1 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2 Sudakshina Das <sudi.das@arm.com>
4 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
6 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
7 (lshrdi3): Generate thumb2_lsrl for valid immediates.
8 * config/arm/constraints.md (Pg): New.
9 * config/arm/predicates.md (long_shift_imm): New.
10 (arm_reg_or_long_shift_imm): Likewise.
11 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
12 (thumb2_lsll): Likewise.
15 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
16 Sudakshina Das <sudi.das@arm.com>
18 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
19 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
20 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
21 register pairs for doubleword quantities for ARMv8.1M-Mainline.
22 * config/arm/thumb2.md (thumb2_asrl): New.
23 (thumb2_lsll): Likewise.
25 2020-01-17 Jakub Jelinek <jakub@redhat.com>
27 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
30 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
32 * gdbinit.in (help-gcc-hooks): New command.
33 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
34 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
37 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
39 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
42 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
44 * config/aarch64/aarch64-protos.h
45 (aarch64_sve_ld1ro_operand_p): New.
46 * config/aarch64/aarch64-sve-builtins-base.cc
47 (class load_replicate): New.
48 (class svld1ro_impl): New.
49 (class svld1rq_impl): Change to inherit from load_replicate.
50 (svld1ro): New sve intrinsic function base.
51 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
53 * config/aarch64/aarch64-sve-builtins-base.h
55 * config/aarch64/aarch64-sve-builtins.cc
56 (function_expander::add_mem_operand): Modify assert to allow
58 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
60 * config/aarch64/aarch64.c
61 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
62 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
63 (aarch64_sve_ld1ro_operand_p): New.
64 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
65 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
66 * config/aarch64/predicates.md
67 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
69 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
71 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
72 Introduce this ACLE specified predefined macro.
73 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
74 (fp): Disabling this disables f64mm.
75 (simd): Disabling this disables f64mm.
76 (fp16): Disabling this disables f64mm.
77 (sve): Disabling this disables f64mm.
78 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
79 (AARCH64_ISA_F64MM): New.
81 * doc/invoke.texi (f64mm): Document new option.
83 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
85 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
86 (neoversen1_tunings): Likewise.
88 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
91 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
92 Add assert to ensure prolog has been emitted.
93 (aarch64_split_atomic_op): Likewise.
94 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
95 Use epilogue_completed rather than reload_completed.
96 (aarch64_atomic_exchange<mode>): Likewise.
97 (aarch64_atomic_<atomic_optab><mode>): Likewise.
98 (atomic_nand<mode>): Likewise.
99 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
100 (atomic_fetch_nand<mode>): Likewise.
101 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
102 (atomic_nand_fetch<mode>): Likewise.
104 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
107 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
109 (REVERSE_CONDITION): Delete.
110 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
111 (CCFP_CCFPE): Likewise.
112 (e): New mode attribute.
113 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
114 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
115 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
116 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
117 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
118 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
119 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
120 name of generator from gen_ccmpdi to gen_ccmpccdi.
121 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
122 the previous comparison but aren't able to, use the new ccmp_rev
125 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
127 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
128 than testing directly for INTEGER_CST.
129 (gimplify_target_expr, gimplify_omp_depend): Likewise.
131 2020-01-17 Jakub Jelinek <jakub@redhat.com>
133 PR tree-optimization/93292
134 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
135 get_vectype_for_scalar_type returns NULL.
137 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
139 * params.opt (-param=max-predicted-iterations): Increase range from 0.
140 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
142 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
144 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
146 * params.opt: (max-predicted-iterations): Set bounds.
147 * predict.c (real_almost_one, real_br_prob_base,
148 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
149 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
150 probabilities; do not truncate to reg_br_prob_bases.
151 (estimate_loops_at_level): Pass max_cyclic_prob.
152 (estimate_loops): Compute max_cyclic_prob.
153 (estimate_bb_frequencies): Do not initialize real_*; update calculation
155 * profile-count.c (profile_probability::to_sreal): New.
156 * profile-count.h (class sreal): Move up in file.
157 (profile_probability::to_sreal): Declare.
159 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
162 (arm_invalid_conversion): New function for target hook.
163 (arm_invalid_unary_op): New function for target hook.
164 (arm_invalid_binary_op): New function for target hook.
166 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
168 * config.gcc: Add arm_bf16.h.
169 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
170 (arm_simd_builtin_std_type): Add BFmode.
171 (arm_init_simd_builtin_types): Define element types for vector types.
172 (arm_init_bf16_types): New function.
173 (arm_init_builtins): Add arm_init_bf16_types function call.
174 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
175 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
176 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
177 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
178 (arm_vector_mode_supported_p): Add V4BF, V8BF.
179 (arm_mangle_type): Add __bf16.
180 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
181 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
182 arm_bf16_ptr_type_node.
183 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
184 define_split between ARM registers.
185 * config/arm/arm_bf16.h: New file.
186 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
187 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
189 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
190 * config/arm/vfp.md: Add BFmode to movhf patterns.
192 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
193 Andre Vieira <andre.simoesdiasvieira@arm.com>
195 * config/arm/arm-cpus.in (mve, mve_float): New features.
196 (dsp, mve, mve.fp): New options.
197 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
198 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
199 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
201 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
202 Thomas Preud'homme <thomas.preudhomme@arm.com>
204 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
206 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
207 error for using -mcmse when targeting Armv8.1-M Mainline.
209 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
210 Thomas Preud'homme <thomas.preudhomme@arm.com>
212 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
213 address in r4 when targeting Armv8.1-M Mainline.
214 (nonsecure_call_value_internal): Likewise.
215 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
216 a register match_operand again. Emit BLXNS when targeting
218 (nonsecure_call_value_reg_thumb2): Likewise.
220 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
221 Thomas Preud'homme <thomas.preudhomme@arm.com>
223 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
224 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
225 variable as true when floating-point ABI is not hard. Replace
226 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
227 Generate VLSTM and VLLDM instruction respectively before and
228 after a function call to cmse_nonsecure_call function.
229 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
230 (VUNSPEC_VLLDM): Likewise.
231 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
232 (lazy_load_multiple_insn): Likewise.
234 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
235 Thomas Preud'homme <thomas.preudhomme@arm.com>
237 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
238 (arm_emit_vfp_multi_reg_pop): Likewise.
239 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
240 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
241 restore callee-saved VFP registers.
243 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
244 Thomas Preud'homme <thomas.preudhomme@arm.com>
246 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
247 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
248 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
249 callee-saved GPRs as well as clear ip register before doing a nonsecure
250 call then restore callee-saved GPRs after it when targeting
252 (arm_reorg): Adapt to function rename.
254 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
255 Thomas Preud'homme <thomas.preudhomme@arm.com>
257 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
258 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
259 clear_vfp_multiple pattern based on a new vfp parameter.
260 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
261 targeting Armv8.1-M Mainline.
262 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
263 unconditionally when targeting Armv8.1-M Mainline architecture. Check
264 whether VFP registers are available before looking call_used_regs for a
266 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
267 of prototype of clear_operation_p.
268 (clear_vfp_multiple_operation): New predicate.
269 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
270 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
272 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
273 Thomas Preud'homme <thomas.preudhomme@arm.com>
275 * config/arm/arm-protos.h (clear_operation_p): Declare.
276 * config/arm/arm.c (clear_operation_p): New function.
277 (cmse_clear_registers): Generate clear_multiple instruction pattern if
278 targeting Armv8.1-M Mainline or successor.
279 (output_return_instruction): Only output APSR register clearing if
280 Armv8.1-M Mainline instructions not available.
281 (thumb_exit): Likewise.
282 * config/arm/predicates.md (clear_multiple_operation): New predicate.
283 * config/arm/thumb2.md (clear_apsr): New define_insn.
284 (clear_multiple): Likewise.
285 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
287 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
288 Thomas Preud'homme <thomas.preudhomme@arm.com>
290 * config/arm/arm.c (fp_sysreg_names): Declare and define.
291 (use_return_insn): Also return false for Armv8.1-M Mainline.
292 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
293 Mainline instructions are available.
294 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
295 when targeting Armv8.1-M Mainline Security Extensions.
296 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
297 Mainline entry function.
298 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
299 targeting Armv8.1-M Mainline or successor.
300 (arm_expand_epilogue): Fix indentation of caller-saved register
301 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
303 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
304 (FP_SYSREGS): Likewise.
305 (enum vfp_sysregs_encoding): Define enum.
306 (fp_sysreg_names): Declare.
307 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
308 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
309 (pop_fpsysreg_insn): Likewise.
311 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
312 Thomas Preud'homme <thomas.preudhomme@arm.com>
314 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
315 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
316 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
317 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
318 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
319 (ARMv8_1m_main): New feature group.
320 (armv8.1-m.main): New architecture.
321 * config/arm/arm-tables.opt: Regenerate.
322 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
323 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
324 (arm_options_perform_arch_sanity_checks): Error out when targeting
325 Armv8.1-M Mainline Security Extensions.
326 * config/arm/arm.h (arm_arch8_1m_main): Declare.
328 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
330 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
331 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
332 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
333 aarch64_bfdot_laneq): New.
334 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
335 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
336 vbfdotq_laneq_f32): New.
337 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
339 (isquadop): Add V4BF, V8BF.
341 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
343 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
344 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
345 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
346 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
347 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
348 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
349 usdot_laneq, sudot_lane,sudot_laneq): New.
350 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
351 (aarch64_<sur>dot_lane): New.
352 * config/aarch64/arm_neon.h (vusdot_s32): New.
354 (vusdot_lane_s32): New.
355 (vsudot_lane_s32): New.
356 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
357 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
359 2020-01-16 Martin Liska <mliska@suse.cz>
361 * value-prof.c (dump_histogram_value): Fix
362 obvious spacing issue.
364 2020-01-16 Andrew Pinski <apinski@marvell.com>
366 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
367 !storage_order_barrier_p.
369 2020-01-16 Andrew Pinski <apinski@marvell.com>
371 * sched-int.h (_dep): Add unused bit-field field for the padding.
372 * sched-deps.c (init_dep_1): Init unused field.
374 2020-01-16 Andrew Pinski <apinski@marvell.com>
376 * optabs.h (create_expand_operand): Initialize target field also.
378 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
380 PR tree-optimization/92429
381 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
382 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
384 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
387 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
389 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
390 aarch64_sve_int_mode to each mode.
392 2020-01-15 David Malcolm <dmalcolm@redhat.com>
394 * doc/analyzer.texi (Overview): Add note about
397 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
399 PR tree-optimization/93231
400 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
401 input_type is unsigned. Use tree_to_shwi for shift constant.
402 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
403 (simplify_count_trailing_zeroes): Add test to handle known non-zero
404 inputs more efficiently.
406 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
408 * config/i386/i386.md (*movsf_internal): Do not require
409 SSE2 ISA for alternatives 14 and 15.
411 2020-01-15 Richard Biener <rguenther@suse.de>
414 * tree-eh.c (sink_clobbers): If we already visited the destination
415 block do not defer insertion.
416 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
417 the purpose of defered insertion.
419 2020-01-15 Jakub Jelinek <jakub@redhat.com>
421 * BASE-VER: Bump to 10.0.1.
423 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
425 PR tree-optimization/93247
426 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
427 type of the stmt that we're going to vectorize.
429 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
431 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
432 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
435 2020-01-15 Martin Liska <mliska@suse.cz>
437 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
438 2 calls of streamer_read_hwi in a function call.
440 2020-01-15 Richard Biener <rguenther@suse.de>
442 * alias.c (record_alias_subset): Avoid redundant work when
443 subset is already recorded.
445 2020-01-14 David Malcolm <dmalcolm@redhat.com>
447 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
448 the analyzer options provide CWE identifiers.
450 2020-01-14 David Malcolm <dmalcolm@redhat.com>
452 * tree-diagnostic-path.cc (path_summary::event_range::print):
453 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
454 using get_pure_location.
456 2020-01-15 Jakub Jelinek <jakub@redhat.com>
458 PR tree-optimization/93262
459 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
460 perform head trimming only if the last argument is constant,
461 either all ones, or larger or equal to head trim, in the latter
462 case decrease the last argument by head_trim.
464 PR tree-optimization/93249
465 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
466 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
467 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
468 perform head trim unless we can prove there are no '\0' chars
469 from the source among the first head_trim chars.
471 2020-01-14 David Malcolm <dmalcolm@redhat.com>
473 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
475 2020-01-15 Jakub Jelinek <jakub@redhat.com>
479 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
480 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
481 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
482 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
483 just a single alternative instead of two, make operands 1 and 2
486 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
489 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
492 2020-01-14 David Malcolm <dmalcolm@redhat.com>
494 * Makefile.in (lang_opt_files): Add analyzer.opt.
495 (ANALYZER_OBJS): New.
496 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
497 tristate.o and ANALYZER_OBJS.
498 (TEXI_GCCINT_FILES): Add analyzer.texi.
499 * common.opt (-fanalyzer): New driver option.
500 * config.in: Regenerate.
501 * configure: Regenerate.
502 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
503 (gccdepdir): Also create depdir for "analyzer" subdir.
504 * digraph.cc: New file.
505 * digraph.h: New file.
506 * doc/analyzer.texi: New file.
507 * doc/gccint.texi ("Static Analyzer") New menu item.
508 (analyzer.texi): Include it.
509 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
510 ("Warning Options"): Add static analysis warnings to the list.
511 (-Wno-analyzer-double-fclose): New option.
512 (-Wno-analyzer-double-free): New option.
513 (-Wno-analyzer-exposure-through-output-file): New option.
514 (-Wno-analyzer-file-leak): New option.
515 (-Wno-analyzer-free-of-non-heap): New option.
516 (-Wno-analyzer-malloc-leak): New option.
517 (-Wno-analyzer-possible-null-argument): New option.
518 (-Wno-analyzer-possible-null-dereference): New option.
519 (-Wno-analyzer-null-argument): New option.
520 (-Wno-analyzer-null-dereference): New option.
521 (-Wno-analyzer-stale-setjmp-buffer): New option.
522 (-Wno-analyzer-tainted-array-index): New option.
523 (-Wno-analyzer-use-after-free): New option.
524 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
525 (-Wno-analyzer-use-of-uninitialized-value): New option.
526 (-Wanalyzer-too-complex): New option.
527 (-fanalyzer-call-summaries): New warning.
528 (-fanalyzer-checker=): New warning.
529 (-fanalyzer-fine-grained): New warning.
530 (-fno-analyzer-state-merge): New warning.
531 (-fno-analyzer-state-purge): New warning.
532 (-fanalyzer-transitivity): New warning.
533 (-fanalyzer-verbose-edges): New warning.
534 (-fanalyzer-verbose-state-changes): New warning.
535 (-fanalyzer-verbosity=): New warning.
536 (-fdump-analyzer): New warning.
537 (-fdump-analyzer-callgraph): New warning.
538 (-fdump-analyzer-exploded-graph): New warning.
539 (-fdump-analyzer-exploded-nodes): New warning.
540 (-fdump-analyzer-exploded-nodes-2): New warning.
541 (-fdump-analyzer-exploded-nodes-3): New warning.
542 (-fdump-analyzer-supergraph): New warning.
543 * doc/sourcebuild.texi (dg-require-dot): New.
545 * gdbinit.in (break-on-saved-diagnostic): New command.
546 * graphviz.cc: New file.
547 * graphviz.h: New file.
548 * ordered-hash-map-tests.cc: New file.
549 * ordered-hash-map.h: New file.
550 * passes.def (pass_analyzer): Add before
551 pass_ipa_whole_program_visibility.
552 * selftest-run-tests.c (selftest::run_tests): Call
553 selftest::ordered_hash_map_tests_cc_tests.
554 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
556 * shortest-paths.h: New file.
557 * timevar.def (TV_ANALYZER): New timevar.
558 (TV_ANALYZER_SUPERGRAPH): Likewise.
559 (TV_ANALYZER_STATE_PURGE): Likewise.
560 (TV_ANALYZER_PLAN): Likewise.
561 (TV_ANALYZER_SCC): Likewise.
562 (TV_ANALYZER_WORKLIST): Likewise.
563 (TV_ANALYZER_DUMP): Likewise.
564 (TV_ANALYZER_DIAGNOSTICS): Likewise.
565 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
566 * tree-pass.h (make_pass_analyzer): New decl.
567 * tristate.cc: New file.
568 * tristate.h: New file.
570 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
573 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
574 alternatives 9 and 10.
576 2020-01-14 David Malcolm <dmalcolm@redhat.com>
578 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
579 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
580 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
581 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
582 (selftest::hash_map_tests_c_tests): Call it.
583 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
584 New static constant, using the value of = H::empty_zero_p.
585 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
586 from default_hash_traits <Value>.
587 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
589 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
590 * hash-table.h (hash_table::alloc_entries): Guard the loop of
591 calls to mark_empty with !Descriptor::empty_zero_p.
592 (hash_table::empty_slow): Conditionalize the memset call with a
593 check that Descriptor::empty_zero_p; otherwise, loop through the
594 entries calling mark_empty on them.
595 * hash-traits.h (int_hash::empty_zero_p): New static constant.
596 (pointer_hash::empty_zero_p): Likewise.
597 (pair_hash::empty_zero_p): Likewise.
598 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
600 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
601 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
602 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
603 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
604 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
605 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
606 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
607 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
609 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
612 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
614 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
615 fix typo on return value.
617 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
620 * cgraph.c (symbol_table::create_edge): Init speculative_id and
622 (cgraph_edge::make_speculative): Add param for setting speculative_id
624 (cgraph_edge::speculative_call_info): Update comments and find reference
625 by speculative_id for multiple indirect targets.
626 (cgraph_edge::resolve_speculation): Decrease the speculations
627 for indirect edge, drop it's speculative if not direct target
628 left. Update comments.
629 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
630 (cgraph_node::dump): Print num_speculative_call_targets.
631 (cgraph_node::verify_node): Don't report error if speculative
632 edge not include statement.
633 (cgraph_edge::num_speculative_call_targets_p): New function.
634 * cgraph.h (int common_target_id): Remove.
635 (int common_target_probability): Remove.
636 (num_speculative_call_targets): New variable.
637 (make_speculative): Add param for setting speculative_id.
638 (cgraph_edge::num_speculative_call_targets_p): New declare.
639 (target_prob): New variable.
640 (speculative_id): New variable.
641 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
642 call summaries for multiple speculative call targets.
643 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
644 * ipa-profile.c (struct speculative_call_target): New struct.
645 (class speculative_call_summary): New class.
646 (class speculative_call_summaries): New class.
647 (call_sums): New variable.
648 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
649 (ipa_profile_write_edge_summary): New function.
650 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
651 (ipa_profile_dump_all_summaries): New function.
652 (ipa_profile_read_edge_summary): New function.
653 (ipa_profile_read_summary_section): New function.
654 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
655 (ipa_profile): Generate num_speculative_call_targets from
657 * ipa-ref.h (speculative_id): New variable.
658 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
659 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
660 common_target_probability. Stream out speculative_id and
661 num_speculative_call_targets.
662 (input_edge): Likewise.
663 * predict.c (dump_prediction): Remove edges count assert to be
665 * symtab.c (symtab_node::create_reference): Init speculative_id.
666 (symtab_node::clone_references): Clone speculative_id.
667 (symtab_node::clone_referring): Clone speculative_id.
668 (symtab_node::clone_reference): Clone speculative_id.
669 (symtab_node::clear_stmts_in_references): Clear speculative_id.
670 * tree-inline.c (copy_bb): Duplicate all the speculative edges
671 if indirect call contains multiple speculative targets.
672 * value-prof.h (check_ic_target): Remove.
673 * value-prof.c (gimple_value_profile_transformations):
674 Use void function gimple_ic_transform.
675 * value-prof.c (gimple_ic_transform): Handle topn case.
676 Fix comment typos. Change it to a void function.
678 2020-01-13 Andrew Pinski <apinski@marvell.com>
680 * config/aarch64/aarch64-cores.def (octeontx2): New define.
681 (octeontx2t98): New define.
682 (octeontx2t96): New define.
683 (octeontx2t93): New define.
684 (octeontx2f95): New define.
685 (octeontx2f95n): New define.
686 (octeontx2f95mm): New define.
687 * config/aarch64/aarch64-tune.md: Regenerate.
688 * doc/invoke.texi (-mcpu=): Document the new cpu types.
690 2020-01-13 Jason Merrill <jason@redhat.com>
692 PR c++/33799 - destroy return value if local cleanup throws.
693 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
695 2020-01-13 Martin Liska <mliska@suse.cz>
697 * ipa-cp.c (get_max_overall_size): Use newly
698 renamed param param_ipa_cp_unit_growth.
699 * params.opt: Remove legacy param name.
701 2020-01-13 Martin Sebor <msebor@redhat.com>
703 PR tree-optimization/93213
704 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
705 stores to be eliminated.
707 2020-01-13 Martin Liska <mliska@suse.cz>
709 * opts.c (print_help): Do not print CL_PARAM
710 and CL_WARNING for CL_OPTIMIZATION.
712 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
715 * doc/invoke.texi (Warning Options): Add caveat about some warnings
716 depending on optimization settings.
718 2020-01-13 Jakub Jelinek <jakub@redhat.com>
720 PR tree-optimization/90838
721 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
722 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
723 argument rather than to initialize temporary for targets that
724 don't use the mode argument at all. Initialize ctzval to avoid
727 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
729 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
730 * tree-core.h: Document it.
731 * gimplify.c (gimplify_omp_workshare): Set it.
732 * omp-low.c (lower_omp_target): Use it.
733 * tree-pretty-print.c (dump_omp_clause): Print it.
735 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
736 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
738 2020-01-10 David Malcolm <dmalcolm@redhat.com>
740 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
741 * common.opt (fdiagnostics-path-format=): New option.
742 (diagnostic_path_format): New enum.
743 (fdiagnostics-show-path-depths): New option.
744 * coretypes.h (diagnostic_event_id_t): New forward decl.
745 * diagnostic-color.c (color_dict): Add "path".
746 * diagnostic-event-id.h: New file.
747 * diagnostic-format-json.cc (json_from_expanded_location): Make
749 (json_end_diagnostic): Call context->make_json_for_path if it
750 exists and the diagnostic has a path.
751 (diagnostic_output_format_init): Clear context->print_path.
752 * diagnostic-path.h: New file.
753 * diagnostic-show-locus.c (colorizer::set_range): Special-case
754 when printing a run of events in a diagnostic_path so that they
755 all get the same color.
756 (layout::m_diagnostic_path_p): New field.
757 (layout::layout): Initialize it.
758 (layout::print_any_labels): Don't colorize the label text for an
759 event in a diagnostic_path.
760 (gcc_rich_location::add_location_if_nearby): Add
761 "restrict_to_current_line_spans" and "label" params. Pass the
762 former to layout.maybe_add_location_range; pass the latter
763 when calling add_range.
764 * diagnostic.c: Include "diagnostic-path.h".
765 (diagnostic_initialize): Initialize context->path_format and
766 context->show_path_depths.
767 (diagnostic_show_any_path): New function.
768 (diagnostic_path::interprocedural_p): New function.
769 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
770 (simple_diagnostic_path::num_events): New function.
771 (simple_diagnostic_path::get_event): New function.
772 (simple_diagnostic_path::add_event): New function.
773 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
774 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
775 (debug): New overload taking a diagnostic_path *.
776 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
777 * diagnostic.h (enum diagnostic_path_format): New enum.
778 (json::value): New forward decl.
779 (diagnostic_context::path_format): New field.
780 (diagnostic_context::show_path_depths): New field.
781 (diagnostic_context::print_path): New callback field.
782 (diagnostic_context::make_json_for_path): New callback field.
783 (diagnostic_show_any_path): New decl.
784 (json_from_expanded_location): New decl.
785 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
786 (-fdiagnostics-show-path-depths): New option.
787 (-fdiagnostics-color): Add "path" to description of default
788 GCC_COLORS; describe it.
789 (-fdiagnostics-format=json): Document how diagnostic paths are
790 represented in the JSON output format.
791 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
792 Add optional params "restrict_to_current_line_spans" and "label".
793 * opts.c (common_handle_option): Handle
794 OPT_fdiagnostics_path_format_ and
795 OPT_fdiagnostics_show_path_depths.
796 * pretty-print.c: Include "diagnostic-event-id.h".
797 (pp_format): Implement "%@" format code for printing
798 diagnostic_event_id_t *.
799 (selftest::test_pp_format): Add tests for "%@".
800 * selftest-run-tests.c (selftest::run_tests): Call
801 selftest::tree_diagnostic_path_cc_tests.
802 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
803 * toplev.c (general_init): Initialize global_dc->path_format and
804 global_dc->show_path_depths.
805 * tree-diagnostic-path.cc: New file.
806 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
807 non-static. Drop "diagnostic" param in favor of storing the
808 original value of "where" and re-using it.
809 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
810 maybe_unwind_expanded_macro_loc.
811 (tree_diagnostics_defaults): Initialize context->print_path and
812 context->make_json_for_path.
813 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
815 (default_tree_make_json_for_path): New decl.
816 (maybe_unwind_expanded_macro_loc): New decl.
818 2020-01-10 Jakub Jelinek <jakub@redhat.com>
820 PR tree-optimization/93210
821 * fold-const.h (native_encode_initializer,
822 can_native_interpret_type_p): Declare.
823 * fold-const.c (native_encode_string): Fix up handling with off != -1,
825 (native_encode_initializer): New function, moved from dwarf2out.c.
826 Adjust to native_encode_expr compatible arguments, including dry-run
827 and partial extraction modes. Don't handle STRING_CST.
828 (can_native_interpret_type_p): No longer static.
829 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
830 offset / BITS_PER_UNIT fits into int and don't call it if
831 can_native_interpret_type_p fails. If suboff is NULL and for
832 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
833 native_encode_initializer.
834 (fold_const_aggregate_ref_1): Formatting fix.
835 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
836 (tree_add_const_value_attribute): Adjust caller.
838 PR tree-optimization/90838
839 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
840 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
841 CTZ_DEFINED_VALUE_AT_ZERO.
843 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
846 * lra-constraints.c (match_reload): Permit input operands have the
847 same mode as output while other input operands have a different
850 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
852 PR tree-optimization/90838
853 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
854 (check_ctz_string): Likewise.
855 (optimize_count_trailing_zeroes): Likewise.
856 (simplify_count_trailing_zeroes): Likewise.
857 (pass_forwprop::execute): Try ctz simplification.
858 * match.pd: Add matching for ctz idioms.
860 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
862 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
864 (aarch64_invalid_unary_op): New function for target hook.
865 (aarch64_invalid_binary_op): New function for target hook.
867 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
869 * config.gcc: Add arm_bf16.h.
870 * config/aarch64/aarch64-builtins.c
871 (aarch64_simd_builtin_std_type): Add BFmode.
872 (aarch64_init_simd_builtin_types): Define element types for vector
874 (aarch64_init_bf16_types): New function.
875 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
876 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
878 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
879 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
881 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
882 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
883 * config/aarch64/aarch64.c
884 (aarch64_classify_vector_mode): Add support for BF types.
885 (aarch64_gimplify_va_arg_expr): Add support for BF types.
886 (aarch64_vq_mode): Add support for BF types.
887 (aarch64_simd_container_mode): Add support for BF types.
888 (aarch64_mangle_type): Add support for BF scalar type.
889 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
890 * config/aarch64/arm_bf16.h: New file.
891 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
892 * config/aarch64/iterators.md: Add BF types to mode attributes.
893 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
895 2020-01-10 Jason Merrill <jason@redhat.com>
897 PR c++/93173 - incorrect tree sharing.
898 * gimplify.c (copy_if_shared): No longer static.
899 * gimplify.h: Declare it.
901 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
903 * doc/invoke.texi (-msve-vector-bits=): Document that
904 -msve-vector-bits=128 now generates VL-specific code for
905 little-endian targets.
906 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
907 build_vector_type_for_mode to construct the data vector types.
908 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
909 VL-specific code for -msve-vector-bits=128 on little-endian targets.
910 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
913 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
915 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
918 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
920 * config/aarch64/aarch64-builtins.c
921 (aarch64_builtin_vectorized_function): Check for specific vector modes,
922 rather than checking the number of elements and the element mode.
924 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
926 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
927 get_related_vectype_for_scalar_type rather than build_vector_type
928 to create the index type for a conditional reduction.
930 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
932 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
933 for any type of gather or scatter, including strided accesses.
935 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
937 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
940 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
942 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
944 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
945 parameter and its use to reset DR_OFFSET's.
946 (vect_transform_loop): Remove orig_drs_init argument.
947 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
948 member of dr_vec_info rather than the offset of the associated
949 data_reference's innermost_loop_behavior.
950 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
951 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
952 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
954 (vectorizable_store): Likewise.
955 (vectorizable_load): Likewise.
957 2020-01-10 Richard Biener <rguenther@suse.de>
959 * gimple-ssa-store-merging
960 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
962 2020-01-10 Martin Liska <mliska@suse.cz>
965 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
966 encapsulation that was there before r280040.
968 2020-01-10 Richard Biener <rguenther@suse.de>
971 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
972 sequences to avoid walking them again for secondary opportunities.
973 (pass_lower_eh_dispatch::execute): Instead actually insert
976 2020-01-10 Richard Biener <rguenther@suse.de>
979 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
980 (cleanup_all_empty_eh): Walk landing pads in reverse order to
983 2020-01-10 Martin Jambor <mjambor@suse.cz>
985 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
986 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
987 to get param_ipa_sra_max_replacements.
988 (param_splitting_across_edge): Pass the caller to
989 pull_accesses_from_callee.
991 2020-01-10 Martin Jambor <mjambor@suse.cz>
993 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
994 * ipa-cp.c (max_new_size): Removed.
995 (orig_overall_size): New variable.
996 (get_max_overall_size): New function.
997 (estimate_local_effects): Use it. Adjust dump.
998 (decide_about_value): Likewise.
999 (ipcp_propagate_stage): Do not calculate max_new_size, just store
1000 orig_overall_size. Adjust dump.
1001 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
1003 2020-01-10 Martin Jambor <mjambor@suse.cz>
1005 * params.opt (param_ipa_max_agg_items): Mark as Optimization
1006 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
1007 instead of param_ipa_max_agg_items.
1008 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
1009 optimization info for the callee.
1011 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
1013 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
1014 markers if debug_inline_points is false.
1016 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1018 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
1020 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
1021 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
1022 aarch64-sve-builtins-sve2.h.
1023 (aarch64-sve-builtins-sve2.o): New rule.
1024 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
1025 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
1026 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
1027 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
1028 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
1029 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
1031 * config/aarch64/aarch64-sve.md: Update comments with SVE2
1032 instructions that are handled here.
1033 (@cond_asrd<mode>): Generalize to...
1034 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
1035 (*cond_asrd<mode>_2): Generalize to...
1036 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
1037 (*cond_asrd<mode>_z): Generalize to...
1038 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
1039 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
1040 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
1041 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
1042 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
1044 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
1045 (@aarch64_scatter_stnt<mode>): Likewise.
1046 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
1047 (@aarch64_mul_lane_<mode>): Likewise.
1048 (@aarch64_sve_suqadd<mode>_const): Likewise.
1049 (*<sur>h<addsub><mode>): Generalize to...
1050 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
1052 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
1053 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
1054 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
1055 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
1056 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
1057 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
1058 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
1059 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
1060 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
1061 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
1062 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
1063 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
1064 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
1065 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
1066 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
1067 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
1068 (@aarch64_sve2_xar<mode>): Likewise.
1069 (@aarch64_sve2_bcax<mode>): Likewise.
1070 (*aarch64_sve2_eor3<mode>): Rename to...
1071 (@aarch64_sve2_eor3<mode>): ...this.
1072 (@aarch64_sve2_bsl<mode>): New expander.
1073 (@aarch64_sve2_nbsl<mode>): Likewise.
1074 (@aarch64_sve2_bsl1n<mode>): Likewise.
1075 (@aarch64_sve2_bsl2n<mode>): Likewise.
1076 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
1077 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
1078 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
1079 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
1080 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
1081 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
1082 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
1083 (<su>mull<bt><Vwide>): Generalize to...
1084 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
1086 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
1087 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
1088 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
1089 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
1090 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
1091 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
1092 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
1093 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
1094 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
1095 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
1096 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
1097 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
1098 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
1099 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
1100 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
1101 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
1102 (<SHRNB:r>shrnb<mode>): Generalize to...
1103 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
1105 (<SHRNT:r>shrnt<mode>): Generalize to...
1106 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
1108 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
1109 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
1110 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
1111 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
1112 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
1113 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
1114 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
1115 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
1116 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
1117 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
1118 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
1119 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
1120 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
1121 (@aarch64_sve2_cvtnt<mode>): Likewise.
1122 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
1123 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
1124 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
1125 (@aarch64_sve2_cvtxnt<mode>): Likewise.
1126 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
1127 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
1128 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
1129 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
1130 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
1131 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
1132 (@aarch64_sve2_pmul<mode>): Likewise.
1133 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
1134 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
1135 (@aarch64_sve2_tbl2<mode>): Likewise.
1136 (@aarch64_sve2_tbx<mode>): Likewise.
1137 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
1138 (@aarch64_sve2_histcnt<mode>): Likewise.
1139 (@aarch64_sve2_histseg<mode>): Likewise.
1140 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
1141 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
1142 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
1143 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
1144 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
1145 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
1146 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
1147 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
1148 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
1149 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
1150 (SVE2_PMULL_PAIR_I): New mode iterators.
1151 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
1152 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
1153 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
1154 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
1155 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
1156 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
1157 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
1158 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
1159 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
1160 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
1161 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
1162 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
1163 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
1164 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
1165 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
1166 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
1167 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
1168 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
1169 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
1170 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
1171 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
1172 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
1173 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
1174 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
1175 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
1176 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
1177 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
1178 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
1179 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
1180 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
1181 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
1183 (VNARROW, Ventype): New mode attributes.
1184 (Vewtype): Handle VNx2DI. Fix typo in comment.
1185 (VDOUBLE): New mode attribute.
1186 (sve_lane_con): Handle VNx8HI.
1187 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
1188 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
1189 (sve_int_op, sve_int_op_rev): Handle the above codes.
1190 (sve_pred_int_rhs2_operand): Likewise.
1191 (MULLBT, SHRNB, SHRNT): Delete.
1192 (SVE_INT_SHIFT_IMM): New int iterator.
1193 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
1194 and UNSPEC_WHILEHS for TARGET_SVE2.
1195 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
1196 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
1197 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
1198 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
1199 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
1200 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
1201 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
1202 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
1203 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
1204 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
1205 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
1206 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
1207 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
1208 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
1209 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
1210 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
1211 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
1212 (optab): Handle the new unspecs.
1213 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
1215 (lr): Handle the new unspecs.
1217 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
1218 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
1219 (sve_int_qsub_op): New int attributes.
1220 (sve_fp_op, rot): Handle the new unspecs.
1221 * config/aarch64/aarch64-sve-builtins.h
1222 (function_resolver::require_matching_pointer_type): Declare.
1223 (function_resolver::resolve_unary): Add an optional boolean argument.
1224 (function_resolver::finish_opt_n_resolution): Add an optional
1225 type_suffix_index argument.
1226 (gimple_folder::redirect_call): Declare.
1227 (gimple_expander::prepare_gather_address_operands): Add an optional
1229 * config/aarch64/aarch64-sve-builtins.cc: Include
1230 aarch64-sve-builtins-sve2.h.
1231 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
1232 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
1233 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
1234 (TYPES_hsd_integer): Use TYPES_hsd_signed.
1235 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
1236 (TYPES_s_unsigned): Likewise.
1237 (TYPES_s_integer): Use TYPES_s_unsigned.
1238 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
1239 (TYPES_sd_integer): Use them.
1240 (TYPES_d_unsigned): New macro.
1241 (TYPES_d_integer): Use it.
1242 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
1243 (TYPES_cvt_narrow): Likewise.
1244 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
1245 (preds_mx): New variable.
1246 (function_builder::add_overloaded_function): Allow the new feature
1247 set to be more restrictive than the original one.
1248 (function_resolver::infer_pointer_type): Remove qualifiers from
1249 the pointer type before printing it.
1250 (function_resolver::require_matching_pointer_type): New function.
1251 (function_resolver::resolve_sv_displacement): Handle functions
1252 that don't support 32-bit vector indices or svint32_t vector offsets.
1253 (function_resolver::finish_opt_n_resolution): Take the inferred type
1254 as a separate argument.
1255 (function_resolver::resolve_unary): Optionally treat all forms in
1256 the same way as normal merging functions.
1257 (gimple_folder::redirect_call): New function.
1258 (function_expander::prepare_gather_address_operands): Add an argument
1259 that says whether scaled forms are available. If they aren't,
1260 handle scaling of vector indices and don't add the extension and
1262 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
1263 fall back to using cond_* instead.
1264 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
1265 Split out the member variables into...
1266 (rtx_code_function_base): ...this new base class.
1267 (rtx_code_function_rotated): Inherit rtx_code_function_base.
1268 (unspec_based_function): Split out the member variables into...
1269 (unspec_based_function_base): ...this new base class.
1270 (unspec_based_function_rotated): Inherit unspec_based_function_base.
1271 (unspec_based_function_exact_insn): New class.
1272 (unspec_based_add_function, unspec_based_add_lane_function)
1273 (unspec_based_lane_function, unspec_based_pred_function)
1274 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
1275 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
1276 (unspec_based_sub_function, unspec_based_sub_lane_function): New
1278 (unspec_based_fused_function): New class.
1279 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
1280 (unspec_based_fused_lane_function): New class.
1281 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
1283 (CODE_FOR_MODE1): New macro.
1284 (fixed_insn_function): New class.
1285 (while_comparison): Likewise.
1286 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
1287 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
1288 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
1289 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
1290 (load_gather_sv_restricted, shift_left_imm_long): Declare.
1291 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
1292 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
1293 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
1294 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
1295 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
1296 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
1297 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
1298 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
1299 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
1300 Also add an initial argument for unary_convert_narrowt, regardless
1301 of the predication type.
1302 (build_32_64): Allow loads and stores to specify MODE_none.
1303 (build_sv_index64, build_sv_uint_offset): New functions.
1304 (long_type_suffix): New function.
1305 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
1306 (binary_imm_long_base, load_gather_sv_base): Likewise.
1307 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
1308 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
1309 (unary_narrowb_base, unary_narrowt_base): Likewise.
1310 (binary_long_lane_def, binary_long_lane): New shape.
1311 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
1312 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
1313 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
1314 (binary_to_uint_def, binary_to_uint): Likewise.
1315 (binary_wide_def, binary_wide): Likewise.
1316 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
1317 (compare_def, compare): Likewise.
1318 (compare_ptr_def, compare_ptr): Likewise.
1319 (load_ext_gather_index_restricted_def,
1320 load_ext_gather_index_restricted): Likewise.
1321 (load_ext_gather_offset_restricted_def,
1322 load_ext_gather_offset_restricted): Likewise.
1323 (load_gather_sv_def): Inherit from load_gather_sv_base.
1324 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
1325 (shift_left_imm_def, shift_left_imm): Likewise.
1326 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
1327 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
1328 (store_scatter_index_restricted_def,
1329 store_scatter_index_restricted): Likewise.
1330 (store_scatter_offset_restricted_def,
1331 store_scatter_offset_restricted): Likewise.
1332 (tbl_tuple_def, tbl_tuple): Likewise.
1333 (ternary_long_lane_def, ternary_long_lane): Likewise.
1334 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
1335 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
1336 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
1337 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
1338 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
1339 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
1340 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
1341 (ternary_uint_def, ternary_uint): Likewise.
1342 (unary_convert): Fix typo in comment.
1343 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
1344 (unary_long_def, unary_long): Likewise.
1345 (unary_narrowb_def, unary_narrowb): Likewise.
1346 (unary_narrowt_def, unary_narrowt): Likewise.
1347 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
1348 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
1349 (unary_to_int_def, unary_to_int): Likewise.
1350 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
1351 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
1352 (svasrd_impl): Delete.
1353 (svcadd_impl::expand): Handle integer operations too.
1354 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
1355 new functions to derive the unspec numbers.
1356 (svmla_svmls_lane_impl): Replace with...
1357 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
1358 integer operations too.
1359 (svwhile_impl): Rename to...
1360 (svwhilelx_impl): ...this and inherit from while_comparison.
1361 (svasrd): Use unspec_based_function.
1362 (svmla_lane): Use svmla_lane_impl.
1363 (svmls_lane): Use svmls_lane_impl.
1364 (svrecpe, svrsqrte): Handle unsigned integer operations too.
1365 (svwhilele, svwhilelt): Use svwhilelx_impl.
1366 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
1367 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
1368 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
1369 * config/aarch64/aarch64-sve-builtins.def: Include
1370 aarch64-sve-builtins-sve2.def.
1372 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1374 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
1375 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
1376 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
1377 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
1378 immediates as well as vector ones.
1379 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
1380 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
1381 (aarch64_sve_qsub_immediate): Update calls accordingly.
1383 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1385 * config/aarch64/aarch64-sve2.md: Add banner comments.
1386 (<su>mulh<r>s<mode>3): Move further up file.
1387 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
1388 (*aarch64_sve2_sra<mode>): Move further down file.
1389 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
1391 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1393 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
1395 (while_optab_cmp): Handle them.
1396 * config/aarch64/aarch64-sve.md
1397 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
1398 and add a "@" marker.
1399 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
1400 instead of gen_aarch64_sve2_while_ptest.
1401 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
1403 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1405 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
1406 (UNSPEC_WHILELE): ...this.
1407 (UNSPEC_WHILE_LO): Rename to...
1408 (UNSPEC_WHILELO): ...this.
1409 (UNSPEC_WHILE_LS): Rename to...
1410 (UNSPEC_WHILELS): ...this.
1411 (UNSPEC_WHILE_LT): Rename to...
1412 (UNSPEC_WHILELT): ...this.
1413 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
1414 (cmp_op, while_optab_cmp): Likewise.
1415 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
1416 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
1417 (svwhilelt): Likewise.
1419 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1421 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
1422 (unary_to_uint): Define.
1423 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
1424 (unary_count): Rename to...
1425 (unary_to_uint_def, unary_to_uint): ...this.
1426 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
1428 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1430 * config/aarch64/aarch64-sve-builtins-functions.h
1431 (code_for_mode_function): New class.
1432 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
1433 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
1434 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
1435 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
1436 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
1438 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1440 * config/aarch64/iterators.md (addsub): New code attribute.
1441 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
1443 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
1444 in the asm string and attributes. Fix indentation.
1445 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
1447 (@aarch64_sve_<optab><mode>): ...this.
1448 * config/aarch64/aarch64-sve-builtins.h
1449 (function_expander::expand_signed_unpred_op): Delete.
1450 * config/aarch64/aarch64-sve-builtins.cc
1451 (function_expander::expand_signed_unpred_op): Likewise.
1452 (function_expander::map_to_rtx_codes): If the optab isn't defined,
1453 try using code_for_aarch64_sve instead.
1454 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
1455 (svqsub_impl): Likewise.
1456 (svqadd, svqsub): Use rtx_code_function instead.
1458 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1460 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
1461 (HADDSUB, sur, addsub): Remove them.
1463 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1465 * tree-nrv.c (pass_return_slot::execute): Handle all internal
1466 functions the same way, rather than singling out those that
1467 aren't mapped directly to optabs.
1469 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
1471 * target.def (compatible_vector_types_p): New target hook.
1472 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
1473 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
1474 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
1475 * doc/tm.texi: Regenerate.
1476 * gimple-expr.c: Include target.h.
1477 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
1478 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
1480 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
1481 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
1482 Use the original predicate if it already has a suitable type.
1484 2020-01-09 Martin Jambor <mjambor@suse.cz>
1486 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
1487 resolve_speculation and redirect_call_stmt_to_callee static. Change
1488 return type of set_call_stmt to cgraph_edge *.
1489 * auto-profile.c (afdo_indirect_call): Adjust call to
1490 redirect_call_stmt_to_callee.
1491 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
1492 make the this pointer explicit, adjust self-recursive calls and the
1493 call top make_direct. Return the resulting edge.
1494 (cgraph_edge::remove): Make this pointer explicit.
1495 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
1496 (cgraph_edge::make_direct): Likewise, adjust call to
1497 resolve_speculation.
1498 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
1499 call to set_call_stmt.
1500 (cgraph_update_edges_for_call_stmt_node): Update call to
1501 set_call_stmt and remove.
1502 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1503 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
1504 (cgraph_node::create_edge_including_clones): Moved "first" definition
1505 of edge to the block where it was used. Adjusted calls to
1507 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
1508 cgraph_edge::remove.
1509 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
1510 make_direct and redirect_call_stmt_to_callee.
1511 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
1512 resolve_speculation and make_direct.
1513 * ipa-inline-transform.c (inline_transform): Adjust call to
1514 redirect_call_stmt_to_callee.
1515 (check_speculations_1):: Adjust call to resolve_speculation.
1516 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
1517 resolve-speculation.
1518 (inline_small_functions): Adjust call to resolve_speculation.
1519 (ipa_inline): Likewise.
1520 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
1522 * ipa-visibility.c (function_and_variable_visibility): Make iteration
1523 safe with regards to edge removal, adjust calls to
1524 redirect_call_stmt_to_callee.
1525 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
1526 and redirect_call_stmt_to_callee.
1527 * multiple_target.c (create_dispatcher_calls): Adjust call to
1528 redirect_call_stmt_to_callee
1529 (redirect_to_specific_clone): Likewise.
1530 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
1531 Adjust calls to cgraph_edge::remove.
1532 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
1533 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
1534 (expand_call_inline): Adjust call to cgraph_edge::remove.
1536 2020-01-09 Martin Liska <mliska@suse.cz>
1538 * params.opt: Set Optimization for
1539 param_max_speculative_devirt_maydefs.
1541 2020-01-09 Martin Sebor <msebor@redhat.com>
1545 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
1547 2020-01-09 Martin Liska <mliska@suse.cz>
1549 * auto-profile.c (auto_profile): Use opt_for_fn
1551 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
1552 (propagate_vals_across_arith_jfunc): Likewise.
1553 (hint_time_bonus): Likewise.
1554 (incorporate_penalties): Likewise.
1555 (good_cloning_opportunity_p): Likewise.
1556 (perform_estimation_of_a_value): Likewise.
1557 (estimate_local_effects): Likewise.
1558 (ipcp_propagate_stage): Likewise.
1559 * ipa-fnsummary.c (decompose_param_expr): Likewise.
1560 (set_switch_stmt_execution_predicate): Likewise.
1561 (analyze_function_body): Likewise.
1562 * ipa-inline-analysis.c (offline_size): Likewise.
1563 * ipa-inline.c (early_inliner): Likewise.
1564 * ipa-prop.c (ipa_analyze_node): Likewise.
1565 (ipcp_transform_function): Likewise.
1566 * ipa-sra.c (process_scan_results): Likewise.
1567 (ipa_sra_summarize_function): Likewise.
1568 * params.opt: Rename ipcp-unit-growth to
1569 ipa-cp-unit-growth. Add Optimization for various
1570 IPA-related parameters.
1572 2020-01-09 Richard Biener <rguenther@suse.de>
1575 * gimplify.c (gimplify_expr): Deal with NOP definitions.
1577 2020-01-09 Richard Biener <rguenther@suse.de>
1579 PR tree-optimization/93040
1580 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
1582 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
1584 * common/config/avr/avr-common.c (avr_option_optimization_table)
1585 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
1587 2020-01-09 Martin Liska <mliska@suse.cz>
1589 * cgraphclones.c (symbol_table::materialize_all_clones):
1590 Use cgraph_node::dump_name.
1592 2020-01-09 Jakub Jelinek <jakub@redhat.com>
1595 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
1596 output_operand_lossage instead of gcc_unreachable.
1597 * doc/md.texi (riscv f constraint): Fix typo.
1600 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
1601 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
1602 CONST_SCALAR_INT_P instead of CONST_INT_P.
1603 (*subv<mode>4_1): Rename to ...
1604 (subv<mode>4_1): ... this.
1605 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
1606 define_insn_and_split patterns.
1607 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
1610 2020-01-08 David Malcolm <dmalcolm@redhat.com>
1612 * vec.c (class selftest::count_dtor): New class.
1613 (selftest::test_auto_delete_vec): New test.
1614 (selftest::vec_c_tests): Call it.
1615 * vec.h (class auto_delete_vec): New class template.
1616 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
1618 2020-01-08 David Malcolm <dmalcolm@redhat.com>
1620 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
1622 2020-01-08 Jim Wilson <jimw@sifive.com>
1624 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
1625 use of TLS_MODEL_LOCAL_EXEC when not pic.
1627 2020-01-08 David Malcolm <dmalcolm@redhat.com>
1629 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
1632 2020-01-08 Jakub Jelinek <jakub@redhat.com>
1635 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
1636 *stack_protect_set_3 peephole2): Also check that the second
1637 insns source is general_operand.
1640 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
1641 predicate for output operand instead of register_operand.
1642 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
1643 memory destination and non-memory operands[2].
1645 2020-01-08 Martin Liska <mliska@suse.cz>
1647 * cgraph.c (cgraph_node::dump): Use ::dump_name or
1648 ::dump_asm_name instead of (::name or ::asm_name).
1649 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
1650 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
1651 (analyze_functions): Likewise.
1652 (expand_all_functions): Likewise.
1653 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1654 (propagate_bits_across_jump_function): Likewise.
1655 (dump_profile_updates): Likewise.
1656 (ipcp_store_bits_results): Likewise.
1657 (ipcp_store_vr_results): Likewise.
1658 * ipa-devirt.c (dump_targets): Likewise.
1659 * ipa-fnsummary.c (analyze_function_body): Likewise.
1660 * ipa-hsa.c (check_warn_node_versionable): Likewise.
1661 (process_hsa_functions): Likewise.
1662 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
1663 (set_alias_uids): Likewise.
1664 * ipa-inline-transform.c (save_inline_function_body): Likewise.
1665 * ipa-inline.c (recursive_inlining): Likewise.
1666 (inline_to_all_callers_1): Likewise.
1667 (ipa_inline): Likewise.
1668 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1669 (ipa_propagate_frequency): Likewise.
1670 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1671 (remove_described_reference): Likewise.
1672 * ipa-pure-const.c (worse_state): Likewise.
1673 (check_retval_uses): Likewise.
1674 (analyze_function): Likewise.
1675 (propagate_pure_const): Likewise.
1676 (propagate_nothrow): Likewise.
1677 (dump_malloc_lattice): Likewise.
1678 (propagate_malloc): Likewise.
1679 (pass_local_pure_const::execute): Likewise.
1680 * ipa-visibility.c (optimize_weakref): Likewise.
1681 (function_and_variable_visibility): Likewise.
1682 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
1683 (ipa_discover_variable_flags): Likewise.
1684 * lto-streamer-out.c (output_function): Likewise.
1685 (output_constructor): Likewise.
1686 * tree-inline.c (copy_bb): Likewise.
1687 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
1688 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
1690 2020-01-08 Richard Biener <rguenther@suse.de>
1693 * tree-eh.c (sink_clobbers): Update virtual operands for
1694 the first and last stmt only. Add a dry-run capability.
1695 (pass_lower_eh_dispatch::execute): Perform clobber sinking
1696 after CFG manipulations and in RPO order to catch all
1697 secondary opportunities reliably.
1699 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
1702 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
1704 2019-01-08 Richard Biener <rguenther@suse.de>
1707 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
1708 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
1709 virtual operand, also updating SSA use.
1710 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
1711 Update stmt after resetting virtual operand.
1712 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
1713 * gimple-iterator.c (gsi_remove): When not removing the stmt
1714 permanently do not delink immediate uses or mark the stmt modified.
1716 2020-01-08 Martin Liska <mliska@suse.cz>
1718 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
1719 (ipa_call_context::estimate_size_and_time): Likewise.
1720 (inline_analyze_function): Likewise.
1722 2020-01-08 Martin Liska <mliska@suse.cz>
1724 * cgraph.c (cgraph_node::dump): Use systematically
1727 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
1729 Add -nodevicespecs option for avr.
1732 * config/avr/avr.opt (-nodevicespecs): New driver option.
1733 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
1734 "-specs=device-specs/..." if that option is not set.
1735 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
1737 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
1739 Implement 64-bit double functions for avr.
1742 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
1743 --with-double-comparison.
1744 * doc/install.texi: Document them.
1745 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
1746 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
1747 <WITH_DOUBLE_COMPARISON>: New built-in defines.
1748 * doc/invoke.texi (AVR Built-in Macros): Document them.
1749 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
1750 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
1751 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
1753 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
1756 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
1757 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
1758 when only building rm-profile multilibs.
1760 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
1763 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
1764 lattice for a value to check.
1765 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
1766 finite propagation in self-recursive scc.
1768 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
1770 * ipa-inline.c (caller_growth_limits): Restore the AND.
1772 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
1774 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
1775 (VEC_ALLREG_ALT): New iterator.
1776 (VEC_ALLREG_INT_MODE): New iterator.
1777 (VCMP_MODE): New iterator.
1778 (VCMP_MODE_INT): New iterator.
1779 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
1780 (vec_cmp<u>v64qidi): New define_expand.
1781 (vec_cmp<mode>di_exec): Use VCMP_MODE.
1782 (vec_cmpu<mode>di_exec): New define_expand.
1783 (vec_cmp<u>v64qidi_exec): New define_expand.
1784 (vec_cmp<mode>di_dup): Use VCMP_MODE.
1785 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
1786 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
1787 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
1788 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
1789 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
1790 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
1791 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
1792 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
1793 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
1795 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
1796 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
1798 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
1800 * config/gcn/constraints.md (DA): Update description and match.
1802 (Db): New constraint.
1803 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
1805 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
1806 Implement 'Db' mixed immediate type.
1807 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
1808 (addcv64si3_dup<exec_vcc>): Delete.
1809 (subcv64si3<exec_vcc>): Rework constraints.
1810 (addv64di3): Rework constraints.
1811 (addv64di3_exec): Rework constraints.
1812 (subv64di3): Rework constraints.
1813 (addv64di3_dup): Delete.
1814 (addv64di3_dup_exec): Delete.
1815 (addv64di3_zext): Rework constraints.
1816 (addv64di3_zext_exec): Rework constraints.
1817 (addv64di3_zext_dup): Rework constraints.
1818 (addv64di3_zext_dup_exec): Rework constraints.
1819 (addv64di3_zext_dup2): Rework constraints.
1820 (addv64di3_zext_dup2_exec): Rework constraints.
1821 (addv64di3_sext_dup2): Rework constraints.
1822 (addv64di3_sext_dup2_exec): Rework constraints.
1824 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
1826 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
1827 existing target checks.
1829 2020-01-07 Richard Biener <rguenther@suse.de>
1831 * doc/install.texi: Bump minimal supported MPC version.
1833 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
1835 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
1836 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
1837 * langhooks.c: Include stor-layout.h.
1838 (lhd_simulate_enum_decl): New function.
1839 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
1840 handle_arm_sve_h for the LTO frontend.
1841 (register_vector_type): Cope with null returns from pushdecl.
1843 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
1845 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
1846 (aarch64_sve::nvectors_if_data_type): Replace with...
1847 (aarch64_sve::builtin_type_p): ...this.
1848 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
1849 (find_vector_type): Delete.
1850 (add_sve_type_attribute): New function.
1851 (lookup_sve_type_attribute): Likewise.
1852 (register_builtin_types): Add an "SVE type" attribute to each type.
1853 (register_tuple_type): Likewise.
1854 (svbool_type_p, nvectors_if_data_type): Delete.
1855 (mangle_builtin_type): Use lookup_sve_type_attribute.
1856 (builtin_type_p): Likewise. Add an overload that returns the
1857 number of constituent vector and predicate registers.
1858 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
1859 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
1860 instead of aarch64_sve_argument_p.
1861 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
1862 (aarch64_pass_by_reference): Likewise.
1863 (aarch64_function_value_1): Likewise.
1864 (aarch64_return_in_memory): Likewise.
1865 (aarch64_layout_arg): Likewise.
1867 2020-01-07 Jakub Jelinek <jakub@redhat.com>
1869 PR tree-optimization/93156
1870 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
1871 least significant bit is always clear.
1873 PR tree-optimization/93118
1874 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
1875 simplifier with two intermediate conversions.
1877 2020-01-07 Martin Liska <mliska@suse.cz>
1879 * params.opt: Add Optimization for various parameters.
1881 2020-01-07 Martin Liska <mliska@suse.cz>
1884 * doc/extend.texi: Explain cloning for target_clone
1887 2020-01-07 Martin Liska <mliska@suse.cz>
1889 PR tree-optimization/92860
1890 * common.opt: Make in Optimization option
1891 as it is affected by -O0, which is an Optimization
1893 * tree-inline.c (tree_inlinable_function_p):
1894 Use opt_for_fn for warn_inline.
1895 (expand_call_inline): Likewise.
1897 2020-01-07 Martin Liska <mliska@suse.cz>
1899 PR tree-optimization/92860
1900 * common.opt: Make flag_ree as optimization
1903 2020-01-07 Martin Liska <mliska@suse.cz>
1905 PR optimization/92860
1906 * params.opt: Mark param_min_crossjump_insns with Optimization
1909 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
1911 * ipa-inline-analysis.c (estimate_growth): Fix typo.
1912 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
1914 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
1916 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
1917 helper function to return the valid addressing formats for a given
1918 hard register and mode.
1919 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
1921 * config/rs6000/constraints.md (Q constraint): Update
1923 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
1926 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
1927 Use 'Q' for doing vector extract from memory.
1928 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
1930 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
1931 doing vector extract from memory.
1932 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
1933 extract from memory.
1935 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
1936 for the offset being 34-bits when -mcpu=future is used.
1938 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
1940 * config/pa/pa.md: Revert change to use ordered_comparison_operator
1941 instead of cmpib_comparison_operator in cmpib patterns.
1942 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
1943 of cmpib_comparison_operator. Revise comment.
1945 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1947 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
1948 in an IFN_DIV_POW2 node to be equal.
1950 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1952 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
1953 (vect_check_scalar_mask): ...this.
1954 (vectorizable_store, vectorizable_load): Update call accordingly.
1955 (vectorizable_call): Use vect_check_scalar_mask to check the mask
1956 argument in calls to conditional internal functions.
1958 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1960 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
1961 '0' matching inputs.
1962 (subv64di3_exec): Likewise.
1964 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
1966 * config/mips/mips.c (vr4130_align_insns): Fix typo.
1967 * doc/md.texi (movstr): Likewise.
1969 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1971 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
1974 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1976 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
1978 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
1979 to a temporary file and use move-if-change to update the real
1980 file where necessary.
1982 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1984 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
1985 rather than Upa for CPY /M.
1987 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1989 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
1992 2020-01-06 Martin Liska <mliska@suse.cz>
1994 PR tree-optimization/92860
1995 * params.opt: Mark param_max_combine_insns with Optimization
1998 2020-01-05 Jakub Jelinek <jakub@redhat.com>
2001 * config/i386/i386.md (SWIDWI): New mode iterator.
2002 (DWI, dwi): Add TImode variants.
2003 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
2004 <general_hilo_operand> instead of <general_operand>. Use
2005 CONST_SCALAR_INT_P instead of CONST_INT_P.
2006 (*addv<mode>4_1): Rename to ...
2007 (addv<mode>4_1): ... this.
2008 (QWI): New mode attribute.
2009 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
2010 define_insn_and_split patterns.
2011 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
2013 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
2014 <general_hilo_operand> instead of <general_operand>.
2015 (*addcarry<mode>_1): New define_insn.
2016 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
2018 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
2020 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
2021 Use "call" instead of "set".
2023 2020-01-03 Martin Jambor <mjambor@suse.cz>
2026 * ipa-cp.c (print_all_lattices): Skip functions without info.
2028 2020-01-03 Jakub Jelinek <jakub@redhat.com>
2031 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
2032 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
2033 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
2034 for 'e' simd clones.
2037 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
2039 (mprefer-vector-width=): Add Save.
2040 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
2041 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
2042 (ix86_debug_options, ix86_function_specific_print): Adjust
2043 ix86_target_string callers.
2044 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
2045 (ix86_valid_target_attribute_tree): Likewise.
2046 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
2047 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
2048 ix86_target_string caller.
2051 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
2052 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
2053 instead of gen_int_shift_amount + convert_modes.
2055 PR rtl-optimization/93088
2056 * loop-iv.c (find_single_def_src): Punt after looking through
2057 128 reg copies for regs with single definitions. Move definitions
2060 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
2062 * config/arm/arm-c.c (arm_cpu_builtins): Define
2063 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
2064 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
2065 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
2066 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
2067 * config/arm/arm-tables.opt: Regenerated.
2068 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
2069 arm_arch_i8mm and arm_arch_bf16 when enabled.
2070 * config/arm/arm.h (TARGET_I8MM): New macro.
2071 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
2072 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
2073 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
2074 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
2075 (v8_6_a_simd_variants): New.
2076 (v8_*_a_simd_variants): Add i8mm and bf16.
2077 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
2079 2020-01-02 Jakub Jelinek <jakub@redhat.com>
2082 * predict.c (compute_function_frequency): Don't call
2083 warn_function_cold on functions that already have cold attribute.
2085 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
2088 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
2089 COMDAT group function labels in .data.rel.ro.local section.
2090 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
2093 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
2094 comparison_operator in B and S integer comparisons. Likewise, use
2095 ordered_comparison_operator instead of cmpib_comparison_operator in
2097 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
2099 2020-01-01 Jakub Jelinek <jakub@redhat.com>
2101 Update copyright years.
2103 * gcc.c (process_command): Update copyright notice dates.
2104 * gcov-dump.c (print_version): Ditto.
2105 * gcov.c (print_version): Ditto.
2106 * gcov-tool.c (print_version): Ditto.
2107 * gengtype.c (create_file): Ditto.
2108 * doc/cpp.texi: Bump @copying's copyright year.
2109 * doc/cppinternals.texi: Ditto.
2110 * doc/gcc.texi: Ditto.
2111 * doc/gccint.texi: Ditto.
2112 * doc/gcov.texi: Ditto.
2113 * doc/install.texi: Ditto.
2114 * doc/invoke.texi: Ditto.
2116 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
2118 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
2121 2020-01-01 Jakub Jelinek <jakub@redhat.com>
2123 PR tree-optimization/93098
2124 * match.pd (popcount): For shift amounts, use integer_onep
2125 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
2126 tests. Make sure that precision is power of two larger than or equal
2127 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
2128 instead of ULL suffixed constants. Formatting fixes.
2130 Copyright (C) 2020 Free Software Foundation, Inc.
2132 Copying and distribution of this file, with or without modification,
2133 are permitted in any medium without royalty provided the copyright
2134 notice and this notice are preserved.