
5160 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5162 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5163 preserve inline entry blocks for the sake of debug inline
5164 entry point markers alone.
5165 (remove_unused_locals): Suggest in comments a better place to
5166 force the preservation of inline entry blocks that are
5167 otherwise unused, but do not preserve them.
5169 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5171 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5173 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5176 * config/i386/constraints.md (Bs): Replace
5177 ix86_indirect_branch_register with
5178 TARGET_INDIRECT_BRANCH_REGISTER.
5180 * config/i386/i386.md (indirect_jump): Likewise.
5181 (tablejump): Likewise.
5182 (*sibcall_memory): Likewise.
5183 (*sibcall_value_memory): Likewise.
5184 Peepholes of indirect call and jump via memory: Likewise.
5185 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5186 (*sibcall_value_GOT_32): Likewise.
5187 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5188 (GOT_memory_operand): Likewise.
5189 (call_insn_operand): Likewise.
5190 (sibcall_insn_operand): Likewise.
5191 (GOT32_symbol_operand): Likewise.
5192 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5194 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5196 PR rtl-optimization/83496
5197 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5198 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5199 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5200 redundant insn, if any.
5201 (relax_delay_slots): Likewise.
5202 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5204 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5206 PR tree-optimization/83965
5207 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5208 that grouped statements are part of a reduction chain. Return
5209 true if the statement is not marked as a reduction itself but
5211 (vect_recog_dot_prod_pattern): Don't check whether the statement
5212 is part of a group here.
5213 (vect_recog_sad_pattern): Likewise.
5214 (vect_recog_widen_sum_pattern): Likewise.
5216 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5219 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5221 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5224 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5226 (ix86_output_indirect_function_return): New prototype.
5227 (ix86_split_simple_return_pop_internal): Likewise.
5228 * config/i386/i386.c (indirect_return_via_cx): New.
5229 (indirect_return_via_cx_bnd): Likewise.
5230 (indirect_thunk_name): Handle return va CX_REG.
5231 (output_indirect_thunk_function): Create alias for
5232 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5233 (ix86_output_indirect_jmp): Remove the bool argument.
5234 (ix86_output_indirect_function_return): New function.
5235 (ix86_split_simple_return_pop_internal): Likewise.
5236 * config/i386/i386.md (*indirect_jump): Don't pass false
5237 to ix86_output_indirect_jmp.
5238 (*tablejump_1): Likewise.
5239 (simple_return_pop_internal): Change it to define_insn_and_split.
5240 Call ix86_split_simple_return_pop_internal to split it for
5242 (simple_return_indirect_internal): Call
5243 ix86_output_indirect_function_return instead of
5244 ix86_output_indirect_jmp.
5246 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5249 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5250 memset and value initialization afterwards.
5252 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5254 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5256 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5259 * common/config/aarch64/aarch64-common.c
5260 (aarch_option_optimization_table[]): Switch
5261 off fomit-frame-pointer
5263 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5264 Chung-Ju Wu <jasonwucj@gmail.com>
5266 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5268 (store_multiple): Ditto.
5270 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5272 * config.gcc: Add --with-cpu support for nds32 target.
5273 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5274 * config/nds32/nds32.opt: Add -mcpu= option.
5276 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5278 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5279 isel=yes): Warn for these deprecated options.
5281 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5283 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5284 ISA_2_5_MASKS_EMBEDDED.
5286 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5288 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5289 p->max as pointers rather than using iterative_hash_expr.
5291 2018-02-23 Carl Love <cel@us.ibm.com>
5293 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5294 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5296 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5297 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5298 P8V_BUILTIN_VEC_VUNSIGNED2.
5300 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5303 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5304 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5306 * lra-constraints.c (curr_insn_transform): Set up
5307 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5309 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5310 * lra-eliminations.c (spill_pseudos): Ditto.
5311 (process_insn_for_elimination): Ditto.
5312 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5313 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5314 LRA_NON_CLOBBERED_ALT.
5316 2018-02-22 Martin Sebor <msebor@redhat.com>
5318 PR tree-optimization/84480
5319 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5320 to maybe_diag_stxncpy_trunc. Call it.
5321 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5322 from gimple_fold_builtin_strcpy. Print inlining stack.
5323 (handle_builtin_stxncpy): Print inlining stack.
5324 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5326 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5329 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5330 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5331 and -fcheck-pointer-bounds are used together.
5332 (indirect_thunk_prefix): New enum.
5333 (indirect_thunk_need_prefix): New function.
5334 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5335 "_nt" instead of "_bnd" for NOTRACK prefix.
5336 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5337 (output_indirect_thunk_function): Likewise.
5339 (ix86_code_end): Update output_indirect_thunk_function calls.
5340 (ix86_output_indirect_branch_via_reg): Replace
5341 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5342 (ix86_output_indirect_branch_via_push): Likewise.
5343 (ix86_output_function_return): Likewise.
5344 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5345 incompatible with -fcf-protection=branch and
5346 -fcheck-pointer-bounds.
5348 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5351 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5352 Change gcc_assert call to output_operand_lossage.
5354 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5356 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5358 2018-02-22 DJ Delorie <dj@redhat.com>
5359 Sebastian Perta <sebastian.perta@renesas.com>
5360 Oleg Endo <olegendo@gcc.gnu.org>
5362 * config/rx/rx.c (rx_rtx_costs): New function.
5363 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5365 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5367 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5369 2018-02-22 Martin Liska <mliska@suse.cz>
5372 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5373 Add "native" as a possible value.
5375 2018-02-22 Martin Liska <mliska@suse.cz>
5378 * config/i386/i386.c (ix86_option_override_internal):
5379 Add "native" as a possible value for -march and -mtune.
5381 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5384 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5385 to all type variants.
5387 PR tree-optimization/84503
5388 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5389 width as info->bitpos + info->bitsize - start.
5390 (merged_store_group::merge_overlapping): Simplify width computation.
5391 (check_no_overlap): New function.
5392 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5393 start + width and last_order of the group, fail if check_no_overlap
5395 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5396 to group if check_no_overlap fails.
5398 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5400 * config/rs6000/altivec.md: Delete contraint arguments to
5401 define_expand, define_split, and define_peephole2, and in
5402 define_insn_and_split if always unused.
5403 * config/rs6000/darwin.md: Ditto.
5404 * config/rs6000/dfp.md: Ditto.
5405 * config/rs6000/rs6000.md: Ditto.
5406 * config/rs6000/sync.md: Ditto.
5407 * config/rs6000/vector.md: Ditto.
5408 * config/rs6000/vsx.md: Ditto.
5410 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5412 * config/rs6000/altivec.md: Write output control strings as braced
5413 blocks instead of double-quoted strings.
5414 * config/rs6000/darwin.md: Ditto.
5415 * config/rs6000/rs6000.md: Ditto.
5416 * config/rs6000/vector.md: Ditto.
5417 * config/rs6000/vsx.md: Ditto.
5419 2018-02-21 Jason Merrill <jason@redhat.com>
5421 PR c++/84314 - ICE with templates and fastcall attribute.
5422 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5424 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5426 * ipa-cp.c (determine_versionability): Fix comment typos.
5428 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5431 * ipa-cp.c (determine_versionability): Do not version functions caling
5434 2018-02-21 Martin Liska <mliska@suse.cz>
5437 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5438 Add "native" as a possible value.
5439 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5440 the macro when native cpu detection is available.
5442 2018-02-21 Martin Liska <mliska@suse.cz>
5445 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5446 Add "native" as a possible value.
5447 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5448 when native cpu detection is available.
5450 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5451 Martin Sebor <msebor@redhat.com>
5453 PR tree-optimization/84478
5454 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5456 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5457 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5458 support which is conservatively correct, for 2 only stay conservative
5459 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5460 argument to the 2 argument get_range_strlen, adjust 6 arg
5461 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5463 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5464 (gimple_fold_builtin_strlen): Pass true as last argument to
5467 2018-02-20 Martin Sebor <msebor@redhat.com>
5470 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5471 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5472 (builtin_memref::builtin_memref): Factor out parts into
5473 set_base_and_offset and call it.
5475 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5478 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5479 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5480 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5481 search at the associated MODE_INT.
5483 2018-02-20 Jeff Law <law@redhat.com>
5486 PR tree-optimization/81592
5488 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5489 for range data rather than using global data.
5490 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5491 range data rather than using global data.
5492 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5493 pass it to children as needed.
5494 (struct directive::fmtresult): Similarly.
5495 (struct directive::set_width): Similarly.
5496 (struct directive::set_precision): Similarly.
5497 (format_integer, format_directive, parse_directive): Similarly.
5498 (format_none): Accept unnamed vr_values parameter.
5499 (format_percent, format_floating, format_character): Similarly.
5500 (format_string, format_plain): Similarly.
5501 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5502 the EVRP range analyzer for range data rather than using global data.
5503 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5504 gimple-ssa-evrp-analyze.h
5505 (class sprintf_dom_walker): Add after_dom_children member function.
5506 Add evrp_range_analyzer member.
5507 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5508 range analyzer as needed.
5509 (sprintf_dom_walker::after_dom_children): New member function.
5510 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5512 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5513 (evrp_range_analyzer::pop_to_marker): Likewise.
5515 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5517 PR tree-optimization/84419
5518 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5519 with the required type if its current type is compatible but
5522 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5525 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5526 after vectorization.
5528 2018-02-20 Martin Liska <mliska@suse.cz>
5531 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5532 possible values if we don't have a hint.
5534 2018-02-20 Martin Liska <mliska@suse.cz>
5538 * final.c (shorten_branches): Build align_tab array with one
5540 * opts.c (finish_options): Add alignment option limit check.
5541 (MAX_CODE_ALIGN): Likewise.
5542 (MAX_CODE_ALIGN_VALUE): Likewise.
5543 * doc/invoke.texi: Document maximum allowed option value for
5544 all -falign-* options.
5546 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5549 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5550 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5551 * var-tracking.c (emit_note_insn_var_location): Remove all references
5552 to NOTE_INSN_CALL_ARG_LOCATION.
5553 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5554 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5555 Use copy_rtx_if_shared.
5556 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5557 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5558 (dwarf2out_var_location): Remove handling of
5559 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5561 * final.c (final_scan_insn): Remove all references to
5562 NOTE_INSN_CALL_ARG_LOCATION.
5563 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5564 before dumping final insns.
5565 * except.c (emit_note_eh_region_end): Remove all references to
5566 NOTE_INSN_CALL_ARG_LOCATION.
5567 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5568 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5569 * config/arc/arc.c (hwloop_optimize): Likewise.
5570 * config/arm/arm.c (create_fix_barrier): Likewise.
5571 * config/s390/s390.c (s390_chunkify_start): Likewise.
5572 * config/sh/sh.c (find_barrier): Likewise.
5573 * config/i386/i386.c (rest_of_insert_endbranch,
5574 ix86_seh_fixup_eh_fallthru): Likewise.
5575 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5576 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5577 * config/frv/frv.c (frv_function_prologue): Likewise.
5578 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5580 (note_outside_basic_block_p): Remove all references to
5581 NOTE_INSN_CALL_ARG_LOCATION.
5582 * gengtype.c (adjust_field_rtx_def): Likewise.
5583 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5585 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5586 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5589 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5592 PR tree-optimization/84452
5593 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5594 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5597 2018-02-19 Martin Liska <mliska@suse.cz>
5600 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5602 2018-02-19 Martin Liska <mliska@suse.cz>
5603 Richard Sandiford <richard.sandiford@linaro.org>
5605 PR tree-optimization/82491
5606 * gimple-fold.c (get_base_constructor): Make earlier bail out
5609 2018-02-19 Carl Love <cel@us.ibm.com>
5611 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5612 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5614 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5615 P8V_BUILTIN_VEC_NEG.
5617 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5619 * config/rl78/rl78.md (movdf): New define expand.
5621 2018-02-19 Martin Liska <mliska@suse.cz>
5624 * doc/invoke.texi: Fix typo.
5625 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5627 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5629 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5630 handle rs6000_single_float and rs6000_double_float specially for
5633 2018-02-16 Jeff Law <law@redhat.com>
5635 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5638 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5641 * ipa-inline.c (inline_small_functions): Fix a typo.
5643 2018-02-16 Nathan Sidwell <nathan@acm.org>
5645 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5647 2018-02-16 Carl Love <cel@us.ibm.com>
5649 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5650 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5651 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5652 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5653 expansion to P8V_BUILTIN_VEC_FLOAT2.
5655 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5657 PR rtl-optimization/70023
5658 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5659 src_regno into account.
5661 2018-02-16 Carl Love <cel@us.ibm.com>
5663 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5664 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5665 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5666 * config/rs6000/rs6000.c: Remove case statements for
5667 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5668 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5669 and P9V_BUILTIN_VEC_VINSERT4B.
5670 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5671 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5672 * config/rs6000/vsx.md:
5673 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5676 2018-02-16 Carl Love <cel@us.ibm.com>
5678 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5680 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5682 * config/rs6000/rs6000-c.c: Add the definitions for
5683 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5684 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5685 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5686 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5687 definition for insert4b and define insn *insert3b_internal.
5688 * doc/extend.texi: Add documentation for vec_extract4b.
5690 2018-02-16 Nathan Sidwell <nathan@acm.org>
5692 * doc/extend.texi (Backwards Compatibility): Mention friend
5693 injection. Note for-scope is deprecated.
5694 * doc/invoke.texi (-ffriend-injection): Deprecate.
5696 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5698 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5699 that moved to I2, also allow destinations that are a paradoxical
5700 subreg (instead of a normal reg).
5702 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5705 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5708 2018-02-16 Richard Biener <rguenther@suse.de>
5710 PR tree-optimization/84037
5711 PR tree-optimization/84016
5713 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5714 Adjust vec_construct for the fact we need additional higher latency
5715 128bit inserts for AVX256 and AVX512 vector builds.
5716 (ix86_add_stmt_cost): Scale vector construction cost for
5719 2018-02-16 Richard Biener <rguenther@suse.de>
5721 PR tree-optimization/84417
5722 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5723 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5724 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5726 2018-02-16 Martin Liska <mliska@suse.cz>
5729 * internal-fn.def (ASAN_CHECK): Set proper flags.
5730 (ASAN_MARK): Likewise.
5732 2018-02-16 Julia Koval <julia.koval@intel.com>
5734 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5735 from PTA_CANNONLAKE.
5737 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5740 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5741 Use ++iter rather than iter++ for std::list iterators.
5742 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5743 defer deleting them until all nodes in the forest are processed. Do
5744 free even leaf nodes. Change to_process into auto_vec.
5747 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5748 * vec.h (vec_default_construct): Use memset instead of placement new
5749 if BROKEN_VALUE_INITIALIZATION is defined.
5750 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5751 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5754 PR rtl-optimization/83723
5755 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5756 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5757 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5759 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5761 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5763 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5765 PR rtl-optimization/81443
5766 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5767 from inner REGs to paradoxical SUBREGs.
5769 2018-02-16 Richard Biener <rguenther@suse.de>
5771 PR tree-optimization/84399
5772 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5773 For operands we can analyze at their definition make sure we can
5774 analyze them at each use as well.
5776 2018-02-16 Richard Biener <rguenther@suse.de>
5778 PR tree-optimization/84190
5779 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5780 volatile accesses if the decl isn't volatile.
5782 2018-02-15 Jason Merrill <jason@redhat.com>
5784 PR c++/84314 - ICE with templates and fastcall attribute.
5785 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5786 TYPE_CANONICAL on an existing type.
5788 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5790 PR tree-optimization/84383
5791 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5792 dstoff nor call operand_equal_p if dstbase is NULL.
5794 PR tree-optimization/84334
5795 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5796 also a CONSTANT_CLASS_P, punt.
5798 2018-02-14 Jim Wilson <jimw@sifive.com>
5800 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5801 first SMALL_OPERAND check. New local min_second_step. Move assert
5802 to where locals are set. Add TARGET_RVC support.
5803 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5805 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5807 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5809 2018-02-14 Martin Sebor <msebor@redhat.com>
5811 PR tree-optimization/83698
5812 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5813 arrays constrain the offset range to their bounds.
5814 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5815 (builtin_access::overlap): Avoid setting the size of overlap if it's
5817 (maybe_diag_overlap): Also consider arrays when deciding what values
5818 of offsets to include in diagnostics.
5820 2018-02-14 Martin Sebor <msebor@redhat.com>
5823 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5824 that correspond to the kind of a declaration.
5826 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5829 * config/pa/pa.md: Load address of PIC label using the linkage table
5830 if the label is nonlocal.
5832 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5834 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5835 warning message if user requests -maltivec=be.
5836 * doc/invoke.texi: Document deprecation of -maltivec=be.
5838 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5841 * config/rs6000/rs6000-c.c: Update definitions for
5842 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5843 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5845 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5848 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5849 add _get_ssp intrinsics. Remove argument from
5850 __builtin_ia32_rdssp[d|q].
5851 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5852 * config/i386/i386-builtin.def: Remove argument from
5853 __builtin_ia32_rdssp[d|q].
5854 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5855 ix86_expand_special_args_builtin for _rdssp[d|q].
5856 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5857 Clear register before usage.
5858 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5859 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5861 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5863 PR tree-optimization/84357
5864 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5865 operand 1 of an ARRAY_REF too.
5867 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5870 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5871 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5873 (set_of_reg): New struct.
5874 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5875 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5876 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5878 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5879 Split into bitclr, bitset, bitinvert patterns if appropriate.
5880 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5881 use rx_fuse_in_memory_bitop.
5882 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5883 to named insn, correct maximum insn length.
5885 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5888 * machmode.def: Define a complex mode for PARTIAL_INT.
5889 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5891 * doc/rtl.texi: Document CSPImode.
5892 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5894 (msp430_hard_regno_nregs_with_padding): Likewise.
5896 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5899 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5901 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5903 PR rtl-optimization/84169
5904 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5905 we generated a parallel as new i3 and we split that to new i2 and i3
5906 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5907 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5908 those to i2, not i1. Partially rewrite this scan code.
5910 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5913 * stor-layout.c (place_field): For variable length fields, adjust
5914 offset_align afterwards not just based on the field's alignment,
5915 but also on the size.
5918 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5919 of exps and logs in the use_exp2 case.
5921 2018-02-13 Jeff Law <law@redhat.com>
5923 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5926 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5929 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5933 * common.opt (gas-loc-support, gas-locview-support): New.
5934 (ginline-points, ginternal-reset-location-views): New.
5935 * doc/invoke.texi: Document them. Use @itemx where intended.
5936 (gvariable-location-views): Adjust.
5937 * target.def (reset_location_view): New.
5938 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5939 (TARGET_RESET_LOCATION_VIEW): New.
5940 * doc/tm.texi: Rebuilt.
5941 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5942 (dwarf2out_default_as_locview_support): New.
5943 (output_asm_line_debug_info): Use option variables.
5944 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5945 (output_loc_list): Likewise.
5946 (add_high_low_attributes): Check option variables.
5947 Don't output entry view attribute in strict mode.
5948 (gen_inlined_subroutine_die): Check option variables.
5949 (dwarf2out_inline_entry): Likewise.
5950 (init_sections_and_labels): Likewise.
5951 (dwarf2out_early_finish): Likewise.
5952 (maybe_reset_location_view): New, from...
5953 (dwarf2out_var_location): ... here. Call it.
5954 * debug.h (dwarf2out_default_as_loc_support): Declare.
5955 (dwarf2out_default_as_locview_support): Declare.
5956 * hooks.c (hook_int_rtx_insn_0): New.
5957 * hooks.h (hook_int_rtx_insn_0): Declare.
5958 * toplev.c (process_options): Take -gas-loc-support and
5959 -gas-locview-support from dwarf2out. Enable
5960 -gvariable-location-views by default only with locview
5961 assembler support. Enable -ginternal-reset-location-views by
5962 default only if the target defines the corresponding hook.
5963 Enable -ginline-points by default if location views are
5964 enabled; force it disabled if statement frontiers are
5966 * tree-inline.c (expand_call_inline): Check option variables.
5967 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5969 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5971 PR tree-optimization/84321
5972 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5973 handling. Also check whether the anti-range contains any values
5974 that satisfy the mask; switch to a VR_RANGE if not.
5976 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5979 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5981 2018-02-13 Martin Jambor <mjambor@suse.cz>
5984 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5985 of call statements, also set location of a load to a temporary.
5987 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5989 * config/rl78/rl78.c (add_vector_labels): New function.
5990 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5991 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5992 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5993 which checks that no arguments are passed.
5994 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5995 * doc/extend.texi: Documentation for the new attribute.
5997 2018-02-13 Andreas Schwab <schwab@suse.de>
5999 * config/riscv/linux.h (CPP_SPEC): Define.
6001 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6004 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6005 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6006 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6007 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6008 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6009 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6010 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6011 temporarily for AES and PCLMUL builtins.
6013 PR tree-optimization/84339
6014 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6015 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6019 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6020 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6021 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6023 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6025 * omp-simd-clone.h: New file.
6026 * omp-simd-clone.c: Include omp-simd-clone.h.
6027 (expand_simd_clones): No longer static.
6028 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6029 cgraph.h and omp-simd-clone.h.
6030 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6031 (vect_recog_widen_shift_pattern): Formatting fix.
6032 (vect_pattern_recog_1): Don't check optab for calls.
6035 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6036 operands[2] into a REG before using gen_lowpart on it.
6038 2018-02-12 Jeff Law <law@redhat.com>
6041 * config/sh/sh.c (find_barrier): Consider a sibling call
6044 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6045 successfully back substituting a reg.
6047 2018-02-12 Richard Biener <rguenther@suse.de>
6049 PR tree-optimization/84037
6050 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6051 parameter, move visited init to caller.
6052 (vect_slp_analyze_operations): Separate cost from validity
6053 check, initialize visited once for all instances.
6054 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6056 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6058 (vect_model_promotion_demotion_cost): Likewise.
6059 (vectorizable_bswap): Guard cost modeling with !slp_node
6060 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6062 (vectorizable_call): Likewise.
6063 (vectorizable_conversion): Likewise.
6064 (vectorizable_assignment): Likewise.
6065 (vectorizable_shift): Likewise.
6066 (vectorizable_operation): Likewise.
6067 (vectorizable_store): Likewise.
6068 (vectorizable_load): Likewise.
6069 (vectorizable_condition): Likewise.
6070 (vectorizable_comparison): Likewise.
6072 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6075 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6076 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6077 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6079 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6082 * params.def (inline-min-speedup): Increase from 8 to 15.
6083 (max-inline-insns-auto): Decrease from 40 to 30.
6084 * ipa-split.c (consider_split): Add some buffer for function to
6085 be considered inlining candidate.
6086 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6089 2018-02-12 Richard Biener <rguenther@suse.de>
6091 PR tree-optimization/84037
6092 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6093 matched stmts if we cannot swap the non-matched ones.
6095 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6097 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6098 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6099 _mm_maskz_scalef_round_ss): New intrinsics.
6100 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6101 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6102 __builtin_ia32_scalefss_round): Remove.
6103 (__builtin_ia32_scalefsd_mask_round,
6104 __builtin_ia32_scalefss_mask_round): New intrinsics.
6105 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6106 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6107 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6108 "<round_constraint>")): Changed to ...
6109 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6110 "<round_scalar_constraint>")): ... this.
6111 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6112 %0, %1, %2<round_op3>}"): Changed to ...
6113 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6114 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6115 %2<round_scalar_mask_op3>}"): ... this.
6116 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6118 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6120 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6121 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6122 (_mm_maskz_sqrt_round_ss): New intrinsics.
6123 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6124 (__builtin_ia32_sqrtsd_mask_round)
6125 (__builtin_ia32_sqrtss_mask_round): New builtins.
6126 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6127 (__builtin_ia32_sqrtss_round): Remove.
6128 (__builtin_ia32_sqrtsd_mask_round)
6129 (__builtin_ia32_sqrtss_mask_round): New builtins.
6130 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6131 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6132 ((match_operand:VF_128 1 "vector_operand"
6133 "xBm,<round_constraint>")): Changed to ...
6134 ((match_operand:VF_128 1 "vector_operand"
6135 "xBm,<round_scalar_constraint>")): ... this.
6136 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6137 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6138 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6139 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6140 %<iptr>1<round_scalar_mask_op3>}): ... this.
6141 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6142 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6144 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6147 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6148 Cast vec_cmpeq result to correct type.
6149 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6150 Cast vec_cmpgt result to correct type.
6152 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6154 * final.c (final_scan_insn_1): Renamed from...
6155 (final_scan_insn): ... this. New wrapper, to recover
6156 seen from the outermost call in recursive ones.
6157 * config/sparc/sparc.c (output_return): Drop seen from call.
6158 (output_sibcall): Likewise.
6159 * config/visium/visium.c (output_branch): Likewise.
6161 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6163 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6166 2018-02-10 Alan Modra <amodra@gmail.com>
6169 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6170 Specify LR as an input.
6172 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6175 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6176 remove_member_access_dummy_vars): New functions.
6177 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6178 lower_omp_1, execute_lower_omp): Use them.
6180 PR rtl-optimization/84308
6181 * shrink-wrap.c (spread_components): Release todo vector.
6183 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6185 PR rtl-optimization/57193
6186 * ira-color.c (struct allocno_color_data): Add member
6187 conflict_allocno_hard_prefs.
6188 (update_conflict_allocno_hard_prefs): New.
6189 (bucket_allocno_compare_func): Add a preference based on
6190 conflict_allocno_hard_prefs.
6191 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6192 (color_allocnos): Remove a dead code. Initiate
6193 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6195 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6198 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6199 constraint from =wa to wa. Avoid a subreg on the output operand,
6200 instead use a pseudo and subreg it in a move.
6201 (p9_xxbrd_<mode>): Changed to ...
6202 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6203 (p9_xxbrd_v2df): New expander.
6204 (p9_xxbrw_<mode>): Changed to ...
6205 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6206 (p9_xxbrw_v4sf): New expander.
6208 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6210 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6212 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6215 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6216 multiply in 32-bit mode.
6217 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6218 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6221 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6223 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6224 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6225 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6226 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6228 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6231 * dwarf2out.c (is_trivial_indirect_ref): New function.
6232 (dwarf2out_late_global_decl): Do not generate a location
6233 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6234 and that are not defined in the current unit.
6236 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6238 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6239 instead of a libcall for UNORDERED.
6241 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6244 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6245 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6247 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6250 * config/s390/s390.c (s390_set_current_function): Invoke
6251 s390_indirect_branch_settings also if fndecl didn't change.
6253 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6255 * config/rs6000/rs6000.md (blockage): Set length to zero.
6257 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6259 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6261 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6264 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6265 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6269 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6270 PARALLEL incoming that failed vt_get_decl_and_offset check.
6273 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6274 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6276 (get_variable_section): For decls in named .bss* sections pass true as
6277 second argument to bss_initializer_p.
6279 2018-02-09 Marek Polacek <polacek@redhat.com>
6280 Jakub Jelinek <jakub@redhat.com>
6283 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6284 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6285 Sync some changes from cxx_fold_indirect_ref.
6287 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6289 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6291 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6292 (BLOCK_INLINE_ENTRY_LABEL): New.
6293 (dwarf2out_var_location): Disregard inline entry markers.
6294 (inline_entry_data): New struct.
6295 (inline_entry_data_hasher): New hashtable type.
6296 (inline_entry_data_hasher::hash): New.
6297 (inline_entry_data_hasher::equal): New.
6298 (inline_entry_data_table): New variable.
6299 (add_high_low_attributes): Add DW_AT_entry_pc and
6300 DW_AT_GNU_entry_view attributes if a pending entry is found
6301 in inline_entry_data_table. Add old entry_pc attribute only
6302 if debug nonbinding markers are disabled.
6303 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6304 markers are enabled.
6305 (block_within_block_p, dwarf2out_inline_entry): New.
6306 (dwarf2out_finish): Check that no entries remained in
6307 inline_entry_data_table.
6308 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6309 (final_scan_insn, notice_source_line): Likewise.
6310 (rest_of_clean_state): Skip inline entry markers.
6311 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6313 * gimple.c (gimple_build_debug_inline_entry): New.
6314 * gimple.h (enum gimple_debug_subcode): Add
6315 GIMPLE_DEBUG_INLINE_ENTRY.
6316 (gimple_build_debug_inline_entry): Declare.
6317 (gimple_debug_inline_entry_p): New.
6318 (gimple_debug_nonbind_marker_p): Adjust.
6319 * insn-notes.def (INLINE_ENTRY): New.
6320 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6321 inline entry marker notes.
6322 (print_insn): Likewise.
6323 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6324 (INSN_DEBUG_MARKER_KIND): Likewise.
6325 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6326 * tree-inline.c (expand_call_inline): Build and insert
6327 debug_inline_entry stmt.
6328 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6329 inline entry blocks early, if nonbind markers are enabled.
6330 (dump_scope_block): Dump fragment info.
6331 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6332 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6333 (gimple_build_debug_inline_entry): New.
6334 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6335 Enable/disable inline entry points too.
6336 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6337 (DEBUG_INSN): Describe inline entry markers.
6339 * common.opt (gvariable-location-views): New.
6340 (gvariable-location-views=incompat5): New.
6341 * config.in: Rebuilt.
6342 * configure: Rebuilt.
6343 * configure.ac: Test assembler for view support.
6344 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6345 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6346 * dwarf2out.c (var_loc_view): New typedef.
6347 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6348 (dwarf2out_locviews_in_attribute): New.
6349 (dwarf2out_locviews_in_loclist): New.
6350 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6351 (enum dw_line_info_opcode): Add LI_adv_address.
6352 (struct dw_line_info_table): Add view.
6353 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6354 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6355 (zero_view_p): New variable.
6356 (ZERO_VIEW_P): New macro.
6357 (output_asm_line_debug_info): New.
6358 (struct var_loc_node): Add view.
6359 (add_AT_view_list, AT_loc_list): New.
6360 (add_var_loc_to_decl): Add view param. Test it against last.
6361 (new_loc_list): Add view params. Record them.
6362 (AT_loc_list_ptr): Handle loc and view lists.
6363 (view_list_to_loc_list_val_node): New.
6364 (print_dw_val): Handle dw_val_class_view_list.
6365 (size_of_die): Likewise.
6366 (value_format): Likewise.
6367 (loc_list_has_views): New.
6368 (gen_llsym): Set vl_symbol too.
6369 (maybe_gen_llsym, skip_loc_list_entry): New.
6370 (dwarf2out_maybe_output_loclist_view_pair): New.
6371 (output_loc_list): Output view list or entries too.
6372 (output_view_list_offset): New.
6373 (output_die): Handle dw_val_class_view_list.
6374 (output_dwarf_version): New.
6375 (output_compilation_unit_header): Use it.
6376 (output_skeleton_debug_sections): Likewise.
6377 (output_rnglists, output_line_info): Likewise.
6378 (output_pubnames, output_aranges): Update version comments.
6379 (output_one_line_info_table): Output view numbers in asm comments.
6380 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6381 Call maybe_gen_llsym.
6382 (loc_list_from_tree_1): Adjust.
6383 (add_AT_location_description): Create view list attribute if
6384 needed, check it's absent otherwise.
6385 (convert_cfa_to_fb_loc_list): Adjust.
6386 (maybe_emit_file): Call output_asm_line_debug_info for test.
6387 (dwarf2out_var_location): Reset views as needed. Precompute
6388 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6389 attribute. Set view.
6390 (new_line_info_table): Reset next view.
6391 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6392 (dwarf2out_source_line): Likewise. Output view resets and labels to
6393 the assembler, or select appropriate line info opcodes.
6394 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6395 (optimize_string_length): Catch it. Adjust.
6396 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6397 dw_val_class_view_list, and remove it if no longer needed.
6398 (hash_loc_list): Hash view numbers.
6399 (loc_list_hasher::equal): Compare them.
6400 (optimize_location_lists): Check whether a view list symbol is
6401 needed, and whether the locview attribute is present, and
6402 whether they match. Remove the locview attribute if no longer
6404 (index_location_lists): Call skip_loc_list_entry for test.
6405 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6406 Use output_dwarf_version.
6407 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6408 (struct dw_val_node): Add val_view_list.
6409 * final.c (SEEN_NEXT_VIEW): New.
6410 (set_next_view_needed): New.
6411 (clear_next_view_needed): New.
6412 (maybe_output_next_view): New.
6413 (final_start_function): Rename to...
6414 (final_start_function_1): ... this. Take pointer to FIRST,
6415 add SEEN parameter. Emit param bindings in the initial view.
6416 (final_start_function): Reintroduce SEEN-less interface.
6417 (final): Rename to...
6418 (final_1): ... this. Take SEEN parameter. Output final pending
6419 next view at the end.
6420 (final): Reintroduce seen-less interface.
6421 (final_scan_insn): Output pending next view before switching
6422 sections or ending a block. Mark the next view as needed when
6423 outputting variable locations. Notify debug backend of section
6424 changes, and of location view changes.
6425 (rest_of_handle_final): Adjust.
6426 * toplev.c (process_options): Autodetect value for debug variable
6427 location views option. Warn on incompat5 without -gdwarf-5.
6428 * doc/invoke.texi (gvariable-location-views): New.
6429 (gvariable-location-views=incompat5): New.
6430 (gno-variable-location-views): New.
6432 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6434 PR tree-optimization/84136
6435 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6436 that the result of find_edge is non-NULL.
6438 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6441 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6442 storing integer register in SImode. Fix cost of 256 and 512
6443 byte aligned SSE register store.
6445 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6447 * config/i386/i386.c (ix86_multiplication_cost): Fix
6448 multiplication cost for TARGET_AVX512DQ.
6450 2018-02-08 Marek Polacek <polacek@redhat.com>
6452 PR tree-optimization/84238
6453 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6456 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6458 PR tree-optimization/84265
6459 * tree-vect-stmts.c (vectorizable_store): Don't treat
6460 VMAT_CONTIGUOUS accesses as grouped.
6461 (vectorizable_load): Likewise.
6463 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6465 PR tree-optimization/81635
6466 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6467 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6468 (test_round_for_mask): New functions.
6469 (wide_int_cc_tests): Call test_round_for_mask.
6470 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6471 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6472 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6473 range returned by get_range_info.
6475 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6478 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6479 * symtab.c: Include builtins.h
6480 (symtab_node::output_to_lto_symbol_table_p): Move here
6481 from lto-streamer-out.c:output_symbol_p.
6482 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6483 (output_symbol_p): Move all logic to symtab.c
6484 (produce_symtab): Update.
6486 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6488 * config/s390/s390-opts.h (enum indirect_branch): Define.
6489 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6490 (s390_indirect_branch_via_thunk)
6491 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6492 (enum s390_indirect_branch_type): Define.
6493 * config/s390/s390.c (struct s390_frame_layout, struct
6494 machine_function): Remove.
6495 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6496 (indirect_branch_table_label_no, indirect_branch_table_name):
6498 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6499 (enum s390_indirect_branch_option): Define.
6500 (s390_return_addr_from_memory): New function.
6501 (s390_handle_string_attribute): New function.
6502 (s390_attribute_table): Add new attribute handler.
6503 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6504 (s390_indirect_branch_via_thunk): New function.
6505 (s390_indirect_branch_via_inline_thunk): New function.
6506 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6507 sibling call optimization for non z10 compiles.
6508 (s390_emit_call): Force indirect branch target to be a single
6509 register. Add r1 clobber for non-z10 compiles.
6510 (s390_emit_epilogue): Emit return jump via return_use expander.
6511 (s390_reorg): Handle JUMP_INSNs as execute targets.
6512 (s390_option_override_internal): Perform validity checks for the
6513 new command line options.
6514 (s390_indirect_branch_attrvalue): New function.
6515 (s390_indirect_branch_settings): New function.
6516 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6517 (s390_output_indirect_thunk_function): New function.
6518 (s390_code_end): Implement target hook.
6519 (s390_case_values_threshold): Implement target hook.
6520 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6522 * config/s390/s390.h (struct s390_frame_layout)
6523 (struct machine_function): Move here from s390.c.
6524 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6525 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6526 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6527 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6528 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6529 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6530 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6531 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6532 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6533 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6534 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6535 (mnemonic attribute): Add values which aren't recognized
6537 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6538 pattern for branch conversion. Fix mnemonic attribute.
6539 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6540 indirect branch via thunk if requested.
6541 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6542 ("*indirect_jump"): Disable for branch conversion using out of
6544 ("indirect_jump_via_thunk<mode>_z10")
6545 ("indirect_jump_via_thunk<mode>")
6546 ("indirect_jump_via_inlinethunk<mode>_z10")
6547 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6548 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6549 ("casesi_jump_via_inlinethunk<mode>_z10")
6550 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6551 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6552 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6553 ("*indirect2_jump"): Disable for branch conversion.
6554 ("casesi_jump"): Turn into expander and expand patterns for branch
6556 ("return_use"): New expander.
6557 ("*return"): Emit return via thunk and rename it to ...
6558 ("*return<mode>"): ... this one.
6559 * config/s390/s390.opt: Add new options and and enum for the
6562 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6564 * lra-constraints.c (match_reload): Unconditionally use
6565 gen_lowpart_SUBREG, rather than selecting between that
6566 and equivalent gen_rtx_SUBREG code.
6568 2018-02-08 Richard Biener <rguenther@suse.de>
6570 PR tree-optimization/84233
6571 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6572 changed flag instead of boguously re-using phi_inserted.
6574 2018-02-08 Martin Jambor <mjambor@suse.cz>
6576 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6577 static local variables.
6579 2018-02-08 Richard Biener <rguenther@suse.de>
6581 PR tree-optimization/84278
6582 * tree-vect-stmts.c (vectorizable_store): When looking for
6583 smaller vector types to perform grouped strided loads/stores
6584 make sure the mode is supported by the target.
6585 (vectorizable_load): Likewise.
6587 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6589 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6590 Increase LDP/STP opportunities by adding adjacent callee-saves.
6592 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6594 PR rtl-optimization/84068
6595 PR rtl-optimization/83459
6596 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6598 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6600 PR tree-optimization/84224
6601 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6602 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6605 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6608 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6609 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6610 count, remove one USE.
6612 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6614 * doc/install.texi (Configuration): Document the
6615 --with-long-double-format={ibm,ieee} PowerPC configuration
6619 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6620 Convert from define_expand to be define_insn_and_split. Rework
6621 float/double/_Float128 conversions to QI/HI/SImode to work with
6622 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6623 conversions to QI/HImode types did a store and then a load to
6624 truncate the value. For conversions to VSX registers, don't split
6625 the insn, instead emit the code directly. Use the code iterator
6626 any_fix to combine signed and unsigned conversions.
6627 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6628 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6629 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6630 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6631 (fix_<mode>di2_hw): Likewise.
6632 (fixuns_<mode>di2_hw): Likewise.
6633 (fix_<mode>si2_hw): Likewise.
6634 (fixuns_<mode>si2_hw): Likewise.
6635 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6636 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6637 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6638 fix<uns>_trunc<SFDF:mode>si2_p8.
6639 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6641 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6642 (fix<uns>_<mode>_mem): Likewise.
6643 (fctiw<u>z_<mode>_mem): Likewise.
6644 (fix<uns>_<mode>_mem): Likewise.
6645 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6646 the register allocator from doing a direct move to the GPRs to do
6647 a store, and instead use the ISA 3.0 store byte/half-word from
6648 vector register instruction. For IEEE 128-bit floating point,
6649 also optimize stores of 32-bit ints.
6650 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6652 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6654 * genextract.c (push_pathstr_operand): New function to support
6656 (walk_rtx): Call push_pathstr_operand.
6657 (print_path): Support [a-zA-Z].
6659 2018-02-07 Richard Biener <rguenther@suse.de>
6661 PR tree-optimization/84037
6662 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6663 (cse_and_gimplify_to_preheader): Declare.
6664 (vect_get_place_in_interleaving_chain): Likewise.
6665 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6667 (_loop_vec_info::~_loop_vec_info): Delete it.
6668 (cse_and_gimplify_to_preheader): New function.
6669 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6670 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6671 (vectorizable_load): Likewise. For grouped stores always base
6672 the IV on the first element.
6673 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6674 condition before gimplifying.
6676 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6678 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6679 *DIV_EXPR and *MOD_EXPR.
6681 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6684 * config/i386/i386.c (ix86_option_override_internal): Mask out
6685 the CF_SET bit when checking -fcf-protection.
6687 2018-02-07 Tom de Vries <tom@codesourcery.com>
6690 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6693 2018-02-07 Richard Biener <rguenther@suse.de>
6695 PR tree-optimization/84204
6696 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6699 PR tree-optimization/84205
6700 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6701 special-case isl_ast_op_zdiv_r.
6703 PR tree-optimization/84223
6704 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6705 Only add conditions from within the region.
6706 (gather_bbs::after_dom_children): Adjust.
6708 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6711 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6712 * config/avr/avr.md: Only post-reload split REG-REG moves if
6713 either register is GENERAL_REG_P.
6715 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6717 PR tree-optimization/84235
6718 * tree-ssa-scopedtables.c
6719 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6720 if the subtraction is performed in floating point type where NaNs are
6721 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6722 build 1. Formatting fix.
6724 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6727 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6728 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6729 and skip it regardless of bb boundaries. Use CALL_P macro,
6730 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6731 unnecessarily, formatting fix.
6733 2018-02-06 Michael Collison <michael.collison@arm.com>
6735 * config/arm/thumb2.md:
6736 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6737 (*thumb_mov_notscc): Ditto.
6739 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6742 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6743 unsigned_fix, not "s".
6745 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6747 * configure.ac (gcc_fn_eh_frame_ro): New function.
6748 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6749 correct .eh_frame permissions.
6750 * configure: Regenerate.
6752 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6754 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6757 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6759 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6760 Display warning message for -mno-speculate-indirect-jumps.
6762 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6764 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6766 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6768 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6770 PR tree-optimization/84225
6771 * tree-eh.c (find_trapping_overflow): Only call
6772 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6774 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6777 * config/i386/i386.c: Reimplement the check of possible options
6778 -mibt/-mshstk conbination. Change error messages.
6779 * doc/invoke.texi: Fix a typo: remove extra '='.
6781 2018-02-06 Marek Polacek <polacek@redhat.com>
6783 PR tree-optimization/84228
6784 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6786 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6789 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6790 emitted arch directives.
6791 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6792 __ARM_FEATURE_COPROC before changing architectures.
6794 2018-02-06 Richard Biener <rguenther@suse.de>
6796 * config/i386/i386.c (print_reg): Fix typo.
6797 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6799 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6801 * configure: Regenerate.
6803 2018-02-05 Martin Sebor <msebor@redhat.com>
6805 PR tree-optimization/83369
6806 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6809 2018-02-05 Martin Liska <mliska@suse.cz>
6811 * doc/invoke.texi: Cherry-pick upstream r323995.
6813 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6815 * ira.c (ira_init_register_move_cost): Adjust comment.
6817 2018-02-05 Martin Liska <mliska@suse.cz>
6819 PR gcov-profile/84137
6820 * doc/gcov.texi: Fix typo in documentation.
6822 2018-02-05 Martin Liska <mliska@suse.cz>
6824 PR gcov-profile/83879
6825 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6826 using dlopen functionality.
6828 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6830 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6831 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6832 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6833 _mm_maskz_range_round_ss): New intrinsics.
6834 (__builtin_ia32_rangesd128_round)
6835 (__builtin_ia32_rangess128_round): Remove.
6836 (__builtin_ia32_rangesd128_mask_round,
6837 __builtin_ia32_rangess128_mask_round): New builtins.
6838 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6839 __builtin_ia32_rangess128_round): Remove.
6840 (__builtin_ia32_rangesd128_mask_round,
6841 __builtin_ia32_rangess128_mask_round): New builtins.
6842 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6843 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6844 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6845 "<round_saeonly_constraint>")): Changed to ...
6846 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6847 "<round_saeonly_scalar_constraint>")): ... this.
6848 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6849 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6850 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6851 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6852 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6854 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6856 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6858 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6859 Remove all values except native, 8540 and 8548.
6861 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6863 * config/i386/i386.c (ix86_output_function_return): Pass
6864 INVALID_REGNUM, instead of -1, as invalid register number to
6865 indirect_thunk_name and output_indirect_thunk.
6867 2018-02-02 Julia Koval <julia.koval@intel.com>
6869 * config.gcc: Add -march=icelake.
6870 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6871 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6872 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6873 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6874 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6875 (processor_target_table): Add icelake.
6876 (ix86_option_override_internal): Handle new PTAs.
6877 (get_builtin_code_for_version): Handle icelake.
6878 (M_INTEL_COREI7_ICELAKE): New.
6879 (fold_builtin_cpu): Handle icelake.
6880 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6881 * doc/invoke.texi: Add -march=icelake.
6883 2018-02-02 Julia Koval <julia.koval@intel.com>
6885 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6886 to wide_int_bitmask.
6887 * wide-int-bitmask.h: New.
6889 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6892 * config/i386/i386.md: Replace Pmode with word_mode in
6893 builtin_setjmp_setup and builtin_longjmp to support x32.
6895 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6899 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6901 (rs6000_supported_cpu_names): New static variable.
6902 (linux_cpu_translation_table): Likewise.
6903 (elf_platform) <cpu>: Define new static variable and use it.
6904 Translate kernel AT_PLATFORM name to canonical name if needed.
6905 Error if platform name is unknown.
6907 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6910 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6912 2018-02-01 Jeff Law <law@redhat.com>
6915 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6916 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6917 the scratch if RELEASE_VIA_POP is false.
6918 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6919 If we have to save a temporary register, decrement SIZE appropriately.
6920 Pass new arguments to release_scratch_register_on_entry.
6921 (ix86_adjust_stack_and_probe): Likewise.
6922 (ix86_emit_probe_stack_range): Pass new arguments to
6923 release_scratch_register_on_entry.
6925 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6927 PR rtl-optimization/84157
6928 * combine.c (change_zero_ext): Use REG_P predicate in
6929 front of HARD_REGISTER_P predicate.
6931 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6933 * config/avr/avr.c (avr_option_override): Move disabling of
6934 -fdelete-null-pointer-checks to...
6935 * common/config/avr/avr-common.c (avr_option_optimization_table):
6938 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6940 PR tree-optimization/81635
6941 * tree-data-ref.c (split_constant_offset_1): For types that
6942 wrap on overflow, try to use range info to prove that wrapping
6945 2018-02-01 Renlin Li <renlin.li@arm.com>
6948 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6950 (aarch64_register_move_cost): Likewise.
6951 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6953 (REG_CLASS_NAMES): Likewise.
6954 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6955 TAILCALL_ADDR_REGS. Remove IP registers.
6956 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6958 2018-02-01 Richard Biener <rguenther@suse.de>
6960 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6961 for specifying RPO order and allow NULL for that.
6962 * domwalk.c (dom_walker::dom_walker): Likewise.
6963 (dom_walker::walk): Handle NULL RPO order.
6964 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6966 (rewrite_update_dom_walker): Likewise.
6967 (mark_def_dom_walker): Likewise.
6969 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6971 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6972 (aarch64_maybe_expand_sve_subreg_move): Declare.
6973 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6974 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6976 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6977 that are semantically a reverse operation.
6978 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6979 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6980 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6982 (aarch64_can_change_mode_class): For big-endian, forbid changes
6983 between two SVE modes if they have different element sizes.
6985 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6987 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6988 the TImode handling for big-endian targets.
6990 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6992 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6993 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6995 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6996 Remove BSWAP handing for big-endian targets and use the form of
6997 LD1RQ appropariate for the mode.
6999 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7001 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7002 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7005 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7008 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7009 check for operands that need to go through aarch64_sve_reload_be.
7011 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7013 PR tree-optimization/81661
7014 PR tree-optimization/84117
7015 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7016 * tree-eh.c: Include gimplify.h.
7017 (find_trapping_overflow, replace_trapping_overflow,
7018 rewrite_to_non_trapping_overflow): New functions.
7019 * tree-vect-loop.c: Include tree-eh.h.
7020 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7021 * tree-data-ref.c: Include tree-eh.h.
7022 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7024 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7026 PR rtl-optimization/84123
7027 * combine.c (change_zero_ext): Check if hard register satisfies
7028 can_change_dest_mode before calling gen_lowpart_SUBREG.
7030 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7033 * ira.c (ira_init_register_move_cost): Remove assert.
7035 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7037 PR rtl-optimization/84071
7038 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7039 * doc/tm.texi: Regenerate.
7041 2018-01-31 Richard Biener <rguenther@suse.de>
7043 PR tree-optimization/84132
7044 * tree-data-ref.c (analyze_miv_subscript): Properly
7045 check whether evolution_function_is_affine_multivariate_p
7046 before calling gcd_of_steps_may_divide_p.
7048 2018-01-31 Julia Koval <julia.koval@intel.com>
7051 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7052 * config/i386/i386.md (rdpid_rex64) New.
7053 (rdpid): Make 32bit only.
7055 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7058 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7059 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7061 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7064 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7066 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7068 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7070 PR rtl-optimization/84071
7071 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7072 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7074 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7076 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7077 (arc_attribute_table): Add 'aux' attribute.
7078 (arc_in_small_data_p): Consider aux like variables.
7079 (arc_is_aux_reg_p): New function.
7080 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7081 (arc_get_aux_arg): New function.
7082 (prepare_move_operands): Handle aux-register access.
7083 (arc_handle_aux_attribute): New function.
7084 * doc/extend.texi (ARC Variable attributes): Add subsection.
7086 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7088 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7089 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7090 (arc_attribute_table): Add 'uncached' attribute.
7091 (arc_print_operand): Print '.di' flag for uncached memory
7093 (arc_in_small_data_p): Do not consider for small data the uncached
7095 (arc_is_uncached_mem_p): New function.
7096 * config/arc/predicates.md (compact_store_memory_operand): Check
7097 for uncached memory accesses.
7098 (nonvol_nonimm_operand): Likewise.
7099 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7101 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7104 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7105 falign-loops=): Add Optimization flag.
7107 2018-01-30 Jeff Law <law@redhat.com>
7110 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7111 INT_REGISTERS_SAVED. Check it prior to calling
7112 get_scratch_register_on_entry.
7113 (ix86_adjust_stack_and_probe): Similarly.
7114 (ix86_emit_probe_stack_range): Similarly.
7115 (ix86_expand_prologue): Corresponding changes.
7117 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7120 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7121 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7123 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7126 * lra-constraints.c (curr_insn_transform): Process AND in the
7129 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7131 PR rtl-optimization/83986
7132 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7133 dependence against last_pending_memory_flush in addition to
7136 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7138 PR tree-optimization/81611
7139 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7142 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7145 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7148 2018-01-30 Richard Biener <rguenther@suse.de>
7149 Jakub Jelinek <jakub@redhat.com>
7151 PR tree-optimization/84111
7152 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7153 inner loops added during recursion, as they don't have up-to-date
7156 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7159 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7160 (can_inline_edge_by_limits_p): ... here.
7161 (can_early_inline_edge_p, check_callers,
7162 update_caller_keys, update_callee_keys, recursive_inlining,
7163 add_new_edges_to_heap, speculation_useful_p,
7164 inline_small_functions,
7165 inline_small_functions, flatten_function,
7166 inline_to_all_callers_1): Update.
7168 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7170 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7173 2018-01-30 Richard Biener <rguenther@suse.de>
7175 PR tree-optimization/83008
7176 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7177 invariant and constant vector uses in stmts when they need
7180 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7183 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7184 * configure: Regenerate.
7186 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7188 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7190 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7191 Use gen_rtx_REG rather than gen_lowpart.
7193 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7195 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7196 rather than 0 when creating partial subregs.
7198 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7200 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7203 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7206 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7207 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7208 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7209 flags. This restores the settings used before the 2017-07-24.
7210 Turning off pre increment/decrement/modify allows IVOPTS to
7211 optimize DF/SF loops where the index is an int.
7213 2018-01-29 Richard Biener <rguenther@suse.de>
7214 Kelvin Nilsen <kelvin@gcc.gnu.org>
7217 * tree-vect-stmts.c (vectorizable_call): Don't call
7218 targetm.vectorize_builtin_md_vectorized_function if callee is
7221 2018-01-22 Carl Love <cel@us.ibm.com>
7223 * doc/extend.tex: Fix typo in second arg in
7224 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7226 2018-01-29 Richard Biener <rguenther@suse.de>
7228 PR tree-optimization/84086
7229 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7230 (flush_ssaname_freelist): When SSA names were released reset
7231 the SCEV hash table.
7233 2018-01-29 Richard Biener <rguenther@suse.de>
7235 PR tree-optimization/84057
7236 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7237 removed paths when removing edges.
7239 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7241 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7242 -mfunction-return=@var{choice}.
7244 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7247 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7249 (layout::print_source_line): Likewise.
7250 (test_get_line_width_without_trailing_whitespace): Add test cases.
7252 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7255 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7258 2018-01-26 Jim Wilson <jimw@sifive.com>
7260 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7262 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7265 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7267 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7268 and CMP + SUB-immediate -> SUBS.
7270 2018-01-26 Martin Sebor <msebor@redhat.com>
7272 PR tree-optimization/83896
7273 * tree-ssa-strlen.c (get_string_len): Rename...
7274 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7275 Avoid assuming length is constant.
7276 (handle_char_store): Use HOST_WIDE_INT for string length.
7278 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7281 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7282 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7284 2018-01-26 Richard Biener <rguenther@suse.de>
7286 PR rtl-optimization/84003
7287 * dse.c (record_store): Only record redundant stores when
7288 the earlier store aliases at least all accesses the later one does.
7290 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7292 PR rtl-optimization/83985
7293 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7294 REG_CFA_RESTORE insns.
7295 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7296 REG_CFA_RESTORE insns here.
7299 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7300 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7302 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7304 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7305 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7306 (arc_init): Likewise.
7307 (arc_override_options): Likewise.
7308 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7310 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7312 * config/arc/arc.h (TARGET_DBNZ): Define.
7313 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7314 properly set the tune attribute.
7315 (dbnz): Use TARGET_DBNZ guard.
7316 * config/arc/arc.opt (mtune): Add core3 option.
7318 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7320 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7321 recognize new pic like addresses.
7322 (arc_delegitimize_address): Clean up.
7324 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7326 * config/arc/arc-arches.def: Option mrf16 valid for all
7328 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7329 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7330 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7331 * config/arc/arc-tables.opt: Regenerate.
7332 * config/arc/arc.c (arc_conditional_register_usage): Handle
7333 reduced register file case.
7334 (arc_file_start): Set must have build attributes.
7335 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7337 * config/arc/arc.opt (mrf16): Add new option.
7338 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7339 * config/arc/genmultilib.awk: Handle new mrf16 option.
7340 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7341 * config/arc/t-multilib: Regenerate.
7342 * doc/invoke.texi (ARC Options): Document mrf16 option.
7344 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7346 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7347 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7348 (arc_attribute_table): Add 'secure_call' attribute.
7349 (arc_print_operand): Print secure call operand.
7350 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7352 (arc_is_secure_call_p): New function. * config/arc/arc.md
7353 (call_i): Add support for sjli instruction.
7354 (call_value_i): Likewise.
7355 * config/arc/constraints.md (Csc): New constraint.
7357 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7358 John Eric Martin <John.Martin@emmicro-us.com>
7360 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7361 * config/arc/arc.c (_arc_jli_section): New struct.
7362 (arc_jli_section): New type.
7363 (rc_jli_sections): New static variable.
7364 (arc_handle_jli_attribute): New function.
7365 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7366 (arc_file_end): New function.
7367 (TARGET_ASM_FILE_END): Define.
7368 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7369 (arc_add_jli_section): New function.
7370 (jli_call_scan): Likewise.
7371 (arc_reorg): Call jli_call_scan.
7372 (arc_output_addsi): Remove 'S' from printing asm operand.
7373 (arc_is_jli_call_p): New function.
7374 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7376 (movhi_insn): Likewise.
7377 (movsi_insn): Likewise.
7378 (movsi_set_cc_insn): Likewise.
7379 (loadqi_update): Likewise.
7380 (load_zeroextendqisi_update): Likewise.
7381 (load_signextendqisi_update): Likewise.
7382 (loadhi_update): Likewise.
7383 (load_zeroextendhisi_update): Likewise.
7384 (load_signextendhisi_update): Likewise.
7385 (loadsi_update): Likewise.
7386 (loadsf_update): Likewise.
7387 (movsicc_insn): Likewise.
7388 (bset_insn): Likewise.
7389 (bxor_insn): Likewise.
7390 (bclr_insn): Likewise.
7391 (bmsk_insn): Likewise.
7392 (bicsi3_insn): Likewise.
7393 (cmpsi_cc_c_insn): Likewise.
7394 (movsi_ne): Likewise.
7395 (movsi_cond_exec): Likewise.
7396 (clrsbsi2): Likewise.
7407 (call_i): Remove 'S' asm letter, add jli instruction.
7408 (call_value_i): Likewise.
7409 * config/arc/arc.op (mjli-always): New option.
7410 * config/arc/constraints.md (Cji): New constraint.
7411 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7413 (subsf3_fpx): Likewise.
7414 (mulsf3_fpx): Likewise.
7415 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7417 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7419 * doc/invoke.texi (ARC): Document mjli-always option.
7421 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7423 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7424 avoid addition with 0 and use incw and decw where possible.
7426 2018-01-26 Richard Biener <rguenther@suse.de>
7428 PR tree-optimization/81082
7429 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7430 association if it requires casting to unsigned.
7431 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7432 from fold_plusminus_mult_expr to catch important cases late when
7433 range info is available.
7435 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7437 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7438 * configure.ac (hidden_linkonce): New test.
7439 * configure: Regenerate.
7440 * config.in: Regenerate.
7442 2018-01-26 Julia Koval <julia.koval@intel.com>
7444 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7445 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7446 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7447 _mm_mask_bitshuffle_epi64_mask): Fix type.
7448 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7449 USI_FTYPE_V4DI_V4DI_USI): Remove.
7450 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7451 __builtin_ia32_vpshufbitqmb256_mask,
7452 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7453 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7454 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7456 2018-01-26 Alan Modra <amodra@gmail.com>
7459 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7460 UNSPEC_VBPERMQ. Sort other unspecs.
7462 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7464 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7466 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7469 * predict.c (drop_profile): Do not push/pop cfun; update also
7471 (handle_missing_profiles): Fix logic looking for zero profiles.
7473 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7476 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7477 on functions with #pragma omp declare simd or functions with simd
7479 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7480 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7481 Remove trailing \n from warning_at calls.
7483 2018-01-25 Tom de Vries <tom@codesourcery.com>
7486 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7487 for neutered workers.
7489 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7492 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7493 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7495 2018-01-24 Jeff Law <law@redhat.com>
7498 * i386.c (get_probe_interval): Move to earlier point.
7499 (ix86_compute_frame_layout): If -fstack-clash-protection and
7500 the frame is larger than the probe interval, then use pushes
7501 to save registers rather than reg->mem moves.
7502 (ix86_expand_prologue): Remove conditional for int_registers_saved
7505 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7508 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7509 min/max for never referenced object.
7511 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7514 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7516 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7517 attributes from DECL_ATTRIBUTES (decl) without affecting
7518 DECL_ATTRIBUTES (current_function_decl).
7519 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7520 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7522 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7524 PR tree-optimization/83979
7525 * fold-const.c (fold_comparison): Use constant_boolean_node
7526 instead of boolean_{true,false}_node.
7528 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7530 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7533 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7535 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7536 Simplify the clause that sets the length attribute.
7537 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7538 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7539 clause that sets the length attribute.
7540 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7542 2018-01-24 Tom de Vries <tom@codesourcery.com>
7545 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7546 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7547 Add strict parameter.
7548 (prevent_branch_around_nothing): Insert dummy insn between branch to
7549 label and label with no ptx insn inbetween.
7550 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7552 2018-01-24 Tom de Vries <tom@codesourcery.com>
7555 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7556 for neutered threads in warp.
7557 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7559 2018-01-24 Richard Biener <rguenther@suse.de>
7561 PR tree-optimization/83176
7562 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7565 2018-01-24 Richard Biener <rguenther@suse.de>
7567 PR tree-optimization/82819
7568 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7569 code generating pluses that are no-ops in the target precision.
7571 2018-01-24 Richard Biener <rguenther@suse.de>
7574 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7576 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7578 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7579 to merge probabilities.
7580 * predict.c (probably_never_executed): Also mark as cold functions
7581 with global 0 profile and guessed local profile.
7582 * profile-count.c (profile_probability::combine_with_count): New
7584 * profile-count.h (profile_probability::operator*,
7585 profile_probability::operator*=, profile_probability::operator/,
7586 profile_probability::operator/=): Reduce precision to adjusted
7587 and set value to guessed on contradictory divisions.
7588 (profile_probability::combine_with_freq): Remove.
7589 (profile_probability::combine_wiht_count): Declare.
7590 (profile_count::force_nonzero):: Set to adjusted.
7591 (profile_count::probability_in):: Set quality to adjusted.
7592 * tree-ssa-tail-merge.c (replace_block_by): Use
7595 2018-01-23 Andrew Waterman <andrew@sifive.com>
7596 Jim Wilson <jimw@sifive.com>
7598 * config/riscv/riscv.c (riscv_stack_boundary): New.
7599 (riscv_option_override): Set riscv_stack_boundary. Handle
7600 riscv_preferred_stack_boundary_arg.
7601 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7602 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7603 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7604 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7605 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7606 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7608 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7611 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7612 of struct ix86_frame.
7613 (ix86_expand_epilogue): Likewise. Add a local variable for
7614 the reg_save_offset field in struct ix86_frame.
7616 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7618 PR tree-optimization/82604
7619 * tree-loop-distribution.c (enum partition_kind): New enum item
7620 PKIND_PARTIAL_MEMSET.
7621 (partition_builtin_p): Support above new enum item.
7622 (generate_code_for_partition): Ditto.
7623 (compute_access_range): Differentiate cases that equality can be
7624 proven at all loops, the innermost loops or no loops.
7625 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7626 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7627 (finalize_partitions, distribute_loop): Don't fuse partition of
7628 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7629 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7632 2018-01-23 Martin Liska <mliska@suse.cz>
7634 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7635 order to ignore the predictor.
7636 (PRED_POLYMORPHIC_CALL): Likewise.
7637 (PRED_RECURSIVE_CALL): Likewise.
7639 2018-01-23 Martin Liska <mliska@suse.cz>
7641 * tree-profile.c (tree_profiling): Print function header to
7642 aware reader which function we are working on.
7643 * value-prof.c (gimple_find_values_to_profile): Do not print
7644 not interesting value histograms.
7646 2018-01-23 Martin Liska <mliska@suse.cz>
7648 * profile-count.h (enum profile_quality): Add
7649 profile_uninitialized as the first value. Do not number values
7650 as they are zero based.
7651 (profile_count::verify): Update sanity check.
7652 (profile_probability::verify): Likewise.
7654 2018-01-23 Nathan Sidwell <nathan@acm.org>
7656 * doc/invoke.texi (ffor-scope): Deprecate.
7658 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7660 PR tree-optimization/83510
7661 * domwalk.c (set_all_edges_as_executable): New function.
7662 (dom_walker::dom_walker): Convert bool param
7663 "skip_unreachable_blocks" to enum reachability. Move setup of
7664 edge flags to set_all_edges_as_executable and only do it when
7665 reachability is REACHABLE_BLOCKS.
7666 * domwalk.h (enum dom_walker::reachability): New enum.
7667 (dom_walker::dom_walker): Convert bool param
7668 "skip_unreachable_blocks" to enum reachability.
7669 (set_all_edges_as_executable): New decl.
7670 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7671 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7673 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7674 but converting true to REACHABLE_BLOCKS.
7675 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7677 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7678 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7679 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7681 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7682 if check_all_array_refs will be called.
7684 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7686 * tree.c (selftest::test_location_wrappers): Add more test
7689 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7691 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7692 (selftest::test_bit_in_range): Likewise.
7694 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7697 * doc/sourcebuild.texi (vect_float): Say that the selector
7698 only describes the situation when -funsafe-math-optimizations is on.
7699 (vect_float_strict): Document.
7701 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7703 PR tree-optimization/83965
7704 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7705 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7706 instead of checking only for a reduction.
7707 (vect_recog_widen_sum_pattern): Likewise.
7709 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7711 * predict.c (probably_never_executed): Only use precise profile info.
7712 (compute_function_frequency): Skip after inlining hack since we now
7713 have quality checking.
7715 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7717 * profile-count.h (profile_probability::very_unlikely,
7718 profile_probability::unlikely, profile_probability::even): Set
7719 precision to guessed.
7721 2018-01-23 Richard Biener <rguenther@suse.de>
7723 PR tree-optimization/83963
7724 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7725 Properly terminate dominator walk when crossing the exit edge not
7726 when visiting its source block.
7728 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7731 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7732 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7734 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7736 PR tree-optimization/83957
7737 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7738 semicolon after for body surrounded by braces.
7740 PR tree-optimization/83081
7741 * profile-count.h (profile_probability::split): New method.
7742 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7743 Use profile_probability::split.
7744 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7745 when splitting a single conditional jump into 2.
7747 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7749 PR tree-optimization/69452
7750 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7753 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7755 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7756 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7757 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7759 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7761 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7762 * config/rl78/rl78.md (movdi): New define_expand.
7763 * config/rl78/rl78.c (rl78_split_movdi): New function.
7765 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7768 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7770 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7771 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7772 128-bit to produce an UNSPEC move to get the double word with the
7773 signbit and then a shift directly to do signbit.
7774 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7775 implementation with a new version that just does either a direct
7776 move or a regular move. Move memory interface to separate insns.
7777 Move insns so they are next to the expander.
7778 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7779 with signbit move. Split big and little endian case.
7780 (signbit<mode>2_dm_mem_le): Likewise.
7781 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7782 (signbit<mode>2_dm2): Likewise.
7784 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7786 * config/rl78/rl78.md (anddi3): New define_expand.
7788 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7790 * config/rl78/rl78.md (umindi3): New define_expand.
7792 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7794 * config/rl78/rl78.md (smindi3): New define_expand.
7796 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7798 * config/rl78/rl78.md (smaxdi3): New define_expand.
7800 2018-01-22 Carl Love <cel@us.ibm.com>
7802 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7803 LVX_V1TI): Add macro expansion.
7804 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7805 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7806 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7807 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7808 Change check to determine if the instruction is a byte reversing
7809 entry. Fix typo in comment.
7810 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7811 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7812 Add def_builtin calls for new builtins.
7813 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7814 Add define_insn expansion.
7816 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7818 * config/rl78/rl78.md (umaxdi3): New define_expand.
7820 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7822 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7823 for non-QImode registers.
7825 2018-01-22 Richard Biener <rguenther@suse.de>
7827 PR tree-optimization/83963
7828 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7829 including the loop exit block.
7830 (scop_detection::merge_sese): Likewise.
7831 (scop_detection::add_scop): Do it here instead.
7833 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7835 * doc/sourcebuild.texi (arm_softfloat): Document.
7837 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7840 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7841 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7842 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7844 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7845 David Edelsohn <dje.gcc@gmail.com>
7848 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7849 Change "crset eq" to "crset 2".
7850 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7851 (*call_indirect_aix<mode>_nospec): Likewise.
7852 (*call_value_indirect_aix<mode>_nospec): Likewise.
7853 (*call_indirect_elfv2<mode>_nospec): Likewise.
7854 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7855 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7856 change assembly output from . to $.
7857 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7858 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7859 (*tablejump<mode>_internal1_nospec): Likewise.
7861 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7864 * config/sh/sh_optimize_sett_clrt.cc:
7865 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7867 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7869 PR tree-optimization/83940
7870 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7871 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7872 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7873 use it to pass back the definition type.
7874 (vect_check_store_rhs): Likewise rhs_dt_out.
7875 (vect_build_gather_load_calls): Add a mask_dt argument and use
7876 it instead of a call to vect_is_simple_use.
7877 (vectorizable_store): Update calls to vect_check_load_store_mask
7878 and vect_check_store_rhs. Use the dt returned by the latter instead
7879 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7880 instead of calls to vect_is_simple_use. Pass the scalar rather
7881 than the vector operand to vect_is_simple_use when handling
7882 second and subsequent copies of an rhs value.
7883 (vectorizable_load): Update calls to vect_check_load_store_mask
7884 and vect_build_gather_load_calls. Use the cached mask_dt and
7885 gs_info.offset_dt instead of calls to vect_is_simple_use.
7887 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7890 * tree-emutls.c: Include gimplify.h.
7891 (lower_emutls_2): New function.
7892 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7893 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7894 it before further processing.
7897 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7898 UINTVAL (trueop1) instead of INTVAL (op1).
7900 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7904 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7905 INCOMING_FRAME_SP_OFFSET if not defined.
7906 (scan_trace): Add ENTRY argument. If true and
7907 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7908 emit a note to adjust the CFA offset.
7909 (create_cfi_notes): Adjust scan_trace callers.
7910 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7911 INCOMING_FRAME_SP_OFFSET in the CIE.
7912 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7913 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7915 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7916 * doc/tm.texi: Regenerated.
7918 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7920 PR rtl-optimization/83147
7921 * lra-constraints.c (remove_inheritance_pseudos): Use
7922 lra_substitute_pseudo_within_insn.
7924 2018-01-19 Tom de Vries <tom@codesourcery.com>
7925 Cesar Philippidis <cesar@codesourcery.com>
7928 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7930 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7933 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7934 spaces for function labels.
7936 2018-01-19 Martin Liska <mliska@suse.cz>
7938 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7939 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7940 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7941 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7942 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7943 (PRED_CONST_RETURN): Change from 69 to 65.
7944 (PRED_NULL_RETURN): Change from 91 to 71.
7945 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7946 (PRED_LOOP_GUARD): Change from 66 to 73.
7948 2018-01-19 Martin Liska <mliska@suse.cz>
7950 * predict.c (predict_insn_def): Add new assert.
7951 (struct branch_predictor): Change type to signed integer.
7952 (test_prediction_value_range): Amend test to cover
7954 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7955 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7956 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7957 (PRED_LOOP_IV_COMPARE): Likewise.
7958 * predict.h (PROB_UNINITIALIZED): Define new constant.
7960 2018-01-19 Martin Liska <mliska@suse.cz>
7962 * predict.c (dump_prediction): Add new format for
7963 analyze_brprob.py script which is enabled with -details
7965 * profile-count.h (precise_p): New function.
7967 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7969 PR tree-optimization/83922
7970 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7971 there are no statements that need masking.
7972 (vect_active_double_reduction_p): New function.
7973 (vect_analyze_loop_operations): Use it when handling phis that
7974 are not in the loop header.
7976 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7978 PR tree-optimization/83914
7979 * tree-vect-loop.c (vectorizable_induction): Don't convert
7980 init_expr or apply the peeling adjustment for inductions
7981 that are nested within the vectorized loop.
7983 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7985 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7988 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7992 * function.h (gimplify_parameters): Add gimple_seq * argument.
7993 * function.c: Include gimple.h and options.h.
7994 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7995 for the added local temporaries if needed.
7996 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7997 if there are any parameter cleanups, wrap whole body into a
7998 try/finally with the cleanups.
8000 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8003 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8004 Use GET_MODE_CLASS for scalar floating point.
8006 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8010 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8011 Fix call of call_cgraph_insertion_hooks.
8013 2018-01-18 Martin Sebor <msebor@redhat.com>
8015 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8017 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8020 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8023 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8026 * common.opt: (-ffile-prefix-map): New option.
8027 * opts.c (common_handle_option): Defer it.
8028 * opts-global.c (handle_common_deferred_options): Handle it.
8029 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8030 * file-prefix-map.h: New file.
8031 (remap_debug_filename, add_debug_prefix_map): ...here.
8032 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8033 * final.c (debug_prefix_map, add_debug_prefix_map
8034 remap_debug_filename): Move to...
8035 * file-prefix-map.c: New file.
8036 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8037 generalize, get rid of alloca(), use strrchr() instead of strchr().
8038 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8039 Implement in terms of add_prefix_map().
8040 (remap_macro_filename, remap_debug_filename): Implement in term of
8042 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8043 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8044 * dbxout.c: Include file-prefix-map.h.
8045 * varasm.c: Likewise.
8046 * vmsdbgout.c: Likewise.
8047 * xcoffout.c: Likewise.
8048 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8049 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8050 * doc/invoke.texi (-ffile-prefix-map): Document.
8051 (-fdebug-prefix-map): Update description.
8053 2018-01-18 Martin Liska <mliska@suse.cz>
8055 * config/i386/i386.c (indirect_thunk_name): Document that also
8057 (output_indirect_thunk): Document why both instructions
8058 (pause and lfence) are generated.
8060 2018-01-18 Richard Biener <rguenther@suse.de>
8062 PR tree-optimization/83887
8063 * graphite-scop-detection.c
8064 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8065 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8066 (scop_detection::merge_sese): Re-implement with a flood-fill
8067 algorithm that properly finds a SESE region if it exists.
8069 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8072 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8073 pointer_diff optimizations use view_convert instead of convert.
8075 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8077 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8078 Generate different code for -mno-speculate-indirect-jumps.
8079 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8080 (*call_indirect_aix<mode>): Disable for
8081 -mno-speculate-indirect-jumps.
8082 (*call_indirect_aix<mode>_nospec): New define_insn.
8083 (*call_value_indirect_aix<mode>): Disable for
8084 -mno-speculate-indirect-jumps.
8085 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8086 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8087 -mno-speculate-indirect-jumps.
8088 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8090 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8092 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8093 long double type, set the flags for noting the default long double
8094 type, even if we don't pass or return a long double type.
8096 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8099 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8102 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8104 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8105 support for merge[hl].
8106 (fold_mergehl_helper): New helper function.
8107 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8108 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8109 (altivec_vmrglw_direct): Add xxmrglw insn.
8111 2018-01-17 Andrew Waterman <andrew@sifive.com>
8113 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8114 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8116 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8119 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8120 call the lto_location_cache before reading the
8121 DECL_SOURCE_LOCATION of the types.
8123 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8124 Richard Sandiford <richard.sandiford@linaro.org>
8126 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8127 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8128 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8129 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8130 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8132 * config/aarch64/constraints.md (aarch64_movti_operand):
8134 * config/aarch64/predicates.md (Uti): Add new constraint.
8136 2018-01-17 Carl Love <cel@us.ibm.com>
8138 * config/rs6000/vsx.md (define_expand xl_len_r,
8139 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8140 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8142 (define_expand, define_insn): Move the shift left from the
8143 define_insn to the define_expand for lxvl and stxvl instructions.
8144 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8145 and XL_LEN_R definitions to PURE.
8147 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8149 * config/i386/i386.c (indirect_thunk_name): Declare regno
8150 as unsigned int. Compare regno with INVALID_REGNUM.
8151 (output_indirect_thunk): Ditto.
8152 (output_indirect_thunk_function): Ditto.
8153 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8154 in the call to output_indirect_thunk_function.
8156 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8159 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8160 rather than the size of inner_type to determine the stack slot size
8161 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8163 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8166 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8169 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8171 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8172 endian Linux systems to optionally enable multilibs for selecting
8173 the long double type if the user configured an explicit type.
8174 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8175 have no long double multilibs if not defined.
8176 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8177 warn if the user used -mabi={ieee,ibm}longdouble and we built
8178 multilibs for long double.
8179 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8180 appropriate multilib option.
8181 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8183 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8184 for building long double multilibs.
8185 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8187 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8189 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8192 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8194 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8197 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8200 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8203 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8205 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8206 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8208 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8210 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8211 different rtl trees depending on TARGET_64BIT.
8212 (rs6000_gen_lvx): Likewise.
8214 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8216 * config/visium/visium.md (nop): Tweak comment.
8217 (hazard_nop): Likewise.
8219 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8221 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8222 -mspeculate-indirect-jumps.
8223 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8224 for -mno-speculate-indirect-jumps.
8225 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8226 (*call_value_indirect_elfv2<mode>): Disable for
8227 -mno-speculate-indirect-jumps.
8228 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8229 (indirect_jump): Emit different RTL for
8230 -mno-speculate-indirect-jumps.
8231 (*indirect_jump<mode>): Disable for
8232 -mno-speculate-indirect-jumps.
8233 (*indirect_jump<mode>_nospec): New define_insn.
8234 (tablejump): Emit different RTL for
8235 -mno-speculate-indirect-jumps.
8236 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8237 (tablejumpsi_nospec): New define_expand.
8238 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8239 (tablejumpdi_nospec): New define_expand.
8240 (*tablejump<mode>_internal1): Disable for
8241 -mno-speculate-indirect-jumps.
8242 (*tablejump<mode>_internal1_nospec): New define_insn.
8243 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8246 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8248 * caller-save.c (insert_save): Drop unnecessary parameter. All
8251 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8252 Richard Biener <rguenth@suse.de>
8255 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8256 return early, inline manually is_gimple_sizepos. Make sure if we
8257 call gimplify_expr we don't end up with a gimple constant.
8258 * tree.c (variably_modified_type_p): Don't return true for
8259 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8260 * gimplify.h (is_gimple_sizepos): Remove.
8262 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8264 PR tree-optimization/83857
8265 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8266 vectorizable_live_operation for pure SLP statements.
8267 (vectorizable_live_operation): Handle PHIs.
8269 2018-01-16 Richard Biener <rguenther@suse.de>
8271 PR tree-optimization/83867
8272 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8273 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8275 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8278 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8279 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8280 If off is not INTEGER_CST, issue a may not be aligned warning
8281 rather than isn't aligned. Use isn%'t rather than isn't.
8282 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8284 <case MULT_EXPR>: Improve the case when bottom and one of the
8285 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8286 operand, in that case check if the other operand is multiple of
8287 bottom divided by the INTEGER_CST operand.
8289 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8292 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8293 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8294 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8295 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8296 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8297 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8298 (pa_function_arg_size): New function.
8300 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8302 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8303 in a separate statement.
8305 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8307 PR tree-optimization/83847
8308 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8309 group gathers and scatters.
8311 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8313 PR rtl-optimization/86620
8314 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8316 PR rtl-optimization/83213
8317 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8318 to last if both are JUMP_INSNs.
8320 PR tree-optimization/83843
8321 * gimple-ssa-store-merging.c
8322 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8323 store_immediate_info for bswap/nop orig_stores.
8325 2018-01-15 Andrew Waterman <andrew@sifive.com>
8327 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8329 <UDIV>: Increase cost if !TARGET_DIV.
8331 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8333 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8334 (define_attr "cr_logical_3op"): New.
8335 (cceq_ior_compare): Adjust.
8336 (cceq_ior_compare_complement): Adjust.
8337 (*cceq_rev_compare): Adjust.
8338 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8339 (is_cracked_insn): Adjust.
8340 (insn_must_be_first_in_group): Adjust.
8341 * config/rs6000/40x.md: Adjust.
8342 * config/rs6000/440.md: Adjust.
8343 * config/rs6000/476.md: Adjust.
8344 * config/rs6000/601.md: Adjust.
8345 * config/rs6000/603.md: Adjust.
8346 * config/rs6000/6xx.md: Adjust.
8347 * config/rs6000/7450.md: Adjust.
8348 * config/rs6000/7xx.md: Adjust.
8349 * config/rs6000/8540.md: Adjust.
8350 * config/rs6000/cell.md: Adjust.
8351 * config/rs6000/e300c2c3.md: Adjust.
8352 * config/rs6000/e500mc.md: Adjust.
8353 * config/rs6000/e500mc64.md: Adjust.
8354 * config/rs6000/e5500.md: Adjust.
8355 * config/rs6000/e6500.md: Adjust.
8356 * config/rs6000/mpc.md: Adjust.
8357 * config/rs6000/power4.md: Adjust.
8358 * config/rs6000/power5.md: Adjust.
8359 * config/rs6000/power6.md: Adjust.
8360 * config/rs6000/power7.md: Adjust.
8361 * config/rs6000/power8.md: Adjust.
8362 * config/rs6000/power9.md: Adjust.
8363 * config/rs6000/rs64.md: Adjust.
8364 * config/rs6000/titan.md: Adjust.
8366 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8368 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8369 ix86_indirect_branch_register logic.
8371 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8373 * config/i386/constraints.md (Bs): Update
8374 ix86_indirect_branch_register check. Don't check
8375 ix86_indirect_branch_register with GOT_memory_operand.
8377 * config/i386/predicates.md (GOT_memory_operand): Don't check
8378 ix86_indirect_branch_register here.
8379 (GOT32_symbol_operand): Likewise.
8381 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8383 * config/i386/predicates.md (constant_call_address_operand):
8384 Rewrite ix86_indirect_branch_register logic.
8385 (sibcall_insn_operand): Likewise.
8387 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8389 * config/i386/constraints.md (Bs): Replace
8390 ix86_indirect_branch_thunk_register with
8391 ix86_indirect_branch_register.
8393 * config/i386/i386.md (indirect_jump): Likewise.
8394 (tablejump): Likewise.
8395 (*sibcall_memory): Likewise.
8396 (*sibcall_value_memory): Likewise.
8397 Peepholes of indirect call and jump via memory: Likewise.
8398 * config/i386/i386.opt: Likewise.
8399 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8400 (GOT_memory_operand): Likewise.
8401 (call_insn_operand): Likewise.
8402 (sibcall_insn_operand): Likewise.
8403 (GOT32_symbol_operand): Likewise.
8405 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8408 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8409 type rather than type addr's type points to.
8410 (expand_omp_atomic_mutex): Likewise.
8411 (expand_omp_atomic): Likewise.
8413 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8416 * config/i386/i386.c (output_indirect_thunk_function): Use
8417 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8418 for __x86_return_thunk.
8420 2018-01-15 Richard Biener <rguenther@suse.de>
8423 * expmed.c (extract_bit_field_1): Fix typo.
8425 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8428 * config/arm/iterators.md (VF): New mode iterator.
8429 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8430 Remove integer-related logic from pattern.
8431 (neon_vabd<mode>_3): Likewise.
8433 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8436 * common.opt (fstrict-overflow): No longer an alias.
8437 (fwrapv-pointer): New option.
8438 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8439 also for pointer types based on flag_wrapv_pointer.
8440 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8441 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8442 opts->x_flag_wrapv got set.
8443 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8444 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8445 POINTER_TYPE_OVERFLOW_UNDEFINED.
8446 * match.pd: Likewise in address comparison pattern.
8447 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8449 2018-01-15 Richard Biener <rguenther@suse.de>
8452 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8453 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8454 Reset type names to their identifier if their TYPE_DECL doesn't
8455 have linkage (and thus is used for ODR and devirt).
8456 (save_debug_info_for_decl): Remove.
8457 (save_debug_info_for_type): Likewise.
8458 (add_tree_to_fld_list): Adjust.
8459 * tree-pretty-print.c (dump_generic_node): Make dumping of
8460 type names more robust.
8462 2018-01-15 Richard Biener <rguenther@suse.de>
8464 * BASE-VER: Bump to 8.0.1.
8466 2018-01-14 Martin Sebor <msebor@redhat.com>
8469 * builtins.c (check_access): Avoid warning when the no-warning bit
8472 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8474 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8475 * ira-color (allocno_hard_regs_compare): Likewise.
8477 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8480 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8481 Use .pushsection/.popsection.
8483 2018-01-14 Martin Sebor <msebor@redhat.com>
8486 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8488 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8490 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8491 entry from extra_headers.
8492 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8493 extra_headers, make the list bitwise identical to the i?86-*-* one.
8495 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8497 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8498 -mcmodel=large with -mindirect-branch=thunk,
8499 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8500 -mfunction-return=thunk-extern.
8501 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8502 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8503 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8505 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8507 * config/i386/i386.c (print_reg): Print the name of the full
8508 integer register without '%'.
8509 (ix86_print_operand): Handle 'V'.
8510 * doc/extend.texi: Document 'V' modifier.
8512 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8514 * config/i386/constraints.md (Bs): Disallow memory operand for
8515 -mindirect-branch-register.
8517 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8518 (GOT_memory_operand): Likewise.
8519 (call_insn_operand): Likewise.
8520 (sibcall_insn_operand): Likewise.
8521 (GOT32_symbol_operand): Likewise.
8522 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8523 for -mindirect-branch-register.
8524 (tablejump): Likewise.
8525 (*sibcall_memory): Likewise.
8526 (*sibcall_value_memory): Likewise.
8527 Disallow peepholes of indirect call and jump via memory for
8528 -mindirect-branch-register.
8529 (*call_pop): Replace m with Bw.
8530 (*call_value_pop): Likewise.
8531 (*sibcall_pop_memory): Replace m with Bs.
8532 * config/i386/i386.opt (mindirect-branch-register): New option.
8533 * doc/invoke.texi: Document -mindirect-branch-register option.
8535 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8537 * config/i386/i386-protos.h (ix86_output_function_return): New.
8538 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8539 set function_return_type.
8540 (indirect_thunk_name): Add ret_p to indicate thunk for function
8542 (output_indirect_thunk_function): Pass false to
8543 indirect_thunk_name.
8544 (ix86_output_indirect_branch_via_reg): Likewise.
8545 (ix86_output_indirect_branch_via_push): Likewise.
8546 (output_indirect_thunk_function): Create alias for function
8547 return thunk if regno < 0.
8548 (ix86_output_function_return): New function.
8549 (ix86_handle_fndecl_attribute): Handle function_return.
8550 (ix86_attribute_table): Add function_return.
8551 * config/i386/i386.h (machine_function): Add
8552 function_return_type.
8553 * config/i386/i386.md (simple_return_internal): Use
8554 ix86_output_function_return.
8555 (simple_return_internal_long): Likewise.
8556 * config/i386/i386.opt (mfunction-return=): New option.
8557 (indirect_branch): Mention -mfunction-return=.
8558 * doc/extend.texi: Document function_return function attribute.
8559 * doc/invoke.texi: Document -mfunction-return= option.
8561 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8563 * config/i386/i386-opts.h (indirect_branch): New.
8564 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8565 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8566 with local indirect jump when converting indirect call and jump.
8567 (ix86_set_indirect_branch_type): New.
8568 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8569 (indirectlabelno): New.
8570 (indirect_thunk_needed): Likewise.
8571 (indirect_thunk_bnd_needed): Likewise.
8572 (indirect_thunks_used): Likewise.
8573 (indirect_thunks_bnd_used): Likewise.
8574 (INDIRECT_LABEL): Likewise.
8575 (indirect_thunk_name): Likewise.
8576 (output_indirect_thunk): Likewise.
8577 (output_indirect_thunk_function): Likewise.
8578 (ix86_output_indirect_branch_via_reg): Likewise.
8579 (ix86_output_indirect_branch_via_push): Likewise.
8580 (ix86_output_indirect_branch): Likewise.
8581 (ix86_output_indirect_jmp): Likewise.
8582 (ix86_code_end): Call output_indirect_thunk_function if needed.
8583 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8585 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8586 (ix86_attribute_table): Add indirect_branch.
8587 * config/i386/i386.h (machine_function): Add indirect_branch_type
8588 and has_local_indirect_jump.
8589 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8591 (tablejump): Likewise.
8592 (*indirect_jump): Use ix86_output_indirect_jmp.
8593 (*tablejump_1): Likewise.
8594 (simple_return_indirect_internal): Likewise.
8595 * config/i386/i386.opt (mindirect-branch=): New option.
8596 (indirect_branch): New.
8599 (thunk-inline): Likewise.
8600 (thunk-extern): Likewise.
8601 * doc/extend.texi: Document indirect_branch function attribute.
8602 * doc/invoke.texi: Document -mindirect-branch= option.
8604 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8607 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8609 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8611 * ipa-inline.c (want_inline_small_function_p): Return false if
8612 inlining has already failed with CIF_FINAL_ERROR.
8613 (update_caller_keys): Call want_inline_small_function_p before
8615 (update_callee_keys): Likewise.
8617 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8619 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8621 (rs6000_quadword_masked_address_p): Likewise.
8622 (quad_aligned_load_p): Likewise.
8623 (quad_aligned_store_p): Likewise.
8624 (const_load_sequence_p): Add comment to describe the outer-most loop.
8625 (mimic_memory_attributes_and_flags): New function.
8626 (rs6000_gen_stvx): Likewise.
8627 (replace_swapped_aligned_store): Likewise.
8628 (rs6000_gen_lvx): Likewise.
8629 (replace_swapped_aligned_load): Likewise.
8630 (replace_swapped_load_constant): Capitalize argument name in
8631 comment describing this function.
8632 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8633 and stores that access quad-word aligned addresses and replace
8634 with stvx or lvx instructions when appropriate.
8635 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8636 New function prototype.
8637 (rs6000_quadword_masked_address_p): Likewise.
8638 (rs6000_gen_lvx): Likewise.
8639 (rs6000_gen_stvx): Likewise.
8640 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8641 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8642 when memory address is aligned.
8643 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8644 this split to select lvx instruction when memory address is aligned.
8645 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8646 instruction when memory address is aligned.
8647 (*vsx_le_perm_load_v16qi): Likewise.
8648 (four unnamed splitters): Modify to select the stvx instruction
8649 when memory is aligned.
8651 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8653 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8654 which appears in the queue multiple times.
8656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8657 Alan Hayward <alan.hayward@arm.com>
8658 David Sherwood <david.sherwood@arm.com>
8660 * tree-vectorizer.h (vec_lower_bound): New structure.
8661 (_loop_vec_info): Add check_nonzero and lower_bounds.
8662 (LOOP_VINFO_CHECK_NONZERO): New macro.
8663 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8664 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8665 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8666 fields. Make seg_len the distance travelled, not including the
8668 (dr_direction_indicator): Declare.
8669 (dr_zero_step_indicator): Likewise.
8670 (dr_known_forward_stride_p): Likewise.
8671 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8673 (runtime_alias_check_p): Allow runtime alias checks with
8675 (operator ==): Compare access_size and align.
8676 (prune_runtime_alias_test_list): Rework for new distinction between
8677 the access_size and seg_len.
8678 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8680 (get_segment_min_max): New function.
8681 (create_intersect_range_checks): Use it.
8682 (dr_step_indicator): New function.
8683 (dr_direction_indicator): Likewise.
8684 (dr_zero_step_indicator): Likewise.
8685 (dr_known_forward_stride_p): Likewise.
8686 * tree-loop-distribution.c (data_ref_segment_size): Return
8687 DR_STEP * (niters - 1).
8688 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8690 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8691 (vect_preserves_scalar_order_p): New function, split out from...
8692 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8693 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8694 (vect_vfa_access_size): New function.
8695 (vect_vfa_align): Likewise.
8696 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8697 (dump_lower_bound): New function.
8698 (vect_check_lower_bound): Likewise.
8699 (vect_small_gap_p): Likewise.
8700 (vectorizable_with_step_bound_p): Likewise.
8701 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8702 depencies if the vectorization factor is 1. Convert the checks
8703 for nonzero steps into checks on the bounds of DR_STEP. Try using
8704 a bunds check for variable steps if the minimum required step is
8705 relatively small. Update calls to the dr_with_seg_len
8706 constructor and to vect_compile_time_alias.
8707 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8709 (vect_loop_versioning): Call it.
8710 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8712 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8714 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8715 Alan Hayward <alan.hayward@arm.com>
8716 David Sherwood <david.sherwood@arm.com>
8718 * doc/sourcebuild.texi (vect_scatter_store): Document.
8719 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8721 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8723 * genopinit.c (main): Add supports_vec_scatter_store and
8724 supports_vec_scatter_store_cached to target_optabs.
8725 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8726 IFN_MASK_SCATTER_STORE.
8727 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8729 * internal-fn.h (internal_store_fn_p): Declare.
8730 (internal_fn_stored_value_index): Likewise.
8731 * internal-fn.c (scatter_store_direct): New macro.
8732 (expand_scatter_store_optab_fn): New function.
8733 (direct_scatter_store_optab_supported_p): New macro.
8734 (internal_store_fn_p): New function.
8735 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8736 IFN_MASK_SCATTER_STORE.
8737 (internal_fn_mask_index): Likewise.
8738 (internal_fn_stored_value_index): New function.
8739 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8741 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8742 * optabs-query.c (supports_vec_scatter_store_p): New function.
8743 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8744 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8745 true for scatter stores.
8746 (vect_gather_scatter_fn_p): Handle scatter stores too.
8747 (vect_check_gather_scatter): Consider using scatter stores if
8748 supports_vec_scatter_store_p.
8749 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8751 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8752 internal_fn_stored_value_index.
8753 (check_load_store_masking): Handle scatter stores too.
8754 (vect_get_store_rhs): Make public.
8755 (vectorizable_call): Use internal_store_fn_p.
8756 (vectorizable_store): Handle scatter store internal functions.
8757 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8758 when deciding whether the end of the group has been reached.
8759 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8760 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8761 (mask_scatter_store<mode>): New insns.
8763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8764 Alan Hayward <alan.hayward@arm.com>
8765 David Sherwood <david.sherwood@arm.com>
8767 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8768 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8769 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8771 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8772 Use vect_truncate_gather_scatter_offset if we can't treat the
8773 operation as a normal gather load or scatter store.
8774 (get_group_load_store_type): Take the gather_scatter_info
8775 as argument. Try using a gather load or scatter store for
8776 single-element groups.
8777 (get_load_store_type): Update calls to get_group_load_store_type
8778 and vect_use_strided_gather_scatters_p.
8780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8781 Alan Hayward <alan.hayward@arm.com>
8782 David Sherwood <david.sherwood@arm.com>
8784 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8785 optional tree argument.
8786 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8788 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8789 but continue to use the current value as a fallback.
8790 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8791 to compare the updates.
8792 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8793 (get_load_store_type): Use it when handling a strided access.
8794 (vect_get_strided_load_store_ops): New function.
8795 (vect_get_data_ptr_increment): Likewise.
8796 (vectorizable_load): Handle strided gather loads. Always pass
8797 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8799 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8800 Alan Hayward <alan.hayward@arm.com>
8801 David Sherwood <david.sherwood@arm.com>
8803 * doc/md.texi (gather_load@var{m}): Document.
8804 (mask_gather_load@var{m}): Likewise.
8805 * genopinit.c (main): Add supports_vec_gather_load and
8806 supports_vec_gather_load_cached to target_optabs.
8807 * optabs-tree.c (init_tree_optimization_optabs): Use
8808 ggc_cleared_alloc to allocate target_optabs.
8809 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8810 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8812 * internal-fn.h (internal_load_fn_p): Declare.
8813 (internal_gather_scatter_fn_p): Likewise.
8814 (internal_fn_mask_index): Likewise.
8815 (internal_gather_scatter_fn_supported_p): Likewise.
8816 * internal-fn.c (gather_load_direct): New macro.
8817 (expand_gather_load_optab_fn): New function.
8818 (direct_gather_load_optab_supported_p): New macro.
8819 (direct_internal_fn_optab): New function.
8820 (internal_load_fn_p): Likewise.
8821 (internal_gather_scatter_fn_p): Likewise.
8822 (internal_fn_mask_index): Likewise.
8823 (internal_gather_scatter_fn_supported_p): Likewise.
8824 * optabs-query.c (supports_at_least_one_mode_p): New function.
8825 (supports_vec_gather_load_p): Likewise.
8826 * optabs-query.h (supports_vec_gather_load_p): Declare.
8827 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8828 and memory_type field.
8829 (NUM_PATTERNS): Bump to 15.
8830 * tree-vect-data-refs.c: Include internal-fn.h.
8831 (vect_gather_scatter_fn_p): New function.
8832 (vect_describe_gather_scatter_call): Likewise.
8833 (vect_check_gather_scatter): Try using internal functions for
8834 gather loads. Recognize existing calls to a gather load function.
8835 (vect_analyze_data_refs): Consider using gather loads if
8836 supports_vec_gather_load_p.
8837 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8838 (vect_get_gather_scatter_offset_type): Likewise.
8839 (vect_convert_mask_for_vectype): Likewise.
8840 (vect_add_conversion_to_patterm): Likewise.
8841 (vect_try_gather_scatter_pattern): Likewise.
8842 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8843 (vect_vect_recog_func_ptrs): Add it.
8844 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8845 internal_fn_mask_index and internal_gather_scatter_fn_p.
8846 (check_load_store_masking): Take the gather_scatter_info as an
8847 argument and handle gather loads.
8848 (vect_get_gather_scatter_ops): New function.
8849 (vectorizable_call): Check internal_load_fn_p.
8850 (vectorizable_load): Likewise. Handle gather load internal
8852 (vectorizable_store): Update call to check_load_store_masking.
8853 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8854 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8855 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8856 (aarch64_gather_scale_operand_d): New predicates.
8857 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8858 (mask_gather_load<mode>): New insns.
8860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8861 Alan Hayward <alan.hayward@arm.com>
8862 David Sherwood <david.sherwood@arm.com>
8864 * optabs.def (fold_left_plus_optab): New optab.
8865 * doc/md.texi (fold_left_plus_@var{m}): Document.
8866 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8867 * internal-fn.c (fold_left_direct): Define.
8868 (expand_fold_left_optab_fn): Likewise.
8869 (direct_fold_left_optab_supported_p): Likewise.
8870 * fold-const-call.c (fold_const_fold_left): New function.
8871 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8872 * tree-parloops.c (valid_reduction_p): New function.
8873 (gather_scalar_reductions): Use it.
8874 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8875 (vect_finish_replace_stmt): Declare.
8876 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8877 (needs_fold_left_reduction_p): New function, split out from...
8878 (vect_is_simple_reduction): ...here. Accept reductions that
8879 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8880 (vect_force_simple_reduction): Also store the reduction type in
8881 the assignment's STMT_VINFO_REDUC_TYPE.
8882 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8883 (merge_with_identity): New function.
8884 (vect_expand_fold_left): Likewise.
8885 (vectorize_fold_left_reduction): Likewise.
8886 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8887 scalar phi in place for it. Check for target support and reject
8888 cases that would reassociate the operation. Defer the transform
8889 phase to vectorize_fold_left_reduction.
8890 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8891 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8892 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8894 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8896 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8897 call to ifc_temp_var.
8899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8900 Alan Hayward <alan.hayward@arm.com>
8901 David Sherwood <david.sherwood@arm.com>
8903 * target.def (legitimize_address_displacement): Take the original
8904 offset as a poly_int.
8905 * targhooks.h (default_legitimize_address_displacement): Update
8907 * targhooks.c (default_legitimize_address_displacement): Likewise.
8908 * doc/tm.texi: Regenerate.
8909 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8910 as an argument, moving assert of ad->disp == ad->disp_term to...
8911 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8912 Try calling targetm.legitimize_address_displacement before expanding
8913 the address rather than afterwards, and adjust for the new interface.
8914 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8915 Match the new hook interface. Handle SVE addresses.
8916 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8919 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8921 * Makefile.in (OBJS): Add early-remat.o.
8922 * target.def (select_early_remat_modes): New hook.
8923 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8924 * doc/tm.texi: Regenerate.
8925 * targhooks.h (default_select_early_remat_modes): Declare.
8926 * targhooks.c (default_select_early_remat_modes): New function.
8927 * timevar.def (TV_EARLY_REMAT): New timevar.
8928 * passes.def (pass_early_remat): New pass.
8929 * tree-pass.h (make_pass_early_remat): Declare.
8930 * early-remat.c: New file.
8931 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8933 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8936 Alan Hayward <alan.hayward@arm.com>
8937 David Sherwood <david.sherwood@arm.com>
8939 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8940 vfm1 with a bound_epilog parameter.
8941 (vect_do_peeling): Update calls accordingly, and move the prologue
8942 call earlier in the function. Treat the base bound_epilog as 0 for
8943 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8944 this base when peeling for gaps.
8945 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8946 with fully-masked loops.
8947 (vect_estimate_min_profitable_iters): Handle the single peeled
8948 iteration in that case.
8950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8951 Alan Hayward <alan.hayward@arm.com>
8952 David Sherwood <david.sherwood@arm.com>
8954 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8955 single-element interleaving even if the size is not a power of 2.
8956 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8957 accesses for single-element interleaving if the group size is
8960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8961 Alan Hayward <alan.hayward@arm.com>
8962 David Sherwood <david.sherwood@arm.com>
8964 * doc/md.texi (fold_extract_last_@var{m}): Document.
8965 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8966 * optabs.def (fold_extract_last_optab): New optab.
8967 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8968 * internal-fn.c (fold_extract_direct): New macro.
8969 (expand_fold_extract_optab_fn): Likewise.
8970 (direct_fold_extract_optab_supported_p): Likewise.
8971 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8972 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8973 EXTRACT_LAST_REDUCTION.
8974 (get_initial_def_for_reduction): Do not create an initial vector
8975 for EXTRACT_LAST_REDUCTION reductions.
8976 (vectorizable_reduction): Leave the scalar phi in place for
8977 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8978 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8979 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8980 transform phase to vectorizable_condition.
8981 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8983 (vect_finish_stmt_generation): ...here.
8984 (vect_finish_replace_stmt): New function.
8985 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8986 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8988 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8991 Alan Hayward <alan.hayward@arm.com>
8992 David Sherwood <david.sherwood@arm.com>
8994 * doc/md.texi (extract_last_@var{m}): Document.
8995 * optabs.def (extract_last_optab): New optab.
8996 * internal-fn.def (EXTRACT_LAST): New internal function.
8997 * internal-fn.c (cond_unary_direct): New macro.
8998 (expand_cond_unary_optab_fn): Likewise.
8999 (direct_cond_unary_optab_supported_p): Likewise.
9000 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9001 loops using EXTRACT_LAST.
9002 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9003 (extract_last_<mode>): ...this optab.
9004 (vec_extract<mode><Vel>): Update accordingly.
9006 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9007 Alan Hayward <alan.hayward@arm.com>
9008 David Sherwood <david.sherwood@arm.com>
9010 * target.def (empty_mask_is_expensive): New hook.
9011 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9012 * doc/tm.texi: Regenerate.
9013 * targhooks.h (default_empty_mask_is_expensive): Declare.
9014 * targhooks.c (default_empty_mask_is_expensive): New function.
9015 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9016 if the target says that empty masks are expensive.
9017 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9019 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9021 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9022 Alan Hayward <alan.hayward@arm.com>
9023 David Sherwood <david.sherwood@arm.com>
9025 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9026 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9027 (vect_use_loop_mask_for_alignment_p): New function.
9028 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9029 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9030 niters_skip argument. Make sure that the first niters_skip elements
9031 of the first iteration are inactive.
9032 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9033 Update call to vect_set_loop_masks_directly.
9034 (get_misalign_in_elems): New function, split out from...
9035 (vect_gen_prolog_loop_niters): ...here.
9036 (vect_update_init_of_dr): Take a code argument that specifies whether
9037 the adjustment should be added or subtracted.
9038 (vect_update_init_of_drs): Likewise.
9039 (vect_prepare_for_masked_peels): New function.
9040 (vect_do_peeling): Skip prologue peeling if we're using a mask
9041 instead. Update call to vect_update_inits_of_drs.
9042 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9044 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9045 alignment. Do not include the number of peeled iterations in
9046 the minimum threshold in that case.
9047 (vectorizable_induction): Adjust the start value down by
9048 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9049 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9050 Take the number of skipped iterations into account when calculating
9052 * tree-vect-stmts.c (vect_gen_while_not): New function.
9054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9055 Alan Hayward <alan.hayward@arm.com>
9056 David Sherwood <david.sherwood@arm.com>
9058 * doc/sourcebuild.texi (vect_fully_masked): Document.
9059 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9061 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9063 (vect_analyze_loop_2): ...here. Don't check the vectorization
9064 factor against the number of loop iterations if the loop is
9067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9068 Alan Hayward <alan.hayward@arm.com>
9069 David Sherwood <david.sherwood@arm.com>
9071 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9072 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9073 (dump_groups): Update accordingly.
9074 (iv_use::mem_type): New member variable.
9075 (address_p): New function.
9076 (record_use): Add a mem_type argument and initialize the new
9078 (record_group_use): Add a mem_type argument. Use address_p.
9079 Remove obsolete null checks of base_object. Update call to record_use.
9080 (find_interesting_uses_op): Update call to record_group_use.
9081 (find_interesting_uses_cond): Likewise.
9082 (find_interesting_uses_address): Likewise.
9083 (get_mem_type_for_internal_fn): New function.
9084 (find_address_like_use): Likewise.
9085 (find_interesting_uses_stmt): Try find_address_like_use before
9086 calling find_interesting_uses_op.
9087 (addr_offset_valid_p): Use the iv mem_type field as the type
9088 of the addressed memory.
9089 (add_autoinc_candidates): Likewise.
9090 (get_address_cost): Likewise.
9091 (split_small_address_groups_p): Use address_p.
9092 (split_address_groups): Likewise.
9093 (add_iv_candidate_for_use): Likewise.
9094 (autoinc_possible_for_pair): Likewise.
9095 (rewrite_groups): Likewise.
9096 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9097 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9098 (get_alias_ptr_type_for_ptr_address): New function.
9099 (rewrite_use_address): Rewrite address uses in calls that were
9100 identified by find_address_like_use.
9102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9103 Alan Hayward <alan.hayward@arm.com>
9104 David Sherwood <david.sherwood@arm.com>
9106 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9108 * gimple-expr.h (is_gimple_addressable: Likewise.
9109 * gimple-expr.c (is_gimple_address): Likewise.
9110 * internal-fn.c (expand_call_mem_ref): New function.
9111 (expand_mask_load_optab_fn): Use it.
9112 (expand_mask_store_optab_fn): Likewise.
9114 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9115 Alan Hayward <alan.hayward@arm.com>
9116 David Sherwood <david.sherwood@arm.com>
9118 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9119 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9120 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9121 (cond_umax@var{mode}): Document.
9122 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9123 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9124 (cond_umin_optab, cond_umax_optab): New optabs.
9125 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9126 (COND_IOR, COND_XOR): New internal functions.
9127 * internal-fn.h (get_conditional_internal_fn): Declare.
9128 * internal-fn.c (cond_binary_direct): New macro.
9129 (expand_cond_binary_optab_fn): Likewise.
9130 (direct_cond_binary_optab_supported_p): Likewise.
9131 (get_conditional_internal_fn): New function.
9132 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9133 Cope with reduction statements that are vectorized as calls rather
9135 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9136 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9137 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9138 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9139 (UNSPEC_COND_EOR): New unspecs.
9140 (optab): Add mappings for them.
9141 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9142 (sve_int_op, sve_fp_op): New int attributes.
9144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9145 Alan Hayward <alan.hayward@arm.com>
9146 David Sherwood <david.sherwood@arm.com>
9148 * optabs.def (while_ult_optab): New optab.
9149 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9150 * internal-fn.def (WHILE_ULT): New internal function.
9151 * internal-fn.h (direct_internal_fn_supported_p): New override
9152 that takes two types as argument.
9153 * internal-fn.c (while_direct): New macro.
9154 (expand_while_optab_fn): New function.
9155 (convert_optab_supported_p): Likewise.
9156 (direct_while_optab_supported_p): New macro.
9157 * wide-int.h (wi::udiv_ceil): New function.
9158 * tree-vectorizer.h (rgroup_masks): New structure.
9159 (vec_loop_masks): New typedef.
9160 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9162 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9163 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9164 (vect_max_vf): New function.
9165 (slpeel_make_loop_iterate_ntimes): Delete.
9166 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9167 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9168 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9169 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9170 internal-fn.h, stor-layout.h and optabs-query.h.
9171 (vect_set_loop_mask): New function.
9172 (add_preheader_seq): Likewise.
9173 (add_header_seq): Likewise.
9174 (interleave_supported_p): Likewise.
9175 (vect_maybe_permute_loop_masks): Likewise.
9176 (vect_set_loop_masks_directly): Likewise.
9177 (vect_set_loop_condition_masked): Likewise.
9178 (vect_set_loop_condition_unmasked): New function, split out from
9179 slpeel_make_loop_iterate_ntimes.
9180 (slpeel_make_loop_iterate_ntimes): Rename to..
9181 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9182 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9183 (vect_do_peeling): Update call accordingly.
9184 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9186 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9187 mask_compare_type, can_fully_mask_p and fully_masked_p.
9188 (release_vec_loop_masks): New function.
9189 (_loop_vec_info): Use it to free the loop masks.
9190 (can_produce_all_loop_masks_p): New function.
9191 (vect_get_max_nscalars_per_iter): Likewise.
9192 (vect_verify_full_masking): Likewise.
9193 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9194 retries, and free the mask rgroups before retrying. Check loop-wide
9195 reasons for disallowing fully-masked loops. Make the final decision
9196 about whether use a fully-masked loop or not.
9197 (vect_estimate_min_profitable_iters): Do not assume that peeling
9198 for the number of iterations will be needed for fully-masked loops.
9199 (vectorizable_reduction): Disable fully-masked loops.
9200 (vectorizable_live_operation): Likewise.
9201 (vect_halve_mask_nunits): New function.
9202 (vect_double_mask_nunits): Likewise.
9203 (vect_record_loop_mask): Likewise.
9204 (vect_get_loop_mask): Likewise.
9205 (vect_transform_loop): Handle the case in which the final loop
9206 iteration might handle a partial vector. Call vect_set_loop_condition
9207 instead of slpeel_make_loop_iterate_ntimes.
9208 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9209 (check_load_store_masking): New function.
9210 (prepare_load_store_mask): Likewise.
9211 (vectorizable_store): Handle fully-masked loops.
9212 (vectorizable_load): Likewise.
9213 (supportable_widening_operation): Use vect_halve_mask_nunits for
9215 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9216 (vect_gen_while): New function.
9217 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9218 (aarch64_uqdec<mode>): New insn.
9220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9221 Alan Hayward <alan.hayward@arm.com>
9222 David Sherwood <david.sherwood@arm.com>
9224 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9225 (reduc_xor_scal_optab): New optabs.
9226 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9227 (reduc_xor_scal_@var{m}): Document.
9228 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9229 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9231 * fold-const-call.c (fold_const_call): Handle them.
9232 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9233 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9234 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9235 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9236 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9237 (UNSPEC_XORV): New unspecs.
9238 (optab): Add entries for them.
9239 (BITWISEV): New int iterator.
9240 (bit_reduc_op): New int attributes.
9242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9243 Alan Hayward <alan.hayward@arm.com>
9244 David Sherwood <david.sherwood@arm.com>
9246 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9247 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9248 * optabs.def (vec_shl_insert_optab): New optab.
9249 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9250 (duplicate_and_interleave): Likewise.
9251 * tree-vect-loop.c: Include internal-fn.h.
9252 (neutral_op_for_slp_reduction): New function, split out from
9253 get_initial_defs_for_reduction.
9254 (get_initial_def_for_reduction): Handle option 2 for variable-length
9255 vectors by loading the neutral value into a vector and then shifting
9256 the initial value into element 0.
9257 (get_initial_defs_for_reduction): Replace the code argument with
9258 the neutral value calculated by neutral_op_for_slp_reduction.
9259 Use gimple_build_vector for constant-length vectors.
9260 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9261 but the first group_size elements have a neutral value.
9262 Use duplicate_and_interleave otherwise.
9263 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9264 Update call to get_initial_defs_for_reduction. Handle SLP
9265 reductions for variable-length vectors by creating one vector
9266 result for each scalar result, with the elements associated
9267 with other scalar results stubbed out with the neutral value.
9268 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9269 Require IFN_VEC_SHL_INSERT for double reductions on
9270 variable-length vectors, or SLP reductions that have
9271 a neutral value. Require can_duplicate_and_interleave_p
9272 support for variable-length unchained SLP reductions if there
9273 is no neutral value, such as for MIN/MAX reductions. Also require
9274 the number of vector elements to be a multiple of the number of
9275 SLP statements when doing variable-length unchained SLP reductions.
9276 Update call to vect_create_epilog_for_reduction.
9277 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9278 and remove initial values.
9279 (duplicate_and_interleave): Make public.
9280 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9281 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9284 Alan Hayward <alan.hayward@arm.com>
9285 David Sherwood <david.sherwood@arm.com>
9287 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9288 (can_duplicate_and_interleave_p): New function.
9289 (vect_get_and_check_slp_defs): Take the vector of statements
9290 rather than just the current one. Remove excess parentheses.
9291 Restriction rejectinon of vect_constant_def and vect_external_def
9292 for variable-length vectors to boolean types, or types for which
9293 can_duplicate_and_interleave_p is false.
9294 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9295 (duplicate_and_interleave): New function.
9296 (vect_get_constant_vectors): Use gimple_build_vector for
9297 constant-length vectors and suitable variable-length constant
9298 vectors. Use duplicate_and_interleave for other variable-length
9299 vectors. Don't defer the update when inserting new statements.
9301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9302 Alan Hayward <alan.hayward@arm.com>
9303 David Sherwood <david.sherwood@arm.com>
9305 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9306 min_profitable_iters doesn't go negative.
9308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9309 Alan Hayward <alan.hayward@arm.com>
9310 David Sherwood <david.sherwood@arm.com>
9312 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9313 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9314 * optabs.def (vec_mask_load_lanes_optab): New optab.
9315 (vec_mask_store_lanes_optab): Likewise.
9316 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9317 (MASK_STORE_LANES): Likewise.
9318 * internal-fn.c (mask_load_lanes_direct): New macro.
9319 (mask_store_lanes_direct): Likewise.
9320 (expand_mask_load_optab_fn): Handle masked operations.
9321 (expand_mask_load_lanes_optab_fn): New macro.
9322 (expand_mask_store_optab_fn): Handle masked operations.
9323 (expand_mask_store_lanes_optab_fn): New macro.
9324 (direct_mask_load_lanes_optab_supported_p): Likewise.
9325 (direct_mask_store_lanes_optab_supported_p): Likewise.
9326 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9328 (vect_load_lanes_supported): Likewise.
9329 * tree-vect-data-refs.c (strip_conversion): New function.
9330 (can_group_stmts_p): Likewise.
9331 (vect_analyze_data_ref_accesses): Use it instead of checking
9332 for a pair of assignments.
9333 (vect_store_lanes_supported): Take a masked_p parameter.
9334 (vect_load_lanes_supported): Likewise.
9335 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9336 vect_store_lanes_supported and vect_load_lanes_supported.
9337 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9338 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9339 parameter. Don't allow gaps for masked accesses.
9340 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9341 and vect_load_lanes_supported.
9342 (get_load_store_type): Take a masked_p parameter and update
9343 call to get_group_load_store_type.
9344 (vectorizable_store): Update call to get_load_store_type.
9345 Handle IFN_MASK_STORE_LANES.
9346 (vectorizable_load): Update call to get_load_store_type.
9347 Handle IFN_MASK_LOAD_LANES.
9349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9350 Alan Hayward <alan.hayward@arm.com>
9351 David Sherwood <david.sherwood@arm.com>
9353 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9355 * config/aarch64/aarch64-protos.h
9356 (aarch64_sve_struct_memory_operand_p): Declare.
9357 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9358 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9359 (VPRED, vpred): Handle SVE structure modes.
9360 * config/aarch64/constraints.md (Utx): New constraint.
9361 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9362 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9363 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9364 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9365 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9366 structure modes. Split into pieces after RA.
9367 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9368 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9370 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9371 SVE structure modes.
9372 (aarch64_classify_address): Likewise.
9373 (sizetochar): Move earlier in file.
9374 (aarch64_print_operand): Handle SVE register lists.
9375 (aarch64_array_mode): New function.
9376 (aarch64_sve_struct_memory_operand_p): Likewise.
9377 (TARGET_ARRAY_MODE): Redefine.
9379 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9380 Alan Hayward <alan.hayward@arm.com>
9381 David Sherwood <david.sherwood@arm.com>
9383 * target.def (array_mode): New target hook.
9384 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9385 * doc/tm.texi: Regenerate.
9386 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9387 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9388 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9390 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9393 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9394 Alan Hayward <alan.hayward@arm.com>
9395 David Sherwood <david.sherwood@arm.com>
9397 * fold-const.c (fold_binary_loc): Check the argument types
9398 rather than the result type when testing for a vector operation.
9400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9402 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9403 * doc/tm.texi: Regenerate.
9405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9406 Alan Hayward <alan.hayward@arm.com>
9407 David Sherwood <david.sherwood@arm.com>
9409 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9410 (sve): Document new AArch64 extension.
9411 * doc/md.texi (w): Extend the description of the AArch64
9412 constraint to include SVE vectors.
9413 (Upl, Upa): Document new AArch64 predicate constraints.
9414 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9416 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9417 (msve-vector-bits=): New option.
9418 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9419 SVE when these are disabled.
9420 (sve): New extension.
9421 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9422 modes. Adjust their number of units based on aarch64_sve_vg.
9423 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9424 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9425 aarch64_addr_query_type.
9426 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9427 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9428 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9429 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9430 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9431 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9432 (aarch64_simd_imm_zero_p): Delete.
9433 (aarch64_check_zero_based_sve_index_immediate): Declare.
9434 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9435 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9436 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9437 (aarch64_sve_float_mul_immediate_p): Likewise.
9438 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9440 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9441 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9442 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9443 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9444 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9445 (aarch64_regmode_natural_size): Likewise.
9446 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9447 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9449 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9450 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9451 for VG and the SVE predicate registers.
9452 (V_ALIASES): Add a "z"-prefixed alias.
9453 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9454 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9455 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9456 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9457 (REG_CLASS_NAMES): Add entries for them.
9458 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9459 and the predicate registers.
9460 (aarch64_sve_vg): Declare.
9461 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9462 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9463 (REGMODE_NATURAL_SIZE): Define.
9464 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9466 * config/aarch64/aarch64.c: Include cfgrtl.h.
9467 (simd_immediate_info): Add a constructor for series vectors,
9468 and an associated step field.
9469 (aarch64_sve_vg): New variable.
9470 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9471 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9472 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9473 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9474 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9475 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9476 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9477 (aarch64_get_mask_mode): New functions.
9478 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9479 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9480 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9481 predicate modes and predicate registers. Explicitly restrict
9482 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9483 to store a vector mode if it is recognized by
9484 aarch64_classify_vector_mode.
9485 (aarch64_regmode_natural_size): New function.
9486 (aarch64_hard_regno_caller_save_mode): Return the original mode
9488 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9489 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9490 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9491 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9493 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9494 does not overlap dest if the function is frame-related. Handle
9496 (aarch64_split_add_offset): New function.
9497 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9498 them aarch64_add_offset.
9499 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9500 and update call to aarch64_sub_sp.
9501 (aarch64_add_cfa_expression): New function.
9502 (aarch64_expand_prologue): Pass extra temporary registers to the
9503 functions above. Handle the case in which we need to emit new
9504 DW_CFA_expressions for registers that were originally saved
9505 relative to the stack pointer, but now have to be expressed
9506 relative to the frame pointer.
9507 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9509 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9510 IP0 and IP1 values for SVE frames.
9511 (aarch64_expand_vec_series): New function.
9512 (aarch64_expand_sve_widened_duplicate): Likewise.
9513 (aarch64_expand_sve_const_vector): Likewise.
9514 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9515 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9516 into the register, rather than emitting a SET directly.
9517 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9518 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9519 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9520 (offset_9bit_signed_scaled_p): New functions.
9521 (aarch64_replicate_bitmask_imm): New function.
9522 (aarch64_bitmask_imm): Use it.
9523 (aarch64_cannot_force_const_mem): Reject expressions involving
9524 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9525 (aarch64_classify_index): Handle SVE indices, by requiring
9526 a plain register index with a scale that matches the element size.
9527 (aarch64_classify_address): Handle SVE addresses. Assert that
9528 the mode of the address is VOIDmode or an integer mode.
9529 Update call to aarch64_classify_symbol.
9530 (aarch64_classify_symbolic_expression): Update call to
9531 aarch64_classify_symbol.
9532 (aarch64_const_vec_all_in_range_p): New function.
9533 (aarch64_print_vector_float_operand): Likewise.
9534 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9535 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9536 and the FP immediates 1.0 and 0.5.
9537 (aarch64_print_address_internal): Handle SVE addresses.
9538 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9539 (aarch64_regno_regclass): Handle predicate registers.
9540 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9542 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9543 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9544 (aarch64_convert_sve_vector_bits): New function.
9545 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9546 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9548 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9549 Handle SVE vector and predicate modes. Accept VL-based constants
9550 that need only one temporary register, and VL offsets that require
9551 no temporary registers.
9552 (aarch64_conditional_register_usage): Mark the predicate registers
9553 as fixed if SVE isn't available.
9554 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9555 Return true for SVE vector and predicate modes.
9556 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9557 rather than an unsigned int. Handle SVE modes.
9558 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9560 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9562 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9563 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9564 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9565 (aarch64_sve_float_mul_immediate_p): New functions.
9566 (aarch64_sve_valid_immediate): New function.
9567 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9568 Explicitly reject structure modes. Check for INDEX constants.
9569 Handle PTRUE and PFALSE constants.
9570 (aarch64_check_zero_based_sve_index_immediate): New function.
9571 (aarch64_simd_imm_zero_p): Delete.
9572 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9573 vector modes. Accept constants in the range of CNT[BHWD].
9574 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9575 ask for an Advanced SIMD mode.
9576 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9577 (aarch64_simd_vector_alignment): Handle SVE predicates.
9578 (aarch64_vectorize_preferred_vector_alignment): New function.
9579 (aarch64_simd_vector_alignment_reachable): Use it instead of
9581 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9582 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9584 (MAX_VECT_LEN): Delete.
9585 (expand_vec_perm_d): Add a vec_flags field.
9586 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9587 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9588 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9590 (aarch64_evpc_rev): Rename to...
9591 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9592 (aarch64_evpc_rev_global): New function.
9593 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9594 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9596 (aarch64_evpc_sve_tbl): New function.
9597 (aarch64_expand_vec_perm_const_1): Update after rename of
9598 aarch64_evpc_rev. Handle SVE permutes too, trying
9599 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9600 than aarch64_evpc_tbl.
9601 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9602 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9603 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9604 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9605 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9606 (aarch64_expand_sve_vcond): New functions.
9607 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9608 of aarch64_vector_mode_p.
9609 (aarch64_dwarf_poly_indeterminate_value): New function.
9610 (aarch64_compute_pressure_classes): Likewise.
9611 (aarch64_can_change_mode_class): Likewise.
9612 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9613 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9614 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9615 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9616 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9617 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9618 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9619 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9621 (Dn, Dl, Dr): Accept const as well as const_vector.
9622 (Dz): Likewise. Compare against CONST0_RTX.
9623 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9624 of "vector" where appropriate.
9625 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9626 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9627 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9628 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9629 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9630 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9631 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9632 (v_int_equiv): Extend to SVE modes.
9633 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9635 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9636 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9637 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9638 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9639 (SVE_COND_FP_CMP): New int iterators.
9640 (perm_hilo): Handle the new unpack unspecs.
9641 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9643 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9644 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9645 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9646 (aarch64_equality_operator, aarch64_constant_vector_operand)
9647 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9648 (aarch64_sve_nonimmediate_operand): Likewise.
9649 (aarch64_sve_general_operand): Likewise.
9650 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9651 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9652 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9653 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9654 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9655 (aarch64_sve_float_arith_immediate): Likewise.
9656 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9657 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9658 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9659 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9660 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9661 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9662 (aarch64_sve_float_arith_operand): Likewise.
9663 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9664 (aarch64_sve_float_mul_operand): Likewise.
9665 (aarch64_sve_vec_perm_operand): Likewise.
9666 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9667 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9668 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9669 as well as const_vector.
9670 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9671 in file. Use CONST0_RTX and CONSTM1_RTX.
9672 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9673 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9674 Use aarch64_simd_imm_zero.
9675 * config/aarch64/aarch64-sve.md: New file.
9676 * config/aarch64/aarch64.md: Include it.
9677 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9678 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9679 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9680 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9681 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9682 (sve): New attribute.
9683 (enabled): Disable instructions with the sve attribute unless
9685 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9686 aarch64_expand_mov_immediate.
9687 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9688 CNT[BHSD] immediates.
9689 (movti): Split CONST_POLY_INT moves into two halves.
9690 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9691 Split additions that need a temporary here if the destination
9692 is the stack pointer.
9693 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9694 (*add<mode>3_poly_1): New instruction.
9695 (set_clobber_cc): New expander.
9697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9699 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9700 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9701 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9702 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9703 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9704 Change innermode from fixed_mode_size to machine_mode.
9705 (simplify_subreg): Update call accordingly. Handle a constant-sized
9706 subreg of a variable-length CONST_VECTOR.
9708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9709 Alan Hayward <alan.hayward@arm.com>
9710 David Sherwood <david.sherwood@arm.com>
9712 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9713 (add_offset_to_base): New function, split out from...
9714 (create_mem_ref): ...here. When handling a scale other than 1,
9715 check first whether the address is valid without the offset.
9716 Add it into the base if so, leaving the index and scale as-is.
9718 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9721 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9722 fold_for_warn before checking if arg2 is INTEGER_CST.
9724 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9726 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9727 (store_multiple_operation): Delete.
9728 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9729 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9730 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9731 guarded by TARGET_STRING.
9732 (rs6000_output_load_multiple): Delete.
9733 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9734 OPTION_MASK_STRING / TARGET_STRING handling.
9735 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9736 (const rs6000_opt_masks) <"string">: Change mask to 0.
9737 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9738 (MASK_STRING): Delete.
9739 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9741 (load_multiple): Delete.
9748 (store_multiple): Delete.
9755 (movmemsi_8reg): Delete.
9756 (corresponding unnamed define_insn): Delete.
9757 (movmemsi_6reg): Delete.
9758 (corresponding unnamed define_insn): Delete.
9759 (movmemsi_4reg): Delete.
9760 (corresponding unnamed define_insn): Delete.
9761 (movmemsi_2reg): Delete.
9762 (corresponding unnamed define_insn): Delete.
9763 (movmemsi_1reg): Delete.
9764 (corresponding unnamed define_insn): Delete.
9765 * config/rs6000/rs6000.opt (mno-string): New.
9766 (mstring): Replace by deprecation warning stub.
9767 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9769 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9771 * regrename.c (regrename_do_replace): If replacing the same
9772 reg multiple times, try to reuse last created gen_raw_REG.
9775 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9776 main to workaround a bug in GDB.
9778 2018-01-12 Tom de Vries <tom@codesourcery.com>
9781 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9783 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9785 PR rtl-optimization/80481
9786 * ira-color.c (get_cap_member): New function.
9787 (allocnos_conflict_by_live_ranges_p): Use it.
9788 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9789 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9791 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9794 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9795 (*saddl_se_1): Ditto.
9797 (*ssubl_se_1): Ditto.
9799 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9801 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9802 rather than wi::to_widest for DR_INITs.
9803 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9804 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9805 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9807 (vect_analyze_group_access_1): Note that here.
9809 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9811 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9812 polynomial type sizes.
9814 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9816 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9817 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9818 (gimple_add_tmp_var): Likewise.
9820 2018-01-12 Martin Liska <mliska@suse.cz>
9822 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9823 (gimple_alloc_sizes): Likewise.
9824 (dump_gimple_statistics): Use PRIu64 in printf format.
9825 * gimple.h: Change uint64_t to int.
9827 2018-01-12 Martin Liska <mliska@suse.cz>
9829 * tree-core.h: Use uint64_t instead of int.
9830 * tree.c (tree_node_counts): Likewise.
9831 (tree_node_sizes): Likewise.
9832 (dump_tree_statistics): Use PRIu64 in printf format.
9834 2018-01-12 Martin Liska <mliska@suse.cz>
9836 * Makefile.in: As qsort_chk is implemented in vec.c, add
9837 vec.o to linkage of gencfn-macros.
9838 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9839 passing the info to record_node_allocation_statistics.
9840 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9842 * ggc-common.c (struct ggc_usage): Add operator== and use
9843 it in operator< and compare function.
9844 * mem-stats.h (struct mem_usage): Likewise.
9845 * vec.c (struct vec_usage): Remove operator< and compare
9846 function. Can be simply inherited.
9848 2018-01-12 Martin Jambor <mjambor@suse.cz>
9851 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9852 * tree-ssa-math-opts.c: Include domwalk.h.
9853 (convert_mult_to_fma_1): New function.
9854 (fma_transformation_info): New type.
9855 (fma_deferring_state): Likewise.
9856 (cancel_fma_deferring): New function.
9857 (result_of_phi): Likewise.
9858 (last_fma_candidate_feeds_initial_phi): Likewise.
9859 (convert_mult_to_fma): Added deferring logic, split actual
9860 transformation to convert_mult_to_fma_1.
9861 (math_opts_dom_walker): New type.
9862 (math_opts_dom_walker::after_dom_children): New method, body moved
9863 here from pass_optimize_widening_mul::execute, added deferring logic
9865 (pass_optimize_widening_mul::execute): Moved most of code to
9866 math_opts_dom_walker::after_dom_children.
9867 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9868 * config/i386/i386.c (ix86_option_override_internal): Added
9869 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9871 2018-01-12 Richard Biener <rguenther@suse.de>
9874 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9875 inline instance vars.
9877 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9880 * config/rx/rx.c (rx_is_restricted_memory_address):
9883 2018-01-12 Richard Biener <rguenther@suse.de>
9885 PR tree-optimization/80846
9886 * target.def (split_reduction): New target hook.
9887 * targhooks.c (default_split_reduction): New function.
9888 * targhooks.h (default_split_reduction): Declare.
9889 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9890 target requests first reduce vectors by combining low and high
9892 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9893 (get_vectype_for_scalar_type_and_size): Export.
9894 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9895 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9896 * doc/tm.texi: Regenerate.
9897 * config/i386/i386.c (ix86_split_reduction): Implement
9898 TARGET_VECTORIZE_SPLIT_REDUCTION.
9900 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9903 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9904 in PIC mode except for TARGET_VXWORKS_RTP.
9905 * config/sparc/sparc.c: Include cfgrtl.h.
9906 (TARGET_INIT_PIC_REG): Define.
9907 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9908 (sparc_pic_register_p): New predicate.
9909 (sparc_legitimate_address_p): Use it.
9910 (sparc_legitimize_pic_address): Likewise.
9911 (sparc_delegitimize_address): Likewise.
9912 (sparc_mode_dependent_address_p): Likewise.
9913 (gen_load_pcrel_sym): Remove 4th parameter.
9914 (load_got_register): Adjust call to above. Remove obsolete stuff.
9915 (sparc_expand_prologue): Do not call load_got_register here.
9916 (sparc_flat_expand_prologue): Likewise.
9917 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9918 (sparc_use_pseudo_pic_reg): New function.
9919 (sparc_init_pic_reg): Likewise.
9920 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9921 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9923 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9925 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9926 Add item for branch_cost.
9928 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9930 PR rtl-optimization/83565
9931 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9932 not extend the result to a larger mode for rotate operations.
9933 (num_sign_bit_copies1): Likewise.
9935 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9938 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9940 Use values-Xc.o for -pedantic.
9941 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9943 2018-01-12 Martin Liska <mliska@suse.cz>
9946 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9948 (possible_polymorphic_call_targets): Use it.
9949 (ipa_devirt): Likewise.
9951 2018-01-12 Martin Liska <mliska@suse.cz>
9953 * profile-count.h (enum profile_quality): Use 0 as invalid
9954 enum value of profile_quality.
9956 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9958 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9959 -mext-string options.
9961 2018-01-12 Richard Biener <rguenther@suse.de>
9963 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9964 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9965 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9967 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9969 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9971 * configure.ac (--with-long-double-format): Add support for the
9972 configuration option to change the default long double format on
9974 * config.gcc (powerpc*-linux*-*): Likewise.
9975 * configure: Regenerate.
9976 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9977 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9978 used without modification.
9980 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9982 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9983 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9984 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9985 MISC_BUILTIN_SPEC_BARRIER.
9986 (rs6000_init_builtins): Likewise.
9987 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9989 (speculation_barrier): New define_insn.
9990 * doc/extend.texi: Document __builtin_speculation_barrier.
9992 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9995 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9996 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9997 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9999 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10000 integral modes instead of "ss" and "sd".
10001 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10002 vectors with 32-bit and 64-bit elements.
10003 (vecdupssescalarmodesuffix): New mode attribute.
10004 (vec_dup<mode>): Use it.
10006 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10009 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10010 frame if argument is passed on stack.
10012 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10015 * ree.c (combine_reaching_defs): Optimize also
10016 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10017 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10019 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10021 PR middle-end/83189
10022 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10024 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10026 PR middle-end/83718
10027 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10028 after they are computed.
10030 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10032 PR tree-optimization/83695
10033 * gimple-loop-linterchange.cc
10034 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10035 reset cached scev information after interchange.
10036 (pass_linterchange::execute): Remove call to scev_reset_htab.
10038 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10040 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10041 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10042 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10043 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10044 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10045 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10046 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10047 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10048 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10049 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10050 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10051 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10052 (V_lane_reg): Likewise.
10053 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10055 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10056 (vfmal_lane_low<mode>_intrinsic,
10057 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10058 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10059 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10060 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10061 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10062 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10064 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10066 * config/arm/arm-cpus.in (fp16fml): New feature.
10067 (ALL_SIMD): Add fp16fml.
10068 (armv8.2-a): Add fp16fml as an option.
10069 (armv8.3-a): Likewise.
10070 (armv8.4-a): Add fp16fml as part of fp16.
10071 * config/arm/arm.h (TARGET_FP16FML): Define.
10072 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10074 * config/arm/arm-modes.def (V2HF): Define.
10075 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10076 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10077 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10078 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10079 vfmsl_low, vfmsl_high): New set of builtins.
10080 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10081 (vfml_op): New code attribute.
10082 (VFMLHALVES): New int iterator.
10083 (VFML, VFMLSEL): New mode attributes.
10084 (V_reg): Define mapping for V2HF.
10085 (V_hi, V_lo): New mode attributes.
10086 (VF_constraint): Likewise.
10087 (vfml_half, vfml_half_selector): New int attributes.
10088 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10090 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10091 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10093 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10094 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10095 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10096 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10098 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10099 Document new effective target and option set.
10101 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10103 * config/arm/arm-cpus.in (armv8_4): New feature.
10104 (ARMv8_4a): New fgroup.
10105 (armv8.4-a): New arch.
10106 * config/arm/arm-tables.opt: Regenerate.
10107 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10108 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10109 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10110 Add matching rules for -march=armv8.4-a and extensions.
10111 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10113 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10116 * config/rx/rx.md (BW): New mode attribute.
10117 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10119 2018-01-11 Richard Biener <rguenther@suse.de>
10121 PR tree-optimization/83435
10122 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10123 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10124 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10126 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10127 Alan Hayward <alan.hayward@arm.com>
10128 David Sherwood <david.sherwood@arm.com>
10130 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10132 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10133 (aarch64_print_address_internal): Use it to check for a zero offset.
10135 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10136 Alan Hayward <alan.hayward@arm.com>
10137 David Sherwood <david.sherwood@arm.com>
10139 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10140 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10141 Return a poly_int64 rather than a HOST_WIDE_INT.
10142 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10143 rather than a HOST_WIDE_INT.
10144 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10145 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10146 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10147 final_offset from HOST_WIDE_INT to poly_int64.
10148 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10149 to_constant when getting the number of units in an Advanced SIMD
10151 (aarch64_builtin_vectorized_function): Check for a constant number
10153 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10155 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10156 attribute instead of GET_MODE_NUNITS.
10157 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10158 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10159 GET_MODE_SIZE for fixed-size registers.
10160 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10161 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10162 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10163 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10164 (aarch64_print_operand, aarch64_print_address_internal)
10165 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10166 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10167 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10168 Handle polynomial GET_MODE_SIZE.
10169 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10170 wider than SImode without modification.
10171 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10172 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10173 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10174 passing and returning SVE modes.
10175 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10176 rather than GEN_INT.
10177 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10178 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10179 (aarch64_allocate_and_probe_stack_space): Likewise.
10180 (aarch64_layout_frame): Cope with polynomial offsets.
10181 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10182 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10183 polynomial offsets.
10184 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10185 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10186 poly_int64 rather than a HOST_WIDE_INT.
10187 (aarch64_get_separate_components, aarch64_process_components)
10188 (aarch64_expand_prologue, aarch64_expand_epilogue)
10189 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10190 (aarch64_anchor_offset): New function, split out from...
10191 (aarch64_legitimize_address): ...here.
10192 (aarch64_builtin_vectorization_cost): Handle polynomial
10193 TYPE_VECTOR_SUBPARTS.
10194 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10196 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10197 number of elements from the PARALLEL rather than the mode.
10198 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10199 rather than GET_MODE_BITSIZE.
10200 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10201 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10202 (aarch64_expand_vec_perm_const_1): Handle polynomial
10203 d->perm.length () and d->perm elements.
10204 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10205 Apply to_constant to d->perm elements.
10206 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10207 polynomial CONST_VECTOR_NUNITS.
10208 (aarch64_move_pointer): Take amount as a poly_int64 rather
10210 (aarch64_progress_pointer): Avoid temporary variable.
10211 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10212 the mode attribute instead of GET_MODE.
10214 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10215 Alan Hayward <alan.hayward@arm.com>
10216 David Sherwood <david.sherwood@arm.com>
10218 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10219 x exists before using it.
10220 (aarch64_add_constant_internal): Rename to...
10221 (aarch64_add_offset_1): ...this. Replace regnum with separate
10222 src and dest rtxes. Handle the case in which they're different,
10223 including when the offset is zero. Replace scratchreg with an rtx.
10224 Use 2 additions if there is no spare register into which we can
10225 move a 16-bit constant.
10226 (aarch64_add_constant): Delete.
10227 (aarch64_add_offset): Replace reg with separate src and dest
10228 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10229 Use aarch64_add_offset_1.
10230 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10231 an rtx rather than an int. Take the delta as a poly_int64
10232 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10233 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10234 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10235 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10236 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10237 and aarch64_add_sp.
10238 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10239 aarch64_add_constant.
10241 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10243 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10244 Use scalar_float_mode.
10246 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10248 * config/aarch64/aarch64-simd.md
10249 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10250 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10251 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10252 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10253 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10254 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10255 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10256 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10257 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10258 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10260 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10263 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10264 targ_options->x_arm_arch_string is non NULL.
10266 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10268 * config/aarch64/aarch64.h
10269 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10271 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10274 * expmed.c (emit_store_flag_force): Swap if const op0
10275 and change VOIDmode to mode of op0.
10277 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10279 PR rtl-optimization/83761
10280 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10281 than bytes to mode_for_size.
10283 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10285 PR middle-end/83189
10286 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10287 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10290 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10292 PR middle-end/83575
10293 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10294 when in layout mode.
10295 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10296 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10299 2018-01-10 Michael Collison <michael.collison@arm.com>
10301 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10302 * config/aarch64/aarch64-option-extension.def: Add
10303 AARCH64_OPT_EXTENSION of 'fp16fml'.
10304 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10305 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10306 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10307 * config/aarch64/constraints.md (Ui7): New constraint.
10308 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10309 (VFMLA_SEL_W): Ditto.
10312 (VFMLA16_LOW): New int iterator.
10313 (VFMLA16_HIGH): Ditto.
10314 (UNSPEC_FMLAL): New unspec.
10315 (UNSPEC_FMLSL): Ditto.
10316 (UNSPEC_FMLAL2): Ditto.
10317 (UNSPEC_FMLSL2): Ditto.
10318 (f16mac): New code attribute.
10319 * config/aarch64/aarch64-simd-builtins.def
10320 (aarch64_fmlal_lowv2sf): Ditto.
10321 (aarch64_fmlsl_lowv2sf): Ditto.
10322 (aarch64_fmlalq_lowv4sf): Ditto.
10323 (aarch64_fmlslq_lowv4sf): Ditto.
10324 (aarch64_fmlal_highv2sf): Ditto.
10325 (aarch64_fmlsl_highv2sf): Ditto.
10326 (aarch64_fmlalq_highv4sf): Ditto.
10327 (aarch64_fmlslq_highv4sf): Ditto.
10328 (aarch64_fmlal_lane_lowv2sf): Ditto.
10329 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10330 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10331 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10332 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10333 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10334 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10335 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10336 (aarch64_fmlal_lane_highv2sf): Ditto.
10337 (aarch64_fmlsl_lane_highv2sf): Ditto.
10338 (aarch64_fmlal_laneq_highv2sf): Ditto.
10339 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10340 (aarch64_fmlalq_lane_highv4sf): Ditto.
10341 (aarch64_fmlsl_lane_highv4sf): Ditto.
10342 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10343 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10344 * config/aarch64/aarch64-simd.md:
10345 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10346 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10347 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10348 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10349 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10350 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10351 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10352 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10353 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10354 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10355 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10356 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10357 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10358 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10359 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10360 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10361 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10362 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10363 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10364 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10365 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10366 (vfmlsl_low_u32): Ditto.
10367 (vfmlalq_low_u32): Ditto.
10368 (vfmlslq_low_u32): Ditto.
10369 (vfmlal_high_u32): Ditto.
10370 (vfmlsl_high_u32): Ditto.
10371 (vfmlalq_high_u32): Ditto.
10372 (vfmlslq_high_u32): Ditto.
10373 (vfmlal_lane_low_u32): Ditto.
10374 (vfmlsl_lane_low_u32): Ditto.
10375 (vfmlal_laneq_low_u32): Ditto.
10376 (vfmlsl_laneq_low_u32): Ditto.
10377 (vfmlalq_lane_low_u32): Ditto.
10378 (vfmlslq_lane_low_u32): Ditto.
10379 (vfmlalq_laneq_low_u32): Ditto.
10380 (vfmlslq_laneq_low_u32): Ditto.
10381 (vfmlal_lane_high_u32): Ditto.
10382 (vfmlsl_lane_high_u32): Ditto.
10383 (vfmlal_laneq_high_u32): Ditto.
10384 (vfmlsl_laneq_high_u32): Ditto.
10385 (vfmlalq_lane_high_u32): Ditto.
10386 (vfmlslq_lane_high_u32): Ditto.
10387 (vfmlalq_laneq_high_u32): Ditto.
10388 (vfmlslq_laneq_high_u32): Ditto.
10389 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10390 (AARCH64_FL_FOR_ARCH8_4): New.
10391 (AARCH64_ISA_F16FML): New ISA flag.
10392 (TARGET_F16FML): New feature flag for fp16fml.
10393 (doc/invoke.texi): Document new fp16fml option.
10395 2018-01-10 Michael Collison <michael.collison@arm.com>
10397 * config/aarch64/aarch64-builtins.c:
10398 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10399 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10400 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10401 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10402 (AARCH64_ISA_SHA3): New ISA flag.
10403 (TARGET_SHA3): New feature flag for sha3.
10404 * config/aarch64/iterators.md (sha512_op): New int attribute.
10405 (CRYPTO_SHA512): New int iterator.
10406 (UNSPEC_SHA512H): New unspec.
10407 (UNSPEC_SHA512H2): Ditto.
10408 (UNSPEC_SHA512SU0): Ditto.
10409 (UNSPEC_SHA512SU1): Ditto.
10410 * config/aarch64/aarch64-simd-builtins.def
10411 (aarch64_crypto_sha512hqv2di): New builtin.
10412 (aarch64_crypto_sha512h2qv2di): Ditto.
10413 (aarch64_crypto_sha512su0qv2di): Ditto.
10414 (aarch64_crypto_sha512su1qv2di): Ditto.
10415 (aarch64_eor3qv8hi): Ditto.
10416 (aarch64_rax1qv2di): Ditto.
10417 (aarch64_xarqv2di): Ditto.
10418 (aarch64_bcaxqv8hi): Ditto.
10419 * config/aarch64/aarch64-simd.md:
10420 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10421 (aarch64_crypto_sha512su0qv2di): Ditto.
10422 (aarch64_crypto_sha512su1qv2di): Ditto.
10423 (aarch64_eor3qv8hi): Ditto.
10424 (aarch64_rax1qv2di): Ditto.
10425 (aarch64_xarqv2di): Ditto.
10426 (aarch64_bcaxqv8hi): Ditto.
10427 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10428 (vsha512h2q_u64): Ditto.
10429 (vsha512su0q_u64): Ditto.
10430 (vsha512su1q_u64): Ditto.
10431 (veor3q_u16): Ditto.
10432 (vrax1q_u64): Ditto.
10433 (vxarq_u64): Ditto.
10434 (vbcaxq_u16): Ditto.
10435 * config/arm/types.md (crypto_sha512): New type attribute.
10436 (crypto_sha3): Ditto.
10437 (doc/invoke.texi): Document new sha3 option.
10439 2018-01-10 Michael Collison <michael.collison@arm.com>
10441 * config/aarch64/aarch64-builtins.c:
10442 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10443 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10444 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10445 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10446 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10447 (AARCH64_ISA_SM4): New ISA flag.
10448 (TARGET_SM4): New feature flag for sm4.
10449 * config/aarch64/aarch64-simd-builtins.def
10450 (aarch64_sm3ss1qv4si): Ditto.
10451 (aarch64_sm3tt1aq4si): Ditto.
10452 (aarch64_sm3tt1bq4si): Ditto.
10453 (aarch64_sm3tt2aq4si): Ditto.
10454 (aarch64_sm3tt2bq4si): Ditto.
10455 (aarch64_sm3partw1qv4si): Ditto.
10456 (aarch64_sm3partw2qv4si): Ditto.
10457 (aarch64_sm4eqv4si): Ditto.
10458 (aarch64_sm4ekeyqv4si): Ditto.
10459 * config/aarch64/aarch64-simd.md:
10460 (aarch64_sm3ss1qv4si): Ditto.
10461 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10462 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10463 (aarch64_sm4eqv4si): Ditto.
10464 (aarch64_sm4ekeyqv4si): Ditto.
10465 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10466 (sm3part_op): Ditto.
10467 (CRYPTO_SM3TT): Ditto.
10468 (CRYPTO_SM3PART): Ditto.
10469 (UNSPEC_SM3SS1): New unspec.
10470 (UNSPEC_SM3TT1A): Ditto.
10471 (UNSPEC_SM3TT1B): Ditto.
10472 (UNSPEC_SM3TT2A): Ditto.
10473 (UNSPEC_SM3TT2B): Ditto.
10474 (UNSPEC_SM3PARTW1): Ditto.
10475 (UNSPEC_SM3PARTW2): Ditto.
10476 (UNSPEC_SM4E): Ditto.
10477 (UNSPEC_SM4EKEY): Ditto.
10478 * config/aarch64/constraints.md (Ui2): New constraint.
10479 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10480 * config/arm/types.md (crypto_sm3): New type attribute.
10481 (crypto_sm4): Ditto.
10482 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10483 (vsm3tt1aq_u32): Ditto.
10484 (vsm3tt1bq_u32): Ditto.
10485 (vsm3tt2aq_u32): Ditto.
10486 (vsm3tt2bq_u32): Ditto.
10487 (vsm3partw1q_u32): Ditto.
10488 (vsm3partw2q_u32): Ditto.
10489 (vsm4eq_u32): Ditto.
10490 (vsm4ekeyq_u32): Ditto.
10491 (doc/invoke.texi): Document new sm4 option.
10493 2018-01-10 Michael Collison <michael.collison@arm.com>
10495 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10496 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10497 (AARCH64_FL_FOR_ARCH8_4): New.
10498 (AARCH64_FL_V8_4): New flag.
10499 (doc/invoke.texi): Document new armv8.4-a option.
10501 2018-01-10 Michael Collison <michael.collison@arm.com>
10503 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10504 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10505 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10506 * config/aarch64/aarch64-option-extension.def: Add
10507 AARCH64_OPT_EXTENSION of 'sha2'.
10508 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10509 (crypto): Disable sha2 and aes if crypto disabled.
10510 (crypto): Enable aes and sha2 if enabled.
10511 (simd): Disable sha2 and aes if simd disabled.
10512 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10514 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10515 (TARGET_SHA2): New feature flag for sha2.
10516 (TARGET_AES): New feature flag for aes.
10517 * config/aarch64/aarch64-simd.md:
10518 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10519 conditional on TARGET_AES.
10520 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10521 (aarch64_crypto_sha1hsi): Make pattern conditional
10523 (aarch64_crypto_sha1hv4si): Ditto.
10524 (aarch64_be_crypto_sha1hv4si): Ditto.
10525 (aarch64_crypto_sha1su1v4si): Ditto.
10526 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10527 (aarch64_crypto_sha1su0v4si): Ditto.
10528 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10529 (aarch64_crypto_sha256su0v4si): Ditto.
10530 (aarch64_crypto_sha256su1v4si): Ditto.
10531 (doc/invoke.texi): Document new aes and sha2 options.
10533 2018-01-10 Martin Sebor <msebor@redhat.com>
10535 PR tree-optimization/83781
10536 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10539 2018-01-11 Martin Sebor <msebor@gmail.com>
10540 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10542 PR tree-optimization/83501
10543 PR tree-optimization/81703
10545 * tree-ssa-strlen.c (get_string_cst): Rename...
10546 (get_string_len): ...to this. Handle global constants.
10547 (handle_char_store): Adjust.
10549 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10550 Jim Wilson <jimw@sifive.com>
10552 * config/riscv/riscv-protos.h (riscv_output_return): New.
10553 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10554 (riscv_attribute_table, riscv_output_return),
10555 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10556 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10557 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10559 (riscv_expand_prologue): Add early return for naked function.
10560 (riscv_expand_epilogue): Likewise.
10561 (riscv_function_ok_for_sibcall): Return false for naked function.
10562 (riscv_set_current_function): New.
10563 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10564 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10565 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10566 * doc/extend.texi (RISC-V Function Attributes): New.
10568 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10570 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10571 check for 128-bit long double before checking TCmode.
10572 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10573 128-bit long doubles before checking TFmode or TCmode.
10574 (FLOAT128_IBM_P): Likewise.
10576 2018-01-10 Martin Sebor <msebor@redhat.com>
10578 PR tree-optimization/83671
10579 * builtins.c (c_strlen): Unconditionally return zero for the empty
10581 Use -Warray-bounds for warnings.
10582 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10583 for non-constant array indices with COMPONENT_REF, arrays of
10584 arrays, and pointers to arrays.
10585 (gimple_fold_builtin_strlen): Determine and set length range for
10586 non-constant character arrays.
10588 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10590 PR middle-end/81897
10591 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10594 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10596 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10598 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10601 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10602 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10603 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10604 indexed_or_indirect_operand predicate.
10605 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10606 (*vsx_le_perm_load_v8hi): Likewise.
10607 (*vsx_le_perm_load_v16qi): Likewise.
10608 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10609 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10610 (*vsx_le_perm_store_v8hi): Likewise.
10611 (*vsx_le_perm_store_v16qi): Likewise.
10612 (eight unnamed splitters): Likewise.
10614 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10616 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10617 * config/rs6000/emmintrin.h: Likewise.
10618 * config/rs6000/mmintrin.h: Likewise.
10619 * config/rs6000/xmmintrin.h: Likewise.
10621 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10624 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10626 * tree.c (tree_nop_conversion): Return true for location wrapper
10628 (maybe_wrap_with_location): New function.
10629 (selftest::check_strip_nops): New function.
10630 (selftest::test_location_wrappers): New function.
10631 (selftest::tree_c_tests): Call it.
10632 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10633 (maybe_wrap_with_location): New decl.
10634 (EXPR_LOCATION_WRAPPER_P): New macro.
10635 (location_wrapper_p): New inline function.
10636 (tree_strip_any_location_wrapper): New inline function.
10638 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10641 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10642 stack_realign_offset for the largest alignment of stack slot
10644 (ix86_find_max_used_stack_alignment): New function.
10645 (ix86_finalize_stack_frame_flags): Use it. Set
10646 max_used_stack_alignment if we don't realign stack.
10647 * config/i386/i386.h (machine_function): Add
10648 max_used_stack_alignment.
10650 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10652 * config/arm/arm.opt (-mbranch-cost): New option.
10653 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10656 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10659 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10660 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10662 2018-01-10 Richard Biener <rguenther@suse.de>
10665 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10666 early out so it also covers the case where we have a non-NULL
10669 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10671 PR tree-optimization/83753
10672 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10673 for non-strided grouped accesses if the number of elements is 1.
10675 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10678 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10679 * i386.h (TARGET_USE_GATHER): Define.
10680 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10682 2018-01-10 Martin Liska <mliska@suse.cz>
10685 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10686 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10688 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10689 CLEANUP_NO_PARTITIONING is not set.
10691 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10693 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10694 for vectors, as a partial revert of r254296.
10695 * rtl.h (const_vec_p): Delete.
10696 (const_vec_duplicate_p): Don't test for vector CONSTs.
10697 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10698 * expmed.c (make_tree): Likewise.
10701 * common.md (E, F): Use CONSTANT_P instead of checking for
10703 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10704 checking for CONST_VECTOR.
10706 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10708 PR middle-end/83575
10709 * predict.c (force_edge_cold): Handle in more sane way edges
10710 with no prediction.
10712 2018-01-09 Carl Love <cel@us.ibm.com>
10714 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10716 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10717 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10718 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10719 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10720 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10721 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10722 * config/rs6000/rs6000-protos.h: Add extern defition for
10723 rs6000_generate_float2_double_code.
10724 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10726 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10727 (float2_v2df): Add define_expand.
10729 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10732 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10733 op_mode in the force_to_mode call.
10735 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10737 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10738 instead of checking each element individually.
10739 (aarch64_evpc_uzp): Likewise.
10740 (aarch64_evpc_zip): Likewise.
10741 (aarch64_evpc_ext): Likewise.
10742 (aarch64_evpc_rev): Likewise.
10743 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10744 instead of checking each element individually. Return true without
10746 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10747 whether all selected elements come from the same input, instead of
10748 checking each element individually. Remove calls to gen_rtx_REG,
10749 start_sequence and end_sequence and instead assert that no rtl is
10752 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10754 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10755 order of HIGH and CONST checks.
10757 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10759 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10760 if the destination isn't an SSA_NAME.
10762 2018-01-09 Richard Biener <rguenther@suse.de>
10764 PR tree-optimization/83668
10765 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10767 (canonicalize_loop_form): ... here, renamed from ...
10768 (canonicalize_loop_closed_ssa_form): ... this and amended to
10769 swap successor edges for loop exit blocks to make us use
10770 the RPO order we need for initial schedule generation.
10772 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10774 PR tree-optimization/64811
10775 * match.pd: When optimizing comparisons with Inf, avoid
10776 introducing or losing exceptions from comparisons with NaN.
10778 2018-01-09 Martin Liska <mliska@suse.cz>
10781 * asan.c (shadow_mem_size): Add gcc_assert.
10783 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10785 Don't save registers in main().
10788 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10789 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10790 * config/avr/avr.c (avr_set_current_function): Don't error if
10791 naked, OS_task or OS_main are specified at the same time.
10792 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10794 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10796 * common/config/avr/avr-common.c (avr_option_optimization_table):
10797 Switch on -mmain-is-OS_task for optimizing compilations.
10799 2018-01-09 Richard Biener <rguenther@suse.de>
10801 PR tree-optimization/83572
10802 * graphite.c: Include cfganal.h.
10803 (graphite_transform_loops): Connect infinite loops to exit
10804 and remove fake edges at the end.
10806 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10808 * ipa-inline.c (edge_badness): Revert accidental checkin.
10810 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10813 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10814 symbols; not inline clones.
10816 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10819 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10820 hard registers. Formatting fixes.
10822 PR preprocessor/83722
10823 * gcc.c (try_generate_repro): Pass
10824 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10825 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10828 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10829 Kito Cheng <kito.cheng@gmail.com>
10831 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10832 (riscv_leaf_function_p): Delete.
10833 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10835 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10837 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10839 (do_ifelse): New function.
10840 (do_isel): New function.
10841 (do_sub3): New function.
10842 (do_add3): New function.
10843 (do_load_mask_compare): New function.
10844 (do_overlap_load_compare): New function.
10845 (expand_compare_loop): New function.
10846 (expand_block_compare): Call expand_compare_loop() when appropriate.
10847 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10848 option description.
10849 (-mblock-compare-inline-loop-limit): New option.
10851 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10854 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10855 Reverse order of second and third operands in first alternative.
10856 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10857 of first and second elements in UNSPEC_VPERMR vector.
10858 (altivec_expand_vec_perm_le): Likewise.
10860 2018-01-08 Jeff Law <law@redhat.com>
10862 PR rtl-optimizatin/81308
10863 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10864 (process_switch): If group_case_labels makes a change, then set
10866 (pass_convert_switch::execute): If a switch is converted, then
10867 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10869 PR rtl-optimization/81308
10870 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10873 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10875 PR target/83663 - Revert r255946
10876 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10877 generation for cases where splatting a value is not useful.
10878 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10879 across a vec_duplicate and a paradoxical subreg forming a vector
10880 mode to a vec_concat.
10882 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10884 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10885 -march=armv8.3-a variants.
10886 * config/arm/t-multilib: Likewise.
10887 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10889 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10891 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10893 (cceq_ior_compare_complement): Give it a name so I can use it, and
10894 change boolean_or_operator predicate to boolean_operator so it can
10895 be used to generate a crand.
10896 (eqne): New code iterator.
10897 (bd/bd_neg): New code_attrs.
10898 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10899 a single define_insn.
10900 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10901 decrement (bdnzt/bdnzf/bdzt/bdzf).
10902 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10903 with the new names of the branch decrement patterns, and added the
10904 names of the branch decrement conditional patterns.
10906 2018-01-08 Richard Biener <rguenther@suse.de>
10908 PR tree-optimization/83563
10909 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10912 2018-01-08 Richard Biener <rguenther@suse.de>
10914 PR middle-end/83713
10915 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10917 2018-01-08 Richard Biener <rguenther@suse.de>
10919 PR tree-optimization/83685
10920 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10921 references to abnormals.
10923 2018-01-08 Richard Biener <rguenther@suse.de>
10926 * dwarf2out.c (output_indirect_strings): Handle empty
10927 skeleton_debug_str_hash.
10928 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10930 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10932 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10933 (emit_store_direct): Likewise.
10934 (arc_trampoline_adjust_address): Likewise.
10935 (arc_asm_trampoline_template): New function.
10936 (arc_initialize_trampoline): Use asm_trampoline_template.
10937 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10938 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10939 * config/arc/arc.md (flush_icache): Delete pattern.
10941 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10943 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10944 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10947 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10950 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10951 by not USED_FOR_TARGET.
10952 (make_pass_resolve_sw_modes): Likewise.
10954 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10956 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10959 2018-01-08 Richard Biener <rguenther@suse.de>
10961 PR middle-end/83580
10962 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10964 2018-01-08 Richard Biener <rguenther@suse.de>
10966 PR middle-end/83517
10967 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10969 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10971 PR middle-end/81897
10972 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10973 basic blocks with a small number of successors.
10974 (convert_control_dep_chain_into_preds): Improve handling of
10976 (dump_predicates): Split apart into...
10977 (dump_pred_chain): ...here...
10978 (dump_pred_info): ...and here.
10979 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10980 (can_chain_union_be_invalidated_p): Improve check for invalidation
10982 (uninit_uses_cannot_happen): Avoid unnecessary if
10983 convert_control_dep_chain_into_preds yielded nothing.
10985 2018-01-06 Martin Sebor <msebor@redhat.com>
10987 PR tree-optimization/83640
10988 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10989 subtracting negative offset from size.
10990 (builtin_access::overlap): Adjust offset bounds of the access to fall
10991 within the size of the object if possible.
10993 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10995 PR rtl-optimization/83699
10996 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10997 extract_bit_field_as_subreg to cases in which the extracted
10998 value is also a vector.
11000 * lra-constraints.c (process_alt_operands): Test for the equivalence
11001 substitutions when detecting a possible reload cycle.
11003 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11006 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11007 by default if flag_selective_schedling{,2}. Formatting fixes.
11009 PR rtl-optimization/83682
11010 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11011 if it has non-VECTOR_MODE element mode.
11012 (vec_duplicate_p): Likewise.
11014 PR middle-end/83694
11015 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11016 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11018 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11021 * config/i386/i386-builtin.def
11022 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11023 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11024 Require also OPTION_MASK_ISA_AVX512F in addition to
11025 OPTION_MASK_ISA_GFNI.
11026 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11027 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11028 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11029 to OPTION_MASK_ISA_GFNI.
11030 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11031 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11032 OPTION_MASK_ISA_AVX512BW.
11033 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11034 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11035 addition to OPTION_MASK_ISA_GFNI.
11036 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11037 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11038 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11039 to OPTION_MASK_ISA_GFNI.
11040 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11041 a requirement for all ISAs rather than any of them with a few
11043 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11045 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11046 bitmasks to be enabled with 3 exceptions, instead of requiring any
11047 enabled ISA with lots of exceptions.
11048 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11049 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11050 Change avx512bw in isa attribute to avx512f.
11051 * config/i386/sgxintrin.h: Add license boilerplate.
11052 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11053 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11054 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11055 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11057 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11058 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11059 temporarily sse2 rather than sse if not enabled already.
11062 * config/i386/sse.md (VI248_VLBW): Rename to ...
11063 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11064 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11065 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11066 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11067 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11068 mode iterator instead of VI248_VLBW.
11070 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11072 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11073 (record_modified): Skip clobbers; add debug output.
11074 (param_change_prob): Use sreal frequencies.
11076 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11078 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11079 punt for user-aligned variables.
11081 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11083 * tree-chrec.c (chrec_contains_symbols): Return true for
11086 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11089 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11090 of (x|y) == x for BICS pattern.
11092 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11094 PR tree-optimization/83605
11095 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11096 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11099 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11101 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11102 * config/epiphany/rtems.h: New file.
11104 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11105 Uros Bizjak <ubizjak@gmail.com>
11108 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11109 QIreg_operand instead of register_operand predicate.
11110 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11111 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11112 comments instead of -fmitigate[-_]rop.
11114 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11117 * cgraphunit.c (symbol_table::compile): Switch to text_section
11118 before calling assembly_start debug hook.
11119 * run-rtl-passes.c (run_rtl_passes): Likewise.
11122 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11124 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11125 range_int_cst_p rather than !symbolic_range_p before calling
11126 extract_range_from_multiplicative_op_1.
11128 2018-01-04 Jeff Law <law@redhat.com>
11130 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11131 redundant test in assertion.
11133 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11135 * doc/rtl.texi: Document machine_mode wrapper classes.
11137 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11139 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11140 using tree_to_uhwi.
11142 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11144 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11145 the VEC_PERM_EXPR fold to fail.
11147 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11150 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11151 to switched_sections.
11153 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11156 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11157 test for d.testing.
11159 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11162 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11163 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11165 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11168 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11169 is BLKmode and bitpos not zero or mode change is needed.
11171 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11174 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11177 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11180 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11181 instead of MULT rtx. Update all corresponding splitters.
11182 (*saddl_se): Ditto.
11183 (*ssub<modesuffix>): Ditto.
11184 (*ssubl_se): Ditto.
11185 (*cmp_sadd_di): Update split patterns.
11186 (*cmp_sadd_si): Ditto.
11187 (*cmp_sadd_sidi): Ditto.
11188 (*cmp_ssub_di): Ditto.
11189 (*cmp_ssub_si): Ditto.
11190 (*cmp_ssub_sidi): Ditto.
11191 * config/alpha/predicates.md (const23_operand): New predicate.
11192 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11193 Look for ASHIFT, not MULT inner operand.
11194 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11196 2018-01-04 Martin Liska <mliska@suse.cz>
11198 PR gcov-profile/83669
11199 * gcov.c (output_intermediate_file): Add version to intermediate
11201 * doc/gcov.texi: Document new field 'version' in intermediate
11202 file format. Fix location of '-k' option of gcov command.
11204 2018-01-04 Martin Liska <mliska@suse.cz>
11207 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11209 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11211 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11213 2018-01-03 Martin Sebor <msebor@redhat.com>
11215 PR tree-optimization/83655
11216 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11217 checking calls with invalid arguments.
11219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11221 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11222 (vectorizable_mask_load_store): Delete.
11223 (vectorizable_call): Return false for masked loads and stores.
11224 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11225 instead of gimple_assign_rhs1.
11226 (vectorizable_load): Handle IFN_MASK_LOAD.
11227 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11231 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11233 (vectorizable_mask_load_store): ...here.
11234 (vectorizable_load): ...and here.
11236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11238 * tree-vect-stmts.c (vect_build_all_ones_mask)
11239 (vect_build_zero_merge_argument): New functions, split out from...
11240 (vectorizable_load): ...here.
11242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11244 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11246 (vectorizable_mask_load_store): ...here.
11247 (vectorizable_store): ...and here.
11249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11251 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11253 (vectorizable_mask_load_store): ...here.
11255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11257 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11258 (vect_model_store_cost): Take a vec_load_store_type instead of a
11260 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11261 (vect_model_store_cost): Take a vec_load_store_type instead of a
11263 (vectorizable_mask_load_store): Update accordingly.
11264 (vectorizable_store): Likewise.
11265 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11269 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11270 IFN_MASK_LOAD calls here rather than...
11271 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11274 Alan Hayward <alan.hayward@arm.com>
11275 David Sherwood <david.sherwood@arm.com>
11277 * expmed.c (extract_bit_field_1): For vector extracts,
11278 fall back to extract_bit_field_as_subreg if vec_extract
11281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11282 Alan Hayward <alan.hayward@arm.com>
11283 David Sherwood <david.sherwood@arm.com>
11285 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11286 they are variable or constant sized.
11287 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11288 slots for constant-sized data.
11290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11291 Alan Hayward <alan.hayward@arm.com>
11292 David Sherwood <david.sherwood@arm.com>
11294 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11295 handling COND_EXPRs with boolean comparisons, try to find a better
11296 basis for the mask type than the boolean itself.
11298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11300 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11301 is calculated and how it can be overridden.
11302 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11303 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11305 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11309 Alan Hayward <alan.hayward@arm.com>
11310 David Sherwood <david.sherwood@arm.com>
11312 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11313 Remove the mode argument.
11314 (aarch64_simd_valid_immediate): Remove the mode and inverse
11316 * config/aarch64/iterators.md (bitsize): New iterator.
11317 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11318 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11319 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11320 aarch64_simd_valid_immediate.
11321 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11322 (aarch64_reg_or_bic_imm): Likewise.
11323 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11324 with an insn_type enum and msl with a modifier_type enum.
11325 Replace element_width with a scalar_mode. Change the shift
11326 to unsigned int. Add constructors for scalar_float_mode and
11327 scalar_int_mode elements.
11328 (aarch64_vect_float_const_representable_p): Delete.
11329 (aarch64_can_const_movi_rtx_p)
11330 (aarch64_simd_scalar_immediate_valid_for_move)
11331 (aarch64_simd_make_constant): Update call to
11332 aarch64_simd_valid_immediate.
11333 (aarch64_advsimd_valid_immediate_hs): New function.
11334 (aarch64_advsimd_valid_immediate): Likewise.
11335 (aarch64_simd_valid_immediate): Remove mode and inverse
11336 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11337 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11338 and aarch64_float_const_representable_p on the result.
11339 (aarch64_output_simd_mov_immediate): Remove mode argument.
11340 Update call to aarch64_simd_valid_immediate and use of
11341 simd_immediate_info.
11342 (aarch64_output_scalar_simd_mov_immediate): Update call
11345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11346 Alan Hayward <alan.hayward@arm.com>
11347 David Sherwood <david.sherwood@arm.com>
11349 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11350 (mode_nunits): Likewise CONST_MODE_NUNITS.
11351 * machmode.def (ADJUST_NUNITS): Document.
11352 * genmodes.c (mode_data::need_nunits_adj): New field.
11353 (blank_mode): Update accordingly.
11354 (adj_nunits): New variable.
11355 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11357 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11358 listed in adj_nunits.
11359 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11360 listed in adj_nunits. Don't emit case statements for such modes.
11361 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11362 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11363 nothing if adj_nunits is nonnull.
11364 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11365 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11366 (emit_mode_fbit): Update use of print_maybe_const_decl.
11367 (emit_move_size): Likewise. Treat the array as non-const
11369 (emit_mode_adjustments): Handle adj_nunits.
11371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11373 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11374 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11375 (VECTOR_MODES): Use it.
11376 (make_vector_modes): Take the prefix as an argument.
11378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11379 Alan Hayward <alan.hayward@arm.com>
11380 David Sherwood <david.sherwood@arm.com>
11382 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11383 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11384 for MODE_VECTOR_BOOL.
11385 * machmode.def (VECTOR_BOOL_MODE): Document.
11386 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11387 (make_vector_bool_mode): New function.
11388 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11390 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11391 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11393 * stor-layout.c (int_mode_for_mode): Likewise.
11394 * tree.c (build_vector_type_for_mode): Likewise.
11395 * varasm.c (output_constant_pool_2): Likewise.
11396 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11397 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11398 for MODE_VECTOR_BOOL.
11399 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11400 of mode class checks.
11401 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11402 instead of a list of mode class checks.
11403 (expand_vector_scalar_condition): Likewise.
11404 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11407 Alan Hayward <alan.hayward@arm.com>
11408 David Sherwood <david.sherwood@arm.com>
11410 * machmode.h (mode_size): Change from unsigned short to
11412 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11413 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11414 or if measurement_type is not polynomial.
11415 (fixed_size_mode::includes_p): Check for constant-sized modes.
11416 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11417 return a poly_uint16 rather than an unsigned short.
11418 (emit_mode_size): Change the type of mode_size from unsigned short
11419 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11420 (emit_mode_adjustments): Cope with polynomial vector sizes.
11421 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11423 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11425 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11426 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11427 * caller-save.c (setup_save_areas): Likewise.
11428 (replace_reg_with_saved_mem): Likewise.
11429 * calls.c (emit_library_call_value_1): Likewise.
11430 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11431 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11432 (gen_lowpart_for_combine): Likewise.
11433 * convert.c (convert_to_integer_1): Likewise.
11434 * cse.c (equiv_constant, cse_insn): Likewise.
11435 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11436 (cselib_subst_to_values): Likewise.
11437 * dce.c (word_dce_process_block): Likewise.
11438 * df-problems.c (df_word_lr_mark_ref): Likewise.
11439 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11440 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11441 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11442 (rtl_for_decl_location): Likewise.
11443 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11444 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11445 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11446 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11447 (expand_expr_real_1): Likewise.
11448 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11449 (pad_below): Likewise.
11450 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11451 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11452 * ira.c (get_subreg_tracking_sizes): Likewise.
11453 * ira-build.c (ira_create_allocno_objects): Likewise.
11454 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11455 (ira_sort_regnos_for_alter_reg): Likewise.
11456 * ira-costs.c (record_operand_costs): Likewise.
11457 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11458 (resolve_simple_move): Likewise.
11459 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11460 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11461 (lra_constraints): Likewise.
11462 (CONST_POOL_OK_P): Reject variable-sized modes.
11463 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11464 (add_pseudo_to_slot, lra_spill): Likewise.
11465 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11466 * optabs-query.c (get_best_extraction_insn): Likewise.
11467 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11468 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11469 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11470 * recog.c (offsettable_address_addr_space_p): Likewise.
11471 * regcprop.c (maybe_mode_change): Likewise.
11472 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11473 * regrename.c (build_def_use): Likewise.
11474 * regstat.c (dump_reg_info): Likewise.
11475 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11476 (find_reloads, find_reloads_subreg_address): Likewise.
11477 * reload1.c (eliminate_regs_1): Likewise.
11478 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11479 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11480 (simplify_binary_operation_1, simplify_subreg): Likewise.
11481 * targhooks.c (default_function_arg_padding): Likewise.
11482 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11483 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11484 (verify_gimple_assign_ternary): Likewise.
11485 * tree-inline.c (estimate_move_cost): Likewise.
11486 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11487 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11488 (get_address_cost_ainc): Likewise.
11489 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11490 (vect_supportable_dr_alignment): Likewise.
11491 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11492 (vectorizable_reduction): Likewise.
11493 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11494 (vectorizable_operation, vectorizable_load): Likewise.
11495 * tree.c (build_same_sized_truth_vector_type): Likewise.
11496 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11497 * var-tracking.c (emit_note_insn_var_location): Likewise.
11498 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11499 (ADDR_VEC_ALIGN): Likewise.
11501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11502 Alan Hayward <alan.hayward@arm.com>
11503 David Sherwood <david.sherwood@arm.com>
11505 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11507 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11508 or if measurement_type is polynomial.
11509 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11510 * combine.c (make_extraction): Likewise.
11511 * dse.c (find_shift_sequence): Likewise.
11512 * dwarf2out.c (mem_loc_descriptor): Likewise.
11513 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11514 (extract_bit_field, extract_low_bits): Likewise.
11515 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11516 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11517 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11518 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11519 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11520 * reload.c (find_reloads): Likewise.
11521 * reload1.c (alter_reg): Likewise.
11522 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11523 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11524 * tree-if-conv.c (predicate_mem_writes): Likewise.
11525 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11526 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11527 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11528 * valtrack.c (dead_debug_insert_temp): Likewise.
11529 * varasm.c (mergeable_constant_section): Likewise.
11530 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11533 Alan Hayward <alan.hayward@arm.com>
11534 David Sherwood <david.sherwood@arm.com>
11536 * expr.c (expand_assignment): Cope with polynomial mode sizes
11537 when assigning to a CONCAT.
11539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11540 Alan Hayward <alan.hayward@arm.com>
11541 David Sherwood <david.sherwood@arm.com>
11543 * machmode.h (mode_precision): Change from unsigned short to
11545 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11547 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11548 or if measurement_type is not polynomial.
11549 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11550 in which the mode is already known to be a scalar_int_mode.
11551 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11552 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11554 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11555 for GET_MODE_PRECISION.
11556 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11557 for GET_MODE_PRECISION.
11558 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11560 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11561 (expand_field_assignment, make_extraction): Likewise.
11562 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11563 (get_last_value): Likewise.
11564 * convert.c (convert_to_integer_1): Likewise.
11565 * cse.c (cse_insn): Likewise.
11566 * expr.c (expand_expr_real_1): Likewise.
11567 * lra-constraints.c (simplify_operand_subreg): Likewise.
11568 * optabs-query.c (can_atomic_load_p): Likewise.
11569 * optabs.c (expand_atomic_load): Likewise.
11570 (expand_atomic_store): Likewise.
11571 * ree.c (combine_reaching_defs): Likewise.
11572 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11573 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11574 * tree.h (type_has_mode_precision_p): Likewise.
11575 * ubsan.c (instrument_si_overflow): Likewise.
11577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11578 Alan Hayward <alan.hayward@arm.com>
11579 David Sherwood <david.sherwood@arm.com>
11581 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11582 polynomial numbers of units.
11583 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11584 (valid_vector_subparts_p): New function.
11585 (build_vector_type): Remove temporary shim and take the number
11586 of units as a poly_uint64 rather than an int.
11587 (build_opaque_vector_type): Take the number of units as a
11588 poly_uint64 rather than an int.
11589 * tree.c (build_vector_from_ctor): Handle polynomial
11590 TYPE_VECTOR_SUBPARTS.
11591 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11592 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11593 (build_vector_from_val): If the number of units is variable,
11594 use build_vec_duplicate_cst for constant operands and
11595 VEC_DUPLICATE_EXPR otherwise.
11596 (make_vector_type): Remove temporary is_constant ().
11597 (build_vector_type, build_opaque_vector_type): Take the number of
11598 units as a poly_uint64 rather than an int.
11599 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11601 * cfgexpand.c (expand_debug_expr): Likewise.
11602 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11603 (store_constructor, expand_expr_real_1): Likewise.
11604 (const_scalar_mask_from_tree): Likewise.
11605 * fold-const-call.c (fold_const_reduction): Likewise.
11606 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11607 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11608 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11609 (fold_relational_const): Likewise.
11610 (native_interpret_vector): Likewise. Change the size from an
11611 int to an unsigned int.
11612 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11613 TYPE_VECTOR_SUBPARTS.
11614 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11615 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11616 duplicating a non-constant operand into a variable-length vector.
11617 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11618 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11619 * ipa-icf.c (sem_variable::equals): Likewise.
11620 * match.pd: Likewise.
11621 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11622 * print-tree.c (print_node): Likewise.
11623 * stor-layout.c (layout_type): Likewise.
11624 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11625 * tree-cfg.c (verify_gimple_comparison): Likewise.
11626 (verify_gimple_assign_binary): Likewise.
11627 (verify_gimple_assign_ternary): Likewise.
11628 (verify_gimple_assign_single): Likewise.
11629 * tree-pretty-print.c (dump_generic_node): Likewise.
11630 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11631 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11632 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11633 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11634 (vect_shift_permute_load_chain): Likewise.
11635 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11636 (expand_vector_condition, optimize_vector_constructor): Likewise.
11637 (lower_vec_perm, get_compute_type): Likewise.
11638 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11639 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11640 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11641 (vect_recog_mask_conversion_pattern): Likewise.
11642 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11643 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11644 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11645 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11646 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11647 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11648 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11649 (supportable_widening_operation): Likewise.
11650 (supportable_narrowing_operation): Likewise.
11651 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11653 * varasm.c (output_constant): Likewise.
11655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11656 Alan Hayward <alan.hayward@arm.com>
11657 David Sherwood <david.sherwood@arm.com>
11659 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11660 so that both the length == 3 and length != 3 cases set up their
11661 own permute vectors. Add comments explaining why we know the
11662 number of elements is constant.
11663 (vect_permute_load_chain): Likewise.
11665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11666 Alan Hayward <alan.hayward@arm.com>
11667 David Sherwood <david.sherwood@arm.com>
11669 * machmode.h (mode_nunits): Change from unsigned char to
11671 (ONLY_FIXED_SIZE_MODES): New macro.
11672 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11673 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11674 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11676 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11677 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11678 or if measurement_type is not polynomial.
11679 * genmodes.c (ZERO_COEFFS): New macro.
11680 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11682 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11683 Use ZERO_COEFFS when emitting initializers.
11684 * data-streamer.h (bp_pack_poly_value): New function.
11685 (bp_unpack_poly_value): Likewise.
11686 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11687 for GET_MODE_NUNITS.
11688 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11689 for GET_MODE_NUNITS.
11690 * tree.c (make_vector_type): Remove temporary shim and make
11691 the real function take the number of units as a poly_uint64
11692 rather than an int.
11693 (build_vector_type_for_mode): Handle polynomial nunits.
11694 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11695 * emit-rtl.c (const_vec_series_p_1): Likewise.
11696 (gen_rtx_CONST_VECTOR): Likewise.
11697 * fold-const.c (test_vec_duplicate_folding): Likewise.
11698 * genrecog.c (validate_pattern): Likewise.
11699 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11700 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11701 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11702 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11703 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11704 * rtlanal.c (subreg_get_info): Likewise.
11705 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11706 (vect_grouped_load_supported): Likewise.
11707 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11708 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11709 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11710 (simplify_const_unary_operation, simplify_binary_operation_1)
11711 (simplify_const_binary_operation, simplify_ternary_operation)
11712 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11713 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11714 instead of CONST_VECTOR_NUNITS.
11715 * varasm.c (output_constant_pool_2): Likewise.
11716 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11717 explicit-encoded elements in the XVEC for variable-length vectors.
11719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11721 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11724 Alan Hayward <alan.hayward@arm.com>
11725 David Sherwood <david.sherwood@arm.com>
11727 * coretypes.h (fixed_size_mode): Declare.
11728 (fixed_size_mode_pod): New typedef.
11729 * builtins.h (target_builtins::x_apply_args_mode)
11730 (target_builtins::x_apply_result_mode): Change type to
11731 fixed_size_mode_pod.
11732 * builtins.c (apply_args_size, apply_result_size, result_vector)
11733 (expand_builtin_apply_args_1, expand_builtin_apply)
11734 (expand_builtin_return): Update accordingly.
11736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11738 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11739 * cselib.c (cselib_hash_rtx): Likewise.
11740 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11741 CONST_VECTOR encoding.
11743 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11744 Jeff Law <law@redhat.com>
11747 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11748 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11749 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11750 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11753 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11754 explicitly probe *sp in a noreturn function if there were any callee
11755 register saves or frame pointer is needed.
11757 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11760 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11761 BLKmode for ternary, binary or unary expressions.
11764 * var-tracking.c (delete_vta_debug_insn): New inline function.
11765 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11766 insns from get_insns () to NULL instead of each bb separately.
11767 Use delete_vta_debug_insn. No longer static.
11768 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11769 delete_vta_debug_insns callers.
11770 * rtl.h (delete_vta_debug_insns): Declare.
11771 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11772 instead of variable_tracking_main.
11774 2018-01-03 Martin Sebor <msebor@redhat.com>
11776 PR tree-optimization/83603
11777 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11778 arguments past the endof the argument list in functions declared
11779 without a prototype.
11780 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11781 Avoid checking when arguments are null.
11783 2018-01-03 Martin Sebor <msebor@redhat.com>
11786 * doc/extend.texi (attribute const): Fix a typo.
11787 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11788 issuing -Wsuggest-attribute for void functions.
11790 2018-01-03 Martin Sebor <msebor@redhat.com>
11792 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11793 offset_int::from instead of wide_int::to_shwi.
11794 (maybe_diag_overlap): Remove assertion.
11795 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11796 * gimple-ssa-sprintf.c (format_directive): Same.
11797 (parse_directive): Same.
11798 (sprintf_dom_walker::compute_format_length): Same.
11799 (try_substitute_return_value): Same.
11801 2018-01-03 Jeff Law <law@redhat.com>
11803 PR middle-end/83654
11804 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11805 non-constant residual for zero at runtime and avoid probing in
11806 that case. Reorganize code for trailing problem to mirror handling
11809 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11811 PR tree-optimization/83501
11812 * tree-ssa-strlen.c (get_string_cst): New.
11813 (handle_char_store): Call get_string_cst.
11815 2018-01-03 Martin Liska <mliska@suse.cz>
11817 PR tree-optimization/83593
11818 * tree-ssa-strlen.c: Include tree-cfg.h.
11819 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11820 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11821 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11823 (strlen_dom_walker::before_dom_children): Call
11824 gimple_purge_dead_eh_edges. Dump tranformation with details
11826 (strlen_dom_walker::before_dom_children): Update call by adding
11827 new argument cleanup_eh.
11828 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11830 2018-01-03 Martin Liska <mliska@suse.cz>
11833 * cif-code.def (VARIADIC_THUNK): New enum value.
11834 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11837 2018-01-03 Jan Beulich <jbeulich@suse.com>
11839 * sse.md (mov<mode>_internal): Tighten condition for when to use
11840 vmovdqu<ssescalarsize> for TI and OI modes.
11842 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11844 Update copyright years.
11846 2018-01-03 Martin Liska <mliska@suse.cz>
11849 * ipa-visibility.c (function_and_variable_visibility): Skip
11850 functions with noipa attribure.
11852 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11854 * gcc.c (process_command): Update copyright notice dates.
11855 * gcov-dump.c (print_version): Ditto.
11856 * gcov.c (print_version): Ditto.
11857 * gcov-tool.c (print_version): Ditto.
11858 * gengtype.c (create_file): Ditto.
11859 * doc/cpp.texi: Bump @copying's copyright year.
11860 * doc/cppinternals.texi: Ditto.
11861 * doc/gcc.texi: Ditto.
11862 * doc/gccint.texi: Ditto.
11863 * doc/gcov.texi: Ditto.
11864 * doc/install.texi: Ditto.
11865 * doc/invoke.texi: Ditto.
11867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11869 * vector-builder.h (vector_builder::m_full_nelts): Change from
11870 unsigned int to poly_uint64.
11871 (vector_builder::full_nelts): Update prototype accordingly.
11872 (vector_builder::new_vector): Likewise.
11873 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11874 (vector_builder::operator ==): Likewise.
11875 (vector_builder::finalize): Likewise.
11876 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11877 Take the number of elements as a poly_uint64 rather than an
11879 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11880 from unsigned int to poly_uint64.
11881 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11882 (vec_perm_indices::new_vector): Likewise.
11883 (vec_perm_indices::length): Likewise.
11884 (vec_perm_indices::nelts_per_input): Likewise.
11885 (vec_perm_indices::input_nelts): Likewise.
11886 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11887 number of elements per input as a poly_uint64 rather than an
11888 unsigned int. Use the original encoding for variable-length
11889 vectors, rather than clamping each individual element.
11890 For the second and subsequent elements in each pattern,
11891 clamp the step and base before clamping their sum.
11892 (vec_perm_indices::series_p): Handle polynomial element counts.
11893 (vec_perm_indices::all_in_range_p): Likewise.
11894 (vec_perm_indices_to_tree): Likewise.
11895 (vec_perm_indices_to_rtx): Likewise.
11896 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11897 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11898 (tree_vector_builder::new_binary_operation): Handle polynomial
11899 element counts. Return false if we need to know the number
11900 of elements at compile time.
11901 * fold-const.c (fold_vec_perm): Punt if the number of elements
11902 isn't known at compile time.
11904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11906 * vec-perm-indices.h (vec_perm_builder): Change element type
11907 from HOST_WIDE_INT to poly_int64.
11908 (vec_perm_indices::element_type): Update accordingly.
11909 (vec_perm_indices::clamp): Handle polynomial element_types.
11910 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11911 (vec_perm_indices::all_in_range_p): Likewise.
11912 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11914 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11915 polynomial vec_perm_indices element types.
11916 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11917 * fold-const.c (fold_vec_perm): Likewise.
11918 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11919 * tree-vect-generic.c (lower_vec_perm): Likewise.
11920 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11921 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11922 element type to HOST_WIDE_INT.
11924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11925 Alan Hayward <alan.hayward@arm.com>
11926 David Sherwood <david.sherwood@arm.com>
11928 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11929 rather than an int. Use plus_constant.
11930 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11931 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11934 Alan Hayward <alan.hayward@arm.com>
11935 David Sherwood <david.sherwood@arm.com>
11937 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11938 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
11945 mode sizes. Require a constant size for BLKmode parameters
11946 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11947 forces a parameter to be padded at the lsb end in order to
11948 fill a complete number of words, require the parameter size
11949 to be ordered wrt UNITS_PER_WORD.
11951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11952 Alan Hayward <alan.hayward@arm.com>
11953 David Sherwood <david.sherwood@arm.com>
11955 * reload1.c (spill_stack_slot_width): Change element type
11956 from unsigned int to poly_uint64_pod.
11957 (alter_reg): Treat mode sizes as polynomial.
11959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11960 Alan Hayward <alan.hayward@arm.com>
11961 David Sherwood <david.sherwood@arm.com>
11963 * reload.c (complex_word_subreg_p): New function.
11964 (reload_inner_reg_of_subreg, push_reload): Use it.
11966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11967 Alan Hayward <alan.hayward@arm.com>
11968 David Sherwood <david.sherwood@arm.com>
11970 * lra-constraints.c (process_alt_operands): Reject matched
11971 operands whose sizes aren't ordered.
11972 (match_reload): Refer to this check here.
11974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11975 Alan Hayward <alan.hayward@arm.com>
11976 David Sherwood <david.sherwood@arm.com>
11978 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11979 that the mode size is in the set {1, 2, 4, 8, 16}.
11981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11982 Alan Hayward <alan.hayward@arm.com>
11983 David Sherwood <david.sherwood@arm.com>
11985 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11986 Use plus_constant instead of gen_rtx_PLUS.
11988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11989 Alan Hayward <alan.hayward@arm.com>
11990 David Sherwood <david.sherwood@arm.com>
11992 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11993 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11994 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11995 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11996 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11997 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11998 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11999 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12000 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12001 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12003 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12004 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12005 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12006 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12007 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12008 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12009 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12010 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12011 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12012 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12014 * expr.c (emit_move_resolve_push): Treat the input and result
12015 of PUSH_ROUNDING as a poly_int64.
12016 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12017 (emit_push_insn): Likewise.
12018 * lra-eliminations.c (mark_not_eliminable): Likewise.
12019 * recog.c (push_operand): Likewise.
12020 * reload1.c (elimination_effects): Likewise.
12021 * rtlanal.c (nonzero_bits1): Likewise.
12022 * calls.c (store_one_arg): Likewise. Require the padding to be
12023 known at compile time.
12025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12026 Alan Hayward <alan.hayward@arm.com>
12027 David Sherwood <david.sherwood@arm.com>
12029 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12030 Use plus_constant instead of gen_rtx_PLUS.
12032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12033 Alan Hayward <alan.hayward@arm.com>
12034 David Sherwood <david.sherwood@arm.com>
12036 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12037 rather than an int.
12039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12040 Alan Hayward <alan.hayward@arm.com>
12041 David Sherwood <david.sherwood@arm.com>
12043 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12044 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12045 via stack temporaries. Treat the mode size as polynomial too.
12047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12048 Alan Hayward <alan.hayward@arm.com>
12049 David Sherwood <david.sherwood@arm.com>
12051 * expr.c (expand_expr_real_2): When handling conversions involving
12052 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12053 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12054 as a poly_uint64 too.
12056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12057 Alan Hayward <alan.hayward@arm.com>
12058 David Sherwood <david.sherwood@arm.com>
12060 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12063 Alan Hayward <alan.hayward@arm.com>
12064 David Sherwood <david.sherwood@arm.com>
12066 * combine.c (can_change_dest_mode): Handle polynomial
12067 REGMODE_NATURAL_SIZE.
12068 * expmed.c (store_bit_field_1): Likewise.
12069 * expr.c (store_constructor): Likewise.
12070 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12071 and polynomial REGMODE_NATURAL_SIZE.
12072 (gen_lowpart_common): Likewise.
12073 * reginfo.c (record_subregs_of_mode): Likewise.
12074 * rtlanal.c (read_modify_subreg_p): Likewise.
12076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12077 Alan Hayward <alan.hayward@arm.com>
12078 David Sherwood <david.sherwood@arm.com>
12080 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12081 numbers of elements.
12083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12084 Alan Hayward <alan.hayward@arm.com>
12085 David Sherwood <david.sherwood@arm.com>
12087 * match.pd: Cope with polynomial numbers of vector elements.
12089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12090 Alan Hayward <alan.hayward@arm.com>
12091 David Sherwood <david.sherwood@arm.com>
12093 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12094 in a POINTER_PLUS_EXPR.
12096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12097 Alan Hayward <alan.hayward@arm.com>
12098 David Sherwood <david.sherwood@arm.com>
12100 * omp-simd-clone.c (simd_clone_subparts): New function.
12101 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12102 (ipa_simd_modify_function_body): Likewise.
12104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12105 Alan Hayward <alan.hayward@arm.com>
12106 David Sherwood <david.sherwood@arm.com>
12108 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12109 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12110 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12111 (expand_vector_condition, vector_element): Likewise.
12112 (subparts_gt): New function.
12113 (get_compute_type): Use subparts_gt.
12114 (count_type_subparts): Delete.
12115 (expand_vector_operations_1): Use subparts_gt instead of
12116 count_type_subparts.
12118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12119 Alan Hayward <alan.hayward@arm.com>
12120 David Sherwood <david.sherwood@arm.com>
12122 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12123 (vect_compile_time_alias): ...this new function. Do the calculation
12124 on poly_ints rather than trees.
12125 (vect_prune_runtime_alias_test_list): Update call accordingly.
12127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12128 Alan Hayward <alan.hayward@arm.com>
12129 David Sherwood <david.sherwood@arm.com>
12131 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12133 (vect_schedule_slp_instance): Likewise.
12135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12136 Alan Hayward <alan.hayward@arm.com>
12137 David Sherwood <david.sherwood@arm.com>
12139 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12140 constant and extern definitions for variable-length vectors.
12141 (vect_get_constant_vectors): Note that the number of units
12142 is known to be constant.
12144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12145 Alan Hayward <alan.hayward@arm.com>
12146 David Sherwood <david.sherwood@arm.com>
12148 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12149 of units as polynomial. Choose between WIDE and NARROW based
12152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12153 Alan Hayward <alan.hayward@arm.com>
12154 David Sherwood <david.sherwood@arm.com>
12156 * tree-vect-stmts.c (simd_clone_subparts): New function.
12157 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12160 Alan Hayward <alan.hayward@arm.com>
12161 David Sherwood <david.sherwood@arm.com>
12163 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12164 vectors as polynomial. Use build_index_vector for
12165 IFN_GOMP_SIMD_LANE.
12167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12168 Alan Hayward <alan.hayward@arm.com>
12169 David Sherwood <david.sherwood@arm.com>
12171 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12172 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12173 for variable-length vectors.
12174 (vectorizable_mask_load_store): Treat the number of units as
12175 polynomial, asserting that it is constant if the condition has
12176 already been enforced.
12177 (vectorizable_store, vectorizable_load): Likewise.
12179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12180 Alan Hayward <alan.hayward@arm.com>
12181 David Sherwood <david.sherwood@arm.com>
12183 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12184 of units as polynomial. Punt if we can't tell at compile time
12185 which vector contains the final result.
12187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12188 Alan Hayward <alan.hayward@arm.com>
12189 David Sherwood <david.sherwood@arm.com>
12191 * tree-vect-loop.c (vectorizable_induction): Treat the number
12192 of units as polynomial. Punt on SLP inductions. Use an integer
12193 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12194 cast of such a series for variable-length floating-point
12197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12198 Alan Hayward <alan.hayward@arm.com>
12199 David Sherwood <david.sherwood@arm.com>
12201 * tree.h (build_index_vector): Declare.
12202 * tree.c (build_index_vector): New function.
12203 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12204 of units as polynomial, forcibly converting it to a constant if
12205 vectorizable_reduction has already enforced the condition.
12206 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12207 to create a {1,2,3,...} vector.
12208 (vectorizable_reduction): Treat the number of units as polynomial.
12209 Choose vectype_in based on the largest scalar element size rather
12210 than the smallest number of units. Enforce the restrictions
12213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12214 Alan Hayward <alan.hayward@arm.com>
12215 David Sherwood <david.sherwood@arm.com>
12217 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12218 number of units as polynomial.
12220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12221 Alan Hayward <alan.hayward@arm.com>
12222 David Sherwood <david.sherwood@arm.com>
12224 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12225 * target.def (autovectorize_vector_sizes): Return the vector sizes
12226 by pointer, using vector_sizes rather than a bitmask.
12227 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12228 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12229 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12231 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12232 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12233 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12234 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12235 * omp-general.c (omp_max_vf): Likewise.
12236 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12237 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12238 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12239 * tree-vect-slp.c (vect_slp_bb): Likewise.
12240 * doc/tm.texi: Regenerate.
12241 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12243 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12244 the vector size as a poly_uint64 rather than an unsigned int.
12245 (current_vector_size): Change from an unsigned int to a poly_uint64.
12246 (get_vectype_for_scalar_type): Update accordingly.
12247 * tree.h (build_truth_vector_type): Take the size and number of
12248 units as a poly_uint64 rather than an unsigned int.
12249 (build_vector_type): Add a temporary overload that takes
12250 the number of units as a poly_uint64 rather than an unsigned int.
12251 * tree.c (make_vector_type): Likewise.
12252 (build_truth_vector_type): Take the number of units as a poly_uint64
12253 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
12260 as poly_uint64s rather than unsigned ints.
12261 * targhooks.h (default_get_mask_mode): Update accordingly.
12262 * targhooks.c (default_get_mask_mode): Likewise.
12263 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12264 * doc/tm.texi: Regenerate.
12266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12267 Alan Hayward <alan.hayward@arm.com>
12268 David Sherwood <david.sherwood@arm.com>
12270 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12271 * omp-general.c (omp_max_vf): Likewise.
12272 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12273 (expand_omp_simd): Handle polynomial safelen.
12274 * omp-low.c (omplow_simd_context): Add a default constructor.
12275 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12276 (lower_rec_simd_input_clauses): Update accordingly.
12277 (lower_rec_input_clauses): 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-vectorizer.h (vect_nunits_for_cost): New function.
12284 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12285 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12286 (vect_analyze_slp_cost): Likewise.
12287 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12288 (vect_model_load_cost): Likewise.
12290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12291 Alan Hayward <alan.hayward@arm.com>
12292 David Sherwood <david.sherwood@arm.com>
12294 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12295 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12296 from an unsigned int * to a poly_uint64_pod *.
12297 (calculate_unrolling_factor): New function.
12298 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12301 Alan Hayward <alan.hayward@arm.com>
12302 David Sherwood <david.sherwood@arm.com>
12304 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12305 from an unsigned int to a poly_uint64.
12306 (_loop_vec_info::slp_unrolling_factor): Likewise.
12307 (_loop_vec_info::vectorization_factor): Change from an int
12309 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12310 (vect_get_num_vectors): New function.
12311 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12312 (vect_get_num_copies): Use vect_get_num_vectors.
12313 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12314 to an unsigned int *.
12315 (vect_analyze_data_refs): Change min_vf from an int * to a
12317 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12318 than an unsigned HOST_WIDE_INT.
12319 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12320 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12321 to an unsigned int *.
12322 (vect_analyze_data_ref_dependences): Likewise.
12323 (vect_compute_data_ref_alignment): Handle polynomial vf.
12324 (vect_enhance_data_refs_alignment): Likewise.
12325 (vect_prune_runtime_alias_test_list): Likewise.
12326 (vect_shift_permute_load_chain): Likewise.
12327 (vect_supportable_dr_alignment): Likewise.
12328 (dependence_distance_ge_vf): Take the vectorization factor as a
12329 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12330 (vect_analyze_data_refs): Change min_vf from an int * to a
12332 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12333 vfm1 as a poly_uint64 rather than an int. Make the same change
12334 for the returned bound_scalar.
12335 (vect_gen_vector_loop_niters): Handle polynomial vf.
12336 (vect_do_peeling): Likewise. Update call to
12337 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12338 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12340 * tree-vect-loop.c (vect_determine_vectorization_factor)
12341 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12342 (vect_get_known_peeling_cost): Likewise.
12343 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12344 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12345 (vect_transform_loop): Likewise. Use the lowest possible VF when
12346 updating the upper bounds of the loop.
12347 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12348 rather than an int.
12349 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12350 polynomial unroll factors.
12351 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12352 (vect_make_slp_decision): Likewise.
12353 (vect_supported_load_permutation_p): Likewise, and polynomial
12355 (vect_analyze_slp_cost): Handle polynomial vf.
12356 (vect_slp_analyze_node_operations): Likewise.
12357 (vect_slp_analyze_bb_1): Likewise.
12358 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12359 than an unsigned HOST_WIDE_INT.
12360 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12361 (vectorizable_load): Handle polynomial vf.
12362 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12364 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12367 Alan Hayward <alan.hayward@arm.com>
12368 David Sherwood <david.sherwood@arm.com>
12370 * match.pd: Handle bit operations involving three constants
12371 and try to fold one pair.
12373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12375 * tree-vect-loop-manip.c: Include gimple-fold.h.
12376 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12377 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12378 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12379 Add a path that uses a step of VF instead of 1, but disable it
12381 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12382 and niters_no_overflow parameters. Update calls to
12383 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12384 Create a new SSA name if the latter choses to use a ste other
12385 than zero, and return it via niters_vector_mult_vf_var.
12386 * tree-vect-loop.c (vect_transform_loop): Update calls to
12387 vect_do_peeling, vect_gen_vector_loop_niters and
12388 slpeel_make_loop_iterate_ntimes.
12389 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12390 (vect_gen_vector_loop_niters): Update declarations after above changes.
12392 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12394 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12395 128-bit round to integer instructions.
12396 (ceil<mode>2): Likewise.
12397 (btrunc<mode>2): Likewise.
12398 (round<mode>2): Likewise.
12400 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12402 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12403 unaligned VSX load/store on P8/P9.
12404 (expand_block_clear): Allow the use of unaligned VSX
12405 load/store on P8/P9.
12407 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12409 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12411 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12412 swap associated with both a load and a store.
12414 2018-01-02 Andrew Waterman <andrew@sifive.com>
12416 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12417 * config/riscv/riscv.md (clear_cache): Use it.
12419 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12421 * web.c: Remove out-of-date comment.
12423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12425 * expr.c (fixup_args_size_notes): Check that any existing
12426 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12427 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12428 (emit_single_push_insn): ...here.
12430 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12432 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12433 (const_vector_encoded_nelts): New function.
12434 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12435 (const_vector_int_elt, const_vector_elt): Declare.
12436 * emit-rtl.c (const_vector_int_elt_1): New function.
12437 (const_vector_elt): Likewise.
12438 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12439 of CONST_VECTOR_ELT.
12441 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12443 * expr.c: Include rtx-vector-builder.h.
12444 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12445 directly on the tree encoding.
12446 (const_vector_from_tree): Likewise.
12447 * optabs.c: Include rtx-vector-builder.h.
12448 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12449 sequence of "u" values.
12450 * vec-perm-indices.c: Include rtx-vector-builder.h.
12451 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12452 directly on the vec_perm_indices encoding.
12454 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12456 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12457 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12458 * rtx-vector-builder.h: New file.
12459 * rtx-vector-builder.c: Likewise.
12460 * rtl.h (rtx_def::u2): Add a const_vector field.
12461 (CONST_VECTOR_NPATTERNS): New macro.
12462 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12463 (CONST_VECTOR_DUPLICATE_P): Likewise.
12464 (CONST_VECTOR_STEPPED_P): Likewise.
12465 (CONST_VECTOR_ENCODED_ELT): Likewise.
12466 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12467 (unwrap_const_vec_duplicate): Likewise.
12468 (const_vec_series_p): Check for a non-duplicated vector encoding.
12469 Say that the function only returns true for integer vectors.
12470 * emit-rtl.c: Include rtx-vector-builder.h.
12471 (gen_const_vec_duplicate_1): Delete.
12472 (gen_const_vector): Call gen_const_vec_duplicate instead of
12473 gen_const_vec_duplicate_1.
12474 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12475 (gen_const_vec_duplicate): Use rtx_vector_builder.
12476 (gen_const_vec_series): Likewise.
12477 (gen_rtx_CONST_VECTOR): Likewise.
12478 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12479 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12480 Build a new vector rather than modifying a CONST_VECTOR in-place.
12481 (handle_special_swappables): Update call accordingly.
12482 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12483 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12484 Build a new vector rather than modifying a CONST_VECTOR in-place.
12485 (handle_special_swappables): Update call accordingly.
12487 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12489 * simplify-rtx.c (simplify_const_binary_operation): Use
12490 CONST_VECTOR_ELT instead of XVECEXP.
12492 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12494 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12495 the selector elements to be different from the data elements
12496 if the selector is a VECTOR_CST.
12497 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12498 ssizetype for the selector.
12500 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12502 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12503 before testing each element individually.
12504 * tree-vect-generic.c (lower_vec_perm): Likewise.
12506 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12508 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12509 * selftest-run-tests.c (selftest::run_tests): Call it.
12510 * vector-builder.h (vector_builder::operator ==): New function.
12511 (vector_builder::operator !=): Likewise.
12512 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12513 (vec_perm_indices::all_from_input_p): New function.
12514 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12515 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12516 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12517 instead of reading the VECTOR_CST directly. Detect whether both
12518 vector inputs are the same before constructing the vec_perm_indices,
12519 and update the number of inputs argument accordingly. Use the
12520 utility functions added above. Only construct sel2 if we need to.
12522 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12524 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12525 the broadcast of the low byte.
12526 (expand_mult_highpart): Use an explicit encoding for the permutes.
12527 * optabs-query.c (can_mult_highpart_p): Likewise.
12528 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12529 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12530 (vectorizable_bswap): Likewise.
12531 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12532 explicit encoding for the power-of-2 permutes.
12533 (vect_permute_store_chain): Likewise.
12534 (vect_grouped_load_supported): Likewise.
12535 (vect_permute_load_chain): Likewise.
12537 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12539 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12540 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12541 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12542 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12543 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12544 (vect_gen_perm_mask_any): Likewise.
12546 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12548 * int-vector-builder.h: New file.
12549 * vec-perm-indices.h: Include int-vector-builder.h.
12550 (vec_perm_indices): Redefine as an int_vector_builder.
12551 (auto_vec_perm_indices): Delete.
12552 (vec_perm_builder): Redefine as a stand-alone class.
12553 (vec_perm_indices::vec_perm_indices): New function.
12554 (vec_perm_indices::clamp): Likewise.
12555 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12556 (vec_perm_indices::new_vector): New function.
12557 (vec_perm_indices::new_expanded_vector): Update for new
12558 vec_perm_indices class.
12559 (vec_perm_indices::rotate_inputs): New function.
12560 (vec_perm_indices::all_in_range_p): Operate directly on the
12561 encoded form, without computing elided elements.
12562 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12563 encoding. Update for new vec_perm_indices class.
12564 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12565 the given vec_perm_builder.
12566 (expand_vec_perm_var): Update vec_perm_builder constructor.
12567 (expand_mult_highpart): Use vec_perm_builder instead of
12568 auto_vec_perm_indices.
12569 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12570 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12571 or double series encoding as appropriate.
12572 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12573 vec_perm_indices instead of auto_vec_perm_indices.
12574 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12575 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12576 (vect_permute_store_chain): Likewise.
12577 (vect_grouped_load_supported): Likewise.
12578 (vect_permute_load_chain): Likewise.
12579 (vect_shift_permute_load_chain): Likewise.
12580 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12581 (vect_transform_slp_perm_load): Likewise.
12582 (vect_schedule_slp_instance): Likewise.
12583 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12584 (vectorizable_mask_load_store): Likewise.
12585 (vectorizable_bswap): Likewise.
12586 (vectorizable_store): Likewise.
12587 (vectorizable_load): Likewise.
12588 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12589 vec_perm_indices instead of auto_vec_perm_indices. Use
12590 tree_to_vec_perm_builder to read the vector from a tree.
12591 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12592 vec_perm_builder instead of a vec_perm_indices.
12593 (have_whole_vector_shift): Use vec_perm_builder and
12594 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12595 truncation to calc_vec_perm_mask_for_shift.
12596 (vect_create_epilog_for_reduction): Likewise.
12597 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12598 from auto_vec_perm_indices to vec_perm_indices.
12599 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12600 instead of changing individual elements.
12601 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12602 the vector in d.perm.
12603 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12604 from auto_vec_perm_indices to vec_perm_indices.
12605 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12606 instead of changing individual elements.
12607 (arm_vectorize_vec_perm_const): Use new_vector to install
12608 the vector in d.perm.
12609 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12610 Update vec_perm_builder constructor.
12611 (rs6000_expand_interleave): Likewise.
12612 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12613 (rs6000_expand_interleave): Likewise.
12615 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12617 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12618 to qimode could truncate the indices.
12619 * optabs.c (expand_vec_perm_var): Likewise.
12621 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12623 * Makefile.in (OBJS): Add vec-perm-indices.o.
12624 * vec-perm-indices.h: New file.
12625 * vec-perm-indices.c: Likewise.
12626 * target.h (vec_perm_indices): Replace with a forward class
12628 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12629 * optabs.h: Include vec-perm-indices.h.
12630 (expand_vec_perm): Delete.
12631 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12632 (expand_vec_perm_const): Declare.
12633 * target.def (vec_perm_const_ok): Replace with...
12634 (vec_perm_const): ...this new hook.
12635 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12636 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12637 * doc/tm.texi: Regenerate.
12638 * optabs.def (vec_perm_const): Delete.
12639 * doc/md.texi (vec_perm_const): Likewise.
12640 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12641 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12642 expand_vec_perm for constant permutation vectors. Assert that
12643 the mode of variable permutation vectors is the integer equivalent
12644 of the mode that is being permuted.
12645 * optabs-query.h (selector_fits_mode_p): Declare.
12646 * optabs-query.c: Include vec-perm-indices.h.
12647 (selector_fits_mode_p): New function.
12648 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12649 is defined, instead of checking whether the vec_perm_const_optab
12650 exists. Use targetm.vectorize.vec_perm_const instead of
12651 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12652 fit in the vector mode before using a variable permute.
12653 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12654 vec_perm_indices instead of an rtx.
12655 (expand_vec_perm): Replace with...
12656 (expand_vec_perm_const): ...this new function. Take the selector
12657 as a vec_perm_indices rather than an rtx. Also take the mode of
12658 the selector. Update call to shift_amt_for_vec_perm_mask.
12659 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12660 Use vec_perm_indices::new_expanded_vector to expand the original
12661 selector into bytes. Check whether the indices fit in the vector
12662 mode before using a variable permute.
12663 (expand_vec_perm_var): Make global.
12664 (expand_mult_highpart): Use expand_vec_perm_const.
12665 * fold-const.c: Includes vec-perm-indices.h.
12666 * tree-ssa-forwprop.c: Likewise.
12667 * tree-vect-data-refs.c: Likewise.
12668 * tree-vect-generic.c: Likewise.
12669 * tree-vect-loop.c: Likewise.
12670 * tree-vect-slp.c: Likewise.
12671 * tree-vect-stmts.c: Likewise.
12672 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12674 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12675 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12676 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12677 (aarch64_vectorize_vec_perm_const): ...this new function.
12678 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12679 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12680 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12681 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12682 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12683 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12684 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12686 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12687 check for NEON modes.
12688 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12689 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12690 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12691 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12693 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12694 the old VEC_PERM_CONST conditions.
12695 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12696 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12697 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12698 (ia64_vectorize_vec_perm_const_ok): Merge into...
12699 (ia64_vectorize_vec_perm_const): ...this new function.
12700 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12701 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12702 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12703 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12704 * config/mips/mips.c (mips_expand_vec_perm_const)
12705 (mips_vectorize_vec_perm_const_ok): Merge into...
12706 (mips_vectorize_vec_perm_const): ...this new function.
12707 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12708 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12709 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12710 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12711 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12712 (rs6000_expand_vec_perm_const): Delete.
12713 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12715 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12716 (altivec_expand_vec_perm_const_le): Take each operand individually.
12717 Operate on constant selectors rather than rtxes.
12718 (altivec_expand_vec_perm_const): Likewise. Update call to
12719 altivec_expand_vec_perm_const_le.
12720 (rs6000_expand_vec_perm_const): Delete.
12721 (rs6000_vectorize_vec_perm_const_ok): Delete.
12722 (rs6000_vectorize_vec_perm_const): New function.
12723 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12724 an element count and rtx array.
12725 (rs6000_expand_extract_even): Update call accordingly.
12726 (rs6000_expand_interleave): Likewise.
12727 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12728 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12729 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12730 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12731 (rs6000_expand_vec_perm_const): Delete.
12732 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12733 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12734 (altivec_expand_vec_perm_const_le): Take each operand individually.
12735 Operate on constant selectors rather than rtxes.
12736 (altivec_expand_vec_perm_const): Likewise. Update call to
12737 altivec_expand_vec_perm_const_le.
12738 (rs6000_expand_vec_perm_const): Delete.
12739 (rs6000_vectorize_vec_perm_const_ok): Delete.
12740 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12741 reference to the SPE evmerge intructions.
12742 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12743 an element count and rtx array.
12744 (rs6000_expand_extract_even): Update call accordingly.
12745 (rs6000_expand_interleave): Likewise.
12746 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12747 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12749 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12751 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12753 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12754 vector mode and that that mode matches the mode of the data
12756 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12757 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12758 directly using expand_vec_perm_1 when forcing selectors into
12760 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12762 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12764 * optabs-query.h (can_vec_perm_p): Delete.
12765 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12766 * optabs-query.c (can_vec_perm_p): Split into...
12767 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12768 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12769 particular selector is valid.
12770 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12771 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12772 (vect_grouped_load_supported): Likewise.
12773 (vect_shift_permute_load_chain): Likewise.
12774 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12775 (vect_transform_slp_perm_load): Likewise.
12776 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12777 (vectorizable_bswap): Likewise.
12778 (vect_gen_perm_mask_checked): Likewise.
12779 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12780 implementations of variable permutation vectors into account
12781 when deciding which selector to use.
12782 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12783 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12784 with a false third argument.
12785 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12786 to test whether the constant selector is valid and can_vec_perm_var_p
12787 to test whether a variable selector is valid.
12789 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12791 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12792 * optabs-query.c (can_vec_perm_p): Likewise.
12793 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12794 instead of vec_perm_indices.
12795 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12796 (vect_gen_perm_mask_checked): Likewise,
12797 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12798 (vect_gen_perm_mask_checked): Likewise,
12800 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12802 * optabs-query.h (qimode_for_vec_perm): Declare.
12803 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12804 (qimode_for_vec_perm): ...this new function.
12805 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12807 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12809 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12810 does not have a conditional at the top.
12812 2018-01-02 Richard Biener <rguenther@suse.de>
12814 * ipa-inline.c (big_speedup_p): Fix expression.
12816 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12819 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12822 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12826 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12827 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12828 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12829 cond_taken_branch_cost 3->4.
12831 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12833 PR tree-optimization/83581
12834 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12835 TODO_cleanup_cfg if any changes have been made.
12837 PR middle-end/83608
12838 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12839 convert_modes if target mode has the right side, but different mode
12842 PR middle-end/83609
12843 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12844 last argument when extracting from CONCAT. If either from_real or
12845 from_imag is NULL, use expansion through memory. If result is not
12846 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12847 the parts directly to inner mode, if even that fails, use expansion
12850 PR middle-end/83623
12851 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12852 check for bswap in mode rather than HImode and use that in expand_unop
12855 Copyright (C) 2018 Free Software Foundation, Inc.
12857 Copying and distribution of this file, with or without modification,
12858 are permitted in any medium without royalty provided the copyright
12859 notice and this notice are preserved.