]> gcc.gnu.org Git - gcc.git/blob - gcc/ChangeLog
[PATCH, GCC/ARM] Fix MVE scalar shift tests
[gcc.git] / gcc / ChangeLog
1 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2
3 * config/arm/arm.md: Prevent scalar shifts from being used when big
4 endian is enabled.
5
6 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
7 Richard Biener <rguenther@suse.de>
8
9 PR tree-optimization/93586
10 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
11 after mismatched array refs; do not sure type size information to
12 recover from unmatched referneces with !flag_strict_aliasing_p.
13
14 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
15
16 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
17 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
18 (scatter_store<mode>): Rename to ...
19 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
20 (scatter<mode>_exec): Delete. Move contents ...
21 (mask_scatter_store<mode>): ... here, and rename that to ...
22 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
23 Remove mode conversion.
24 (mask_gather_load<mode>): Rename to ...
25 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
26 Remove mode conversion.
27 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
28
29 2020-02-21 Martin Jambor <mjambor@suse.cz>
30
31 PR tree-optimization/93845
32 * tree-sra.c (verify_sra_access_forest): Only test access size of
33 scalar types.
34
35 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
36
37 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
38 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
39 (addv64di3_exec): Likewise.
40 (subv64di3): Likewise.
41 (subv64di3_exec): Likewise.
42 (addv64di3_zext): Likewise.
43 (addv64di3_zext_exec): Likewise.
44 (addv64di3_zext_dup): Likewise.
45 (addv64di3_zext_dup_exec): Likewise.
46 (addv64di3_zext_dup2): Likewise.
47 (addv64di3_zext_dup2_exec): Likewise.
48 (addv64di3_sext_dup2): Likewise.
49 (addv64di3_sext_dup2_exec): Likewise.
50 (<expander>v64di3): Likewise.
51 (<expander>v64di3_exec): Likewise.
52 (*<reduc_op>_dpp_shr_v64di): Likewise.
53 (*plus_carry_dpp_shr_v64di): Likewise.
54 * config/gcn/gcn.md (adddi3): Likewise.
55 (addptrdi3): Likewise.
56 (<expander>di3): Likewise.
57
58 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
59
60 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
61
62 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
63
64 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
65 support. Use aarch64_emit_mult instead of emitting multiplication
66 instructions directly.
67 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
68 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
69
70 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
71
72 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
73 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
74 instead of emitting multiplication instructions directly.
75 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
76 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
77 (@aarch64_frecps<mode>): New expanders.
78
79 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
80
81 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
82 on and produce uint64_ts rather than ints.
83 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
84 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
85
86 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
87
88 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
89 an unused xmsk register when handling approximate rsqrt.
90
91 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
92
93 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
94 flag_finite_math_only condition.
95
96 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
97
98 PR target/93828
99 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
100 to destination operand for shufps alternative.
101 (*vec_extractv2si_1): Ditto.
102
103 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
104
105 PR target/93658
106 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
107 vector modes.
108
109 2020-02-20 Martin Liska <mliska@suse.cz>
110
111 PR translation/93831
112 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
113
114 2020-02-20 Martin Liska <mliska@suse.cz>
115
116 PR translation/93830
117 * common/config/avr/avr-common.c: Remote trailing "|".
118
119 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
120
121 * collect2.c (maybe_run_lto_and_relink): Fix typo in
122 comment.
123
124 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
125
126 PR tree-optimization/93767
127 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
128 access-size bias from the offset calculations for negative strides.
129
130 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
131
132 * collect2.c (c_file, o_file): Make const again.
133 (ldout,lderrout, dump_ld_file): Remove.
134 (tool_cleanup): Avoid calling not signal-safe functions.
135 (maybe_run_lto_and_relink): Avoid possible signal handler
136 access to unintialzed memory (lto_o_files).
137 (main): Avoid leaking temp files in $TMPDIR.
138 Initialize c_file/o_file with concat, which avoids exposing
139 uninitialized memory to signal handler, which calls unlink(!).
140 Avoid calling maybe_unlink when the main function returns,
141 since the atexit handler is already doing this.
142 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
143
144 2020-02-19 Martin Jambor <mjambor@suse.cz>
145
146 PR tree-optimization/93776
147 * tree-sra.c (create_access): Do not create zero size accesses.
148 (get_access_for_expr): Do not search for zero sized accesses.
149
150 2020-02-19 Martin Jambor <mjambor@suse.cz>
151
152 PR tree-optimization/93667
153 * tree-sra.c (scalarizable_type_p): Return false if record fields
154 do not follow wach other.
155
156 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
157
158 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
159 rather than fmv.x.s/fmv.s.x.
160
161 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
162
163 * config/aarch64/aarch64-simd-builtins.def
164 (intrinsic_vec_smult_lo_): New.
165 (intrinsic_vec_umult_lo_): Likewise.
166 (vec_widen_smult_hi_): Likewise.
167 (vec_widen_umult_hi_): Likewise.
168 * config/aarch64/aarch64-simd.md
169 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
170 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
171 (vmull_high_s16): Likewise.
172 (vmull_high_s32): Likewise.
173 (vmull_high_u8): Likewise.
174 (vmull_high_u16): Likewise.
175 (vmull_high_u32): Likewise.
176 (vmull_s8): Likewise.
177 (vmull_s16): Likewise.
178 (vmull_s32): Likewise.
179 (vmull_u8): Likewise.
180 (vmull_u16): Likewise.
181 (vmull_u32): Likewise.
182
183 2020-02-18 Martin Liska <mliska@suse.cz>
184
185 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
186 bootstrap by missing removal of invalid sanity check.
187
188 2020-02-18 Martin Liska <mliska@suse.cz>
189
190 PR ipa/92518
191 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
192 Always compare LHS of gimple_assign.
193
194 2020-02-18 Martin Liska <mliska@suse.cz>
195
196 PR ipa/93583
197 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
198 and return type of functions.
199 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
200 Drop MALLOC attribute for void functions.
201 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
202 malloc_state for a new VOID clone.
203
204 2020-02-18 Martin Liska <mliska@suse.cz>
205
206 PR ipa/92924
207 * common.opt: Add -fprofile-reproducibility.
208 * doc/invoke.texi: Document it.
209 * value-prof.c (dump_histogram_value):
210 Document and support behavior for counters[0]
211 being a negative value.
212 (get_nth_most_common_value): Handle negative
213 counters[0] in respect to flag_profile_reproducible.
214
215 2020-02-18 Jakub Jelinek <jakub@redhat.com>
216
217 PR ipa/93797
218 * cgraph.c (verify_speculative_call): Use speculative_id instead of
219 speculative_uid in messages. Remove trailing whitespace from error
220 message. Use num_speculative_call_targets instead of
221 num_speculative_targets in a message.
222 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
223 edge messages and stmt instead of cal_stmt in reference message.
224
225 PR tree-optimization/93780
226 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
227 before calling build_vector_type.
228 (execute_update_addresses_taken): Likewise.
229
230 PR driver/93796
231 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
232 typo, functoin -> function.
233 * tree.c (free_lang_data_in_decl): Fix comment typo,
234 functoin -> function.
235 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
236
237 2020-02-17 David Malcolm <dmalcolm@redhat.com>
238
239 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
240 won't be printed.
241 (print_option_information): Don't call get_option_url if URLs
242 won't be printed.
243
244 2020-02-17 Alexandre Oliva <oliva@adacore.com>
245
246 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
247 handling of register_common-less targets.
248
249 2020-02-17 Martin Liska <mliska@suse.cz>
250
251 PR ipa/93760
252 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
253
254 2020-02-17 Martin Liska <mliska@suse.cz>
255
256 PR translation/93755
257 * config/rs6000/rs6000.c (rs6000_option_override_internal):
258 Fix double quotes.
259
260 2020-02-17 Martin Liska <mliska@suse.cz>
261
262 PR other/93756
263 * config/rx/elf.opt: Fix typo.
264
265 2020-02-17 Richard Biener <rguenther@suse.de>
266
267 PR c/86134
268 * opts-global.c (print_ignored_options): Use inform and
269 amend message.
270
271 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
272
273 PR target/93047
274 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
275
276 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
277
278 PR target/93743
279 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
280 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
281
282 2020-02-15 Jason Merrill <jason@redhat.com>
283
284 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
285
286 2020-02-15 Jakub Jelinek <jakub@redhat.com>
287
288 PR tree-optimization/93744
289 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
290 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
291 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
292 sure @2 in the first and @1 in the other patterns has no side-effects.
293
294 2020-02-15 David Malcolm <dmalcolm@redhat.com>
295 Bernd Edlinger <bernd.edlinger@hotmail.de>
296
297 PR 87488
298 PR other/93168
299 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
300 * configure.ac (--with-diagnostics-urls): New configuration
301 option, based on --with-diagnostics-color.
302 (DIAGNOSTICS_URLS_DEFAULT): New define.
303 * config.h: Regenerate.
304 * configure: Regenerate.
305 * diagnostic.c (diagnostic_urls_init): Handle -1 for
306 DIAGNOSTICS_URLS_DEFAULT from configure-time
307 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
308 and TERM_URLS environment variable.
309 * diagnostic-url.h (diagnostic_url_format): New enum type.
310 (diagnostic_urls_enabled_p): rename to...
311 (determine_url_format): ... this, and change return type.
312 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
313 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
314 the linux console, and mingw.
315 (diagnostic_urls_enabled_p): rename to...
316 (determine_url_format): ... this, and adjust.
317 * pretty-print.h (pretty_printer::show_urls): rename to...
318 (pretty_printer::url_format): ... this, and change to enum.
319 * pretty-print.c (pretty_printer::pretty_printer,
320 pp_begin_url, pp_end_url, test_urls): Adjust.
321 * doc/install.texi (--with-diagnostics-urls): Document the new
322 configuration option.
323 (--with-diagnostics-color): Document the existing interaction
324 with GCC_COLORS better.
325 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
326 vindex reference. Update description of defaults based on the above.
327 (-fdiagnostics-color): Update description of how -fdiagnostics-color
328 interacts with GCC_COLORS.
329
330 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
331
332 PR target/93704
333 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
334 conjunction with TARGET_GNU_TLS in early return.
335
336 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
337
338 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
339 the mode is not wider than UNITS_PER_WORD.
340
341 2020-02-14 Martin Jambor <mjambor@suse.cz>
342
343 PR tree-optimization/93516
344 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
345 access of the same type as the parent.
346 (propagate_subaccesses_from_lhs): Likewise.
347
348 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
349
350 PR target/93724
351 * config/i386/avx512vbmi2intrin.h
352 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
353 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
354 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
355 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
356 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
357 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
358 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
359 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
360 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
361 of lacking a closing parenthesis.
362 * config/i386/avx512vbmi2vlintrin.h
363 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
364 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
365 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
366 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
367 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
368 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
369 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
370 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
371 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
372 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
373 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
374 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
375 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
376 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
377 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
378 _mm_shldi_epi32, _mm_mask_shldi_epi32,
379 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
380 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
381
382 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
383
384 PR target/93656
385 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
386 the target function entry.
387
388 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
389
390 * common/config/arc/arc-common.c (arc_option_optimization_table):
391 Disable if-conversion step when optimized for size.
392
393 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
394
395 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
396 R12-R15 are always in ARCOMPACT16_REGS register class.
397 * config/arc/arc.opt (mq-class): Deprecate.
398 * config/arc/constraint.md ("q"): Remove dependency on mq-class
399 option.
400 * doc/invoke.texi (mq-class): Update text.
401 * common/config/arc/arc-common.c (arc_option_optimization_table):
402 Update list.
403
404 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
405
406 * config/arc/arc.c (arc_insn_cost): New function.
407 (TARGET_INSN_COST): Define.
408 * config/arc/arc.md (cost): New attribute.
409 (add_n): Use arc_nonmemory_operand.
410 (ashlsi3_insn): Likewise, also update constraints.
411 (ashrsi3_insn): Likewise.
412 (rotrsi3): Likewise.
413 (add_shift): Likewise.
414 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
415
416 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
417
418 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
419 registers.
420 (umulsidi_600): Likewise.
421
422 2020-02-13 Jakub Jelinek <jakub@redhat.com>
423
424 PR target/93696
425 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
426 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
427 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
428 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
429 pass __A to the builtin followed by __W instead of __A followed by
430 __B.
431 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
432 _mm512_mask_popcnt_epi64): Likewise.
433 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
434 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
435 _mm256_mask_popcnt_epi64): Likewise.
436
437 PR tree-optimization/93582
438 * fold-const.h (shift_bytes_in_array_left,
439 shift_bytes_in_array_right): Declare.
440 * fold-const.c (shift_bytes_in_array_left,
441 shift_bytes_in_array_right): New function, moved from
442 gimple-ssa-store-merging.c, no longer static.
443 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
444 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
445 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
446 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
447 shift_bytes_in_array.
448 (verify_shift_bytes_in_array): Rename to ...
449 (verify_shift_bytes_in_array_left): ... this. Use
450 shift_bytes_in_array_left instead of shift_bytes_in_array.
451 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
452 instead of verify_shift_bytes_in_array.
453 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
454 / native_interpret_expr where the store covers all needed bits,
455 punt on PDP-endian, otherwise allow all involved offsets and sizes
456 not to be byte-aligned.
457
458 PR target/93673
459 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
460 use const_0_to_255_operand predicate instead of immediate_operand.
461 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
462 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
463 vgf2p8affineinvqb_<mode><mask_name>,
464 vgf2p8affineqb_<mode><mask_name>): Drop mode from
465 const_0_to_255_operand predicated operands.
466
467 2020-02-12 Jeff Law <law@redhat.com>
468
469 * config/h8300/h8300.md (comparison shortening peepholes): Use
470 a mode iterator to merge the HImode and SImode peepholes.
471
472 2020-02-12 Jakub Jelinek <jakub@redhat.com>
473
474 PR middle-end/93663
475 * real.c (is_even): Make static. Function comment fix.
476 (is_halfway_below): Make static, don't assert R is not inf/nan,
477 instead return false for those. Small formatting fixes.
478
479 2020-02-12 Martin Sebor <msebor@redhat.com>
480
481 PR middle-end/93646
482 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
483 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
484 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
485 (strlen_check_and_optimize_call): Adjust callee name.
486
487 2020-02-12 Jeff Law <law@redhat.com>
488
489 * config/h8300/h8300.md (comparison shortening peepholes): Drop
490 (and (xor)) variant. Combine other two into single peephole.
491
492 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
493
494 PR rtl-optimization/93565
495 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
496
497 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
498
499 * config/aarch64/aarch64-simd.md
500 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
501 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
502 generating separate ADDV and zero_extend patterns.
503 * config/aarch64/iterators.md (VDQV_E): New iterator.
504
505 2020-02-12 Jeff Law <law@redhat.com>
506
507 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
508 expanders, splits, etc.
509 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
510 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
511 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
512 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
513 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
514 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
515 function prototype.
516 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
517
518 2020-02-12 Jakub Jelinek <jakub@redhat.com>
519
520 PR target/93670
521 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
522 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
523 TARGET_AVX512DQ from condition.
524 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
525 instead of <mask_mode512bit_condition> in condition. If
526 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
527 vextract*32x8.
528 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
529 from condition.
530
531 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
532
533 PR target/91052
534 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
535
536 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
537
538 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
539 where strlen is more legible.
540 (rs6000_builtin_vectorized_libmass): Ditto.
541 (rs6000_print_options_internal): Ditto.
542
543 2020-02-11 Martin Sebor <msebor@redhat.com>
544
545 PR tree-optimization/93683
546 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
547
548 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
549
550 * config/rs6000/predicates.md (cint34_operand): Rename the
551 -mprefixed-addr option to be -mprefixed.
552 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
553 the -mprefixed-addr option to be -mprefixed.
554 (OTHER_FUTURE_MASKS): Likewise.
555 (POWERPC_MASKS): Likewise.
556 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
557 the -mprefixed-addr option to be -mprefixed. Change error
558 messages to refer to -mprefixed.
559 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
560 -mprefixed.
561 (rs6000_legitimate_offset_address_p): Likewise.
562 (rs6000_mode_dependent_address): Likewise.
563 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
564 "-mprefixed" for target attributes and pragmas.
565 (address_to_insn_form): Rename the -mprefixed-addr option to be
566 -mprefixed.
567 (rs6000_adjust_insn_length): Likewise.
568 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
569 -mprefixed-addr option to be -mprefixed.
570 (ASM_OUTPUT_OPCODE): Likewise.
571 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
572 -mprefixed-addr option to be -mprefixed.
573 * config/rs6000/rs6000.opt (-mprefixed): Rename the
574 -mprefixed-addr option to be prefixed. Change the option from
575 being undocumented to being documented.
576 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
577 -mprefixed option. Update the -mpcrel documentation to mention
578 -mprefixed.
579
580 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
581
582 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
583 including FIRST_PSEUDO_REGISTER - 1.
584 * ira-color.c (print_hard_reg_set): Ditto.
585
586 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
587
588 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
589 (USTERNOP_QUALIFIERS): New define.
590 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
591 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
592 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
593 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
594 * config/arm/arm_neon.h (vusdot_s32): New.
595 (vusdot_lane_s32): New.
596 (vusdotq_lane_s32): New.
597 (vsudot_lane_s32): New.
598 (vsudotq_lane_s32): New.
599 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
600 * config/arm/iterators.md (DOTPROD_I8MM): New.
601 (sup, opsuffix): Add <us/su>.
602 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
603 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
604
605 2020-02-11 Richard Biener <rguenther@suse.de>
606
607 PR tree-optimization/93661
608 PR tree-optimization/93662
609 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
610 tree_to_poly_int64.
611 * tree-sra.c (get_access_for_expr): Likewise.
612
613 2020-02-10 Jakub Jelinek <jakub@redhat.com>
614
615 PR target/93637
616 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
617 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
618 Change condition from TARGET_AVX2 to TARGET_AVX.
619
620 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
621
622 PR other/93641
623 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
624 argument of strncmp.
625
626 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
627
628 Try to generate zero-based comparisons.
629 * config/cris/cris.c (cris_reduce_compare): New function.
630 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
631 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
632 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
633
634 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
635
636 PR target/91913
637 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
638 in Thumb state and also as a destination in Arm state. Add T16
639 variants.
640
641 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
642
643 * md.texi (Define Subst): Match closing paren in example.
644
645 2020-02-10 Jakub Jelinek <jakub@redhat.com>
646
647 PR target/58218
648 PR other/93641
649 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
650 arguments of strncmp.
651
652 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
653
654 PR ipa/93203
655 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
656 but different source value.
657 (adjust_callers_for_value_intersection): New function.
658 (gather_edges_for_value): Adjust order of callers to let a
659 non-self-recursive caller be the first element.
660 (self_recursive_pass_through_p): Add a new parameter "simple", and
661 check generalized self-recursive pass-through jump function.
662 (self_recursive_agg_pass_through_p): Likewise.
663 (find_more_scalar_values_for_callers_subset): Compute value from
664 pass-through jump function for self-recursive.
665 (intersect_with_plats): Cleanup previous implementation code for value
666 itersection with self-recursive call edge.
667 (intersect_with_agg_replacements): Likewise.
668 (intersect_aggregates_with_edge): Deduce value from pass-through jump
669 function for self-recursive call edge. Cleanup previous implementation
670 code for value intersection with self-recursive call edge.
671 (decide_whether_version_node): Remove dead callers and adjust order
672 to let a non-self-recursive caller be the first element.
673
674 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
675
676 * recog.c: Move pass_split_before_sched2 code in front of
677 pass_split_before_regstack.
678 (pass_data_split_before_sched2): Rename pass to split3 from split4.
679 (pass_data_split_before_regstack): Rename pass to split4 from split3.
680 (rest_of_handle_split_before_sched2): Remove.
681 (pass_split_before_sched2::execute): Unconditionally call
682 split_all_insns.
683 (enable_split_before_sched2): New function.
684 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
685 (pass_split_before_regstack::gate): Ditto.
686 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
687 Update name check for renamed split4 pass.
688 * config/sh/sh.c (register_sh_passes): Update pass insertion
689 point for renamed split4 pass.
690
691 2020-02-09 Jakub Jelinek <jakub@redhat.com>
692
693 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
694 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
695 copying them around between host and target.
696
697 2020-02-08 Andrew Pinski <apinski@marvell.com>
698
699 PR target/91927
700 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
701 STRICT_ALIGNMENT also.
702
703 2020-02-08 Jim Wilson <jimw@sifive.com>
704
705 PR target/93532
706 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
707
708 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
709 Jakub Jelinek <jakub@redhat.com>
710
711 PR target/65782
712 * config/i386/i386.h (CALL_USED_REGISTERS): Make
713 xmm16-xmm31 call-used even in 64-bit ms-abi.
714
715 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
716
717 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
718 (simd_ummla, simd_usmmla): Likewise.
719 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
720 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
721 (vusmmlaq_s32): New.
722
723 2020-02-07 Richard Biener <rguenther@suse.de>
724
725 PR middle-end/93519
726 * tree-inline.c (fold_marked_statements): Do a PRE walk,
727 skipping unreachable regions.
728 (optimize_inline_calls): Skip folding stmts when we didn't
729 inline.
730
731 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
732
733 PR target/85667
734 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
735 Don't return aggregates with only SFmode and DFmode in SSE
736 register.
737 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
738
739 2020-02-07 Jakub Jelinek <jakub@redhat.com>
740
741 PR target/93122
742 * config/rs6000/rs6000-logue.c
743 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
744 if it fails, move rs into end_addr and retry. Add
745 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
746 the insn pattern doesn't describe well what exactly happens to
747 dwarf2cfi.c.
748
749 PR target/93594
750 * config/i386/predicates.md (avx_identity_operand): Remove.
751 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
752 (avx_<castmode><avxsizesuffix>_<castmode>,
753 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
754 a VEC_CONCAT of the operand and UNSPEC_CAST.
755 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
756 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
757 UNSPEC_CAST.
758
759 PR target/93611
760 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
761 recog_data.insn if distance_non_agu_define changed it.
762
763 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
764
765 PR target/93569
766 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
767 we only had X-FORM (reg+reg) addressing for vectors. Also before
768 ISA 3.0, we only had X-FORM addressing for scalars in the
769 traditional Altivec registers.
770
771 2020-02-06 <zhongyunde@huawei.com>
772 Vladimir Makarov <vmakarov@redhat.com>
773
774 PR rtl-optimization/93561
775 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
776 hard register range.
777
778 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
779
780 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
781 attribute.
782
783 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
784
785 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
786 where the low and the high 32 bits are equal to each other specially,
787 with an rldimi instruction.
788
789 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
790
791 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
792
793 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
794
795 * config/arm/arm-tables.opt: Regenerate.
796
797 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
798
799 PR target/87763
800 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
801 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
802 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
803
804 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
805
806 PR rtl-optimization/87763
807 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
808
809 2020-02-06 Delia Burduv <delia.burduv@arm.com>
810
811 * config/aarch64/aarch64-simd-builtins.def
812 (bfmlaq): New built-in function.
813 (bfmlalb): New built-in function.
814 (bfmlalt): New built-in function.
815 (bfmlalb_lane): New built-in function.
816 (bfmlalt_lane): New built-in function.
817 * config/aarch64/aarch64-simd.md
818 (aarch64_bfmmlaqv4sf): New pattern.
819 (aarch64_bfmlal<bt>v4sf): New pattern.
820 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
821 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
822 (vbfmlalbq_f32): New intrinsic.
823 (vbfmlaltq_f32): New intrinsic.
824 (vbfmlalbq_lane_f32): New intrinsic.
825 (vbfmlaltq_lane_f32): New intrinsic.
826 (vbfmlalbq_laneq_f32): New intrinsic.
827 (vbfmlaltq_laneq_f32): New intrinsic.
828 * config/aarch64/iterators.md (BF_MLA): New int iterator.
829 (bt): New int attribute.
830
831 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
832
833 * config/i386/i386.md (*pushtf): Emit "#" instead of
834 calling gcc_unreachable in insn output.
835 (*pushxf): Ditto.
836 (*pushdf): Ditto.
837 (*pushsf_rex64): Ditto for alternatives other than 1.
838 (*pushsf): Ditto for alternatives other than 1.
839
840 2020-02-06 Martin Liska <mliska@suse.cz>
841
842 PR gcov-profile/91971
843 PR gcov-profile/93466
844 * coverage.c (coverage_init): Revert mangling of
845 path into filename. It can lead to huge filename length.
846 Creation of subfolders seem more natural.
847
848 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
849
850 PR target/93300
851 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
852 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
853 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
854
855 2020-02-06 Jakub Jelinek <jakub@redhat.com>
856
857 PR target/93594
858 * config/i386/predicates.md (avx_identity_operand): New predicate.
859 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
860 define_insn_and_split.
861
862 PR libgomp/93515
863 * omp-low.c (use_pointer_for_field): For nested constructs, also
864 look for map clauses on target construct.
865 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
866 taskreg_nesting_level.
867
868 PR libgomp/93515
869 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
870 shared clause, call omp_notice_variable on outer context if any.
871
872 2020-02-05 Jason Merrill <jason@redhat.com>
873
874 PR c++/92003
875 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
876 non-zero address even if weak and not yet defined.
877
878 2020-02-05 Martin Sebor <msebor@redhat.com>
879
880 PR tree-optimization/92765
881 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
882 * tree-ssa-strlen.c (compute_string_length): Remove.
883 (determine_min_objsize): Remove.
884 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
885 Avoid using type size as the upper bound on string length.
886 (handle_builtin_string_cmp): Add an argument. Adjust.
887 (strlen_check_and_optimize_call): Pass additional argument to
888 handle_builtin_string_cmp.
889
890 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
891
892 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
893 (*pushdi2_rex64 peephole2): Unconditionally split after
894 epilogue_completed.
895 (*ashl<mode>3_doubleword): Ditto.
896 (*<shift_insn><mode>3_doubleword): Ditto.
897
898 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
899
900 PR target/93568
901 * config/rs6000/rs6000.c (get_vector_offset): Fix
902
903 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
904
905 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
906
907 2020-02-05 David Malcolm <dmalcolm@redhat.com>
908
909 * doc/analyzer.texi
910 (Special Functions for Debugging the Analyzer): Update description
911 of __analyzer_dump_exploded_nodes.
912
913 2020-02-05 Jakub Jelinek <jakub@redhat.com>
914
915 PR target/92190
916 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
917 include sets and not clobbers in the vzeroupper pattern.
918 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
919 the parallel has 17 (64-bit) or 9 (32-bit) elts.
920 (*avx_vzeroupper_1): New define_insn_and_split.
921
922 PR target/92190
923 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
924 don't run when !optimize.
925 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
926 when !optimize.
927
928 2020-02-05 Richard Biener <rguenther@suse.de>
929
930 PR middle-end/90648
931 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
932 checks before matching calls.
933
934 2020-02-05 Jakub Jelinek <jakub@redhat.com>
935
936 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
937 function comment typo.
938
939 PR middle-end/93555
940 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
941 simd_clone_create failed when i == 0, adjust clone->nargs by
942 clone->inbranch.
943
944 2020-02-05 Martin Liska <mliska@suse.cz>
945
946 PR c++/92717
947 * doc/invoke.texi: Document that one should
948 not combine ASLR and -fpch.
949
950 2020-02-04 Richard Biener <rguenther@suse.de>
951
952 PR tree-optimization/93538
953 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
954
955 2020-02-04 Richard Biener <rguenther@suse.de>
956
957 PR tree-optimization/91123
958 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
959 (vn_walk_cb_data::last_vuse): New member.
960 (vn_walk_cb_data::saved_operands): Likewsie.
961 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
962 (vn_walk_cb_data::push_partial_def): Use finish.
963 (vn_reference_lookup_2): Update last_vuse and use finish if
964 we've saved operands.
965 (vn_reference_lookup_3): Use finish and update calls to
966 push_partial_defs everywhere. When translating through
967 memcpy or aggregate copies save off operands and alias-set.
968 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
969 operation for redundant store removal.
970
971 2020-02-04 Richard Biener <rguenther@suse.de>
972
973 PR tree-optimization/92819
974 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
975 generating more stmts than before.
976
977 2020-02-04 Martin Liska <mliska@suse.cz>
978
979 * config/arm/arm.c (arm_gen_far_branch): Move the function
980 outside of selftests.
981
982 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
983
984 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
985 function to adjust PC-relative vector addresses.
986 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
987 handle vectors with PC-relative addresses.
988
989 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
990
991 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
992 reference.
993 (hard_reg_and_mode_to_addr_mask): Delete.
994 (rs6000_adjust_vec_address): If the original vector address
995 was REG+REG or REG+OFFSET and the element is not zero, do the add
996 of the elements in the original address before adding the offset
997 for the vector element. Use address_to_insn_form to validate the
998 address using the register being loaded, rather than guessing
999 whether the address is a DS-FORM or DQ-FORM address.
1000
1001 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1002
1003 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1004 to calculate the offset in memory from the start of a vector of a
1005 particular element. Add code to keep the element number in
1006 bounds if the element number is variable.
1007 (rs6000_adjust_vec_address): Move calculation of offset of the
1008 vector element to get_vector_offset.
1009 (rs6000_split_vec_extract_var): Do not do the initial AND of
1010 element here, move the code to get_vector_offset.
1011
1012 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1013
1014 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1015 gcc_asserts.
1016
1017 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1018
1019 * config/rs6000/constraints.md: Improve documentation.
1020
1021 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1022
1023 PR target/93548
1024 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1025 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1026
1027 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1028
1029 * config.gcc: Remove "carrizo" support.
1030 * config/gcn/gcn-opts.h (processor_type): Likewise.
1031 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1032 * config/gcn/gcn.opt (gpu_type): Likewise.
1033 * config/gcn/t-omp-device: Likewise.
1034
1035 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1036
1037 PR target/91816
1038 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1039 * config/arm/arm.c (arm_gen_far_branch): New function
1040 arm_gen_far_branch.
1041 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1042
1043 2020-02-03 Julian Brown <julian@codesourcery.com>
1044 Tobias Burnus <tobias@codesourcery.com>
1045
1046 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1047
1048 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1049
1050 PR target/93533
1051 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1052 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1053 result.
1054
1055 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1056
1057 PR rtl-optimization/91333
1058 * ira-color.c (struct allocno_color_data): Add member
1059 hard_reg_prefs.
1060 (init_allocno_threads): Set the member up.
1061 (bucket_allocno_compare_func): Add compare hard reg
1062 prefs.
1063
1064 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1065
1066 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1067
1068 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1069 * config.in: Regenerated.
1070 * configure: Regenerated.
1071 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1072 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1073 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1074
1075 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1076
1077 * configure: Regenerate.
1078
1079 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1080
1081 PR rtl-optimization/91333
1082 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1083 reg preferences comparison up.
1084
1085 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1086
1087 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1088 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1089 aarch64-sve-builtins-base.h.
1090 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1091 aarch64-sve-builtins-base.cc.
1092 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1093 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1094 (svcvtnt): Declare.
1095 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1096 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1097 (svcvtnt): New functions.
1098 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1099 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1100 (svcvtnt): New functions.
1101 (svcvt): Add a form that converts f32 to bf16.
1102 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1103 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1104 Declare.
1105 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1106 Treat B as bfloat16_t.
1107 (ternary_bfloat_lane_base): New class.
1108 (ternary_bfloat_def): Likewise.
1109 (ternary_bfloat): New shape.
1110 (ternary_bfloat_lane_def): New class.
1111 (ternary_bfloat_lane): New shape.
1112 (ternary_bfloat_lanex2_def): New class.
1113 (ternary_bfloat_lanex2): New shape.
1114 (ternary_bfloat_opt_n_def): New class.
1115 (ternary_bfloat_opt_n): New shape.
1116 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1117 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1118 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1119 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1120 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1121 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1122 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1123 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1124 the pattern off the narrow mode instead of the wider one.
1125 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1126 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1127 (sve_fp_op): Handle them.
1128 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1129 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1130
1131 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1132
1133 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1134 * config/aarch64/aarch64-modes.def (BF): Move definition before
1135 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1136 (SVE_MODES): Handle BF modes.
1137 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1138 BF modes.
1139 (aarch64_full_sve_mode): Likewise.
1140 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1141 and VNx32BF.
1142 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1143 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1144 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1145 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1146 new SVE BF modes.
1147 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1148 type_class_index.
1149 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1150 (TYPES_all_data): Add bf16.
1151 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1152 (register_tuple_type): Increase buffer size.
1153 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1154 (bf16): New type suffix.
1155 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1156 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1157 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1158 Change type from all_data to all_arith.
1159 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1160 (svminp): Likewise.
1161
1162 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1163 Matthew Malcomson <matthew.malcomson@arm.com>
1164 Richard Sandiford <richard.sandiford@arm.com>
1165
1166 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1167 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1168 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1169 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1170 __ARM_FEATURE_MATMUL_FP64.
1171 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1172 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1173 be disabled at the same time.
1174 (f32mm): New extension.
1175 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1176 (AARCH64_FL_F64MM): Bump to the next bit up.
1177 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1178 (TARGET_SVE_F64MM): New macros.
1179 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1180 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1181 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1182 (UNSPEC_ZIP2Q): New unspeccs.
1183 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1184 (optab, sur, perm_insn): Handle the new unspecs.
1185 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1186 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1187 TARGET_SVE_F64MM instead of separate tests.
1188 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1189 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1190 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1191 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1192 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1193 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1194 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1195 (TYPES_s_signed): New macro.
1196 (TYPES_s_integer): Use it.
1197 (TYPES_d_float): New macro.
1198 (TYPES_d_data): Use it.
1199 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1200 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1201 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1202 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1203 (svmmla): New shape.
1204 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1205 template parameters.
1206 (ternary_resize2_lane_base): Likewise.
1207 (ternary_resize2_base): New class.
1208 (ternary_qq_lane_base): Likewise.
1209 (ternary_intq_uintq_lane_def): Likewise.
1210 (ternary_intq_uintq_lane): New shape.
1211 (ternary_intq_uintq_opt_n_def): New class
1212 (ternary_intq_uintq_opt_n): New shape.
1213 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1214 (ternary_uintq_intq_def): New class.
1215 (ternary_uintq_intq): New shape.
1216 (ternary_uintq_intq_lane_def): New class.
1217 (ternary_uintq_intq_lane): New shape.
1218 (ternary_uintq_intq_opt_n_def): New class.
1219 (ternary_uintq_intq_opt_n): New shape.
1220 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1221 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1222 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1223 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1224 Generalize to...
1225 (svdotprod_lane_impl): ...this new class.
1226 (svmmla_impl, svusdot_impl): New classes.
1227 (svdot_lane): Update to use svdotprod_lane_impl.
1228 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1229 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1230 functions.
1231 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1232 function, with no types defined.
1233 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1234 AARCH64_FL_I8MM functions.
1235 (svmmla): New AARCH64_FL_F32MM function.
1236 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1237 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1238 AARCH64_FL_F64MM function.
1239 (REQUIRED_EXTENSIONS):
1240
1241 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1242
1243 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1244 alternative only.
1245
1246 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1247
1248 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1249 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1250 (*movti_internal): Do not check for
1251 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1252 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1253 just after check for TARGET_AVX.
1254 (*movdf_internal): Ditto.
1255 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1256 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1257 * config/i386/sse.md (mov<mode>_internal): Only check
1258 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1259 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1260 (<sse>_andnot<mode>3<mask_name>): Move check for
1261 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1262 (<code><mode>3<mask_name>): Ditto.
1263 (*andnot<mode>3): Ditto.
1264 (*andnottf3): Ditto.
1265 (*<code><mode>3): Ditto.
1266 (*<code>tf3): Ditto.
1267 (*andnot<VI:mode>3): Remove
1268 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1269 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1270 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1271 (sse4_1_blendv<ssemodesuffix>): Ditto.
1272 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1273 Explain that tune applies to 128bit instructions only.
1274
1275 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1276
1277 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1278 to definition of hsa_kernel_description. Parse assembly to find SGPR
1279 and VGPR count of kernel and store in hsa_kernel_description.
1280
1281 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1282
1283 PR rtl-optimization/91838
1284 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1285 to truncate if allowed or reject combination.
1286
1287 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1288
1289 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1290 (find_inv_vars_cb): Likewise.
1291
1292 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1293
1294 * calls.c (special_function_p): Split out the check for DECL_NAME
1295 being non-NULL and fndecl being extern at file scope into a
1296 new maybe_special_function_p and call it. Drop check for fndecl
1297 being non-NULL that was after a usage of DECL_NAME (fndecl).
1298 * tree.h (maybe_special_function_p): New inline function.
1299
1300 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1301
1302 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1303 (mask_gather_load<mode>): ... here, and zero-initialize the
1304 destination.
1305 (maskload<mode>di): Zero-initialize the destination.
1306 * config/gcn/gcn.c:
1307
1308 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1309
1310 PR analyzer/93356
1311 * doc/analyzer.texi (Limitations): Note that constraints on
1312 floating-point values are currently ignored.
1313
1314 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1315
1316 PR lto/93384
1317 * symtab.c (symtab_node::noninterposable_alias): If localalias
1318 already exists, but is not usable, append numbers after it until
1319 a unique name is found. Formatting fix.
1320
1321 PR middle-end/93505
1322 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1323 rotate counts.
1324
1325 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1326
1327 * config/gcn/gcn.c (print_operand): Handle LTGT.
1328 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1329
1330 2020-01-30 Richard Biener <rguenther@suse.de>
1331
1332 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1333 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1334
1335 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1336
1337 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1338 without a DECL in .data.rel.ro.local.
1339
1340 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1341
1342 PR target/93494
1343 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1344 returned.
1345
1346 PR target/91824
1347 * config/i386/sse.md
1348 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1349 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1350 any_extend code iterator instead of always zero_extend.
1351 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1352 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1353 Use any_extend code iterator instead of always zero_extend.
1354 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1355 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1356 Use any_extend code iterator instead of always zero_extend.
1357 (*sse2_pmovmskb_ext): New define_insn.
1358 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1359
1360 PR target/91824
1361 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1362 (*popcountsi2_zext_falsedep): New define_insn.
1363
1364 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1365
1366 * config.in: Regenerated.
1367 * configure: Regenerated.
1368
1369 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1370
1371 PR bootstrap/93409
1372 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1373 LLVM's assembler changed the default in version 9.
1374
1375 2020-01-24 Jeff Law <law@redhat.com>
1376
1377 PR tree-optimization/89689
1378 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1379
1380 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1381
1382 Revert:
1383
1384 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1385
1386 PR rtl-optimization/87763
1387 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1388 simplification to handle subregs as well as bare regs.
1389 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1390
1391 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1392
1393 PR target/93221
1394 * ira.c (ira): Revert use of simplified LRA algorithm.
1395
1396 2020-01-29 Martin Jambor <mjambor@suse.cz>
1397
1398 PR tree-optimization/92706
1399 * tree-sra.c (struct access): Fields first_link, last_link,
1400 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1401 next_rhs_queued and grp_rhs_queued respectively, new fields
1402 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1403 (struct assign_link): Field next renamed to next_rhs, new field
1404 next_lhs. Updated comment.
1405 (work_queue_head): Renamed to rhs_work_queue_head.
1406 (lhs_work_queue_head): New variable.
1407 (add_link_to_lhs): New function.
1408 (relink_to_new_repr): Also relink LHS lists.
1409 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1410 (add_access_to_lhs_work_queue): New function.
1411 (pop_access_from_work_queue): Renamed to
1412 pop_access_from_rhs_work_queue.
1413 (pop_access_from_lhs_work_queue): New function.
1414 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1415 work_queue.
1416 (child_would_conflict_in_lacc): Renamed to
1417 child_would_conflict_in_acc. Adjusted parameter names.
1418 (create_artificial_child_access): New parameter set_grp_read, use it.
1419 (subtree_mark_written_and_enqueue): Renamed to
1420 subtree_mark_written_and_rhs_enqueue.
1421 (propagate_subaccesses_across_link): Renamed to
1422 propagate_subaccesses_from_rhs.
1423 (propagate_subaccesses_from_lhs): New function.
1424 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1425 RHSs.
1426
1427 2020-01-29 Martin Jambor <mjambor@suse.cz>
1428
1429 PR tree-optimization/92706
1430 * tree-sra.c (struct access): Adjust comment of
1431 grp_total_scalarization.
1432 (find_access_in_subtree): Look for single children spanning an entire
1433 access.
1434 (scalarizable_type_p): Allow register accesses, adjust callers.
1435 (completely_scalarize): Remove function.
1436 (scalarize_elem): Likewise.
1437 (create_total_scalarization_access): Likewise.
1438 (sort_and_splice_var_accesses): Do not track total scalarization
1439 flags.
1440 (analyze_access_subtree): New parameter totally, adjust to new meaning
1441 of grp_total_scalarization.
1442 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1443 (can_totally_scalarize_forest_p): New function.
1444 (create_total_scalarization_access): Likewise.
1445 (create_total_access_and_reshape): Likewise.
1446 (total_should_skip_creating_access): Likewise.
1447 (totally_scalarize_subtree): Likewise.
1448 (analyze_all_variable_accesses): Perform total scalarization after
1449 subaccess propagation using the new functions above.
1450 (initialize_constant_pool_replacements): Output initializers by
1451 traversing the access tree.
1452
1453 2020-01-29 Martin Jambor <mjambor@suse.cz>
1454
1455 * tree-sra.c (verify_sra_access_forest): New function.
1456 (verify_all_sra_access_forests): Likewise.
1457 (create_artificial_child_access): Set parent.
1458 (analyze_all_variable_accesses): Call the verifier.
1459
1460 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1461
1462 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1463 if called on indirect edge.
1464 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1465 speculative call if needed.
1466
1467 2020-01-29 Richard Biener <rguenther@suse.de>
1468
1469 PR tree-optimization/93428
1470 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1471 permutation when the load node is created.
1472 (vect_analyze_slp_instance): Re-use it here.
1473
1474 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1475
1476 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1477
1478 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1479
1480 PR rtl-optimization/93272
1481 * ira-lives.c (process_out_of_region_eh_regs): New function.
1482 (process_bb_node_lives): Call it.
1483
1484 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1485
1486 * coverage.c (read_counts_file): Make error message lowercase.
1487
1488 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1489
1490 * profile-count.c (profile_quality_display_names): Fix ordering.
1491
1492 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1493
1494 PR lto/93318
1495 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1496 hash only when edge is first within the sequence.
1497 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1498 (symbol_table::create_edge): Do not set target_prob.
1499 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1500 the call site hash.
1501 (cgraph_edge::make_speculative): Drop target_prob parameter.
1502 (cgraph_edge::speculative_call_info): Remove.
1503 (cgraph_edge::first_speculative_call_target): New member function.
1504 (update_call_stmt_hash_for_removing_direct_edge): New function.
1505 (cgraph_edge::resolve_speculation): Rewrite to new API.
1506 (cgraph_edge::speculative_call_for_target): New member function.
1507 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1508 multiple speculation targets.
1509 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1510 of profile.
1511 (verify_speculative_call): Verify that targets form an interval.
1512 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1513 (cgraph_edge::first_speculative_call_target): New member function.
1514 (cgraph_edge::next_speculative_call_target): New member function.
1515 (cgraph_edge::speculative_call_target_ref): New member function.
1516 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1517 (cgraph_edge): Remove target_prob.
1518 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1519 Fix handling of speculative calls.
1520 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1521 * ipa-fnsummary.c (analyze_function_body): Likewise.
1522 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1523 * ipa-profile.c (dump_histogram): Fix formating.
1524 (ipa_profile_generate_summary): Watch for overflows.
1525 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1526 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1527 (update_indirect_edges_after_inlining): Update to new API.
1528 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1529 profiles.
1530 * profile-count.h: (profile_probability::adjusted): New.
1531 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1532 updating of profile.
1533 * value-prof.c (gimple_ic_transform): Rename to ...
1534 (dump_ic_profile): ... this one; update dumping.
1535 (stream_in_histogram_value): Fix formating.
1536 (gimple_value_profile_transformations): Update.
1537
1538 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1539
1540 PR target/91461
1541 * config/i386/i386.md (*movoi_internal_avx): Remove
1542 TARGET_SSE_TYPELESS_STORES check.
1543 (*movti_internal): Prefer TARGET_AVX over
1544 TARGET_SSE_TYPELESS_STORES.
1545 (*movtf_internal): Likewise.
1546 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1547 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1548 from TARGET_SSE_TYPELESS_STORES.
1549
1550 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1551
1552 * diagnostic-core.h (warning_at): Rename overload to...
1553 (warning_meta): ...this.
1554 (emit_diagnostic_valist): Delete decl of overload taking
1555 diagnostic_metadata.
1556 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1557 (warning_at): Rename overload taking diagnostic_metadata to...
1558 (warning_meta): ...this.
1559
1560 2020-01-28 Richard Biener <rguenther@suse.de>
1561
1562 PR tree-optimization/93439
1563 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1564 * tree-cfg.c (move_sese_region_to_fn): ... here.
1565 (verify_types_in_gimple_reference): Verify used cliques are
1566 tracked.
1567
1568 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1569
1570 PR target/91399
1571 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1572 argument of a pointer to struct gcc_options and pass it to
1573 parse_mtune_ctrl_str.
1574 (ix86_function_specific_restore): Pass opts to
1575 set_ix86_tune_features.
1576 (ix86_option_override_internal): Likewise.
1577 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1578 gcc_options and use it for x_ix86_tune_ctrl_string.
1579
1580 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1581
1582 PR rtl-optimization/87763
1583 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1584 simplification to handle subregs as well as bare regs.
1585 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1586
1587 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1588
1589 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1590 for reduction chains that (now) include a call.
1591
1592 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1593
1594 PR tree-optimization/92822
1595 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1596 out the don't-care elements of a vector whose significant elements
1597 are duplicates, make the don't-care elements duplicates too.
1598
1599 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1600
1601 PR tree-optimization/93434
1602 * tree-predcom.c (split_data_refs_to_components): Record which
1603 components have had aliasing loads removed. Prevent store-store
1604 commoning for all such components.
1605
1606 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1607
1608 PR target/93418
1609 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1610 -1 or is_vshift is true, use new_vector with number of elts npatterns
1611 rather than new_unary_operation.
1612
1613 PR tree-optimization/93454
1614 * gimple-fold.c (fold_array_ctor_reference): Perform
1615 elt_size.to_uhwi () just once, instead of calling it in every
1616 iteration. Punt if that value is above size of the temporary
1617 buffer. Decrease third native_encode_expr argument when
1618 bufoff + elt_sz is above size of buf.
1619
1620 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1621
1622 * config/mips/mips.c (mips_declare_object_name)
1623 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1624
1625 2020-01-27 Martin Liska <mliska@suse.cz>
1626
1627 PR gcov-profile/93403
1628 * tree-profile.c (gimple_init_gcov_profiler): Generate
1629 both __gcov_indirect_call_profiler_v4 and
1630 __gcov_indirect_call_profiler_v4_atomic.
1631
1632 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1633
1634 PR target/92822
1635 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1636 expander.
1637 (@aarch64_split_simd_mov<mode>): Use it.
1638 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1639 Leave the vec_extract patterns to handle 2-element vectors.
1640 (aarch64_simd_mov_from_<mode>high): Likewise.
1641 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1642 (vec_extractv2dfv1df): Likewise.
1643
1644 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1645
1646 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1647 jump conditions for *compare_condjump<GPI:mode>.
1648
1649 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1650
1651 PR analyzer/93276
1652 * digraph.cc (test_edge::test_edge): Specify template for base
1653 class initializer.
1654
1655 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1656
1657 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1658
1659 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1660
1661 * config/arc/arc-protos.h (gen_mlo): Remove.
1662 (gen_mhi): Likewise.
1663 * config/arc/arc.c (AUX_MULHI): Define.
1664 (arc_must_save_reister): Special handling for r58/59.
1665 (arc_compute_frame_size): Consider mlo/mhi registers.
1666 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1667 paramter is true.
1668 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1669 mlo/mhi name selection.
1670 (arc_restore_callee_saves): Don't early restore blink when ISR.
1671 (arc_expand_prologue): Add mlo/mhi saving.
1672 (arc_expand_epilogue): Add mlo/mhi restoring.
1673 (gen_mlo): Remove.
1674 (gen_mhi): Remove.
1675 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1676 numbering when MUL64 option is used.
1677 (DWARF2_FRAME_REG_OUT): Define.
1678 * config/arc/arc.md (arc600_stall): New pattern.
1679 (VUNSPEC_ARC_ARC600_STALL): Define.
1680 (mulsi64): Use correct mlo/mhi registers.
1681 (mulsi_600): Clean it up.
1682 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1683 TARGET_BIG_ENDIAN.
1684 (mhi_operand): Likewise.
1685
1686 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1687 Petro Karashchenko <petro.karashchenko@ring.com>
1688
1689 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1690 attributes if needed.
1691 (prepare_move_operands): Generate special unspec instruction for
1692 direct access.
1693 (arc_isuncached_mem_p): Propagate uncached attribute to each
1694 structure member.
1695 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1696 (VUNSPEC_ARC_STDI): Likewise.
1697 (ALLI): New mode iterator.
1698 (mALLI): New mode attribute.
1699 (lddi): New instruction pattern.
1700 (stdi): Likewise.
1701 (stdidi_split): Split instruction for architectures which are not
1702 supporting ll64 option.
1703 (lddidi_split): Likewise.
1704
1705 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1706
1707 PR rtl-optimization/92989
1708 * lra-lives.c (process_bb_lives): Update the live-in set before
1709 processing additional clobbers.
1710
1711 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1712
1713 PR rtl-optimization/93170
1714 * cselib.c (cselib_invalidate_regno_val): New function, split out
1715 from...
1716 (cselib_invalidate_regno): ...here.
1717 (cselib_invalidated_by_call_p): New function.
1718 (cselib_process_insn): Iterate over all the hard-register entries in
1719 REG_VALUES and invalidate any that cross call-clobbered registers.
1720
1721 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1722
1723 * dojump.c (split_comparison): Use HONOR_NANS rather than
1724 HONOR_SNANS when splitting LTGT.
1725
1726 2020-01-27 Martin Liska <mliska@suse.cz>
1727
1728 PR driver/91220
1729 * opts.c (print_filtered_help): Exclude language-specific
1730 options from --help=common unless enabled in all FEs.
1731
1732 2020-01-27 Martin Liska <mliska@suse.cz>
1733
1734 * opts.c (print_help): Exclude params from
1735 all except --help=param.
1736
1737 2020-01-27 Martin Liska <mliska@suse.cz>
1738
1739 PR target/93274
1740 * config/i386/i386-features.c (make_resolver_func):
1741 Align the code with ppc64 target implementation.
1742 Do not generate a unique name for resolver function.
1743
1744 2020-01-27 Richard Biener <rguenther@suse.de>
1745
1746 PR tree-optimization/93397
1747 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1748 converted reduction chain SLP graph adjustment.
1749
1750 2020-01-26 Marek Polacek <polacek@redhat.com>
1751
1752 PR sanitizer/93436
1753 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1754 null DECL_NAME.
1755
1756 2020-01-26 Jason Merrill <jason@redhat.com>
1757
1758 PR c++/92601
1759 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1760 of complete types.
1761
1762 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1763
1764 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1765 (rx_setmem): Likewise.
1766
1767 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1768
1769 PR target/93412
1770 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1771 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1772 drop <di> from constraint of last operand.
1773
1774 PR target/93430
1775 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1776 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1777 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1778
1779 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1780
1781 PR ipa/93166
1782 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1783 check assertion.
1784
1785 2020-01-24 Jeff Law <law@redhat.com>
1786
1787 PR tree-optimization/92788
1788 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1789 not EDGE_ABNORMAL.
1790
1791 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1792
1793 PR target/93395
1794 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1795 *avx_vperm_broadcast_<mode>,
1796 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1797 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1798 Move before avx2_perm<mode>/avx512f_perm<mode>.
1799
1800 PR target/93376
1801 * simplify-rtx.c (simplify_const_unary_operation,
1802 simplify_const_binary_operation): Punt for mode precision above
1803 MAX_BITSIZE_MODE_ANY_INT.
1804
1805 2020-01-24 Andrew Pinski <apinski@marvell.com>
1806
1807 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1808 alu.shift_reg to 0.
1809
1810 2020-01-24 Jeff Law <law@redhat.com>
1811
1812 PR target/13721
1813 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1814 for REGs. Call output_operand_lossage to get more reasonable
1815 diagnostics.
1816
1817 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1818
1819 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1820 gcn_fp_compare_operator.
1821 (vec_cmpu<mode>di): Use gcn_compare_operator.
1822 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1823 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1824 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1825 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1826 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1827 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1828 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1829 gcn_fp_compare_operator.
1830 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1831 gcn_fp_compare_operator.
1832 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1833 gcn_fp_compare_operator.
1834 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1835 gcn_fp_compare_operator.
1836
1837 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1838
1839 * doc/install.texi (Cross-Compiler-Specific Options): Document
1840 `--with-toolexeclibdir' option.
1841
1842 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1843
1844 * target.def (flags_regnum): Also mention effect on delay slot filling.
1845 * doc/tm.texi: Regenerate.
1846
1847 2020-01-23 Jeff Law <law@redhat.com>
1848
1849 PR translation/90162
1850 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1851
1852 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1853
1854 PR target/92269
1855 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1856 profiling label
1857
1858 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1859
1860 PR rtl-optimization/93402
1861 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1862 USE insns.
1863
1864 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1865
1866 * config.in: Regenerated.
1867 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1868 for TARGET_LIBC_GNUSTACK.
1869 * configure: Regenerated.
1870 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1871 found to be 2.31 or greater.
1872
1873 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1874
1875 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1876 TARGET_SOFT_FLOAT.
1877 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1878 (mips_asm_file_end): New function. Delegate to
1879 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1880 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1881
1882 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1883
1884 PR target/93376
1885 * config/i386/i386-modes.def (POImode): New mode.
1886 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1887 * config/i386/i386.md (DPWI): New mode attribute.
1888 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1889 (QWI): Rename to...
1890 (QPWI): ... this. Use POI instead of OI for TImode.
1891 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1892 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1893 instead of <QWI>.
1894
1895 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1896
1897 PR target/93341
1898 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1899 unspec.
1900 (speculation_tracker_rev): New pattern.
1901 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1902 Use speculation_tracker_rev to track the inverse condition.
1903
1904 2020-01-23 Richard Biener <rguenther@suse.de>
1905
1906 PR tree-optimization/93381
1907 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1908 alias-set of the def as argument and record the first one.
1909 (vn_walk_cb_data::first_set): New member.
1910 (vn_reference_lookup_3): Pass the alias-set of the current def
1911 to push_partial_def. Fix alias-set used in the aggregate copy
1912 case.
1913 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1914 * real.c (clear_significand_below): Fix out-of-bound access.
1915
1916 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1917
1918 PR target/93346
1919 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1920 New define_insn patterns.
1921
1922 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1923
1924 * doc/sourcebuild.texi (check-function-bodies): Add an
1925 optional target/xfail selector.
1926
1927 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1928
1929 PR rtl-optimization/93124
1930 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1931 bare USE and CLOBBER insns.
1932
1933 2020-01-22 Andrew Pinski <apinski@marvell.com>
1934
1935 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1936
1937 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1938
1939 PR analyzer/93307
1940 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1941 diagnostic_manager into "ana" namespace.
1942 * selftest-run-tests.c (selftest::run_tests): Update for move of
1943 selftest::run_analyzer_selftests to
1944 ana::selftest::run_analyzer_selftests.
1945
1946 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1947
1948 * cfgexpand.c (union_stack_vars): Update the size.
1949
1950 2020-01-22 Richard Biener <rguenther@suse.de>
1951
1952 PR tree-optimization/93381
1953 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1954 throughout, handle all conversions the same.
1955
1956 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1957
1958 PR target/93335
1959 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1960 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1961 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1962 Call force_reg on high_in2 unconditionally.
1963
1964 2020-01-22 Martin Liska <mliska@suse.cz>
1965
1966 PR tree-optimization/92924
1967 * profile.c (compute_value_histograms): Divide
1968 all counter values.
1969
1970 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1971
1972 PR target/91298
1973 * output.h (assemble_name_resolve): Declare.
1974 * varasm.c (assemble_name_resolve): New function.
1975 (assemble_name): Use it.
1976 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1977
1978 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1979
1980 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1981 update_web_docs_git instead of update_web_docs_svn.
1982
1983 2020-01-21 Andrew Pinski <apinski@marvell.com>
1984
1985 PR target/9311
1986 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1987 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1988 (*tlsgd_small_<mode>): Likewise.
1989 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1990 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1991 register. Convert that register back to dest using convert_mode.
1992
1993 2020-01-21 Jim Wilson <jimw@sifive.com>
1994
1995 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1996 instead of XINT.
1997
1998 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1999 Uros Bizjak <ubizjak@gmail.com>
2000
2001 PR target/93319
2002 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2003 with ptr_mode.
2004 (legitimize_tls_address): Do GNU2 TLS address computation in
2005 ptr_mode and zero-extend result to Pmode.
2006 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2007 :P with :PTR and Pmode with ptr_mode.
2008 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2009 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2010 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2011
2012 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2013
2014 PR target/93333
2015 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2016 the last two operands are CONST_INT_P before using them as such.
2017
2018 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2019
2020 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2021 to get the integer element types.
2022
2023 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2024
2025 * config/aarch64/aarch64-sve-builtins.h
2026 (function_expander::convert_to_pmode): Declare.
2027 * config/aarch64/aarch64-sve-builtins.cc
2028 (function_expander::convert_to_pmode): New function.
2029 (function_expander::get_contiguous_base): Use it.
2030 (function_expander::prepare_gather_address_operands): Likewise.
2031 * config/aarch64/aarch64-sve-builtins-sve2.cc
2032 (svwhilerw_svwhilewr_impl::expand): Likewise.
2033
2034 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2035
2036 PR target/92424
2037 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2038 cfun->machine->label_is_assembled.
2039 (aarch64_print_patchable_function_entry): New.
2040 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2041 * config/aarch64/aarch64.h (struct machine_function): New field,
2042 label_is_assembled.
2043
2044 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2045
2046 PR ipa/93315
2047 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2048 NULL on exit.
2049
2050 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2051
2052 PR lto/93318
2053 * cgraph.c (cgraph_edge::resolve_speculation,
2054 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2055 call_stmt_site_hash.
2056
2057 2020-01-21 Martin Liska <mliska@suse.cz>
2058
2059 * config/rs6000/rs6000.c (common_mode_defined): Remove
2060 unused variable.
2061
2062 2020-01-21 Richard Biener <rguenther@suse.de>
2063
2064 PR tree-optimization/92328
2065 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2066 type when value-numbering same-sized store by inserting a
2067 VIEW_CONVERT_EXPR.
2068 (eliminate_dom_walker::eliminate_stmt): When eliminating
2069 a redundant store handle bit-reinterpretation of the same value.
2070
2071 2020-01-21 Andrew Pinski <apinski@marvel.com>
2072
2073 PR tree-opt/93321
2074 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2075 from ...
2076 (prepare_block_for_update): This. Use a worklist instead of
2077 recursing.
2078
2079 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2080
2081 * gcc/config/arm/arm.c (clear_operation_p):
2082 Initialise last_regno, skip first iteration
2083 based on the first_set value and use ints instead
2084 of the unnecessary HOST_WIDE_INTs.
2085
2086 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2087
2088 PR target/93073
2089 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2090 compare_mode other than SFmode or DFmode.
2091
2092 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2093
2094 PR target/93304
2095 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2096 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2097 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2098
2099 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2100
2101 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2102
2103 2020-01-20 Andrew Pinski <apinski@marvell.com>
2104
2105 PR middle-end/93242
2106 * targhooks.c (default_print_patchable_function_entry): Use
2107 output_asm_insn to emit the nop instruction.
2108
2109 2020-01-20 Fangrui Song <maskray@google.com>
2110
2111 PR middle-end/93194
2112 * targhooks.c (default_print_patchable_function_entry): Align to
2113 POINTER_SIZE.
2114
2115 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2116
2117 PR target/93319
2118 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2119 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2120 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2121 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2122 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2123 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2124 Remove the {q} suffix from lea.
2125 (*tls_dynamic_gnu2_call_64): Renamed to ...
2126 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2127 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2128 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2129 Pass Pmode to gen_tls_dynamic_gnu2_64.
2130
2131 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2132
2133 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2134
2135 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2136
2137 * config/aarch64/aarch64-sve-builtins-base.cc
2138 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2139
2140 2020-01-20 Richard Biener <rguenther@suse.de>
2141
2142 PR debug/92763
2143 * dwarf2out.c (prune_unused_types): Unconditionally mark
2144 called function DIEs.
2145
2146 2020-01-20 Martin Liska <mliska@suse.cz>
2147
2148 PR tree-optimization/93199
2149 * tree-eh.c (struct leh_state): Add
2150 new field outer_non_cleanup.
2151 (cleanup_is_dead_in): Pass leh_state instead
2152 of eh_region. Add a checking that state->outer_non_cleanup
2153 points to outer non-clean up region.
2154 (lower_try_finally): Record outer_non_cleanup
2155 for this_state.
2156 (lower_catch): Likewise.
2157 (lower_eh_filter): Likewise.
2158 (lower_eh_must_not_throw): Likewise.
2159 (lower_cleanup): Likewise.
2160
2161 2020-01-20 Richard Biener <rguenther@suse.de>
2162
2163 PR tree-optimization/93094
2164 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2165 (vect_transform_loop): Likewise.
2166 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2167 loop_vectorized_call to vect_transform_loop.
2168 * tree-vect-loop.c (vect_transform_loop): Pass down
2169 loop_vectorized_call to vect_loop_versioning.
2170 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2171 the earlier discovered loop_vectorized_call.
2172
2173 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2174
2175 * doc/contribute.texi: Update for SVN -> Git transition.
2176 * doc/install.texi: Likewise.
2177
2178 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2179
2180 PR lto/93318
2181 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2182 speculative targets.
2183 (verify_speculative_call): New function
2184 (cgraph_node::verify_node): Use it.
2185 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2186 speculations.
2187
2188 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2189
2190 PR lto/93318
2191 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2192 (cgraph_edge::make_direct): Remove all indirect targets.
2193 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2194 (cgraph_node::verify_node): Verify that only one call_stmt or
2195 lto_stmt_uid is set.
2196 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2197 lto_stmt_uid.
2198 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2199 (lto_output_ref): Simplify streaming of stmt.
2200 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2201
2202 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2203
2204 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2205 Mark parameter unused.
2206
2207 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2208
2209 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2210
2211 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2212
2213 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2214
2215 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2216
2217 * Makefile.in: Add coroutine-passes.o.
2218 * builtin-types.def (BT_CONST_SIZE): New.
2219 (BT_FN_BOOL_PTR): New.
2220 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2221 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2222 * coroutine-builtins.def: New file.
2223 * coroutine-passes.cc: New file.
2224 * function.h (struct GTY function): Add a bit to indicate that the
2225 function is a coroutine component.
2226 * internal-fn.c (expand_CO_FRAME): New.
2227 (expand_CO_YIELD): New.
2228 (expand_CO_SUSPN): New.
2229 (expand_CO_ACTOR): New.
2230 * internal-fn.def (CO_ACTOR): New.
2231 (CO_YIELD): New.
2232 (CO_SUSPN): New.
2233 (CO_FRAME): New.
2234 * passes.def: Add pass_coroutine_lower_builtins,
2235 pass_coroutine_early_expand_ifns.
2236 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2237 (make_pass_coroutine_early_expand_ifns): New.
2238 * doc/invoke.texi: Document the fcoroutines command line
2239 switch.
2240
2241 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2242
2243 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2244
2245 PR target/93312
2246 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2247 after checking the argument is a REG. Don't use REGNO (reg)
2248 again to set last_regno, reuse regno variable instead.
2249
2250 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2251
2252 * doc/analyzer.texi (Limitations): Add note about NaN.
2253
2254 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2255 Sudakshina Das <sudi.das@arm.com>
2256
2257 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2258 and valid immediate.
2259 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2260 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2261 * config/arm/constraints.md (Pg): New.
2262 * config/arm/predicates.md (long_shift_imm): New.
2263 (arm_reg_or_long_shift_imm): Likewise.
2264 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2265 (thumb2_lsll): Likewise.
2266 (thumb2_lsrl): New.
2267
2268 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2269 Sudakshina Das <sudi.das@arm.com>
2270
2271 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2272 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2273 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2274 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2275 * config/arm/thumb2.md (thumb2_asrl): New.
2276 (thumb2_lsll): Likewise.
2277
2278 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2279
2280 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2281 unused variable.
2282
2283 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2284
2285 * gdbinit.in (help-gcc-hooks): New command.
2286 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2287 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2288 documentation.
2289
2290 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2291
2292 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2293 correct target macro.
2294
2295 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2296
2297 * config/aarch64/aarch64-protos.h
2298 (aarch64_sve_ld1ro_operand_p): New.
2299 * config/aarch64/aarch64-sve-builtins-base.cc
2300 (class load_replicate): New.
2301 (class svld1ro_impl): New.
2302 (class svld1rq_impl): Change to inherit from load_replicate.
2303 (svld1ro): New sve intrinsic function base.
2304 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2305 New DEF_SVE_FUNCTION.
2306 * config/aarch64/aarch64-sve-builtins-base.h
2307 (svld1ro): New decl.
2308 * config/aarch64/aarch64-sve-builtins.cc
2309 (function_expander::add_mem_operand): Modify assert to allow
2310 OImode.
2311 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2312 pattern.
2313 * config/aarch64/aarch64.c
2314 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2315 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2316 (aarch64_sve_ld1ro_operand_p): New.
2317 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2318 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2319 * config/aarch64/predicates.md
2320 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2321
2322 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2323
2324 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2325 Introduce this ACLE specified predefined macro.
2326 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2327 (fp): Disabling this disables f64mm.
2328 (simd): Disabling this disables f64mm.
2329 (fp16): Disabling this disables f64mm.
2330 (sve): Disabling this disables f64mm.
2331 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2332 (AARCH64_ISA_F64MM): New.
2333 (TARGET_F64MM): New.
2334 * doc/invoke.texi (f64mm): Document new option.
2335
2336 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2337
2338 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2339 (neoversen1_tunings): Likewise.
2340
2341 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2342
2343 PR target/92692
2344 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2345 Add assert to ensure prolog has been emitted.
2346 (aarch64_split_atomic_op): Likewise.
2347 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2348 Use epilogue_completed rather than reload_completed.
2349 (aarch64_atomic_exchange<mode>): Likewise.
2350 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2351 (atomic_nand<mode>): Likewise.
2352 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2353 (atomic_fetch_nand<mode>): Likewise.
2354 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2355 (atomic_nand_fetch<mode>): Likewise.
2356
2357 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2358
2359 PR target/93133
2360 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2361 for FP modes.
2362 (REVERSE_CONDITION): Delete.
2363 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2364 (CCFP_CCFPE): Likewise.
2365 (e): New mode attribute.
2366 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2367 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2368 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2369 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2370 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2371 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2372 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2373 name of generator from gen_ccmpdi to gen_ccmpccdi.
2374 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2375 the previous comparison but aren't able to, use the new ccmp_rev
2376 patterns instead.
2377
2378 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2379
2380 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2381 than testing directly for INTEGER_CST.
2382 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2383
2384 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2385
2386 PR tree-optimization/93292
2387 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2388 get_vectype_for_scalar_type returns NULL.
2389
2390 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2391
2392 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2393 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2394
2395 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2396
2397 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2398 dump.
2399 * params.opt: (max-predicted-iterations): Set bounds.
2400 * predict.c (real_almost_one, real_br_prob_base,
2401 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2402 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2403 probabilities; do not truncate to reg_br_prob_bases.
2404 (estimate_loops_at_level): Pass max_cyclic_prob.
2405 (estimate_loops): Compute max_cyclic_prob.
2406 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2407 of back edge prob.
2408 * profile-count.c (profile_probability::to_sreal): New.
2409 * profile-count.h (class sreal): Move up in file.
2410 (profile_probability::to_sreal): Declare.
2411
2412 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2413
2414 * config/arm/arm.c
2415 (arm_invalid_conversion): New function for target hook.
2416 (arm_invalid_unary_op): New function for target hook.
2417 (arm_invalid_binary_op): New function for target hook.
2418
2419 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2420
2421 * config.gcc: Add arm_bf16.h.
2422 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2423 (arm_simd_builtin_std_type): Add BFmode.
2424 (arm_init_simd_builtin_types): Define element types for vector types.
2425 (arm_init_bf16_types): New function.
2426 (arm_init_builtins): Add arm_init_bf16_types function call.
2427 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2428 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2429 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2430 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2431 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2432 (arm_mangle_type): Add __bf16.
2433 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2434 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2435 arm_bf16_ptr_type_node.
2436 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2437 define_split between ARM registers.
2438 * config/arm/arm_bf16.h: New file.
2439 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2440 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2441 (VQXMOV): Add V8BF.
2442 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2443 * config/arm/vfp.md: Add BFmode to movhf patterns.
2444
2445 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2446 Andre Vieira <andre.simoesdiasvieira@arm.com>
2447
2448 * config/arm/arm-cpus.in (mve, mve_float): New features.
2449 (dsp, mve, mve.fp): New options.
2450 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2451 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2452 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2453
2454 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2455 Thomas Preud'homme <thomas.preudhomme@arm.com>
2456
2457 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2458 Armv8-M Mainline.
2459 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2460 error for using -mcmse when targeting Armv8.1-M Mainline.
2461
2462 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2463 Thomas Preud'homme <thomas.preudhomme@arm.com>
2464
2465 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2466 address in r4 when targeting Armv8.1-M Mainline.
2467 (nonsecure_call_value_internal): Likewise.
2468 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2469 a register match_operand again. Emit BLXNS when targeting
2470 Armv8.1-M Mainline.
2471 (nonsecure_call_value_reg_thumb2): Likewise.
2472
2473 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2474 Thomas Preud'homme <thomas.preudhomme@arm.com>
2475
2476 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2477 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2478 variable as true when floating-point ABI is not hard. Replace
2479 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2480 Generate VLSTM and VLLDM instruction respectively before and
2481 after a function call to cmse_nonsecure_call function.
2482 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2483 (VUNSPEC_VLLDM): Likewise.
2484 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2485 (lazy_load_multiple_insn): Likewise.
2486
2487 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2488 Thomas Preud'homme <thomas.preudhomme@arm.com>
2489
2490 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2491 (arm_emit_vfp_multi_reg_pop): Likewise.
2492 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2493 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2494 restore callee-saved VFP registers.
2495
2496 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2497 Thomas Preud'homme <thomas.preudhomme@arm.com>
2498
2499 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2500 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2501 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2502 callee-saved GPRs as well as clear ip register before doing a nonsecure
2503 call then restore callee-saved GPRs after it when targeting
2504 Armv8.1-M Mainline.
2505 (arm_reorg): Adapt to function rename.
2506
2507 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2508 Thomas Preud'homme <thomas.preudhomme@arm.com>
2509
2510 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2511 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2512 clear_vfp_multiple pattern based on a new vfp parameter.
2513 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2514 targeting Armv8.1-M Mainline.
2515 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2516 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2517 whether VFP registers are available before looking call_used_regs for a
2518 VFP register.
2519 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2520 of prototype of clear_operation_p.
2521 (clear_vfp_multiple_operation): New predicate.
2522 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2523 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2524
2525 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2526 Thomas Preud'homme <thomas.preudhomme@arm.com>
2527
2528 * config/arm/arm-protos.h (clear_operation_p): Declare.
2529 * config/arm/arm.c (clear_operation_p): New function.
2530 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2531 targeting Armv8.1-M Mainline or successor.
2532 (output_return_instruction): Only output APSR register clearing if
2533 Armv8.1-M Mainline instructions not available.
2534 (thumb_exit): Likewise.
2535 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2536 * config/arm/thumb2.md (clear_apsr): New define_insn.
2537 (clear_multiple): Likewise.
2538 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2539
2540 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2541 Thomas Preud'homme <thomas.preudhomme@arm.com>
2542
2543 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2544 (use_return_insn): Also return false for Armv8.1-M Mainline.
2545 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2546 Mainline instructions are available.
2547 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2548 when targeting Armv8.1-M Mainline Security Extensions.
2549 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2550 Mainline entry function.
2551 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2552 targeting Armv8.1-M Mainline or successor.
2553 (arm_expand_epilogue): Fix indentation of caller-saved register
2554 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2555 entry function.
2556 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2557 (FP_SYSREGS): Likewise.
2558 (enum vfp_sysregs_encoding): Define enum.
2559 (fp_sysreg_names): Declare.
2560 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2561 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2562 (pop_fpsysreg_insn): Likewise.
2563
2564 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2565 Thomas Preud'homme <thomas.preudhomme@arm.com>
2566
2567 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2568 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2569 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2570 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2571 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2572 (ARMv8_1m_main): New feature group.
2573 (armv8.1-m.main): New architecture.
2574 * config/arm/arm-tables.opt: Regenerate.
2575 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2576 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2577 (arm_options_perform_arch_sanity_checks): Error out when targeting
2578 Armv8.1-M Mainline Security Extensions.
2579 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2580
2581 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2582
2583 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2584 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2585 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2586 aarch64_bfdot_laneq): New.
2587 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2588 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2589 vbfdotq_laneq_f32): New.
2590 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2591 VBFMLA_W, VBF): New.
2592 (isquadop): Add V4BF, V8BF.
2593
2594 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2595
2596 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2597 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2598 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2599 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2600 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2601 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2602 usdot_laneq, sudot_lane,sudot_laneq): New.
2603 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2604 (aarch64_<sur>dot_lane): New.
2605 * config/aarch64/arm_neon.h (vusdot_s32): New.
2606 (vusdotq_s32): New.
2607 (vusdot_lane_s32): New.
2608 (vsudot_lane_s32): New.
2609 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2610 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2611
2612 2020-01-16 Martin Liska <mliska@suse.cz>
2613
2614 * value-prof.c (dump_histogram_value): Fix
2615 obvious spacing issue.
2616
2617 2020-01-16 Andrew Pinski <apinski@marvell.com>
2618
2619 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2620 !storage_order_barrier_p.
2621
2622 2020-01-16 Andrew Pinski <apinski@marvell.com>
2623
2624 * sched-int.h (_dep): Add unused bit-field field for the padding.
2625 * sched-deps.c (init_dep_1): Init unused field.
2626
2627 2020-01-16 Andrew Pinski <apinski@marvell.com>
2628
2629 * optabs.h (create_expand_operand): Initialize target field also.
2630
2631 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2632
2633 PR tree-optimization/92429
2634 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2635 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2636 control folding.
2637 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2638 tree.
2639
2640 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2641
2642 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2643 aarch64_sve_int_mode to each mode.
2644
2645 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2646
2647 * doc/analyzer.texi (Overview): Add note about
2648 -fdump-ipa-analyzer.
2649
2650 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2651
2652 PR tree-optimization/93231
2653 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2654 input_type is unsigned. Use tree_to_shwi for shift constant.
2655 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2656 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2657 inputs more efficiently.
2658
2659 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2660
2661 * config/i386/i386.md (*movsf_internal): Do not require
2662 SSE2 ISA for alternatives 14 and 15.
2663
2664 2020-01-15 Richard Biener <rguenther@suse.de>
2665
2666 PR middle-end/93273
2667 * tree-eh.c (sink_clobbers): If we already visited the destination
2668 block do not defer insertion.
2669 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2670 the purpose of defered insertion.
2671
2672 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2673
2674 * BASE-VER: Bump to 10.0.1.
2675
2676 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2677
2678 PR tree-optimization/93247
2679 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2680 type of the stmt that we're going to vectorize.
2681
2682 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2683
2684 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2685 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2686 type from the lhs.
2687
2688 2020-01-15 Martin Liska <mliska@suse.cz>
2689
2690 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2691 2 calls of streamer_read_hwi in a function call.
2692
2693 2020-01-15 Richard Biener <rguenther@suse.de>
2694
2695 * alias.c (record_alias_subset): Avoid redundant work when
2696 subset is already recorded.
2697
2698 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2699
2700 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2701 the analyzer options provide CWE identifiers.
2702
2703 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2704
2705 * tree-diagnostic-path.cc (path_summary::event_range::print):
2706 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2707 using get_pure_location.
2708
2709 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2710
2711 PR tree-optimization/93262
2712 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2713 perform head trimming only if the last argument is constant,
2714 either all ones, or larger or equal to head trim, in the latter
2715 case decrease the last argument by head_trim.
2716
2717 PR tree-optimization/93249
2718 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2719 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2720 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2721 perform head trim unless we can prove there are no '\0' chars
2722 from the source among the first head_trim chars.
2723
2724 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2725
2726 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2727
2728 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2729
2730 PR target/93009
2731 * config/i386/sse.md
2732 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2733 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2734 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2735 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2736 just a single alternative instead of two, make operands 1 and 2
2737 commutative.
2738
2739 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2740
2741 PR lto/91576
2742 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2743 TYPE_MODE.
2744
2745 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2746
2747 * Makefile.in (lang_opt_files): Add analyzer.opt.
2748 (ANALYZER_OBJS): New.
2749 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2750 tristate.o and ANALYZER_OBJS.
2751 (TEXI_GCCINT_FILES): Add analyzer.texi.
2752 * common.opt (-fanalyzer): New driver option.
2753 * config.in: Regenerate.
2754 * configure: Regenerate.
2755 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2756 (gccdepdir): Also create depdir for "analyzer" subdir.
2757 * digraph.cc: New file.
2758 * digraph.h: New file.
2759 * doc/analyzer.texi: New file.
2760 * doc/gccint.texi ("Static Analyzer") New menu item.
2761 (analyzer.texi): Include it.
2762 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2763 ("Warning Options"): Add static analysis warnings to the list.
2764 (-Wno-analyzer-double-fclose): New option.
2765 (-Wno-analyzer-double-free): New option.
2766 (-Wno-analyzer-exposure-through-output-file): New option.
2767 (-Wno-analyzer-file-leak): New option.
2768 (-Wno-analyzer-free-of-non-heap): New option.
2769 (-Wno-analyzer-malloc-leak): New option.
2770 (-Wno-analyzer-possible-null-argument): New option.
2771 (-Wno-analyzer-possible-null-dereference): New option.
2772 (-Wno-analyzer-null-argument): New option.
2773 (-Wno-analyzer-null-dereference): New option.
2774 (-Wno-analyzer-stale-setjmp-buffer): New option.
2775 (-Wno-analyzer-tainted-array-index): New option.
2776 (-Wno-analyzer-use-after-free): New option.
2777 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2778 (-Wno-analyzer-use-of-uninitialized-value): New option.
2779 (-Wanalyzer-too-complex): New option.
2780 (-fanalyzer-call-summaries): New warning.
2781 (-fanalyzer-checker=): New warning.
2782 (-fanalyzer-fine-grained): New warning.
2783 (-fno-analyzer-state-merge): New warning.
2784 (-fno-analyzer-state-purge): New warning.
2785 (-fanalyzer-transitivity): New warning.
2786 (-fanalyzer-verbose-edges): New warning.
2787 (-fanalyzer-verbose-state-changes): New warning.
2788 (-fanalyzer-verbosity=): New warning.
2789 (-fdump-analyzer): New warning.
2790 (-fdump-analyzer-callgraph): New warning.
2791 (-fdump-analyzer-exploded-graph): New warning.
2792 (-fdump-analyzer-exploded-nodes): New warning.
2793 (-fdump-analyzer-exploded-nodes-2): New warning.
2794 (-fdump-analyzer-exploded-nodes-3): New warning.
2795 (-fdump-analyzer-supergraph): New warning.
2796 * doc/sourcebuild.texi (dg-require-dot): New.
2797 (dg-check-dot): New.
2798 * gdbinit.in (break-on-saved-diagnostic): New command.
2799 * graphviz.cc: New file.
2800 * graphviz.h: New file.
2801 * ordered-hash-map-tests.cc: New file.
2802 * ordered-hash-map.h: New file.
2803 * passes.def (pass_analyzer): Add before
2804 pass_ipa_whole_program_visibility.
2805 * selftest-run-tests.c (selftest::run_tests): Call
2806 selftest::ordered_hash_map_tests_cc_tests.
2807 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2808 decl.
2809 * shortest-paths.h: New file.
2810 * timevar.def (TV_ANALYZER): New timevar.
2811 (TV_ANALYZER_SUPERGRAPH): Likewise.
2812 (TV_ANALYZER_STATE_PURGE): Likewise.
2813 (TV_ANALYZER_PLAN): Likewise.
2814 (TV_ANALYZER_SCC): Likewise.
2815 (TV_ANALYZER_WORKLIST): Likewise.
2816 (TV_ANALYZER_DUMP): Likewise.
2817 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2818 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2819 * tree-pass.h (make_pass_analyzer): New decl.
2820 * tristate.cc: New file.
2821 * tristate.h: New file.
2822
2823 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2824
2825 PR target/93254
2826 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2827 alternatives 9 and 10.
2828
2829 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2830
2831 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2832 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2833 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2834 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2835 (selftest::hash_map_tests_c_tests): Call it.
2836 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2837 New static constant, using the value of = H::empty_zero_p.
2838 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2839 from default_hash_traits <Value>.
2840 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2841 from Traits.
2842 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2843 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2844 calls to mark_empty with !Descriptor::empty_zero_p.
2845 (hash_table::empty_slow): Conditionalize the memset call with a
2846 check that Descriptor::empty_zero_p; otherwise, loop through the
2847 entries calling mark_empty on them.
2848 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2849 (pointer_hash::empty_zero_p): Likewise.
2850 (pair_hash::empty_zero_p): Likewise.
2851 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2852 Likewise.
2853 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2854 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2855 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2856 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2857 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2858 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2859 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2860 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2861 * tree-vectorizer.h
2862 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2863 Likewise.
2864
2865 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2866
2867 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2868 fix typo on return value.
2869
2870 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2871
2872 PR ipa/69678
2873 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2874 target_prob.
2875 (cgraph_edge::make_speculative): Add param for setting speculative_id
2876 and target_prob.
2877 (cgraph_edge::speculative_call_info): Update comments and find reference
2878 by speculative_id for multiple indirect targets.
2879 (cgraph_edge::resolve_speculation): Decrease the speculations
2880 for indirect edge, drop it's speculative if not direct target
2881 left. Update comments.
2882 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2883 (cgraph_node::dump): Print num_speculative_call_targets.
2884 (cgraph_node::verify_node): Don't report error if speculative
2885 edge not include statement.
2886 (cgraph_edge::num_speculative_call_targets_p): New function.
2887 * cgraph.h (int common_target_id): Remove.
2888 (int common_target_probability): Remove.
2889 (num_speculative_call_targets): New variable.
2890 (make_speculative): Add param for setting speculative_id.
2891 (cgraph_edge::num_speculative_call_targets_p): New declare.
2892 (target_prob): New variable.
2893 (speculative_id): New variable.
2894 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2895 call summaries for multiple speculative call targets.
2896 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2897 * ipa-profile.c (struct speculative_call_target): New struct.
2898 (class speculative_call_summary): New class.
2899 (class speculative_call_summaries): New class.
2900 (call_sums): New variable.
2901 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2902 (ipa_profile_write_edge_summary): New function.
2903 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2904 (ipa_profile_dump_all_summaries): New function.
2905 (ipa_profile_read_edge_summary): New function.
2906 (ipa_profile_read_summary_section): New function.
2907 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2908 (ipa_profile): Generate num_speculative_call_targets from
2909 profile summaries.
2910 * ipa-ref.h (speculative_id): New variable.
2911 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2912 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2913 common_target_probability. Stream out speculative_id and
2914 num_speculative_call_targets.
2915 (input_edge): Likewise.
2916 * predict.c (dump_prediction): Remove edges count assert to be
2917 precise.
2918 * symtab.c (symtab_node::create_reference): Init speculative_id.
2919 (symtab_node::clone_references): Clone speculative_id.
2920 (symtab_node::clone_referring): Clone speculative_id.
2921 (symtab_node::clone_reference): Clone speculative_id.
2922 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2923 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2924 if indirect call contains multiple speculative targets.
2925 * value-prof.h (check_ic_target): Remove.
2926 * value-prof.c (gimple_value_profile_transformations):
2927 Use void function gimple_ic_transform.
2928 * value-prof.c (gimple_ic_transform): Handle topn case.
2929 Fix comment typos. Change it to a void function.
2930
2931 2020-01-13 Andrew Pinski <apinski@marvell.com>
2932
2933 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2934 (octeontx2t98): New define.
2935 (octeontx2t96): New define.
2936 (octeontx2t93): New define.
2937 (octeontx2f95): New define.
2938 (octeontx2f95n): New define.
2939 (octeontx2f95mm): New define.
2940 * config/aarch64/aarch64-tune.md: Regenerate.
2941 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2942
2943 2020-01-13 Jason Merrill <jason@redhat.com>
2944
2945 PR c++/33799 - destroy return value if local cleanup throws.
2946 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2947
2948 2020-01-13 Martin Liska <mliska@suse.cz>
2949
2950 * ipa-cp.c (get_max_overall_size): Use newly
2951 renamed param param_ipa_cp_unit_growth.
2952 * params.opt: Remove legacy param name.
2953
2954 2020-01-13 Martin Sebor <msebor@redhat.com>
2955
2956 PR tree-optimization/93213
2957 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2958 stores to be eliminated.
2959
2960 2020-01-13 Martin Liska <mliska@suse.cz>
2961
2962 * opts.c (print_help): Do not print CL_PARAM
2963 and CL_WARNING for CL_OPTIMIZATION.
2964
2965 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2966
2967 PR driver/92757
2968 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2969 depending on optimization settings.
2970
2971 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2972
2973 PR tree-optimization/90838
2974 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2975 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2976 argument rather than to initialize temporary for targets that
2977 don't use the mode argument at all. Initialize ctzval to avoid
2978 warning at -O0.
2979
2980 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2981
2982 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2983 * tree-core.h: Document it.
2984 * gimplify.c (gimplify_omp_workshare): Set it.
2985 * omp-low.c (lower_omp_target): Use it.
2986 * tree-pretty-print.c (dump_omp_clause): Print it.
2987
2988 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2989 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2990
2991 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2992
2993 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2994 * common.opt (fdiagnostics-path-format=): New option.
2995 (diagnostic_path_format): New enum.
2996 (fdiagnostics-show-path-depths): New option.
2997 * coretypes.h (diagnostic_event_id_t): New forward decl.
2998 * diagnostic-color.c (color_dict): Add "path".
2999 * diagnostic-event-id.h: New file.
3000 * diagnostic-format-json.cc (json_from_expanded_location): Make
3001 non-static.
3002 (json_end_diagnostic): Call context->make_json_for_path if it
3003 exists and the diagnostic has a path.
3004 (diagnostic_output_format_init): Clear context->print_path.
3005 * diagnostic-path.h: New file.
3006 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3007 when printing a run of events in a diagnostic_path so that they
3008 all get the same color.
3009 (layout::m_diagnostic_path_p): New field.
3010 (layout::layout): Initialize it.
3011 (layout::print_any_labels): Don't colorize the label text for an
3012 event in a diagnostic_path.
3013 (gcc_rich_location::add_location_if_nearby): Add
3014 "restrict_to_current_line_spans" and "label" params. Pass the
3015 former to layout.maybe_add_location_range; pass the latter
3016 when calling add_range.
3017 * diagnostic.c: Include "diagnostic-path.h".
3018 (diagnostic_initialize): Initialize context->path_format and
3019 context->show_path_depths.
3020 (diagnostic_show_any_path): New function.
3021 (diagnostic_path::interprocedural_p): New function.
3022 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3023 (simple_diagnostic_path::num_events): New function.
3024 (simple_diagnostic_path::get_event): New function.
3025 (simple_diagnostic_path::add_event): New function.
3026 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3027 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3028 (debug): New overload taking a diagnostic_path *.
3029 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3030 * diagnostic.h (enum diagnostic_path_format): New enum.
3031 (json::value): New forward decl.
3032 (diagnostic_context::path_format): New field.
3033 (diagnostic_context::show_path_depths): New field.
3034 (diagnostic_context::print_path): New callback field.
3035 (diagnostic_context::make_json_for_path): New callback field.
3036 (diagnostic_show_any_path): New decl.
3037 (json_from_expanded_location): New decl.
3038 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3039 (-fdiagnostics-show-path-depths): New option.
3040 (-fdiagnostics-color): Add "path" to description of default
3041 GCC_COLORS; describe it.
3042 (-fdiagnostics-format=json): Document how diagnostic paths are
3043 represented in the JSON output format.
3044 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3045 Add optional params "restrict_to_current_line_spans" and "label".
3046 * opts.c (common_handle_option): Handle
3047 OPT_fdiagnostics_path_format_ and
3048 OPT_fdiagnostics_show_path_depths.
3049 * pretty-print.c: Include "diagnostic-event-id.h".
3050 (pp_format): Implement "%@" format code for printing
3051 diagnostic_event_id_t *.
3052 (selftest::test_pp_format): Add tests for "%@".
3053 * selftest-run-tests.c (selftest::run_tests): Call
3054 selftest::tree_diagnostic_path_cc_tests.
3055 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3056 * toplev.c (general_init): Initialize global_dc->path_format and
3057 global_dc->show_path_depths.
3058 * tree-diagnostic-path.cc: New file.
3059 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3060 non-static. Drop "diagnostic" param in favor of storing the
3061 original value of "where" and re-using it.
3062 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3063 maybe_unwind_expanded_macro_loc.
3064 (tree_diagnostics_defaults): Initialize context->print_path and
3065 context->make_json_for_path.
3066 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3067 decl.
3068 (default_tree_make_json_for_path): New decl.
3069 (maybe_unwind_expanded_macro_loc): New decl.
3070
3071 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3072
3073 PR tree-optimization/93210
3074 * fold-const.h (native_encode_initializer,
3075 can_native_interpret_type_p): Declare.
3076 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3077 simplify.
3078 (native_encode_initializer): New function, moved from dwarf2out.c.
3079 Adjust to native_encode_expr compatible arguments, including dry-run
3080 and partial extraction modes. Don't handle STRING_CST.
3081 (can_native_interpret_type_p): No longer static.
3082 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3083 offset / BITS_PER_UNIT fits into int and don't call it if
3084 can_native_interpret_type_p fails. If suboff is NULL and for
3085 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3086 native_encode_initializer.
3087 (fold_const_aggregate_ref_1): Formatting fix.
3088 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3089 (tree_add_const_value_attribute): Adjust caller.
3090
3091 PR tree-optimization/90838
3092 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3093 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3094 CTZ_DEFINED_VALUE_AT_ZERO.
3095
3096 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3097
3098 PR inline-asm/93027
3099 * lra-constraints.c (match_reload): Permit input operands have the
3100 same mode as output while other input operands have a different
3101 mode.
3102
3103 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3104
3105 PR tree-optimization/90838
3106 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3107 (check_ctz_string): Likewise.
3108 (optimize_count_trailing_zeroes): Likewise.
3109 (simplify_count_trailing_zeroes): Likewise.
3110 (pass_forwprop::execute): Try ctz simplification.
3111 * match.pd: Add matching for ctz idioms.
3112
3113 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3114
3115 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3116 for target hook.
3117 (aarch64_invalid_unary_op): New function for target hook.
3118 (aarch64_invalid_binary_op): New function for target hook.
3119
3120 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3121
3122 * config.gcc: Add arm_bf16.h.
3123 * config/aarch64/aarch64-builtins.c
3124 (aarch64_simd_builtin_std_type): Add BFmode.
3125 (aarch64_init_simd_builtin_types): Define element types for vector
3126 types.
3127 (aarch64_init_bf16_types): New function.
3128 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3129 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3130 modes.
3131 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3132 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3133 patterns.
3134 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3135 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3136 * config/aarch64/aarch64.c
3137 (aarch64_classify_vector_mode): Add support for BF types.
3138 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3139 (aarch64_vq_mode): Add support for BF types.
3140 (aarch64_simd_container_mode): Add support for BF types.
3141 (aarch64_mangle_type): Add support for BF scalar type.
3142 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3143 * config/aarch64/arm_bf16.h: New file.
3144 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3145 * config/aarch64/iterators.md: Add BF types to mode attributes.
3146 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3147
3148 2020-01-10 Jason Merrill <jason@redhat.com>
3149
3150 PR c++/93173 - incorrect tree sharing.
3151 * gimplify.c (copy_if_shared): No longer static.
3152 * gimplify.h: Declare it.
3153
3154 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3155
3156 * doc/invoke.texi (-msve-vector-bits=): Document that
3157 -msve-vector-bits=128 now generates VL-specific code for
3158 little-endian targets.
3159 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3160 build_vector_type_for_mode to construct the data vector types.
3161 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3162 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3163 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3164 for 128-bit vectors.
3165
3166 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3167
3168 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3169 invocation.
3170
3171 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3172
3173 * config/aarch64/aarch64-builtins.c
3174 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3175 rather than checking the number of elements and the element mode.
3176
3177 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3178
3179 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3180 get_related_vectype_for_scalar_type rather than build_vector_type
3181 to create the index type for a conditional reduction.
3182
3183 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3184
3185 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3186 for any type of gather or scatter, including strided accesses.
3187
3188 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3189
3190 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3191 comment.
3192
3193 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3194
3195 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3196 get_dr_vinfo_offset
3197 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3198 parameter and its use to reset DR_OFFSET's.
3199 (vect_transform_loop): Remove orig_drs_init argument.
3200 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3201 member of dr_vec_info rather than the offset of the associated
3202 data_reference's innermost_loop_behavior.
3203 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3204 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3205 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3206 get_dr_vinfo_offset.
3207 (vectorizable_store): Likewise.
3208 (vectorizable_load): Likewise.
3209
3210 2020-01-10 Richard Biener <rguenther@suse.de>
3211
3212 * gimple-ssa-store-merging
3213 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3214
3215 2020-01-10 Martin Liska <mliska@suse.cz>
3216
3217 PR ipa/93217
3218 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3219 encapsulation that was there before r280040.
3220
3221 2020-01-10 Richard Biener <rguenther@suse.de>
3222
3223 PR middle-end/93199
3224 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3225 sequences to avoid walking them again for secondary opportunities.
3226 (pass_lower_eh_dispatch::execute): Instead actually insert
3227 them here.
3228
3229 2020-01-10 Richard Biener <rguenther@suse.de>
3230
3231 PR middle-end/93199
3232 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3233 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3234 avoid quadraticness.
3235
3236 2020-01-10 Martin Jambor <mjambor@suse.cz>
3237
3238 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3239 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3240 to get param_ipa_sra_max_replacements.
3241 (param_splitting_across_edge): Pass the caller to
3242 pull_accesses_from_callee.
3243
3244 2020-01-10 Martin Jambor <mjambor@suse.cz>
3245
3246 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3247 * ipa-cp.c (max_new_size): Removed.
3248 (orig_overall_size): New variable.
3249 (get_max_overall_size): New function.
3250 (estimate_local_effects): Use it. Adjust dump.
3251 (decide_about_value): Likewise.
3252 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3253 orig_overall_size. Adjust dump.
3254 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3255
3256 2020-01-10 Martin Jambor <mjambor@suse.cz>
3257
3258 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3259 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3260 instead of param_ipa_max_agg_items.
3261 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3262 optimization info for the callee.
3263
3264 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3265
3266 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3267 markers if debug_inline_points is false.
3268
3269 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3270
3271 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3272 extra_objs.
3273 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3274 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3275 aarch64-sve-builtins-sve2.h.
3276 (aarch64-sve-builtins-sve2.o): New rule.
3277 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3278 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3279 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3280 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3281 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3282 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3283 TARGET_SVE2_SM4.
3284 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3285 instructions that are handled here.
3286 (@cond_asrd<mode>): Generalize to...
3287 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3288 (*cond_asrd<mode>_2): Generalize to...
3289 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3290 (*cond_asrd<mode>_z): Generalize to...
3291 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3292 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3293 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3294 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3295 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3296 pattern.
3297 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3298 (@aarch64_scatter_stnt<mode>): Likewise.
3299 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3300 (@aarch64_mul_lane_<mode>): Likewise.
3301 (@aarch64_sve_suqadd<mode>_const): Likewise.
3302 (*<sur>h<addsub><mode>): Generalize to...
3303 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3304 new pattern.
3305 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3306 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3307 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3308 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3309 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3310 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3311 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3312 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3313 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3314 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3315 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3316 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3317 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3318 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3319 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3320 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3321 (@aarch64_sve2_xar<mode>): Likewise.
3322 (@aarch64_sve2_bcax<mode>): Likewise.
3323 (*aarch64_sve2_eor3<mode>): Rename to...
3324 (@aarch64_sve2_eor3<mode>): ...this.
3325 (@aarch64_sve2_bsl<mode>): New expander.
3326 (@aarch64_sve2_nbsl<mode>): Likewise.
3327 (@aarch64_sve2_bsl1n<mode>): Likewise.
3328 (@aarch64_sve2_bsl2n<mode>): Likewise.
3329 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3330 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3331 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3332 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3333 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3334 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3335 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3336 (<su>mull<bt><Vwide>): Generalize to...
3337 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3338 pattern.
3339 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3340 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3341 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3342 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3343 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3344 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3345 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3346 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3347 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3348 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3349 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3350 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3351 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3352 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3353 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3354 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3355 (<SHRNB:r>shrnb<mode>): Generalize to...
3356 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3357 new pattern.
3358 (<SHRNT:r>shrnt<mode>): Generalize to...
3359 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3360 new pattern.
3361 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3362 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3363 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3364 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3365 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3366 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3367 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3368 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3369 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3370 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3371 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3372 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3373 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3374 (@aarch64_sve2_cvtnt<mode>): Likewise.
3375 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3376 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3377 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3378 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3379 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3380 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3381 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3382 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3383 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3384 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3385 (@aarch64_sve2_pmul<mode>): Likewise.
3386 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3387 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3388 (@aarch64_sve2_tbl2<mode>): Likewise.
3389 (@aarch64_sve2_tbx<mode>): Likewise.
3390 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3391 (@aarch64_sve2_histcnt<mode>): Likewise.
3392 (@aarch64_sve2_histseg<mode>): Likewise.
3393 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3394 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3395 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3396 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3397 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3398 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3399 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3400 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3401 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3402 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3403 (SVE2_PMULL_PAIR_I): New mode iterators.
3404 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3405 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3406 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3407 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3408 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3409 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3410 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3411 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3412 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3413 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3414 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3415 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3416 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3417 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3418 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3419 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3420 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3421 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3422 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3423 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3424 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3425 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3426 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3427 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3428 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3429 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3430 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3431 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3432 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3433 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3434 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3435 further down file.
3436 (VNARROW, Ventype): New mode attributes.
3437 (Vewtype): Handle VNx2DI. Fix typo in comment.
3438 (VDOUBLE): New mode attribute.
3439 (sve_lane_con): Handle VNx8HI.
3440 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3441 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3442 (sve_int_op, sve_int_op_rev): Handle the above codes.
3443 (sve_pred_int_rhs2_operand): Likewise.
3444 (MULLBT, SHRNB, SHRNT): Delete.
3445 (SVE_INT_SHIFT_IMM): New int iterator.
3446 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3447 and UNSPEC_WHILEHS for TARGET_SVE2.
3448 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3449 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3450 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3451 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3452 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3453 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3454 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3455 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3456 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3457 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3458 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3459 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3460 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3461 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3462 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3463 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3464 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3465 (optab): Handle the new unspecs.
3466 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3467 and UNSPEC_RSHRNT.
3468 (lr): Handle the new unspecs.
3469 (bt): Delete.
3470 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3471 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3472 (sve_int_qsub_op): New int attributes.
3473 (sve_fp_op, rot): Handle the new unspecs.
3474 * config/aarch64/aarch64-sve-builtins.h
3475 (function_resolver::require_matching_pointer_type): Declare.
3476 (function_resolver::resolve_unary): Add an optional boolean argument.
3477 (function_resolver::finish_opt_n_resolution): Add an optional
3478 type_suffix_index argument.
3479 (gimple_folder::redirect_call): Declare.
3480 (gimple_expander::prepare_gather_address_operands): Add an optional
3481 bool parameter.
3482 * config/aarch64/aarch64-sve-builtins.cc: Include
3483 aarch64-sve-builtins-sve2.h.
3484 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3485 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3486 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3487 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3488 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3489 (TYPES_s_unsigned): Likewise.
3490 (TYPES_s_integer): Use TYPES_s_unsigned.
3491 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3492 (TYPES_sd_integer): Use them.
3493 (TYPES_d_unsigned): New macro.
3494 (TYPES_d_integer): Use it.
3495 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3496 (TYPES_cvt_narrow): Likewise.
3497 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3498 (preds_mx): New variable.
3499 (function_builder::add_overloaded_function): Allow the new feature
3500 set to be more restrictive than the original one.
3501 (function_resolver::infer_pointer_type): Remove qualifiers from
3502 the pointer type before printing it.
3503 (function_resolver::require_matching_pointer_type): New function.
3504 (function_resolver::resolve_sv_displacement): Handle functions
3505 that don't support 32-bit vector indices or svint32_t vector offsets.
3506 (function_resolver::finish_opt_n_resolution): Take the inferred type
3507 as a separate argument.
3508 (function_resolver::resolve_unary): Optionally treat all forms in
3509 the same way as normal merging functions.
3510 (gimple_folder::redirect_call): New function.
3511 (function_expander::prepare_gather_address_operands): Add an argument
3512 that says whether scaled forms are available. If they aren't,
3513 handle scaling of vector indices and don't add the extension and
3514 scaling operands.
3515 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3516 fall back to using cond_* instead.
3517 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3518 Split out the member variables into...
3519 (rtx_code_function_base): ...this new base class.
3520 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3521 (unspec_based_function): Split out the member variables into...
3522 (unspec_based_function_base): ...this new base class.
3523 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3524 (unspec_based_function_exact_insn): New class.
3525 (unspec_based_add_function, unspec_based_add_lane_function)
3526 (unspec_based_lane_function, unspec_based_pred_function)
3527 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3528 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3529 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3530 typedefs.
3531 (unspec_based_fused_function): New class.
3532 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3533 (unspec_based_fused_lane_function): New class.
3534 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3535 typedefs.
3536 (CODE_FOR_MODE1): New macro.
3537 (fixed_insn_function): New class.
3538 (while_comparison): Likewise.
3539 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3540 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3541 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3542 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3543 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3544 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3545 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3546 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3547 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3548 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3549 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3550 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3551 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3552 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3553 Also add an initial argument for unary_convert_narrowt, regardless
3554 of the predication type.
3555 (build_32_64): Allow loads and stores to specify MODE_none.
3556 (build_sv_index64, build_sv_uint_offset): New functions.
3557 (long_type_suffix): New function.
3558 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3559 (binary_imm_long_base, load_gather_sv_base): Likewise.
3560 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3561 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3562 (unary_narrowb_base, unary_narrowt_base): Likewise.
3563 (binary_long_lane_def, binary_long_lane): New shape.
3564 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3565 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3566 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3567 (binary_to_uint_def, binary_to_uint): Likewise.
3568 (binary_wide_def, binary_wide): Likewise.
3569 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3570 (compare_def, compare): Likewise.
3571 (compare_ptr_def, compare_ptr): Likewise.
3572 (load_ext_gather_index_restricted_def,
3573 load_ext_gather_index_restricted): Likewise.
3574 (load_ext_gather_offset_restricted_def,
3575 load_ext_gather_offset_restricted): Likewise.
3576 (load_gather_sv_def): Inherit from load_gather_sv_base.
3577 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3578 (shift_left_imm_def, shift_left_imm): Likewise.
3579 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3580 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3581 (store_scatter_index_restricted_def,
3582 store_scatter_index_restricted): Likewise.
3583 (store_scatter_offset_restricted_def,
3584 store_scatter_offset_restricted): Likewise.
3585 (tbl_tuple_def, tbl_tuple): Likewise.
3586 (ternary_long_lane_def, ternary_long_lane): Likewise.
3587 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3588 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3589 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3590 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3591 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3592 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3593 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3594 (ternary_uint_def, ternary_uint): Likewise.
3595 (unary_convert): Fix typo in comment.
3596 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3597 (unary_long_def, unary_long): Likewise.
3598 (unary_narrowb_def, unary_narrowb): Likewise.
3599 (unary_narrowt_def, unary_narrowt): Likewise.
3600 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3601 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3602 (unary_to_int_def, unary_to_int): Likewise.
3603 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3604 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3605 (svasrd_impl): Delete.
3606 (svcadd_impl::expand): Handle integer operations too.
3607 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3608 new functions to derive the unspec numbers.
3609 (svmla_svmls_lane_impl): Replace with...
3610 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3611 integer operations too.
3612 (svwhile_impl): Rename to...
3613 (svwhilelx_impl): ...this and inherit from while_comparison.
3614 (svasrd): Use unspec_based_function.
3615 (svmla_lane): Use svmla_lane_impl.
3616 (svmls_lane): Use svmls_lane_impl.
3617 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3618 (svwhilele, svwhilelt): Use svwhilelx_impl.
3619 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3620 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3621 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3622 * config/aarch64/aarch64-sve-builtins.def: Include
3623 aarch64-sve-builtins-sve2.def.
3624
3625 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3626
3627 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3628 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3629 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3630 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3631 immediates as well as vector ones.
3632 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3633 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3634 (aarch64_sve_qsub_immediate): Update calls accordingly.
3635
3636 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3637
3638 * config/aarch64/aarch64-sve2.md: Add banner comments.
3639 (<su>mulh<r>s<mode>3): Move further up file.
3640 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3641 (*aarch64_sve2_sra<mode>): Move further down file.
3642 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3643
3644 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3645
3646 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3647 and UNSPEC_WHILEWR.
3648 (while_optab_cmp): Handle them.
3649 * config/aarch64/aarch64-sve.md
3650 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3651 and add a "@" marker.
3652 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3653 instead of gen_aarch64_sve2_while_ptest.
3654 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3655
3656 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3657
3658 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3659 (UNSPEC_WHILELE): ...this.
3660 (UNSPEC_WHILE_LO): Rename to...
3661 (UNSPEC_WHILELO): ...this.
3662 (UNSPEC_WHILE_LS): Rename to...
3663 (UNSPEC_WHILELS): ...this.
3664 (UNSPEC_WHILE_LT): Rename to...
3665 (UNSPEC_WHILELT): ...this.
3666 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3667 (cmp_op, while_optab_cmp): Likewise.
3668 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3669 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3670 (svwhilelt): Likewise.
3671
3672 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3673
3674 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3675 (unary_to_uint): Define.
3676 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3677 (unary_count): Rename to...
3678 (unary_to_uint_def, unary_to_uint): ...this.
3679 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3680
3681 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3682
3683 * config/aarch64/aarch64-sve-builtins-functions.h
3684 (code_for_mode_function): New class.
3685 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3686 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3687 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3688 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3689 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3690
3691 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3692
3693 * config/aarch64/iterators.md (addsub): New code attribute.
3694 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3695 Re-express as...
3696 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3697 in the asm string and attributes. Fix indentation.
3698 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3699 Re-express as...
3700 (@aarch64_sve_<optab><mode>): ...this.
3701 * config/aarch64/aarch64-sve-builtins.h
3702 (function_expander::expand_signed_unpred_op): Delete.
3703 * config/aarch64/aarch64-sve-builtins.cc
3704 (function_expander::expand_signed_unpred_op): Likewise.
3705 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3706 try using code_for_aarch64_sve instead.
3707 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3708 (svqsub_impl): Likewise.
3709 (svqadd, svqsub): Use rtx_code_function instead.
3710
3711 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3712
3713 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3714 (HADDSUB, sur, addsub): Remove them.
3715
3716 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3717
3718 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3719 functions the same way, rather than singling out those that
3720 aren't mapped directly to optabs.
3721
3722 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3723
3724 * target.def (compatible_vector_types_p): New target hook.
3725 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3726 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3727 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3728 * doc/tm.texi: Regenerate.
3729 * gimple-expr.c: Include target.h.
3730 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3731 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3732 function.
3733 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3734 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3735 Use the original predicate if it already has a suitable type.
3736
3737 2020-01-09 Martin Jambor <mjambor@suse.cz>
3738
3739 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3740 resolve_speculation and redirect_call_stmt_to_callee static. Change
3741 return type of set_call_stmt to cgraph_edge *.
3742 * auto-profile.c (afdo_indirect_call): Adjust call to
3743 redirect_call_stmt_to_callee.
3744 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3745 make the this pointer explicit, adjust self-recursive calls and the
3746 call top make_direct. Return the resulting edge.
3747 (cgraph_edge::remove): Make this pointer explicit.
3748 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3749 (cgraph_edge::make_direct): Likewise, adjust call to
3750 resolve_speculation.
3751 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3752 call to set_call_stmt.
3753 (cgraph_update_edges_for_call_stmt_node): Update call to
3754 set_call_stmt and remove.
3755 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3756 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3757 (cgraph_node::create_edge_including_clones): Moved "first" definition
3758 of edge to the block where it was used. Adjusted calls to
3759 set_call_stmt.
3760 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3761 cgraph_edge::remove.
3762 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3763 make_direct and redirect_call_stmt_to_callee.
3764 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3765 resolve_speculation and make_direct.
3766 * ipa-inline-transform.c (inline_transform): Adjust call to
3767 redirect_call_stmt_to_callee.
3768 (check_speculations_1):: Adjust call to resolve_speculation.
3769 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3770 resolve-speculation.
3771 (inline_small_functions): Adjust call to resolve_speculation.
3772 (ipa_inline): Likewise.
3773 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3774 make_direct.
3775 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3776 safe with regards to edge removal, adjust calls to
3777 redirect_call_stmt_to_callee.
3778 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3779 and redirect_call_stmt_to_callee.
3780 * multiple_target.c (create_dispatcher_calls): Adjust call to
3781 redirect_call_stmt_to_callee
3782 (redirect_to_specific_clone): Likewise.
3783 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3784 Adjust calls to cgraph_edge::remove.
3785 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3786 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3787 (expand_call_inline): Adjust call to cgraph_edge::remove.
3788
3789 2020-01-09 Martin Liska <mliska@suse.cz>
3790
3791 * params.opt: Set Optimization for
3792 param_max_speculative_devirt_maydefs.
3793
3794 2020-01-09 Martin Sebor <msebor@redhat.com>
3795
3796 PR middle-end/93200
3797 PR fortran/92956
3798 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3799
3800 2020-01-09 Martin Liska <mliska@suse.cz>
3801
3802 * auto-profile.c (auto_profile): Use opt_for_fn
3803 for a parameter.
3804 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3805 (propagate_vals_across_arith_jfunc): Likewise.
3806 (hint_time_bonus): Likewise.
3807 (incorporate_penalties): Likewise.
3808 (good_cloning_opportunity_p): Likewise.
3809 (perform_estimation_of_a_value): Likewise.
3810 (estimate_local_effects): Likewise.
3811 (ipcp_propagate_stage): Likewise.
3812 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3813 (set_switch_stmt_execution_predicate): Likewise.
3814 (analyze_function_body): Likewise.
3815 * ipa-inline-analysis.c (offline_size): Likewise.
3816 * ipa-inline.c (early_inliner): Likewise.
3817 * ipa-prop.c (ipa_analyze_node): Likewise.
3818 (ipcp_transform_function): Likewise.
3819 * ipa-sra.c (process_scan_results): Likewise.
3820 (ipa_sra_summarize_function): Likewise.
3821 * params.opt: Rename ipcp-unit-growth to
3822 ipa-cp-unit-growth. Add Optimization for various
3823 IPA-related parameters.
3824
3825 2020-01-09 Richard Biener <rguenther@suse.de>
3826
3827 PR middle-end/93054
3828 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3829
3830 2020-01-09 Richard Biener <rguenther@suse.de>
3831
3832 PR tree-optimization/93040
3833 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3834
3835 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3836
3837 * common/config/avr/avr-common.c (avr_option_optimization_table)
3838 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3839
3840 2020-01-09 Martin Liska <mliska@suse.cz>
3841
3842 * cgraphclones.c (symbol_table::materialize_all_clones):
3843 Use cgraph_node::dump_name.
3844
3845 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3846
3847 PR inline-asm/93202
3848 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3849 output_operand_lossage instead of gcc_unreachable.
3850 * doc/md.texi (riscv f constraint): Fix typo.
3851
3852 PR target/93141
3853 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3854 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3855 CONST_SCALAR_INT_P instead of CONST_INT_P.
3856 (*subv<mode>4_1): Rename to ...
3857 (subv<mode>4_1): ... this.
3858 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3859 define_insn_and_split patterns.
3860 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3861 patterns.
3862
3863 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3864
3865 * vec.c (class selftest::count_dtor): New class.
3866 (selftest::test_auto_delete_vec): New test.
3867 (selftest::vec_c_tests): Call it.
3868 * vec.h (class auto_delete_vec): New class template.
3869 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3870
3871 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3872
3873 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3874
3875 2020-01-08 Jim Wilson <jimw@sifive.com>
3876
3877 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3878 use of TLS_MODEL_LOCAL_EXEC when not pic.
3879
3880 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3881
3882 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3883 memory leak.
3884
3885 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3886
3887 PR target/93187
3888 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3889 *stack_protect_set_3 peephole2): Also check that the second
3890 insns source is general_operand.
3891
3892 PR target/93174
3893 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3894 predicate for output operand instead of register_operand.
3895 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3896 memory destination and non-memory operands[2].
3897
3898 2020-01-08 Martin Liska <mliska@suse.cz>
3899
3900 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3901 ::dump_asm_name instead of (::name or ::asm_name).
3902 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3903 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3904 (analyze_functions): Likewise.
3905 (expand_all_functions): Likewise.
3906 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3907 (propagate_bits_across_jump_function): Likewise.
3908 (dump_profile_updates): Likewise.
3909 (ipcp_store_bits_results): Likewise.
3910 (ipcp_store_vr_results): Likewise.
3911 * ipa-devirt.c (dump_targets): Likewise.
3912 * ipa-fnsummary.c (analyze_function_body): Likewise.
3913 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3914 (process_hsa_functions): Likewise.
3915 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3916 (set_alias_uids): Likewise.
3917 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3918 * ipa-inline.c (recursive_inlining): Likewise.
3919 (inline_to_all_callers_1): Likewise.
3920 (ipa_inline): Likewise.
3921 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3922 (ipa_propagate_frequency): Likewise.
3923 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3924 (remove_described_reference): Likewise.
3925 * ipa-pure-const.c (worse_state): Likewise.
3926 (check_retval_uses): Likewise.
3927 (analyze_function): Likewise.
3928 (propagate_pure_const): Likewise.
3929 (propagate_nothrow): Likewise.
3930 (dump_malloc_lattice): Likewise.
3931 (propagate_malloc): Likewise.
3932 (pass_local_pure_const::execute): Likewise.
3933 * ipa-visibility.c (optimize_weakref): Likewise.
3934 (function_and_variable_visibility): Likewise.
3935 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3936 (ipa_discover_variable_flags): Likewise.
3937 * lto-streamer-out.c (output_function): Likewise.
3938 (output_constructor): Likewise.
3939 * tree-inline.c (copy_bb): Likewise.
3940 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3941 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3942
3943 2020-01-08 Richard Biener <rguenther@suse.de>
3944
3945 PR middle-end/93199
3946 * tree-eh.c (sink_clobbers): Update virtual operands for
3947 the first and last stmt only. Add a dry-run capability.
3948 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3949 after CFG manipulations and in RPO order to catch all
3950 secondary opportunities reliably.
3951
3952 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3953
3954 PR target/93182
3955 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3956
3957 2019-01-08 Richard Biener <rguenther@suse.de>
3958
3959 PR middle-end/93199
3960 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3961 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3962 virtual operand, also updating SSA use.
3963 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3964 Update stmt after resetting virtual operand.
3965 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3966 * gimple-iterator.c (gsi_remove): When not removing the stmt
3967 permanently do not delink immediate uses or mark the stmt modified.
3968
3969 2020-01-08 Martin Liska <mliska@suse.cz>
3970
3971 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3972 (ipa_call_context::estimate_size_and_time): Likewise.
3973 (inline_analyze_function): Likewise.
3974
3975 2020-01-08 Martin Liska <mliska@suse.cz>
3976
3977 * cgraph.c (cgraph_node::dump): Use systematically
3978 dump_asm_name.
3979
3980 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3981
3982 Add -nodevicespecs option for avr.
3983
3984 PR target/93182
3985 * config/avr/avr.opt (-nodevicespecs): New driver option.
3986 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3987 "-specs=device-specs/..." if that option is not set.
3988 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3989
3990 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3991
3992 Implement 64-bit double functions for avr.
3993
3994 PR target/92055
3995 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3996 --with-double-comparison.
3997 * doc/install.texi: Document them.
3998 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3999 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4000 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4001 * doc/invoke.texi (AVR Built-in Macros): Document them.
4002 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4003 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4004 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4005
4006 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4007
4008 PR target/93188
4009 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4010 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4011 when only building rm-profile multilibs.
4012
4013 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4014
4015 PR ipa/93084
4016 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4017 lattice for a value to check.
4018 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4019 finite propagation in self-recursive scc.
4020
4021 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4022
4023 * ipa-inline.c (caller_growth_limits): Restore the AND.
4024
4025 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4026
4027 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4028 (VEC_ALLREG_ALT): New iterator.
4029 (VEC_ALLREG_INT_MODE): New iterator.
4030 (VCMP_MODE): New iterator.
4031 (VCMP_MODE_INT): New iterator.
4032 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4033 (vec_cmp<u>v64qidi): New define_expand.
4034 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4035 (vec_cmpu<mode>di_exec): New define_expand.
4036 (vec_cmp<u>v64qidi_exec): New define_expand.
4037 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4038 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4039 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4040 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4041 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4042 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4043 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4044 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4045 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4046 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4047 this.
4048 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4049 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4050
4051 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4052
4053 * config/gcn/constraints.md (DA): Update description and match.
4054 (DB): Likewise.
4055 (Db): New constraint.
4056 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4057 parameter.
4058 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4059 Implement 'Db' mixed immediate type.
4060 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4061 (addcv64si3_dup<exec_vcc>): Delete.
4062 (subcv64si3<exec_vcc>): Rework constraints.
4063 (addv64di3): Rework constraints.
4064 (addv64di3_exec): Rework constraints.
4065 (subv64di3): Rework constraints.
4066 (addv64di3_dup): Delete.
4067 (addv64di3_dup_exec): Delete.
4068 (addv64di3_zext): Rework constraints.
4069 (addv64di3_zext_exec): Rework constraints.
4070 (addv64di3_zext_dup): Rework constraints.
4071 (addv64di3_zext_dup_exec): Rework constraints.
4072 (addv64di3_zext_dup2): Rework constraints.
4073 (addv64di3_zext_dup2_exec): Rework constraints.
4074 (addv64di3_sext_dup2): Rework constraints.
4075 (addv64di3_sext_dup2_exec): Rework constraints.
4076
4077 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4078
4079 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4080 existing target checks.
4081
4082 2020-01-07 Richard Biener <rguenther@suse.de>
4083
4084 * doc/install.texi: Bump minimal supported MPC version.
4085
4086 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4087
4088 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4089 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4090 * langhooks.c: Include stor-layout.h.
4091 (lhd_simulate_enum_decl): New function.
4092 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4093 handle_arm_sve_h for the LTO frontend.
4094 (register_vector_type): Cope with null returns from pushdecl.
4095
4096 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4097
4098 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4099 (aarch64_sve::nvectors_if_data_type): Replace with...
4100 (aarch64_sve::builtin_type_p): ...this.
4101 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4102 (find_vector_type): Delete.
4103 (add_sve_type_attribute): New function.
4104 (lookup_sve_type_attribute): Likewise.
4105 (register_builtin_types): Add an "SVE type" attribute to each type.
4106 (register_tuple_type): Likewise.
4107 (svbool_type_p, nvectors_if_data_type): Delete.
4108 (mangle_builtin_type): Use lookup_sve_type_attribute.
4109 (builtin_type_p): Likewise. Add an overload that returns the
4110 number of constituent vector and predicate registers.
4111 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4112 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4113 instead of aarch64_sve_argument_p.
4114 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4115 (aarch64_pass_by_reference): Likewise.
4116 (aarch64_function_value_1): Likewise.
4117 (aarch64_return_in_memory): Likewise.
4118 (aarch64_layout_arg): Likewise.
4119
4120 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4121
4122 PR tree-optimization/93156
4123 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4124 least significant bit is always clear.
4125
4126 PR tree-optimization/93118
4127 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4128 simplifier with two intermediate conversions.
4129
4130 2020-01-07 Martin Liska <mliska@suse.cz>
4131
4132 * params.opt: Add Optimization for various parameters.
4133
4134 2020-01-07 Martin Liska <mliska@suse.cz>
4135
4136 PR ipa/83411
4137 * doc/extend.texi: Explain cloning for target_clone
4138 attribute.
4139
4140 2020-01-07 Martin Liska <mliska@suse.cz>
4141
4142 PR tree-optimization/92860
4143 * common.opt: Make in Optimization option
4144 as it is affected by -O0, which is an Optimization
4145 option.
4146 * tree-inline.c (tree_inlinable_function_p):
4147 Use opt_for_fn for warn_inline.
4148 (expand_call_inline): Likewise.
4149
4150 2020-01-07 Martin Liska <mliska@suse.cz>
4151
4152 PR tree-optimization/92860
4153 * common.opt: Make flag_ree as optimization
4154 attribute.
4155
4156 2020-01-07 Martin Liska <mliska@suse.cz>
4157
4158 PR optimization/92860
4159 * params.opt: Mark param_min_crossjump_insns with Optimization
4160 keyword.
4161
4162 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4163
4164 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4165 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4166
4167 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4168
4169 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4170 helper function to return the valid addressing formats for a given
4171 hard register and mode.
4172 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4173
4174 * config/rs6000/constraints.md (Q constraint): Update
4175 documentation.
4176 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4177 documentation.
4178
4179 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4180 Use 'Q' for doing vector extract from memory.
4181 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4182 memory.
4183 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4184 doing vector extract from memory.
4185 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4186 extract from memory.
4187
4188 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4189 for the offset being 34-bits when -mcpu=future is used.
4190
4191 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4192
4193 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4194 instead of cmpib_comparison_operator in cmpib patterns.
4195 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4196 of cmpib_comparison_operator. Revise comment.
4197
4198 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4199
4200 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4201 in an IFN_DIV_POW2 node to be equal.
4202
4203 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4204
4205 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4206 (vect_check_scalar_mask): ...this.
4207 (vectorizable_store, vectorizable_load): Update call accordingly.
4208 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4209 argument in calls to conditional internal functions.
4210
4211 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4212
4213 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4214 '0' matching inputs.
4215 (subv64di3_exec): Likewise.
4216
4217 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4218
4219 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4220 * doc/md.texi (movstr): Likewise.
4221
4222 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4223
4224 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4225 clobber.
4226
4227 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4228
4229 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4230 Depend on...
4231 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4232 to a temporary file and use move-if-change to update the real
4233 file where necessary.
4234
4235 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4236
4237 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4238 rather than Upa for CPY /M.
4239
4240 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4241
4242 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4243 immediate.
4244
4245 2020-01-06 Martin Liska <mliska@suse.cz>
4246
4247 PR tree-optimization/92860
4248 * params.opt: Mark param_max_combine_insns with Optimization
4249 keyword.
4250
4251 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4252
4253 PR target/93141
4254 * config/i386/i386.md (SWIDWI): New mode iterator.
4255 (DWI, dwi): Add TImode variants.
4256 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4257 <general_hilo_operand> instead of <general_operand>. Use
4258 CONST_SCALAR_INT_P instead of CONST_INT_P.
4259 (*addv<mode>4_1): Rename to ...
4260 (addv<mode>4_1): ... this.
4261 (QWI): New mode attribute.
4262 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4263 define_insn_and_split patterns.
4264 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4265 patterns.
4266 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4267 <general_hilo_operand> instead of <general_operand>.
4268 (*addcarry<mode>_1): New define_insn.
4269 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4270
4271 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4272
4273 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4274 Use "call" instead of "set".
4275
4276 2020-01-03 Martin Jambor <mjambor@suse.cz>
4277
4278 PR ipa/92917
4279 * ipa-cp.c (print_all_lattices): Skip functions without info.
4280
4281 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4282
4283 PR target/93089
4284 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4285 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4286 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4287 for 'e' simd clones.
4288
4289 PR target/93089
4290 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4291 entry.
4292 (mprefer-vector-width=): Add Save.
4293 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4294 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4295 (ix86_debug_options, ix86_function_specific_print): Adjust
4296 ix86_target_string callers.
4297 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4298 (ix86_valid_target_attribute_tree): Likewise.
4299 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4300 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4301 ix86_target_string caller.
4302
4303 PR target/93110
4304 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4305 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4306 instead of gen_int_shift_amount + convert_modes.
4307
4308 PR rtl-optimization/93088
4309 * loop-iv.c (find_single_def_src): Punt after looking through
4310 128 reg copies for regs with single definitions. Move definitions
4311 to first uses.
4312
4313 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4314
4315 * config/arm/arm-c.c (arm_cpu_builtins): Define
4316 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4317 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4318 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4319 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4320 * config/arm/arm-tables.opt: Regenerated.
4321 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4322 arm_arch_i8mm and arm_arch_bf16 when enabled.
4323 * config/arm/arm.h (TARGET_I8MM): New macro.
4324 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4325 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4326 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4327 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4328 (v8_6_a_simd_variants): New.
4329 (v8_*_a_simd_variants): Add i8mm and bf16.
4330 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4331
4332 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4333
4334 PR ipa/93087
4335 * predict.c (compute_function_frequency): Don't call
4336 warn_function_cold on functions that already have cold attribute.
4337
4338 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4339
4340 PR target/67834
4341 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4342 COMDAT group function labels in .data.rel.ro.local section.
4343 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4344
4345 PR target/93111
4346 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4347 comparison_operator in B and S integer comparisons. Likewise, use
4348 ordered_comparison_operator instead of cmpib_comparison_operator in
4349 cmpib patterns.
4350 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4351
4352 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4353
4354 Update copyright years.
4355
4356 * gcc.c (process_command): Update copyright notice dates.
4357 * gcov-dump.c (print_version): Ditto.
4358 * gcov.c (print_version): Ditto.
4359 * gcov-tool.c (print_version): Ditto.
4360 * gengtype.c (create_file): Ditto.
4361 * doc/cpp.texi: Bump @copying's copyright year.
4362 * doc/cppinternals.texi: Ditto.
4363 * doc/gcc.texi: Ditto.
4364 * doc/gccint.texi: Ditto.
4365 * doc/gcov.texi: Ditto.
4366 * doc/install.texi: Ditto.
4367 * doc/invoke.texi: Ditto.
4368
4369 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4370
4371 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4372 summary.
4373
4374 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4375
4376 PR tree-optimization/93098
4377 * match.pd (popcount): For shift amounts, use integer_onep
4378 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4379 tests. Make sure that precision is power of two larger than or equal
4380 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4381 instead of ULL suffixed constants. Formatting fixes.
4382 \f
4383 Copyright (C) 2020 Free Software Foundation, Inc.
4384
4385 Copying and distribution of this file, with or without modification,
4386 are permitted in any medium without royalty provided the copyright
4387 notice and this notice are preserved.
This page took 0.211341 seconds and 6 git commands to generate.