1 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
3 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
6 2021-11-04 Tamar Christina <tamar.christina@arm.com>
8 * config/aarch64/aarch64-builtins.c
9 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
10 ashr_simd, lshr, lshr_simd.
11 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
12 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
13 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
15 2021-11-04 Tamar Christina <tamar.christina@arm.com>
17 * match.pd: New negate+shift pattern.
19 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
21 PR tree-optimization/103079
22 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
24 (gimple_range_calc_op2): Ditto.
26 2021-11-04 Martin Jambor <mjambor@suse.cz>
29 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
30 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
31 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
32 parameter to mark_dead_statements.
33 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
34 (ipa_param_body_adjustments::mark_dead_statements): New parameter
35 debugstack, push into it all SSA names used in debug statements,
36 produce m_dead_ssa_debug_equiv mapping for the removed param.
37 (replace_with_mapped_expr): New function.
38 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
39 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
40 (ipa_param_body_adjustments::common_initialization): Gather and
41 procecc SSA which will be removed but are in debug statements. Simplify.
42 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
44 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
45 when avoiding a copy of an unnecessary statement. Remap removed SSA
46 names in existing debug statements.
47 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
48 parameters if we have already done so.
50 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
53 * gimple.c (gimple_call_static_chain_flags): Handle case when
54 nested function does not bind locally.
56 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
58 * config/aarch64/aarch64.c (aarch64_function_value): Generate
59 a register rtx for Neon vector-tuple modes.
60 (aarch64_layout_arg): Likewise.
62 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
64 * lower-subreg.c (simple_move): Prevent decomposition if
65 modes are not tieable.
67 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
68 Richard Sandiford <richard.sandiford@arm.com>
70 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
71 (v2x4hi_UP): Likewise.
72 (v2x4hf_UP): Likewise.
73 (v2x4bf_UP): Likewise.
74 (v2x2si_UP): Likewise.
75 (v2x2sf_UP): Likewise.
76 (v2x1di_UP): Likewise.
77 (v2x1df_UP): Likewise.
78 (v2x16qi_UP): Likewise.
79 (v2x8hi_UP): Likewise.
80 (v2x8hf_UP): Likewise.
81 (v2x8bf_UP): Likewise.
82 (v2x4si_UP): Likewise.
83 (v2x4sf_UP): Likewise.
84 (v2x2di_UP): Likewise.
85 (v2x2df_UP): Likewise.
86 (v3x8qi_UP): Likewise.
87 (v3x4hi_UP): Likewise.
88 (v3x4hf_UP): Likewise.
89 (v3x4bf_UP): Likewise.
90 (v3x2si_UP): Likewise.
91 (v3x2sf_UP): Likewise.
92 (v3x1di_UP): Likewise.
93 (v3x1df_UP): Likewise.
94 (v3x16qi_UP): Likewise.
95 (v3x8hi_UP): Likewise.
96 (v3x8hf_UP): Likewise.
97 (v3x8bf_UP): Likewise.
98 (v3x4si_UP): Likewise.
99 (v3x4sf_UP): Likewise.
100 (v3x2di_UP): Likewise.
101 (v3x2df_UP): Likewise.
102 (v4x8qi_UP): Likewise.
103 (v4x4hi_UP): Likewise.
104 (v4x4hf_UP): Likewise.
105 (v4x4bf_UP): Likewise.
106 (v4x2si_UP): Likewise.
107 (v4x2sf_UP): Likewise.
108 (v4x1di_UP): Likewise.
109 (v4x1df_UP): Likewise.
110 (v4x16qi_UP): Likewise.
111 (v4x8hi_UP): Likewise.
112 (v4x8hf_UP): Likewise.
113 (v4x8bf_UP): Likewise.
114 (v4x4si_UP): Likewise.
115 (v4x4sf_UP): Likewise.
116 (v4x2di_UP): Likewise.
117 (v4x2df_UP): Likewise.
118 (TYPES_GETREGP): Delete.
119 (TYPES_SETREGP): Likewise.
120 (TYPES_LOADSTRUCT_U): Define.
121 (TYPES_LOADSTRUCT_P): Likewise.
122 (TYPES_LOADSTRUCT_LANE_U): Likewise.
123 (TYPES_LOADSTRUCT_LANE_P): Likewise.
124 (TYPES_STORE1P): Move for consistency.
125 (TYPES_STORESTRUCT_U): Define.
126 (TYPES_STORESTRUCT_P): Likewise.
127 (TYPES_STORESTRUCT_LANE_U): Likewise.
128 (TYPES_STORESTRUCT_LANE_P): Likewise.
129 (aarch64_simd_tuple_types): Define.
130 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
131 (aarch64_init_simd_builtin_functions): Update frontend lookup
132 for builtin functions after handling arm_neon.h pragma.
133 (register_tuple_type): Manually set modes of single-integer
134 tuple types. Record tuple types.
135 * config/aarch64/aarch64-modes.def
136 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
137 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
138 (SVE_MODES): Give single-vector modes priority over vector-
140 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
141 be after all single-vector modes.
142 * config/aarch64/aarch64-simd-builtins.def: Update builtin
143 generator macros to reflect modifications to the backend
145 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
146 Use vector-tuple mode iterator and rename to...
147 (aarch64_simd_ld2<vstruct_elt>): This.
148 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
150 (aarch64_simd_ld2r<vstruct_elt>): This.
151 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
152 iterator and rename to...
153 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
154 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
156 (vec_load_lanes<mode><vstruct_elt>): This.
157 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
159 (aarch64_simd_st2<vstruct_elt>): This.
160 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
161 iterator and rename to...
162 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
163 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
165 (vec_store_lanes<mode><vstruct_elt>): This.
166 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
168 (aarch64_simd_ld3<vstruct_elt>): This.
169 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
171 (aarch64_simd_ld3r<vstruct_elt>): This.
172 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
173 iterator and rename to...
174 (vec_load_lanesci<mode>): This.
175 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
177 (aarch64_simd_st3<vstruct_elt>): This.
178 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
179 iterator and rename to...
180 (vec_store_lanesci<mode>): This.
181 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
183 (aarch64_simd_ld4<vstruct_elt>): This.
184 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
186 (aarch64_simd_ld4r<vstruct_elt>): This.
187 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
188 iterator and rename to...
189 (vec_load_lanesxi<mode>): This.
190 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
192 (aarch64_simd_st4<vstruct_elt>): This.
193 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
194 iterator and rename to...
195 (vec_store_lanesxi<mode>): This.
196 (mov<mode>): Define for Neon vector-tuple modes.
197 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
199 (aarch64_ld1x3<vstruct_elt>): This.
200 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
202 (aarch64_ld1_x3_<vstruct_elt>): This.
203 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
205 (aarch64_ld1x4<vstruct_elt>): This.
206 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
208 (aarch64_ld1_x4_<vstruct_elt>): This.
209 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
211 (aarch64_st1x2<vstruct_elt>): This.
212 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
214 (aarch64_st1_x2_<vstruct_elt>): This.
215 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
217 (aarch64_st1x3<vstruct_elt>): This.
218 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
220 (aarch64_st1_x3_<vstruct_elt>): This.
221 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
223 (aarch64_st1x4<vstruct_elt>): This.
224 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
226 (aarch64_st1_x4_<vstruct_elt>): This.
227 (*aarch64_mov<mode>): Define for vector-tuple modes.
228 (*aarch64_be_mov<mode>): Likewise.
229 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
230 mode iterator and rename to...
231 (aarch64_ld<nregs>r<vstruct_elt>): This.
232 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
234 (aarch64_ld2<vstruct_elt>_dreg): This.
235 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
237 (aarch64_ld3<vstruct_elt>_dreg): This.
238 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
240 (aarch64_ld4<vstruct_elt>_dreg): This.
241 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
242 iterator and rename to...
243 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
244 iterator and rename to...
245 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
246 (aarch64_ld1x2<VQ:mode>): Delete.
247 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
249 (aarch64_ld1x2<vstruct_elt>): This.
250 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
251 tuple mode iterator and rename to...
252 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
253 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
254 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
255 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
257 (aarch64_st2<vstruct_elt>_dreg): This.
258 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
260 (aarch64_st3<vstruct_elt>_dreg): This.
261 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
263 (aarch64_st4<vstruct_elt>_dreg): This.
264 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
265 iterator and rename to...
266 (aarch64_st<nregs><vstruct_elt>): This.
267 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
268 iterator and rename to aarch64_st<nregs><vstruct_elt>.
269 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
270 tuple mode iterator and rename to...
271 (aarch64_st<nregs>_lane<vstruct_elt>): This.
272 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
273 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
275 (aarch64_simd_ld1<vstruct_elt>_x2): This.
276 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
277 Refactor to include new vector-tuple modes.
278 (aarch64_classify_vector_mode): Add cases for new vector-
280 (aarch64_advsimd_partial_struct_mode_p): Define.
281 (aarch64_advsimd_full_struct_mode_p): Likewise.
282 (aarch64_advsimd_vector_array_mode): Likewise.
283 (aarch64_sve_data_mode): Change location in file.
284 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
285 (aarch64_hard_regno_nregs): Handle case of partial Neon
287 (aarch64_classify_address): Refactor to include handling of
288 Neon vector-tuple modes.
289 (aarch64_print_operand): Print "d" for "%R" for a partial
290 Neon vector structure.
291 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
292 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
293 modes with scalar machines modes larger than 8 bytes.
294 (aarch64_can_change_mode_class): Don't allow changes between
295 partial and full Neon vector-structure modes.
296 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
297 builtin and remove boiler-plate code for opaque mode.
298 (vst2_lane_f32): Likewise.
299 (vst2_lane_f64): Likewise.
300 (vst2_lane_p8): Likewise.
301 (vst2_lane_p16): Likewise.
302 (vst2_lane_p64): Likewise.
303 (vst2_lane_s8): Likewise.
304 (vst2_lane_s16): Likewise.
305 (vst2_lane_s32): Likewise.
306 (vst2_lane_s64): Likewise.
307 (vst2_lane_u8): Likewise.
308 (vst2_lane_u16): Likewise.
309 (vst2_lane_u32): Likewise.
310 (vst2_lane_u64): Likewise.
311 (vst2q_lane_f16): Likewise.
312 (vst2q_lane_f32): Likewise.
313 (vst2q_lane_f64): Likewise.
314 (vst2q_lane_p8): Likewise.
315 (vst2q_lane_p16): Likewise.
316 (vst2q_lane_p64): Likewise.
317 (vst2q_lane_s8): Likewise.
318 (vst2q_lane_s16): Likewise.
319 (vst2q_lane_s32): Likewise.
320 (vst2q_lane_s64): Likewise.
321 (vst2q_lane_u8): Likewise.
322 (vst2q_lane_u16): Likewise.
323 (vst2q_lane_u32): Likewise.
324 (vst2q_lane_u64): Likewise.
325 (vst3_lane_f16): Likewise.
326 (vst3_lane_f32): Likewise.
327 (vst3_lane_f64): Likewise.
328 (vst3_lane_p8): Likewise.
329 (vst3_lane_p16): Likewise.
330 (vst3_lane_p64): Likewise.
331 (vst3_lane_s8): Likewise.
332 (vst3_lane_s16): Likewise.
333 (vst3_lane_s32): Likewise.
334 (vst3_lane_s64): Likewise.
335 (vst3_lane_u8): Likewise.
336 (vst3_lane_u16): Likewise.
337 (vst3_lane_u32): Likewise.
338 (vst3_lane_u64): Likewise.
339 (vst3q_lane_f16): Likewise.
340 (vst3q_lane_f32): Likewise.
341 (vst3q_lane_f64): Likewise.
342 (vst3q_lane_p8): Likewise.
343 (vst3q_lane_p16): Likewise.
344 (vst3q_lane_p64): Likewise.
345 (vst3q_lane_s8): Likewise.
346 (vst3q_lane_s16): Likewise.
347 (vst3q_lane_s32): Likewise.
348 (vst3q_lane_s64): Likewise.
349 (vst3q_lane_u8): Likewise.
350 (vst3q_lane_u16): Likewise.
351 (vst3q_lane_u32): Likewise.
352 (vst3q_lane_u64): Likewise.
353 (vst4_lane_f16): Likewise.
354 (vst4_lane_f32): Likewise.
355 (vst4_lane_f64): Likewise.
356 (vst4_lane_p8): Likewise.
357 (vst4_lane_p16): Likewise.
358 (vst4_lane_p64): Likewise.
359 (vst4_lane_s8): Likewise.
360 (vst4_lane_s16): Likewise.
361 (vst4_lane_s32): Likewise.
362 (vst4_lane_s64): Likewise.
363 (vst4_lane_u8): Likewise.
364 (vst4_lane_u16): Likewise.
365 (vst4_lane_u32): Likewise.
366 (vst4_lane_u64): Likewise.
367 (vst4q_lane_f16): Likewise.
368 (vst4q_lane_f32): Likewise.
369 (vst4q_lane_f64): Likewise.
370 (vst4q_lane_p8): Likewise.
371 (vst4q_lane_p16): Likewise.
372 (vst4q_lane_p64): Likewise.
373 (vst4q_lane_s8): Likewise.
374 (vst4q_lane_s16): Likewise.
375 (vst4q_lane_s32): Likewise.
376 (vst4q_lane_s64): Likewise.
377 (vst4q_lane_u8): Likewise.
378 (vst4q_lane_u16): Likewise.
379 (vst4q_lane_u32): Likewise.
380 (vst4q_lane_u64): Likewise.
381 (vtbl3_s8): Likewise.
382 (vtbl3_u8): Likewise.
383 (vtbl3_p8): Likewise.
384 (vtbl4_s8): Likewise.
385 (vtbl4_u8): Likewise.
386 (vtbl4_p8): Likewise.
387 (vld1_u8_x3): Likewise.
388 (vld1_s8_x3): Likewise.
389 (vld1_u16_x3): Likewise.
390 (vld1_s16_x3): Likewise.
391 (vld1_u32_x3): Likewise.
392 (vld1_s32_x3): Likewise.
393 (vld1_u64_x3): Likewise.
394 (vld1_s64_x3): Likewise.
395 (vld1_f16_x3): Likewise.
396 (vld1_f32_x3): Likewise.
397 (vld1_f64_x3): Likewise.
398 (vld1_p8_x3): Likewise.
399 (vld1_p16_x3): Likewise.
400 (vld1_p64_x3): Likewise.
401 (vld1q_u8_x3): Likewise.
402 (vld1q_s8_x3): Likewise.
403 (vld1q_u16_x3): Likewise.
404 (vld1q_s16_x3): Likewise.
405 (vld1q_u32_x3): Likewise.
406 (vld1q_s32_x3): Likewise.
407 (vld1q_u64_x3): Likewise.
408 (vld1q_s64_x3): Likewise.
409 (vld1q_f16_x3): Likewise.
410 (vld1q_f32_x3): Likewise.
411 (vld1q_f64_x3): Likewise.
412 (vld1q_p8_x3): Likewise.
413 (vld1q_p16_x3): Likewise.
414 (vld1q_p64_x3): Likewise.
415 (vld1_u8_x2): Likewise.
416 (vld1_s8_x2): Likewise.
417 (vld1_u16_x2): Likewise.
418 (vld1_s16_x2): Likewise.
419 (vld1_u32_x2): Likewise.
420 (vld1_s32_x2): Likewise.
421 (vld1_u64_x2): Likewise.
422 (vld1_s64_x2): Likewise.
423 (vld1_f16_x2): Likewise.
424 (vld1_f32_x2): Likewise.
425 (vld1_f64_x2): Likewise.
426 (vld1_p8_x2): Likewise.
427 (vld1_p16_x2): Likewise.
428 (vld1_p64_x2): Likewise.
429 (vld1q_u8_x2): Likewise.
430 (vld1q_s8_x2): Likewise.
431 (vld1q_u16_x2): Likewise.
432 (vld1q_s16_x2): Likewise.
433 (vld1q_u32_x2): Likewise.
434 (vld1q_s32_x2): Likewise.
435 (vld1q_u64_x2): Likewise.
436 (vld1q_s64_x2): Likewise.
437 (vld1q_f16_x2): Likewise.
438 (vld1q_f32_x2): Likewise.
439 (vld1q_f64_x2): Likewise.
440 (vld1q_p8_x2): Likewise.
441 (vld1q_p16_x2): Likewise.
442 (vld1q_p64_x2): Likewise.
443 (vld1_s8_x4): Likewise.
444 (vld1q_s8_x4): Likewise.
445 (vld1_s16_x4): Likewise.
446 (vld1q_s16_x4): Likewise.
447 (vld1_s32_x4): Likewise.
448 (vld1q_s32_x4): Likewise.
449 (vld1_u8_x4): Likewise.
450 (vld1q_u8_x4): Likewise.
451 (vld1_u16_x4): Likewise.
452 (vld1q_u16_x4): Likewise.
453 (vld1_u32_x4): Likewise.
454 (vld1q_u32_x4): Likewise.
455 (vld1_f16_x4): Likewise.
456 (vld1q_f16_x4): Likewise.
457 (vld1_f32_x4): Likewise.
458 (vld1q_f32_x4): Likewise.
459 (vld1_p8_x4): Likewise.
460 (vld1q_p8_x4): Likewise.
461 (vld1_p16_x4): Likewise.
462 (vld1q_p16_x4): Likewise.
463 (vld1_s64_x4): Likewise.
464 (vld1_u64_x4): Likewise.
465 (vld1_p64_x4): Likewise.
466 (vld1q_s64_x4): Likewise.
467 (vld1q_u64_x4): Likewise.
468 (vld1q_p64_x4): Likewise.
469 (vld1_f64_x4): Likewise.
470 (vld1q_f64_x4): Likewise.
471 (vld2_s64): Likewise.
472 (vld2_u64): Likewise.
473 (vld2_f64): Likewise.
476 (vld2_p64): Likewise.
477 (vld2_s16): Likewise.
478 (vld2_p16): Likewise.
479 (vld2_s32): Likewise.
481 (vld2_u16): Likewise.
482 (vld2_u32): Likewise.
483 (vld2_f16): Likewise.
484 (vld2_f32): Likewise.
485 (vld2q_s8): Likewise.
486 (vld2q_p8): Likewise.
487 (vld2q_s16): Likewise.
488 (vld2q_p16): Likewise.
489 (vld2q_p64): Likewise.
490 (vld2q_s32): Likewise.
491 (vld2q_s64): Likewise.
492 (vld2q_u8): Likewise.
493 (vld2q_u16): Likewise.
494 (vld2q_u32): Likewise.
495 (vld2q_u64): Likewise.
496 (vld2q_f16): Likewise.
497 (vld2q_f32): Likewise.
498 (vld2q_f64): Likewise.
499 (vld3_s64): Likewise.
500 (vld3_u64): Likewise.
501 (vld3_f64): Likewise.
504 (vld3_s16): Likewise.
505 (vld3_p16): Likewise.
506 (vld3_s32): Likewise.
508 (vld3_u16): Likewise.
509 (vld3_u32): Likewise.
510 (vld3_f16): Likewise.
511 (vld3_f32): Likewise.
512 (vld3_p64): Likewise.
513 (vld3q_s8): Likewise.
514 (vld3q_p8): Likewise.
515 (vld3q_s16): Likewise.
516 (vld3q_p16): Likewise.
517 (vld3q_s32): Likewise.
518 (vld3q_s64): Likewise.
519 (vld3q_u8): Likewise.
520 (vld3q_u16): Likewise.
521 (vld3q_u32): Likewise.
522 (vld3q_u64): Likewise.
523 (vld3q_f16): Likewise.
524 (vld3q_f32): Likewise.
525 (vld3q_f64): Likewise.
526 (vld3q_p64): Likewise.
527 (vld4_s64): Likewise.
528 (vld4_u64): Likewise.
529 (vld4_f64): Likewise.
532 (vld4_s16): Likewise.
533 (vld4_p16): Likewise.
534 (vld4_s32): Likewise.
536 (vld4_u16): Likewise.
537 (vld4_u32): Likewise.
538 (vld4_f16): Likewise.
539 (vld4_f32): Likewise.
540 (vld4_p64): Likewise.
541 (vld4q_s8): Likewise.
542 (vld4q_p8): Likewise.
543 (vld4q_s16): Likewise.
544 (vld4q_p16): Likewise.
545 (vld4q_s32): Likewise.
546 (vld4q_s64): Likewise.
547 (vld4q_u8): Likewise.
548 (vld4q_u16): Likewise.
549 (vld4q_u32): Likewise.
550 (vld4q_u64): Likewise.
551 (vld4q_f16): Likewise.
552 (vld4q_f32): Likewise.
553 (vld4q_f64): Likewise.
554 (vld4q_p64): Likewise.
555 (vld2_dup_s8): Likewise.
556 (vld2_dup_s16): Likewise.
557 (vld2_dup_s32): Likewise.
558 (vld2_dup_f16): Likewise.
559 (vld2_dup_f32): Likewise.
560 (vld2_dup_f64): Likewise.
561 (vld2_dup_u8): Likewise.
562 (vld2_dup_u16): Likewise.
563 (vld2_dup_u32): Likewise.
564 (vld2_dup_p8): Likewise.
565 (vld2_dup_p16): Likewise.
566 (vld2_dup_p64): Likewise.
567 (vld2_dup_s64): Likewise.
568 (vld2_dup_u64): Likewise.
569 (vld2q_dup_s8): Likewise.
570 (vld2q_dup_p8): Likewise.
571 (vld2q_dup_s16): Likewise.
572 (vld2q_dup_p16): Likewise.
573 (vld2q_dup_s32): Likewise.
574 (vld2q_dup_s64): Likewise.
575 (vld2q_dup_u8): Likewise.
576 (vld2q_dup_u16): Likewise.
577 (vld2q_dup_u32): Likewise.
578 (vld2q_dup_u64): Likewise.
579 (vld2q_dup_f16): Likewise.
580 (vld2q_dup_f32): Likewise.
581 (vld2q_dup_f64): Likewise.
582 (vld2q_dup_p64): Likewise.
583 (vld3_dup_s64): Likewise.
584 (vld3_dup_u64): Likewise.
585 (vld3_dup_f64): Likewise.
586 (vld3_dup_s8): Likewise.
587 (vld3_dup_p8): Likewise.
588 (vld3_dup_s16): Likewise.
589 (vld3_dup_p16): Likewise.
590 (vld3_dup_s32): Likewise.
591 (vld3_dup_u8): Likewise.
592 (vld3_dup_u16): Likewise.
593 (vld3_dup_u32): Likewise.
594 (vld3_dup_f16): Likewise.
595 (vld3_dup_f32): Likewise.
596 (vld3_dup_p64): Likewise.
597 (vld3q_dup_s8): Likewise.
598 (vld3q_dup_p8): Likewise.
599 (vld3q_dup_s16): Likewise.
600 (vld3q_dup_p16): Likewise.
601 (vld3q_dup_s32): Likewise.
602 (vld3q_dup_s64): Likewise.
603 (vld3q_dup_u8): Likewise.
604 (vld3q_dup_u16): Likewise.
605 (vld3q_dup_u32): Likewise.
606 (vld3q_dup_u64): Likewise.
607 (vld3q_dup_f16): Likewise.
608 (vld3q_dup_f32): Likewise.
609 (vld3q_dup_f64): Likewise.
610 (vld3q_dup_p64): Likewise.
611 (vld4_dup_s64): Likewise.
612 (vld4_dup_u64): Likewise.
613 (vld4_dup_f64): Likewise.
614 (vld4_dup_s8): Likewise.
615 (vld4_dup_p8): Likewise.
616 (vld4_dup_s16): Likewise.
617 (vld4_dup_p16): Likewise.
618 (vld4_dup_s32): Likewise.
619 (vld4_dup_u8): Likewise.
620 (vld4_dup_u16): Likewise.
621 (vld4_dup_u32): Likewise.
622 (vld4_dup_f16): Likewise.
623 (vld4_dup_f32): Likewise.
624 (vld4_dup_p64): Likewise.
625 (vld4q_dup_s8): Likewise.
626 (vld4q_dup_p8): Likewise.
627 (vld4q_dup_s16): Likewise.
628 (vld4q_dup_p16): Likewise.
629 (vld4q_dup_s32): Likewise.
630 (vld4q_dup_s64): Likewise.
631 (vld4q_dup_u8): Likewise.
632 (vld4q_dup_u16): Likewise.
633 (vld4q_dup_u32): Likewise.
634 (vld4q_dup_u64): Likewise.
635 (vld4q_dup_f16): Likewise.
636 (vld4q_dup_f32): Likewise.
637 (vld4q_dup_f64): Likewise.
638 (vld4q_dup_p64): Likewise.
639 (vld2_lane_u8): Likewise.
640 (vld2_lane_u16): Likewise.
641 (vld2_lane_u32): Likewise.
642 (vld2_lane_u64): Likewise.
643 (vld2_lane_s8): Likewise.
644 (vld2_lane_s16): Likewise.
645 (vld2_lane_s32): Likewise.
646 (vld2_lane_s64): Likewise.
647 (vld2_lane_f16): Likewise.
648 (vld2_lane_f32): Likewise.
649 (vld2_lane_f64): Likewise.
650 (vld2_lane_p8): Likewise.
651 (vld2_lane_p16): Likewise.
652 (vld2_lane_p64): Likewise.
653 (vld2q_lane_u8): Likewise.
654 (vld2q_lane_u16): Likewise.
655 (vld2q_lane_u32): Likewise.
656 (vld2q_lane_u64): Likewise.
657 (vld2q_lane_s8): Likewise.
658 (vld2q_lane_s16): Likewise.
659 (vld2q_lane_s32): Likewise.
660 (vld2q_lane_s64): Likewise.
661 (vld2q_lane_f16): Likewise.
662 (vld2q_lane_f32): Likewise.
663 (vld2q_lane_f64): Likewise.
664 (vld2q_lane_p8): Likewise.
665 (vld2q_lane_p16): Likewise.
666 (vld2q_lane_p64): Likewise.
667 (vld3_lane_u8): Likewise.
668 (vld3_lane_u16): Likewise.
669 (vld3_lane_u32): Likewise.
670 (vld3_lane_u64): Likewise.
671 (vld3_lane_s8): Likewise.
672 (vld3_lane_s16): Likewise.
673 (vld3_lane_s32): Likewise.
674 (vld3_lane_s64): Likewise.
675 (vld3_lane_f16): Likewise.
676 (vld3_lane_f32): Likewise.
677 (vld3_lane_f64): Likewise.
678 (vld3_lane_p8): Likewise.
679 (vld3_lane_p16): Likewise.
680 (vld3_lane_p64): Likewise.
681 (vld3q_lane_u8): Likewise.
682 (vld3q_lane_u16): Likewise.
683 (vld3q_lane_u32): Likewise.
684 (vld3q_lane_u64): Likewise.
685 (vld3q_lane_s8): Likewise.
686 (vld3q_lane_s16): Likewise.
687 (vld3q_lane_s32): Likewise.
688 (vld3q_lane_s64): Likewise.
689 (vld3q_lane_f16): Likewise.
690 (vld3q_lane_f32): Likewise.
691 (vld3q_lane_f64): Likewise.
692 (vld3q_lane_p8): Likewise.
693 (vld3q_lane_p16): Likewise.
694 (vld3q_lane_p64): Likewise.
695 (vld4_lane_u8): Likewise.
696 (vld4_lane_u16): Likewise.
697 (vld4_lane_u32): Likewise.
698 (vld4_lane_u64): Likewise.
699 (vld4_lane_s8): Likewise.
700 (vld4_lane_s16): Likewise.
701 (vld4_lane_s32): Likewise.
702 (vld4_lane_s64): Likewise.
703 (vld4_lane_f16): Likewise.
704 (vld4_lane_f32): Likewise.
705 (vld4_lane_f64): Likewise.
706 (vld4_lane_p8): Likewise.
707 (vld4_lane_p16): Likewise.
708 (vld4_lane_p64): Likewise.
709 (vld4q_lane_u8): Likewise.
710 (vld4q_lane_u16): Likewise.
711 (vld4q_lane_u32): Likewise.
712 (vld4q_lane_u64): Likewise.
713 (vld4q_lane_s8): Likewise.
714 (vld4q_lane_s16): Likewise.
715 (vld4q_lane_s32): Likewise.
716 (vld4q_lane_s64): Likewise.
717 (vld4q_lane_f16): Likewise.
718 (vld4q_lane_f32): Likewise.
719 (vld4q_lane_f64): Likewise.
720 (vld4q_lane_p8): Likewise.
721 (vld4q_lane_p16): Likewise.
722 (vld4q_lane_p64): Likewise.
723 (vqtbl2_s8): Likewise.
724 (vqtbl2_u8): Likewise.
725 (vqtbl2_p8): Likewise.
726 (vqtbl2q_s8): Likewise.
727 (vqtbl2q_u8): Likewise.
728 (vqtbl2q_p8): Likewise.
729 (vqtbl3_s8): Likewise.
730 (vqtbl3_u8): Likewise.
731 (vqtbl3_p8): Likewise.
732 (vqtbl3q_s8): Likewise.
733 (vqtbl3q_u8): Likewise.
734 (vqtbl3q_p8): Likewise.
735 (vqtbl4_s8): Likewise.
736 (vqtbl4_u8): Likewise.
737 (vqtbl4_p8): Likewise.
738 (vqtbl4q_s8): Likewise.
739 (vqtbl4q_u8): Likewise.
740 (vqtbl4q_p8): Likewise.
741 (vqtbx2_s8): Likewise.
742 (vqtbx2_u8): Likewise.
743 (vqtbx2_p8): Likewise.
744 (vqtbx2q_s8): Likewise.
745 (vqtbx2q_u8): Likewise.
746 (vqtbx2q_p8): Likewise.
747 (vqtbx3_s8): Likewise.
748 (vqtbx3_u8): Likewise.
749 (vqtbx3_p8): Likewise.
750 (vqtbx3q_s8): Likewise.
751 (vqtbx3q_u8): Likewise.
752 (vqtbx3q_p8): Likewise.
753 (vqtbx4_s8): Likewise.
754 (vqtbx4_u8): Likewise.
755 (vqtbx4_p8): Likewise.
756 (vqtbx4q_s8): Likewise.
757 (vqtbx4q_u8): Likewise.
758 (vqtbx4q_p8): Likewise.
759 (vst1_s64_x2): Likewise.
760 (vst1_u64_x2): Likewise.
761 (vst1_f64_x2): Likewise.
762 (vst1_s8_x2): Likewise.
763 (vst1_p8_x2): Likewise.
764 (vst1_s16_x2): Likewise.
765 (vst1_p16_x2): Likewise.
766 (vst1_s32_x2): Likewise.
767 (vst1_u8_x2): Likewise.
768 (vst1_u16_x2): Likewise.
769 (vst1_u32_x2): Likewise.
770 (vst1_f16_x2): Likewise.
771 (vst1_f32_x2): Likewise.
772 (vst1_p64_x2): Likewise.
773 (vst1q_s8_x2): Likewise.
774 (vst1q_p8_x2): Likewise.
775 (vst1q_s16_x2): Likewise.
776 (vst1q_p16_x2): Likewise.
777 (vst1q_s32_x2): Likewise.
778 (vst1q_s64_x2): Likewise.
779 (vst1q_u8_x2): Likewise.
780 (vst1q_u16_x2): Likewise.
781 (vst1q_u32_x2): Likewise.
782 (vst1q_u64_x2): Likewise.
783 (vst1q_f16_x2): Likewise.
784 (vst1q_f32_x2): Likewise.
785 (vst1q_f64_x2): Likewise.
786 (vst1q_p64_x2): Likewise.
787 (vst1_s64_x3): Likewise.
788 (vst1_u64_x3): Likewise.
789 (vst1_f64_x3): Likewise.
790 (vst1_s8_x3): Likewise.
791 (vst1_p8_x3): Likewise.
792 (vst1_s16_x3): Likewise.
793 (vst1_p16_x3): Likewise.
794 (vst1_s32_x3): Likewise.
795 (vst1_u8_x3): Likewise.
796 (vst1_u16_x3): Likewise.
797 (vst1_u32_x3): Likewise.
798 (vst1_f16_x3): Likewise.
799 (vst1_f32_x3): Likewise.
800 (vst1_p64_x3): Likewise.
801 (vst1q_s8_x3): Likewise.
802 (vst1q_p8_x3): Likewise.
803 (vst1q_s16_x3): Likewise.
804 (vst1q_p16_x3): Likewise.
805 (vst1q_s32_x3): Likewise.
806 (vst1q_s64_x3): Likewise.
807 (vst1q_u8_x3): Likewise.
808 (vst1q_u16_x3): Likewise.
809 (vst1q_u32_x3): Likewise.
810 (vst1q_u64_x3): Likewise.
811 (vst1q_f16_x3): Likewise.
812 (vst1q_f32_x3): Likewise.
813 (vst1q_f64_x3): Likewise.
814 (vst1q_p64_x3): Likewise.
815 (vst1_s8_x4): Likewise.
816 (vst1q_s8_x4): Likewise.
817 (vst1_s16_x4): Likewise.
818 (vst1q_s16_x4): Likewise.
819 (vst1_s32_x4): Likewise.
820 (vst1q_s32_x4): Likewise.
821 (vst1_u8_x4): Likewise.
822 (vst1q_u8_x4): Likewise.
823 (vst1_u16_x4): Likewise.
824 (vst1q_u16_x4): Likewise.
825 (vst1_u32_x4): Likewise.
826 (vst1q_u32_x4): Likewise.
827 (vst1_f16_x4): Likewise.
828 (vst1q_f16_x4): Likewise.
829 (vst1_f32_x4): Likewise.
830 (vst1q_f32_x4): Likewise.
831 (vst1_p8_x4): Likewise.
832 (vst1q_p8_x4): Likewise.
833 (vst1_p16_x4): Likewise.
834 (vst1q_p16_x4): Likewise.
835 (vst1_s64_x4): Likewise.
836 (vst1_u64_x4): Likewise.
837 (vst1_p64_x4): Likewise.
838 (vst1q_s64_x4): Likewise.
839 (vst1q_u64_x4): Likewise.
840 (vst1q_p64_x4): Likewise.
841 (vst1_f64_x4): Likewise.
842 (vst1q_f64_x4): Likewise.
843 (vst2_s64): Likewise.
844 (vst2_u64): Likewise.
845 (vst2_f64): Likewise.
848 (vst2_s16): Likewise.
849 (vst2_p16): Likewise.
850 (vst2_s32): Likewise.
852 (vst2_u16): Likewise.
853 (vst2_u32): Likewise.
854 (vst2_f16): Likewise.
855 (vst2_f32): Likewise.
856 (vst2_p64): Likewise.
857 (vst2q_s8): Likewise.
858 (vst2q_p8): Likewise.
859 (vst2q_s16): Likewise.
860 (vst2q_p16): Likewise.
861 (vst2q_s32): Likewise.
862 (vst2q_s64): Likewise.
863 (vst2q_u8): Likewise.
864 (vst2q_u16): Likewise.
865 (vst2q_u32): Likewise.
866 (vst2q_u64): Likewise.
867 (vst2q_f16): Likewise.
868 (vst2q_f32): Likewise.
869 (vst2q_f64): Likewise.
870 (vst2q_p64): Likewise.
871 (vst3_s64): Likewise.
872 (vst3_u64): Likewise.
873 (vst3_f64): Likewise.
876 (vst3_s16): Likewise.
877 (vst3_p16): Likewise.
878 (vst3_s32): Likewise.
880 (vst3_u16): Likewise.
881 (vst3_u32): Likewise.
882 (vst3_f16): Likewise.
883 (vst3_f32): Likewise.
884 (vst3_p64): Likewise.
885 (vst3q_s8): Likewise.
886 (vst3q_p8): Likewise.
887 (vst3q_s16): Likewise.
888 (vst3q_p16): Likewise.
889 (vst3q_s32): Likewise.
890 (vst3q_s64): Likewise.
891 (vst3q_u8): Likewise.
892 (vst3q_u16): Likewise.
893 (vst3q_u32): Likewise.
894 (vst3q_u64): Likewise.
895 (vst3q_f16): Likewise.
896 (vst3q_f32): Likewise.
897 (vst3q_f64): Likewise.
898 (vst3q_p64): Likewise.
899 (vst4_s64): Likewise.
900 (vst4_u64): Likewise.
901 (vst4_f64): Likewise.
904 (vst4_s16): Likewise.
905 (vst4_p16): Likewise.
906 (vst4_s32): Likewise.
908 (vst4_u16): Likewise.
909 (vst4_u32): Likewise.
910 (vst4_f16): Likewise.
911 (vst4_f32): Likewise.
912 (vst4_p64): Likewise.
913 (vst4q_s8): Likewise.
914 (vst4q_p8): Likewise.
915 (vst4q_s16): Likewise.
916 (vst4q_p16): Likewise.
917 (vst4q_s32): Likewise.
918 (vst4q_s64): Likewise.
919 (vst4q_u8): Likewise.
920 (vst4q_u16): Likewise.
921 (vst4q_u32): Likewise.
922 (vst4q_u64): Likewise.
923 (vst4q_f16): Likewise.
924 (vst4q_f32): Likewise.
925 (vst4q_f64): Likewise.
926 (vst4q_p64): Likewise.
927 (vtbx4_s8): Likewise.
928 (vtbx4_u8): Likewise.
929 (vtbx4_p8): Likewise.
930 (vld1_bf16_x2): Likewise.
931 (vld1q_bf16_x2): Likewise.
932 (vld1_bf16_x3): Likewise.
933 (vld1q_bf16_x3): Likewise.
934 (vld1_bf16_x4): Likewise.
935 (vld1q_bf16_x4): Likewise.
936 (vld2_bf16): Likewise.
937 (vld2q_bf16): Likewise.
938 (vld2_dup_bf16): Likewise.
939 (vld2q_dup_bf16): Likewise.
940 (vld3_bf16): Likewise.
941 (vld3q_bf16): Likewise.
942 (vld3_dup_bf16): Likewise.
943 (vld3q_dup_bf16): Likewise.
944 (vld4_bf16): Likewise.
945 (vld4q_bf16): Likewise.
946 (vld4_dup_bf16): Likewise.
947 (vld4q_dup_bf16): Likewise.
948 (vst1_bf16_x2): Likewise.
949 (vst1q_bf16_x2): Likewise.
950 (vst1_bf16_x3): Likewise.
951 (vst1q_bf16_x3): Likewise.
952 (vst1_bf16_x4): Likewise.
953 (vst1q_bf16_x4): Likewise.
954 (vst2_bf16): Likewise.
955 (vst2q_bf16): Likewise.
956 (vst3_bf16): Likewise.
957 (vst3q_bf16): Likewise.
958 (vst4_bf16): Likewise.
959 (vst4q_bf16): Likewise.
960 (vld2_lane_bf16): Likewise.
961 (vld2q_lane_bf16): Likewise.
962 (vld3_lane_bf16): Likewise.
963 (vld3q_lane_bf16): Likewise.
964 (vld4_lane_bf16): Likewise.
965 (vld4q_lane_bf16): Likewise.
966 (vst2_lane_bf16): Likewise.
967 (vst2q_lane_bf16): Likewise.
968 (vst3_lane_bf16): Likewise.
969 (vst3q_lane_bf16): Likewise.
970 (vst4_lane_bf16): Likewise.
971 (vst4q_lane_bf16): Likewise.
972 * config/aarch64/geniterators.sh: Modify iterator regex to
973 match new vector-tuple modes.
974 * config/aarch64/iterators.md (insn_count): Extend mode
975 attribute with vector-tuple type information.
980 (VSTRUCT_2D): New mode iterator.
981 (VSTRUCT_2DNX): Likewise.
982 (VSTRUCT_2DX): Likewise.
983 (VSTRUCT_2Q): Likewise.
984 (VSTRUCT_2QD): Likewise.
985 (VSTRUCT_3D): Likewise.
986 (VSTRUCT_3DNX): Likewise.
987 (VSTRUCT_3DX): Likewise.
988 (VSTRUCT_3Q): Likewise.
989 (VSTRUCT_3QD): Likewise.
990 (VSTRUCT_4D): Likewise.
991 (VSTRUCT_4DNX): Likewise.
992 (VSTRUCT_4DX): Likewise.
993 (VSTRUCT_4Q): Likewise.
994 (VSTRUCT_4QD): Likewise.
995 (VSTRUCT_D): Likewise.
996 (VSTRUCT_Q): Likewise.
997 (VSTRUCT_QD): Likewise.
998 (VSTRUCT_ELT): New mode attribute.
999 (vstruct_elt): Likewise.
1000 * genmodes.c (VECTOR_MODE): Add default prefix and order
1002 (VECTOR_MODE_WITH_PREFIX): Define.
1003 (make_vector_mode): Add mode prefix and order parameters.
1005 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1007 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
1009 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1011 * expr.c (emit_group_load_1): Remove historic workaround.
1013 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1015 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
1016 Factor out main loop to...
1017 (aarch64_init_simd_builtin_functions): This new function.
1018 (register_tuple_type): Define.
1019 (aarch64_scalar_builtin_type_p): Define.
1020 (handle_arm_neon_h): Define.
1021 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
1022 pragma for arm_neon.h.
1023 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
1025 (handle_arm_neon_h): Likewise.
1026 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
1027 Remove static modifier.
1028 * config/aarch64/arm_neon.h (target): Remove Neon vector
1029 structure type definitions.
1031 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1033 PR tree-optimization/102943
1034 * gimple-range-path.cc (path_range_query::range_on_path_entry):
1035 Prefer range_of_expr unless there are no statements in the BB.
1037 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1039 PR tree-optimization/102943
1040 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
1041 Avoid duplicate calculation of paths.
1043 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1045 PR tree-optimization/102943
1046 * gimple-range-path.cc (path_range_query::compute_phi_relations):
1047 Only compute relations for SSA names in the import list.
1048 (path_range_query::compute_outgoing_relations): Same.
1049 * gimple-range-path.h (path_range_query::import_p): New.
1051 2021-11-04 Richard Biener <rguenther@suse.de>
1053 PR rtl-optimization/103075
1054 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
1055 false for a VOIDmode operand.
1057 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1059 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
1060 variables private and add "m_" to their names. Remove is_loop.
1061 (aarch64_record_potential_advsimd_unrolling): Replace with...
1062 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
1063 (aarch64_analyze_loop_vinfo): Replace with...
1064 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
1065 Move initialization of (m_)vec_flags to add_stmt_cost.
1066 (aarch64_analyze_bb_vinfo): Delete.
1067 (aarch64_count_ops): Replace with...
1068 (aarch64_vector_costs::count_ops): ...this.
1069 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
1070 using m_costing_for_scalar to test whether we're costing
1071 scalar or vector code.
1072 (aarch64_adjust_body_cost_sve): Replace with...
1073 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
1074 (aarch64_adjust_body_cost): Replace with...
1075 (aarch64_vector_costs::adjust_body_cost): ...this.
1076 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
1078 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1080 * target.def (targetm.vectorize.init_cost): Replace with...
1081 (targetm.vectorize.create_costs): ...this.
1082 (targetm.vectorize.add_stmt_cost): Delete.
1083 (targetm.vectorize.finish_cost): Likewise.
1084 (targetm.vectorize.destroy_cost_data): Likewise.
1085 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
1086 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1087 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1088 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1089 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1090 * doc/tm.texi: Regenerate.
1091 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
1093 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
1094 (vector_costs): New class.
1095 (init_cost): Take a vec_info and return a vector_costs.
1096 (dump_stmt_cost): Remove data parameter.
1097 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
1098 (add_stmt_costs): Likewise.
1099 (finish_cost): Replace data parameter with a vector_costs.
1100 (destroy_cost_data): Delete.
1101 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
1103 (vec_info::vec_info): Remove the target_cost_data parameter and
1104 initialize the member variable to null instead.
1105 (vec_info::~vec_info): Delete target_cost_data instead of calling
1107 (vector_costs::add_stmt_cost): New function.
1108 (vector_costs::finish_cost): Likewise.
1109 (vector_costs::record_stmt_cost): Likewise.
1110 (vector_costs::adjust_cost_for_freq): Likewise.
1111 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
1112 call to vec_info::vec_info.
1113 (vect_compute_single_scalar_iteration_cost): Update after above
1114 changes to costing interface.
1115 (vect_analyze_loop_operations): Likewise.
1116 (vect_estimate_min_profitable_iters): Likewise.
1117 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
1118 at the start_over point, where it needs to be recreated after
1119 trying without slp. Update retry code accordingly.
1120 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
1121 to vec_info::vec_info.
1122 (vect_slp_analyze_operation): Update after above changes to costing
1124 (vect_bb_vectorization_profitable_p): Likewise.
1125 * targhooks.h (default_init_cost): Replace with...
1126 (default_vectorize_create_costs): ...this.
1127 (default_add_stmt_cost): Delete.
1128 (default_finish_cost, default_destroy_cost_data): Likewise.
1129 * targhooks.c (default_init_cost): Replace with...
1130 (default_vectorize_create_costs): ...this.
1131 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
1132 (default_finish_cost, default_destroy_cost_data): Delete.
1133 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
1134 vector_costs. Add a constructor.
1135 (aarch64_init_cost): Replace with...
1136 (aarch64_vectorize_create_costs): ...this.
1137 (aarch64_add_stmt_cost): Replace with...
1138 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
1139 to adjust the cost for inner loops.
1140 (aarch64_finish_cost): Replace with...
1141 (aarch64_vector_costs::finish_cost): ...this.
1142 (aarch64_destroy_cost_data): Delete.
1143 (TARGET_VECTORIZE_INIT_COST): Replace with...
1144 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1145 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1146 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1147 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1148 * config/i386/i386.c (ix86_vector_costs): New structure.
1149 (ix86_init_cost): Replace with...
1150 (ix86_vectorize_create_costs): ...this.
1151 (ix86_add_stmt_cost): Replace with...
1152 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
1153 to adjust the cost for inner loops.
1154 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
1155 (TARGET_VECTORIZE_INIT_COST): Replace with...
1156 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1157 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1158 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1159 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1160 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
1161 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1162 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1163 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1164 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1165 (rs6000_cost_data): Inherit from vector_costs.
1166 Add a constructor. Drop loop_info, cost and costing_for_scalar
1167 in favor of the corresponding vector_costs member variables.
1168 Add "m_" to the names of the remaining member variables and
1170 (rs6000_density_test): Replace with...
1171 (rs6000_cost_data::density_test): ...this.
1172 (rs6000_init_cost): Replace with...
1173 (rs6000_vectorize_create_costs): ...this.
1174 (rs6000_update_target_cost_per_stmt): Replace with...
1175 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
1176 (rs6000_add_stmt_cost): Replace with...
1177 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
1178 to adjust the cost for inner loops.
1179 (rs6000_adjust_vect_cost_per_loop): Replace with...
1180 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
1181 (rs6000_finish_cost): Replace with...
1182 (rs6000_cost_data::finish_cost): ...this. Group loop code
1183 into a single if statement and pass the loop_vinfo down to
1185 (rs6000_destroy_cost_data): Delete.
1187 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1189 PR tree-optimization/103062
1190 PR tree-optimization/103062
1191 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
1192 Increase size of allocation by 1.
1193 (ssa_equiv_stack::push_replacement): Grow as needed.
1194 (ssa_equiv_stack::get_replacement): Same.
1195 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
1196 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
1197 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
1198 (pointer_equiv_analyzer::get_equiv): Same.
1199 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
1200 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
1201 const markers. Use auto_vec instead of tree *.
1203 2021-11-04 Richard Biener <rguenther@suse.de>
1205 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
1206 true parameter and inline valueization.
1207 (vn_nary_op_lookup_1): Inline valueization from ...
1208 (vn_nary_op_compute_hash): ... here and remove it here.
1209 * tree-ssa-pre.c (phi_translate_1): Do not valueize
1210 before vn_nary_lookup_pieces.
1211 (get_representative_for): Mark created SSA representatives
1214 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1216 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
1217 that the operand has a vector mode. Use subreg_lowpart_offset
1218 to test whether an index corresponds to the low part.
1220 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1222 * read-rtl.c: Remove dead !GENERATOR_FILE block.
1223 * read-rtl-function.c (function_reader::consolidate_singletons):
1224 Generate canonical CONST_VECTORs.
1226 2021-11-04 liuhongt <hongtao.liu@intel.com>
1229 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
1230 (regmem_or_bitnot_regmem_operand): .. and extend to handle
1232 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
1233 the operands which are required to be register_operand.
1234 (*<avx512>_vpternlog<mode>_2): Ditto.
1235 (*<avx512>_vpternlog<mode>_3): Ditto.
1236 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
1237 vector HFmodes since it's not a real AVX512FP16 instruction.
1239 2021-11-04 liuhongt <hongtao.liu@intel.com>
1242 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
1243 .COPYSIGN (a,b) when a and b are same type as the truncation
1244 type and has less precision than extend type.
1246 2021-11-04 Richard Biener <rguenther@suse.de>
1248 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
1251 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
1253 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
1254 (VI1_AVX512VNNI): Likewise.
1255 (SDOT_VPDP_SUF): New mode_attr.
1258 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
1259 vpdpwssd when VNNI targets available.
1260 (usdot_prod<mode>): New expander for vector QImode.
1262 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
1264 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
1265 parentheses to base and stride.
1266 (_tile_stream_loadd_internal): Likewise.
1267 (_tile_stored_internal): Likewise.
1269 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
1271 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
1272 arguments to `reg_class_subset_p'.
1274 2021-11-03 Joseph Myers <joseph@codesourcery.com>
1277 * fold-const.c (fold_init): New function.
1278 * fold-const.h (fold_init): New prototype.
1280 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1282 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
1284 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1286 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
1287 a range from edge if arg == phidef.
1289 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1291 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
1292 for constant before any other processing.
1294 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1296 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
1298 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1300 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
1301 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
1302 no explcit context and current_bb is set.
1303 (gimple_ranger::fold_stmt): New.
1304 * gimple-range.h (current_bb, fold_stmt): New.
1305 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
1307 2021-11-03 Richard Biener <rguenther@suse.de>
1309 PR tree-optimization/102970
1310 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
1311 when translating a MEM_REF over a backedge.
1313 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
1315 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
1316 * config/aarch64/aarch64-tune.md: Regenerate.
1317 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
1318 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
1319 * doc/invoke.texi: Add documentation for Ampere-1 core.
1321 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
1323 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
1325 (ldr_got_small_<mode>): Remove pattern.
1326 (ldr_got_small_sidi): Likewise.
1327 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
1328 GOT accesses as moves.
1329 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
1330 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
1331 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
1333 2021-11-03 Martin Liska <mliska@suse.cz>
1335 * gcov.c (read_line): Remove dead variable.
1337 2021-11-03 Martin Liska <mliska@suse.cz>
1340 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
1341 (ipa_fn_summary::account_size_time): Likewise.
1342 (edge_set_predicate): Likewise.
1343 (set_hint_predicate): Likewise.
1344 (add_freqcounting_predicate): Likewise.
1345 (evaluate_conditions_for_known_args): Likewise.
1346 (evaluate_properties_for_edge): Likewise.
1347 (remap_freqcounting_preds_after_dup): Likewise.
1348 (ipa_fn_summary_t::duplicate): Likewise.
1349 (set_cond_stmt_execution_predicate): Likewise.
1350 (set_switch_stmt_execution_predicate): Likewise.
1351 (compute_bb_predicates): Likewise.
1352 (will_be_nonconstant_expr_predicate): Likewise.
1353 (will_be_nonconstant_predicate): Likewise.
1354 (phi_result_unknown_predicate): Likewise.
1355 (predicate_for_phi_result): Likewise.
1356 (analyze_function_body): Likewise.
1357 (compute_fn_summary): Likewise.
1358 (summarize_calls_size_and_time): Likewise.
1359 (estimate_calls_size_and_time): Likewise.
1360 (ipa_call_context::estimate_size_and_time): Likewise.
1361 (remap_edge_summaries): Likewise.
1362 (remap_freqcounting_predicate): Likewise.
1363 (ipa_merge_fn_summary_after_inlining): Likewise.
1364 (ipa_update_overall_fn_summary): Likewise.
1365 (read_ipa_call_summary): Likewise.
1366 (inline_read_section): Likewise.
1367 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
1368 * ipa-predicate.c (predicate::add_clause): Likewise.
1369 (ipa_predicate::add_clause): Likewise.
1370 (predicate::or_with): Likewise.
1371 (ipa_predicate::or_with): Likewise.
1372 (predicate::evaluate): Likewise.
1373 (ipa_predicate::evaluate): Likewise.
1374 (predicate::probability): Likewise.
1375 (ipa_predicate::probability): Likewise.
1376 (dump_condition): Likewise.
1377 (dump_clause): Likewise.
1378 (predicate::dump): Likewise.
1379 (ipa_predicate::dump): Likewise.
1380 (predicate::debug): Likewise.
1381 (ipa_predicate::debug): Likewise.
1382 (predicate::remap_after_duplication): Likewise.
1383 (ipa_predicate::remap_after_duplication): Likewise.
1384 (predicate::remap_after_inlining): Likewise.
1385 (ipa_predicate::remap_after_inlining): Likewise.
1386 (predicate::stream_in): Likewise.
1387 (ipa_predicate::stream_in): Likewise.
1388 (predicate::stream_out): Likewise.
1389 (ipa_predicate::stream_out): Likewise.
1390 (add_condition): Likewise.
1391 * ipa-predicate.h (class predicate): Likewise.
1392 (class ipa_predicate): Likewise.
1393 (add_condition): Likewise.
1395 2021-11-03 Richard Biener <rguenther@suse.de>
1397 * bitmap.h (bitmap_bit_p): Change the return type to bool.
1398 * bitmap.c (bitmap_bit_p): Likewise.
1399 * sbitmap.h (bitmap_bit_p): Likewise.
1400 (bitmap_set_bit): Return whether the bit changed.
1401 (bitmap_clear_bit): Likewise.
1402 * tree-ssa.c (verify_vssa): Make use of the changed state
1403 from bitmap_set_bit.
1405 2021-11-03 Richard Biener <rguenther@suse.de>
1407 PR middle-end/103033
1408 * internal-fn.c (expand_DEFERRED_INIT): Elide the
1409 native_interpret_expr path in favor of folding the
1410 VIEW_CONVERT_EXPR generated when punning the RHS.
1412 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
1414 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
1417 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
1420 * ipa-modref.c (callee_to_caller_flags): New function.
1421 (modref_eaf_analysis::analyze_ssa_name): Use it.
1422 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
1424 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1426 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
1429 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
1430 Uroš Bizjak <ubizjak@gmail.com>
1432 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
1433 rotations by non-constant amounts.
1435 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1437 * ipa-modref.c: Fix anonymous namespace placement.
1438 (class modref_eaf_analysis): New class.
1439 (analyze_ssa_name_flags): Turn to ...
1440 (modref_eaf_analysis::analyze_ssa_name): ... this one.
1441 (merge_call_lhs_flags): Turn to ...
1442 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
1443 (modref_eaf_analysis::merge_with_ssa_name): New member function.
1444 (record_escape_points): Turn to ...
1445 (modref_eaf_analysis::record_escape_points): ... this one.
1446 (analyze_parms): Updat
1447 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
1449 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1451 * gimple.c (gimple_call_static_chain_flags): New function.
1452 * gimple.h (gimple_call_static_chain_flags): Declare
1453 * ipa-modref.c (modref_summary::modref_summary): Initialize
1455 (modref_summary_lto::modref_summary_lto): Likewise.
1456 (modref_summary::useful_p): Test static_chain_flags.
1457 (modref_summary_lto::useful_p): Likewise.
1458 (struct modref_summary_lto): Add static_chain_flags.
1459 (modref_summary::dump): Dump static_chain_flags.
1460 (modref_summary_lto::dump): Likewise.
1461 (struct escape_point): Add static_cahin_arg.
1462 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
1463 (analyze_parms): Handle static chains.
1464 (modref_summaries::duplicate): Duplicate static_chain_flags.
1465 (modref_summaries_lto::duplicate): Likewise.
1466 (modref_write): Stream static_chain_flags.
1467 (read_section): Likewise.
1468 (modref_merge_call_site_flags): Handle static_chain_flags.
1469 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
1470 * tree-ssa-structalias.c (handle_rhs_call): Use
1471 gimple_static_chain_flags.
1473 2021-11-02 Richard Biener <rguenther@suse.de>
1475 PR tree-optimization/103029
1476 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
1477 the PHI nodes in the loop maintain their original operand
1480 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1482 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
1483 (EAF_NOREAD): Renumber.
1484 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
1485 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
1486 (deref_flags): Likewise.
1487 (modref_lattice::init): Likewise.
1488 (modref_lattice::merge): Likewise.
1489 (merge_call_lhs_flags): Likewise.
1490 (analyze_ssa_name_flags): Likewise.
1491 (modref_merge_call_site_flags): Likewise.
1492 * tree-ssa-structalias.c (handle_call_arg): Likewise.
1494 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
1496 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
1497 check and remove an unused local variable with shNadd/shNadd.uw
1500 2021-11-02 Tamar Christina <tamar.christina@arm.com>
1502 PR tree-optimization/103007
1503 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
1506 2021-11-02 Richard Biener <rguenther@suse.de>
1508 PR middle-end/103038
1509 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
1510 (can_native_interpret_type_p): Likewise.
1511 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
1512 RHS if the LHS is an SSA name.
1514 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
1516 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
1517 * langhooks-def.h (lhd_simulate_record_decl): Declare.
1518 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
1519 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
1520 * langhooks.c (lhd_simulate_record_decl): New function.
1522 2021-11-02 Jakub Jelinek <jakub@redhat.com>
1525 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
1526 modes with TImode inner mode if 32-bit.
1528 2021-11-02 liuhongt <hongtao.liu@intel.com>
1530 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
1531 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
1532 vect_slp_v4si_store_unalign): Document efficient target.
1533 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
1534 vect_slp_v16qi_store_unalign_1): Ditto.
1535 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
1536 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
1537 (struct_4char_block_move, struct_8char_block_move,
1538 struct_16char_block_move): Ditto.
1540 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
1541 Jakub Jelinek <jakub@redhat.com>
1544 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
1545 ix86_expand_ti_to_v1ti): New helper functions.
1546 (ix86_expand_v1ti_shift): Check if the amount operand is an
1547 integer constant, and expand as a TImode shift if it isn't.
1548 (ix86_expand_v1ti_rotate): Check if the amount operand is an
1549 integer constant, and expand as a TImode rotate if it isn't.
1550 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
1551 right shifts of V1TImode quantities.
1552 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
1553 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
1554 to QImode general_operand, and let the helper functions lower
1555 shifts by non-constant operands, as TImode shifts. Make
1556 conditional on TARGET_64BIT.
1557 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
1558 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
1559 Make conditional on TARGET_64BIT.
1561 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
1563 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
1564 "*tabort_1_plus"): Remove operands non-null check.
1566 2021-11-01 Martin Liska <mliska@suse.cz>
1568 * opt-functions.awk: Add new sanity checking.
1569 * optc-gen.awk: Add new argument to integer_range_info.
1570 * params.opt: Update 2 params which have negative IntegerRange.
1572 2021-11-01 qing zhao <qing.zhao@oracle.com>
1574 * gimplify.c (gimplify_decl_expr): Do not add call to
1575 __builtin_clear_padding when a variable is a gimple register
1576 or it might not have padding.
1577 (gimplify_init_constructor): Likewise.
1579 2021-11-01 Tamar Christina <tamar.christina@arm.com>
1581 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
1582 movi, dup and extract costing fields.
1583 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
1584 thunderx_extra_costs, thunderx2t99_extra_costs,
1585 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
1587 * config/arm/aarch-cost-tables.h (generic_extra_costs,
1588 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
1589 exynosm1_extra_costs, xgene1_extra_costs): Likewise
1590 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
1591 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
1593 2021-11-01 Tamar Christina <tamar.christina@arm.com>
1595 * cse.c (add_to_set): New.
1596 (find_sets_in_insn): Register constants in sets.
1597 (canonicalize_insn): Use auto_vec instead.
1598 (cse_insn): Try materializing using vec_dup.
1599 * rtl.h (simplify_context::simplify_gen_vec_select,
1600 simplify_gen_vec_select): New.
1601 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
1603 2021-11-01 David Malcolm <dmalcolm@redhat.com>
1605 * common.opt (fdiagnostics-escape-format=): New.
1606 (diagnostics_escape_format): New enum.
1607 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
1608 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
1609 * diagnostic-format-json.cc (json_end_diagnostic): Add
1610 "escape-source" attribute.
1611 * diagnostic-show-locus.c
1612 (exploc_with_display_col::exploc_with_display_col): Replace
1613 "tabstop" param with a cpp_char_column_policy and add an "aspect"
1614 param. Use these to compute m_display_col accordingly.
1615 (struct char_display_policy): New struct.
1616 (layout::m_policy): New field.
1617 (layout::m_escape_on_output): New field.
1618 (def_policy): New function.
1619 (make_range): Update for changes to exploc_with_display_col ctor.
1620 (default_print_decoded_ch): New.
1621 (width_per_escaped_byte): New.
1622 (escape_as_bytes_width): New.
1623 (escape_as_bytes_print): New.
1624 (escape_as_unicode_width): New.
1625 (escape_as_unicode_print): New.
1627 (layout::layout): Initialize new fields. Update m_exploc ctor
1628 call for above change to ctor.
1629 (layout::maybe_add_location_range): Update for changes to
1630 exploc_with_display_col ctor.
1631 (layout::calculate_x_offset_display): Update for change to
1633 (layout::print_source_line): Pass policy
1634 to cpp_display_width_computation. Capture cpp_decoded_char when
1635 calling process_next_codepoint. Move printing of source code to
1636 m_policy.m_print_cb.
1637 (line_label::line_label): Pass in policy rather than context.
1638 (layout::print_any_labels): Update for change to line_label ctor.
1639 (get_affected_range): Pass in policy rather than context, updating
1640 calls to location_compute_display_column accordingly.
1641 (get_printed_columns): Likewise, also for cpp_display_width.
1642 (correction::correction): Pass in policy rather than tabstop.
1643 (correction::compute_display_cols): Pass m_policy rather than
1644 m_tabstop to cpp_display_width.
1645 (correction::m_tabstop): Replace with...
1646 (correction::m_policy): ...this.
1647 (line_corrections::line_corrections): Pass in policy rather than
1649 (line_corrections::m_context): Replace with...
1650 (line_corrections::m_policy): ...this.
1651 (line_corrections::add_hint): Update to use m_policy rather than
1653 (line_corrections::add_hint): Likewise.
1654 (layout::print_trailing_fixits): Likewise.
1655 (selftest::test_display_widths): New.
1656 (selftest::test_layout_x_offset_display_utf8): Update to use
1657 policy rather than tabstop.
1658 (selftest::test_one_liner_labels_utf8): Add test of escaping
1660 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
1661 use policy rather than tabstop.
1662 (selftest::test_overlapped_fixit_printing): Likewise.
1663 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
1664 (selftest::test_overlapped_fixit_printing_2): Likewise.
1665 (selftest::test_tab_expansion): Likewise.
1666 (selftest::test_escaping_bytes_1): New.
1667 (selftest::test_escaping_bytes_2): New.
1668 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
1669 * diagnostic.c (diagnostic_initialize): Initialize
1670 context->escape_format.
1671 (convert_column_unit): Update to use default character width policy.
1672 (selftest::test_diagnostic_get_location_text): Likewise.
1673 * diagnostic.h (enum diagnostics_escape_format): New enum.
1674 (diagnostic_context::escape_format): New field.
1675 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
1676 (-fdiagnostics-format=): Add "escape-source" attribute to examples
1677 of JSON output, and document it.
1678 * input.c (location_compute_display_column): Pass in "policy"
1679 rather than "tabstop", passing to
1680 cpp_byte_column_to_display_column.
1681 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
1682 * input.h (class cpp_char_column_policy): New forward decl.
1683 (location_compute_display_column): Pass in "policy" rather than
1685 * opts.c (common_handle_option): Handle
1686 OPT_fdiagnostics_escape_format_.
1687 * selftest.c (temp_source_file::temp_source_file): New ctor
1688 overload taking a size_t.
1689 * selftest.h (temp_source_file::temp_source_file): Likewise.
1691 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
1693 * dbgcnt.def: Add debug counter for back_thread[12] and
1694 back_threadfull[12].
1695 * passes.def: Pass "first" argument to each back threading pass.
1696 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
1698 (back_threader::debug_counter): New.
1699 (back_threader::maybe_register_path): Call debug_counter.
1701 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
1703 * tree-ssa-threadbackward.c (BT_NONE): New.
1706 (back_threader::back_threader): Add flags.
1707 Move loop initialization here.
1708 (back_threader::~back_threader): New.
1709 (back_threader::find_taken_edge_switch): Change solver and ranger
1711 (back_threader::find_taken_edge_cond): Same.
1712 (back_threader::find_paths_to_names): Same.
1713 (back_threader::find_paths): Same.
1714 (back_threader::dump): Same.
1715 (try_thread_blocks): Merge into thread_blocks.
1716 (back_threader::thread_blocks): New.
1717 (do_early_thread_jumps): Merge into thread_blocks.
1718 (do_thread_jumps): Merge into thread_blocks.
1719 (back_threader::thread_through_all_blocks): Remove.
1721 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
1723 PR tree-optimization/103003
1724 * value-relation.cc (dom_oracle::register_relation): If the 2
1725 ssa names are the same, don't register any relation.
1727 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
1729 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
1730 redundant check for calls_eh_return.
1731 * config/aarch64/aarch64.md (*do_return): Likewise.
1733 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
1735 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
1736 duplicate_loop_to_header_edge to
1737 duplicate_loop_body_to_header_edge.
1738 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
1739 * cfghooks.h (struct cfg_hooks): Likewise.
1740 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
1741 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
1742 (clone_loop_to_header_edge): Likewise.
1743 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
1744 * cfgrtl.c (struct cfg_hooks): Likewise.
1745 * doc/loop.texi: Likewise.
1746 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
1747 (unroll_loop_runtime_iterations): Likewise.
1748 (unroll_loop_stupid): Likewise.
1749 (apply_opt_in_copies): Likewise.
1750 * tree-cfg.c (struct cfg_hooks): Likewise.
1751 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
1752 (try_peel_loop): Likewise.
1753 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
1754 (gimple_duplicate_loop_body_to_header_edge): Likewise.
1755 (tree_transform_and_unroll_loop): Likewise.
1756 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
1759 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
1761 * cfgloopmanip.c (loop_version): Refactor loopify to
1762 loop_version. Move condition generation after loopify.
1764 * cfgloopmanip.h (loopify): Delete.
1766 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
1768 * ipa-fnsummary.c: Include tree-dfa.h.
1769 (points_to_local_or_readonly_memory_p): Return true on return
1771 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
1774 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
1776 PR middle-end/102972
1777 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
1778 internal Fortran name; new permit_num_teams arg to permit
1779 omp_get_num_teams and omp_get_team_num.
1780 (scan_omp_1_stmt): Update call to it, add missing call for
1781 reverse offload, and check for strictly nested API calls in teams.
1783 2021-10-30 Jakub Jelinek <jakub@redhat.com>
1785 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
1787 2021-10-29 David Malcolm <dmalcolm@redhat.com>
1789 * selftest.c (assert_streq): Add newlines when emitting non-equal
1792 2021-10-29 David Malcolm <dmalcolm@redhat.com>
1794 * Makefile.in: Fix syntax for reference to LIBDEPS in
1797 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
1799 * doc/install.texi: Bump required minimum DejaGnu version.
1801 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
1802 Andrew MacLeod <amacleod@redhat.com>
1804 * value-relation.cc (path_oracle::killing_def): Add a
1805 self-equivalence so we don't look to the root oracle.
1807 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
1809 * passes.def: Replace the pass_thread_jumps before VRP* with
1810 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
1811 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
1812 Remove hyphen from "thread-full" name.
1814 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
1816 PR tree-optimization/102983
1817 * gimple-range-cache.h (propagate_updated_value): Make public.
1818 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
1819 when processing gcond stmts.
1821 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
1823 * gimple.c (gimple_call_retslot_flags): New function.
1824 * gimple.h (gimple_call_retslot_flags): Declare.
1825 * ipa-modref.c: Include tree-cfg.h.
1826 (struct escape_entry): Turn parm_index to signed.
1827 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
1828 (modref_summary::modref_summary): Initialize retslot_flags.
1829 (struct modref_summary_lto): Likewise.
1830 (modref_summary::useful_p): Check retslot_flags.
1831 (modref_summary_lto::useful_p): Likewise.
1832 (modref_summary::dump): Dump retslot_flags.
1833 (modref_summary_lto::dump): Likewise.
1834 (struct escape_point): Add hidden_args enum.
1835 (analyze_ssa_name_flags): Ignore return slot return;
1836 use gimple_call_retslot_flags.
1837 (record_escape_points): Break out from ...
1838 (analyze_parms): ... here; handle retslot_flags.
1839 (modref_summaries::duplicate): Duplicate retslot_flags.
1840 (modref_summaries_lto::duplicate): Likewise.
1841 (modref_write_escape_summary): Stream parm_index as signed.
1842 (modref_read_escape_summary): Likewise.
1843 (modref_write): Stream retslot_flags.
1844 (read_section): Likewise.
1845 (struct escape_map): Fix typo in comment.
1846 (update_escape_summary_1): Fix whitespace.
1847 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
1848 (modref_merge_call_site_flags): Merge retslot_flags.
1849 * ipa-modref.h (struct modref_summary): Add retslot_flags.
1850 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
1852 2021-10-29 Tamar Christina <tamar.christina@arm.com>
1854 PR tree-optimization/102977
1855 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
1856 (vect_detect_pair_op): Add crosslane check.
1857 (vect_match_call_complex_mla): Remove.
1858 (class complex_mul_pattern): Update comment.
1859 (complex_mul_pattern::matches): Update detection.
1860 (class complex_fma_pattern): Remove.
1861 (complex_fma_pattern::matches): Remove.
1862 (complex_fma_pattern::recognize): Remove.
1863 (complex_fma_pattern::build): Remove.
1864 (class complex_fms_pattern): Update comment.
1865 (complex_fms_pattern::matches): Remove.
1866 (complex_operations_pattern::recognize): Remove complex_fma_pattern
1868 2021-10-29 Jakub Jelinek <jakub@redhat.com>
1870 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
1873 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
1876 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
1877 patterns match and emit for VSX xxpermdi.
1879 2021-10-29 liuhongt <hongtao.liu@intel.com>
1882 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
1884 (V16HF_FTYPE_V16HF): Ditto.
1885 (V32HF_FTYPE_V32HF): Ditto.
1886 (V8HF_FTYPE_V8HF_ROUND): Ditto.
1887 (V16HF_FTYPE_V16HF_ROUND): Ditto.
1888 (V32HF_FTYPE_V32HF_ROUND): Ditto.
1889 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
1890 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
1891 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
1892 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
1893 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
1894 * config/i386/i386-builtins.c
1895 (ix86_builtin_vectorized_function): Enable vectorization for
1896 HFmode FLOOR/CEIL/TRUNC operation.
1897 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
1899 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
1902 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
1903 Andrew MacLeod <amacleod@redhat.com>
1905 * value-relation.cc (path_oracle::killing_def): Walk the
1906 equivalency list and remove SSA from any equivalencies.
1908 2021-10-28 Stafford Horne <shorne@gmail.com>
1910 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
1913 2021-10-28 Jakub Jelinek <jakub@redhat.com>
1915 PR tree-optimization/102951
1916 * fold-const.h (address_compare): Declare.
1917 * fold-const.c (address_compare): New function.
1918 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
1919 address_compare helper.
1920 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
1923 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
1925 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
1927 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
1929 PR tree-optimization/102940
1930 * tree-vrp.c (execute_ranger_vrp): Reset scev.
1932 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
1934 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
1936 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
1938 * configure: Regenerate.
1939 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
1941 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
1943 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
1945 2021-10-28 Richard Biener <rguenther@suse.de>
1948 * fold-const.c (fold_convert_const): Avoid int to float
1949 constant folding with -frounding-math and inexact result.
1950 * simplify-rtx.c (simplify_const_unary_operation): Likewise
1951 for both float and unsigned_float.
1953 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
1955 * tree-ssa-threadbackward.c
1956 (back_threader::find_taken_edge_switch): Use find_case_label_range
1957 instead of find_taken_edge.
1959 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
1961 * tree-ssa-threadbackward.c (class back_threader_registry):
1962 Inherit from back_jt_path_registry.
1963 (back_threader_registry::thread_through_all_blocks): Remove.
1964 (back_threader_registry::register_path): Remove
1965 m_lowlevel_registry prefix.
1967 2021-10-28 Richard Biener <rguenther@suse.de>
1970 * fold-const.c (fold_convert_const_real_from_real): Honor
1971 -frounding-math if the conversion is not exact.
1972 * simplify-rtx.c (simplify_const_unary_operation): Do not
1973 simplify FLOAT_TRUNCATE with sign dependent rounding.
1975 2021-10-28 Richard Biener <rguenther@suse.de>
1977 PR tree-optimization/102949
1978 * tree-vect-stmts.c (ensure_base_align): Look at the
1979 dr_info of a group leader and assert we are looking at
1980 one with analyzed alignment.
1982 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
1985 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
1986 V1T1 mode for unaligned load and store.
1988 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
1990 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
1993 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
1995 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
1998 2021-10-28 Alexandre Oliva <oliva@adacore.com>
2000 * common.opt (fharden-compares): New.
2001 (fharden-conditional-branches): New.
2002 * doc/invoke.texi: Document new options.
2003 * gimple-harden-conditionals.cc: New.
2004 * Makefile.in (OBJS): Build it.
2005 * passes.def: Add new passes.
2006 * tree-pass.h (make_pass_harden_compares): Declare.
2007 (make_pass_harden_conditional_branches): Declare.
2009 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
2012 * config/rs6000/altivec.md: Add vsx register constraints.
2013 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
2014 (vsx_xxsel<mode>2): Likewise.
2015 (vsx_xxsel<mode>3): Likewise.
2016 (vsx_xxsel<mode>4): Likewise.
2018 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
2021 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
2022 (altivec_vsel<mode>): ... this and update define.
2023 (*altivec_vsel<mode>_uns): Delete.
2024 (altivec_vsel<mode>2): New define_insn.
2025 (altivec_vsel<mode>3): Likewise.
2026 (altivec_vsel<mode>4): Likewise.
2027 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
2028 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
2030 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
2031 selection instead of per element.
2032 * config/rs6000/vector.md:
2033 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
2034 (vsx_xxsel<mode>): ... this and update define.
2035 (*vsx_xxsel<mode>_uns): Delete.
2036 (vsx_xxsel<mode>2): New define_insn.
2037 (vsx_xxsel<mode>3): Likewise.
2038 (vsx_xxsel<mode>4): Likewise.
2040 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
2042 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
2043 HFmode rsqrt without TARGET_SSE_MATH.
2044 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
2045 btrunc condition to be restricted by -ftrapping-math, adjust
2046 use_rsqrt_p function call.
2047 * config/i386/i386.md (rcphf2): New define_insn.
2048 (rsqrthf2): Likewise.
2049 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
2050 (div<mode>3): New expander for HF mode.
2051 (rsqrt<mode>2): Likewise.
2052 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
2053 (*avx512fp16_vmrsqrtv8hf2): Likewise.
2055 2021-10-27 Saagar Jha <saagar@saagarjha.com>
2057 * config.gcc: Adjust for Darwin21.
2058 * config/darwin-c.c (macosx_version_as_macro): Likewise.
2059 * config/darwin-driver.c (validate_macosx_version_min):
2061 (darwin_find_version_from_kernel): Likewise.
2063 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2065 * gimple-range-path.cc
2066 (path_range_query::range_defined_in_block): Call killing_def.
2068 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2070 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
2071 operands as well as relation.
2072 * gimple-range-path.cc
2073 (path_range_query::compute_ranges_in_block): Compute PHI relations
2074 first. Compute outgoing relations at the end.
2075 (path_range_query::compute_ranges): Remove call to compute_relations.
2076 (path_range_query::compute_relations): Remove.
2077 (path_range_query::maybe_register_phi_relation): New.
2078 (path_range_query::compute_phi_relations): Abstract out
2079 registering one PHI relation to...
2080 (path_range_query::compute_outgoing_relations): ...here.
2081 * gimple-range-path.h (class path_range_query): Remove
2083 Add maybe_register_phi_relation.
2085 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2086 Andrew MacLeod <amacleod@redhat.com>
2088 * value-relation.cc (path_oracle::killing_def): Kill second
2091 2021-10-27 Martin Jambor <mjambor@suse.cz>
2093 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
2094 profile feedback depending on their local availability.
2096 2021-10-27 Martin Jambor <mjambor@suse.cz>
2098 * params.opt (param_ipa_cp_profile_count_base): New parameter.
2099 * doc/invoke.texi (Optimize Options): Add entry for
2100 ipa-cp-profile-count-base.
2101 * ipa-cp.c (max_count): Replace with base_count, replace all
2102 occurrences too, unless otherwise stated.
2103 (ipcp_cloning_candidate_p): identify mostly-directly called
2104 functions based on their counts, not max_count.
2105 (compare_edge_profile_counts): New function.
2106 (ipcp_propagate_stage): Instead of setting max_count, find the
2107 appropriate edge count in a sorted vector of counts of eligible
2108 edges and make it the base_count.
2110 2021-10-27 Martin Jambor <mjambor@suse.cz>
2112 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
2113 n_nonrec_calls and itself, document all fields.
2114 (init_caller_stats): Initialize the above new fields.
2115 (gather_caller_stats): Gather self-recursive counts and calls number.
2116 (get_info_about_necessary_edges): Gather counts of self-recursive and
2117 other edges bringing in the requested value separately.
2118 (dump_profile_updates): Rework to dump info about a single node only.
2119 (lenient_count_portion_handling): New function.
2120 (struct gather_other_count_struct): New type.
2121 (gather_count_of_non_rec_edges): New function.
2122 (struct desc_incoming_count_struct): New type.
2123 (analyze_clone_icoming_counts): New function.
2124 (adjust_clone_incoming_counts): Likewise.
2125 (update_counts_for_self_gen_clones): Likewise.
2126 (update_profiling_info): Rewritten.
2127 (update_specialized_profile): Adjust call to dump_profile_updates.
2128 (create_specialized_node): Do not update profiling info.
2129 (decide_about_value): New parameter self_gen_clones, either push new
2130 clones into it or updat their profile counts. For self-recursively
2131 generated values, use a portion of the node count instead of count
2132 from self-recursive edges to estimate goodness.
2133 (decide_whether_version_node): Gather clones for self-generated values
2134 in a new vector, update their profiles at once at the end.
2136 2021-10-27 Richard Biener <rguenther@suse.de>
2138 * tree-vectorizer.c (vect_transform_loops): New function,
2140 (try_vectorize_loop_1): ... here. Simplify as epilogues
2141 are now fully handled in the split part.
2143 2021-10-27 Jakub Jelinek <jakub@redhat.com>
2145 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
2146 iterators with pointer types.
2147 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
2149 2021-10-26 Martin Sebor <msebor@redhat.com>
2151 PR tree-optimization/102238
2152 PR tree-optimization/102919
2153 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
2154 (array_elt_at_offset): Move to pointer-query.
2155 (set_aggregate_size_and_offset): New function.
2156 (field_at_offset): Move to pointer-query.
2157 (get_origin_and_offset): Rename...
2158 (get_origin_and_offset_r): this. Add an argument. Make aggregate
2159 handling more robust.
2160 (get_origin_and_offset): New.
2161 (alias_offset): Add an argument.
2162 (format_string): Use subobject size determined by get_origin_and_offset.
2163 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
2164 Improve/correct handling of aggregates.
2165 (array_elt_at_offset): Same.
2166 * pointer-query.h (field_at_offset): Declare.
2167 (array_elt_at_offset): Declare.
2169 2021-10-26 Martin Sebor <msebor@redhat.com>
2171 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
2173 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
2174 (expand_builtin_fork_or_exec): Same.
2175 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
2176 compute_objsize additional arguments.
2177 (inbounds_memaccess_p): Same.
2178 (array_bounds_checker::check_array_bounds): Add an assert. Stash
2179 statement in a member.
2180 (check_array_bounds_dom_walker::before_dom_children): Same.
2181 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
2182 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
2183 (handle_printf_call): Pass a new argument.
2184 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
2185 (check_access): Add an argument and pass it along to callees.
2186 (check_read_access): Make a member function.
2187 (pass_waccess::check_strcat): Pass access_data ctor additional
2189 (pass_waccess::check_strncat): Same.
2190 (pass_waccess::check_stxcpy): Same.
2191 (pass_waccess::check_stxncpy): Same.
2192 (pass_waccess::check_strncmp): Same.
2193 (pass_waccess::check_read_access): Same.
2194 (pass_waccess::check_builtin): Same.
2195 (pass_waccess::maybe_check_access_sizes): Same.
2196 (pass_waccess::maybe_check_dealloc_call): Same.
2197 * gimple-ssa-warn-access.h (check_read_access): Declare a new
2199 * pointer-query.cc (compute_objsize_r): Add an argument.
2200 (gimple_call_return_array): Same.
2201 (gimple_call_alloc_size): Same.
2202 (access_ref::access_ref): Same.
2203 (access_ref::get_ref): Same.
2204 (pointer_query::get_ref): Same.
2205 (handle_min_max_size): Pass an arguments to callees.
2206 (handle_array_ref): Add an argument.
2207 (handle_mem_ref): Same.
2208 (compute_objsize): Same.
2209 * pointer-query.h (struct access_ref): Adjust signatures.
2210 (struct access_data): Same.
2211 (gimple_call_alloc_size): Add an argument.
2212 (gimple_parm_array_size): Same.
2213 (compute_objsize): Same.
2214 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
2215 argument to compute_objsize.
2216 (strlen_pass::maybe_warn_overflow): Same.
2217 (maybe_diag_stxncpy_trunc): Same.
2219 2021-10-26 Martin Sebor <msebor@redhat.com>
2221 PR middle-end/102453
2222 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
2223 (pass_waccess::check_atomic_builtin): Call it.
2225 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
2227 PR rtl-optimization/102842
2228 * lra-constraints.c (match_reload): Ignore out in checking values
2230 (curr_insn_transform): Collect outputs before doing reloads of operands.
2232 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
2235 * config/rs6000/x86intrin.h: Move some included headers to new
2236 headers. Include new immintrin.h instead of those headers.
2237 * config/rs6000/immintrin.h: New.
2238 * config/rs6000/x86gprintrin.h: New.
2239 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
2241 2021-10-26 Richard Biener <rguenther@suse.de>
2243 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
2244 Remove byte_offset parameter.
2245 (vect_create_data_ref_ptr): Likewise.
2246 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
2248 (vect_create_data_ref_ptr): Likewise.
2249 * tree-vect-stmts.c (vectorizable_store): Adjust.
2250 (vectorizable_load): Likewise.
2252 2021-10-26 Richard Biener <rguenther@suse.de>
2254 PR tree-optimization/96109
2255 * tree-vectorizer.h (dr_misalignment): Add optional offset
2257 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
2258 offset applied for negative stride accesses.
2259 (vect_enhance_data_refs_alignment): Compute negative stride
2260 access offset and pass it to dr_misalignment.
2261 * tree-vect-stmts.c (get_negative_load_store_type): Pass
2262 negative offset to dr_misalignment.
2263 (get_group_load_store_type): Likewise.
2264 (get_load_store_type): Likewise.
2265 (vectorizable_store): Remove asserts about alignment.
2266 (vectorizable_load): Likewise.
2268 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
2270 PR tree-optimization/102897
2271 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
2273 2021-10-26 Richard Biener <rguenther@suse.de>
2275 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
2276 Take offset in bytes.
2277 (vect_create_data_ref_ptr): Likewise.
2278 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
2279 offset by element size.
2280 (vect_create_cond_for_align_checks): Likewise.
2281 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
2282 (vectorizable_load): Remove duplicate leftover from merge
2285 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
2287 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
2288 function to expand V1TI mode logical shifts by integer constants.
2289 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
2290 mode rotations by integer constants.
2291 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
2292 ix86_expand_v1ti_rotate): Prototype new functions here.
2293 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
2294 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
2296 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2298 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
2299 Avoid threading circular paths.
2301 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2303 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
2304 Attempt to resolve all incoming paths to a PHI.
2305 (back_threader::resolve_def): Always return true for PHIs.
2307 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2309 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
2310 Always try to resolve path without looking back.
2311 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
2312 edge is a back edge.
2314 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
2316 PR tree-optimization/102789
2317 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2318 update inits of simd_lane_access.
2320 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2322 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
2323 move to vr-values.c.
2324 (simplify_casted_conds): Move to vrp_folder class.
2325 (execute_vrp): Call via vrp_folder now.
2326 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
2327 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
2328 * vr-values.h (simplify_casted_cond): Add prototype.
2330 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2332 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
2335 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
2337 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
2339 (_mm_testc_si128): Likewise.
2340 (_mm_testnzc_si128): Likewise.
2341 (_mm_blend_ps): Likewise.
2342 (_mm_blendv_ps): Likewise.
2343 (_mm_blend_pd): Likewise.
2344 (_mm_blendv_pd): Likewise.
2345 (_mm_ceil_pd): Likewise.
2346 (_mm_ceil_sd): Likewise.
2347 (_mm_ceil_ps): Likewise.
2348 (_mm_ceil_ss): Likewise.
2349 (_mm_floor_pd): Likewise.
2350 (_mm_floor_sd): Likewise.
2351 (_mm_floor_ps): Likewise.
2352 (_mm_floor_ss): Likewise.
2353 (_mm_minpos_epu16): Likewise.
2354 (_mm_mul_epi32): Likewise.
2355 (_mm_cvtepi8_epi16): Likewise.
2356 (_mm_packus_epi32): Likewise.
2357 (_mm_cmpgt_epi64): Likewise.
2359 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
2361 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
2362 shifts of the mode's smin_value and smax_value when the bit count
2363 operand doesn't have side-effects.
2364 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
2365 when the bit count operand doesn't have side-effects.
2366 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
2367 compile-time evaluation of saturating left shifts with constant
2370 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2372 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
2375 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2377 * flag-types.h (enum ranger_debug): Adjust values.
2378 * params.opt (ranger_debug): Ditto.
2380 2021-10-25 Martin Jambor <mjambor@suse.cz>
2382 PR tree-optimization/102886
2383 * tree-sra.c (totally_scalarize_subtree): Fix the out of
2386 2021-10-25 Andrew Pinski <apinski@marvell.com>
2388 * tree-ssa-dce.c (simple_dce_from_worklist):
2389 Check stmt_unremovable_because_of_non_call_eh_p also
2390 before removing the statement.
2392 2021-10-25 Richard Biener <rguenther@suse.de>
2394 PR tree-optimization/102905
2395 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
2396 Use vect_supportable_dr_alignment again to determine whether
2397 an access is supported when not aligned.
2399 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2401 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
2404 2021-10-25 Jim Wilson <jimw@sifive.com>
2405 Kito Cheng <kito.cheng@sifive.com>
2406 Jia-Wei Chen <jiawei@iscas.ac.cn>
2407 Shi-Hua Liao <shihua@iscas.ac.cn>
2409 * config/riscv/bitmanip.md (shiftm1): New.
2410 (*bset<mode>): Ditto.
2411 (*bset<mode>_mask): Ditto.
2412 (*bset<mode>_1): Ditto.
2413 (*bset<mode>_1_mask): Ditto.
2414 (*bseti<mode>): Ditto.
2415 (*bclr<mode>): Ditto.
2416 (*bclri<mode>): Ditto.
2417 (*binv<mode>): Ditto.
2418 (*binvi<mode>): Ditto.
2419 (*bext<mode>): Ditto.
2421 * config/riscv/predicates.md (splittable_const_int_operand):
2423 (single_bit_mask_operand): New.
2424 (not_single_bit_mask_operand): Ditto.
2425 (const31_operand): Ditto.
2426 (const63_operand): Ditto.
2427 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
2428 (riscv_output_move): Ditto.
2429 (riscv_print_operand): Handle new operand type: T and S.
2430 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
2432 2021-10-25 Jim Wilson <jimw@sifive.com>
2434 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
2437 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2439 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
2441 (riscv_rtx_costs): Ditto.
2443 2021-10-25 Jim Wilson <jimw@sifive.com>
2444 Kito Cheng <kito.cheng@sifive.com>
2445 Jia-Wei Chen <jiawei@iscas.ac.cn>
2447 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
2448 (bitmanip_minmax): New.
2449 (clz_ctz_pcnt): New.
2450 (bitmanip_optab): New.
2451 (bitmanip_insn): New.
2452 (*<optab>_not<mode>): New.
2453 (*xor_not<mode>): New.
2454 (<bitmanip_optab>si2): New.
2455 (*<bitmanip_optab>disi2): New.
2456 (<bitmanip_optab>di2): New.
2457 (*zero_extendhi<GPR:mode>2_bitmanip): New.
2458 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
2459 (*zero_extendhi<GPR:mode>2_zbb): New.
2462 (rotrsi3_sext): New.
2465 (rotlsi3_sext): New.
2466 (bswap<mode>2): New.
2467 (<bitmanip_optab><mode>3): New.
2468 * config/riscv/riscv.md (type): Add rotate.
2469 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
2470 (*zero_extendhi<GPR:mode>2): New.
2471 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
2472 (*extend<SHORT:mode><SUPERQI:mode>2): New.
2474 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2476 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
2478 (riscv_rtx_costs): Ditto.
2480 2021-10-25 Jim Wilson <jimw@sifive.com>
2481 Kito Cheng <kito.cheng@sifive.com>
2482 Jia-Wei Chen <jiawei@iscas.ac.cn>
2484 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
2489 (riscv_rtx_costs): Ditto.
2490 * config/riscv/riscv.md: Include bitmanip.md
2491 (type): Add bitmanip bype.
2492 (zero_extendsidi2): Change to define_expand pattern.
2493 (*zero_extendsidi2_internal): New.
2494 (zero_extendsidi2_shifted): Disable for ZBA.
2496 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2498 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
2499 Add zba, zbb, zbc and zbs.
2500 (riscv_ext_flag_table): Ditto.
2501 * config/riscv/riscv-opts.h (MASK_ZBA): New.
2505 (TARGET_ZBA): Ditto.
2506 (TARGET_ZBB): Ditto.
2507 (TARGET_ZBC): Ditto.
2508 (TARGET_ZBS): Ditto.
2509 * config/riscv/riscv.opt (riscv_zb_subext): New.
2511 2021-10-25 liuhongt <hongtao.liu@intel.com>
2514 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
2515 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
2517 2021-10-25 Richard Biener <rguenther@suse.de>
2519 PR tree-optimization/102920
2520 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
2521 controlling VN_TOP matching behavior.
2522 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
2523 (vn_phi_eq): Do not optimistically match VN_TOP.
2525 2021-10-25 konglin1 <lingling.kong@intel.com>
2527 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
2528 define_insn_and_split.
2529 (fma_<mode>_fadd_fcmul):Likewise
2530 (fma_<complexopname>_<mode>_fma_zero):Likewise
2532 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
2534 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
2535 TARGET_DISABLE_FPREGS.
2536 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
2537 MASK_DISABLE_FPREGS.
2538 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
2539 cost of hardware integer multiplication.
2540 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
2541 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
2542 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
2543 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
2544 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
2545 SOFT_FLOAT. Revise comment.
2546 (msoft-float): New option.
2548 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
2550 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
2552 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
2554 * doc/install.texi (Specific): Remove obsolete details
2555 around GNU/Linux on Itanium.
2556 (Specific): Remove reference to Windows for Itanium.
2558 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
2560 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
2563 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
2565 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
2567 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
2569 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
2570 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
2571 (one_cmplv1ti2): New define expand.
2573 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
2576 * Makefile.in: Handle dvidir and install-dvi target.
2577 * configure: Regenerate.
2578 * configure.ac: Add install-dvi to target_list.
2580 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
2582 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
2585 2021-10-22 Richard Biener <rguenther@suse.de>
2587 PR tree-optimization/102893
2588 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
2589 test for an exit edge.
2591 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
2592 Andrew MacLeod <amacleod@redhat.com>
2594 * gimple-range-path.cc (path_range_query::compute_phi_relations):
2595 Kill any global relations we may know before registering a new
2597 * value-relation.cc (path_oracle::killing_def): New.
2598 * value-relation.h (path_oracle::killing_def): New.
2600 2021-10-22 Richard Biener <rguenther@suse.de>
2603 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
2605 (vn_phi_lookup): Likewise.
2607 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
2610 * doc/invoke.texi: Document -fcf-protection requires i686 or
2613 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
2615 PR middle-end/102764
2616 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
2618 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
2620 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
2621 macro and all invocations.
2623 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
2625 * doc/invoke.texi (ranger-debug): Document.
2626 * flag-types.h (enum ranger_debug): New.
2627 (enum evrp_mode): Remove debug values.
2628 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
2629 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
2630 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
2631 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
2632 (execute_early_vrp): Use evrp-mode directly.
2633 * params.opt (enum evrp_mode): Remove debug values.
2634 (ranger-debug): New.
2635 (ranger-logical-depth): Relocate to be in alphabetical order.
2637 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
2639 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
2640 * flag-types.h: (enum vrp_mode): New.
2641 * params.opt: (vrp1-mode, vrp2-mode): New.
2642 * tree-vrp.c (vrp_pass_num): New.
2643 (pass_vrp::pass_vrp): Set pass number.
2644 (pass_vrp::execute): Choose which VRP mode to execute.
2646 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
2648 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
2649 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
2650 * tree-vrp.c (class rvrp_folder): Relocate here.
2651 (execute_ranger_vrp): New.
2652 * tree-vrp.h (execute_ranger_vrp): Export.
2654 2021-10-21 Martin Liska <mliska@suse.cz>
2658 * opts.c (finish_options): Process flag_var_tracking* options
2659 here as they can be adjusted by optimize attribute.
2660 Process also flag_syntax_only and flag_gtoggle.
2661 * toplev.c (process_options): Remove it here.
2662 * common.opt: Make debug_nonbind_markers_p as PerFunction
2663 attribute as it depends on optimization level.
2665 2021-10-21 Martin Jambor <mjambor@suse.cz>
2667 PR tree-optimization/102505
2668 * tree-sra.c (totally_scalarize_subtree): Check that the
2669 encountered field fits within the acces we would like to put it
2672 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
2674 * tree-ssa-threadbackward.c
2675 (back_threader::maybe_register_path): Remove circular paths check.
2677 2021-10-21 Richard Biener <rguenther@suse.de>
2679 * toplev.c (process_options): Move the initial debug_hooks
2681 (toplev::main): ... before the call of the post_options
2684 2021-10-21 Richard Biener <rguenther@suse.de>
2686 PR tree-optimization/102847
2687 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
2688 load cost in the prologue for VMAT_INVARIANT.
2690 2021-10-21 Richard Biener <rguenther@suse.de>
2692 PR tree-optimization/102847
2693 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
2694 handle VMAT_INVARIANT as a splat in the prologue.
2696 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
2699 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
2700 move to use the same logic as HImode.
2702 2021-10-21 Richard Biener <rguenther@suse.de>
2704 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
2705 superfluous gimple_call_nothrow_p check.
2707 2021-10-21 Andrew Pinski <apinski@marvell.com>
2709 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
2710 Mark the ssa-name of the rhs as one to be removed.
2711 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
2712 Call simple_dce_from_worklist at the end to a simple dce.
2714 2021-10-21 Andrew Pinski <apinski@marvell.com>
2716 * tree-cfg.c (maybe_remove_writeonly_store): New function
2717 factored out from ...
2718 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
2720 2021-10-21 Andrew Pinski <apinski@marvell.com>
2722 * tree-cfg.c (execute_fixup_cfg): Remove comment
2723 about standalone pass.
2725 2021-10-21 Andrew Pinski <apinski@marvell.com>
2727 * tree-cfg.c (execute_fixup_cfg): Output when the statement
2728 is removed when it is a write only var.
2730 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
2732 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
2733 Avoid threading circular paths.
2735 2021-10-20 Alex Coplan <alex.coplan@arm.com>
2737 * calls.c (initialize_argument_information): Remove some dead
2738 code, remove handling for function_arg returning const_int.
2739 * doc/tm.texi: Delete documentation for unused target hooks.
2740 * doc/tm.texi.in: Likewise.
2741 * target.def (load_bounds_for_arg): Delete.
2742 (store_bounds_for_arg): Delete.
2743 (load_returned_bounds): Delete.
2744 (store_returned_bounds): Delete.
2745 * targhooks.c (default_load_bounds_for_arg): Delete.
2746 (default_store_bounds_for_arg): Delete.
2747 (default_load_returned_bounds): Delete.
2748 (default_store_returned_bounds): Delete.
2749 * targhooks.h (default_load_bounds_for_arg): Delete.
2750 (default_store_bounds_for_arg): Delete.
2751 (default_load_returned_bounds): Delete.
2752 (default_store_returned_bounds): Delete.
2754 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
2756 * doc/extend.texi (Basic Asm): Clarify that asm is not an
2758 * doc/invoke.texi (-fno-asm): Fix description for C++.
2760 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
2762 * doc/install.texi: Remove link to old.html
2764 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2766 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
2768 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2770 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
2772 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2774 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
2776 * config/aarch64/constraints.md (D1): New.
2778 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2780 * config/aarch64/aarch64-simd.md
2781 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
2782 (*aarch64_topbits_shuffle<mode>_le): New.
2783 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
2784 (*aarch64_topbits_shuffle<mode>_be): New.
2785 * config/aarch64/predicates.md
2786 (aarch64_simd_shift_imm_vec_exact_top): New.
2788 2021-10-20 Tamar Christina <tamar.christina@arm.com>
2790 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
2791 *aarch64_<srn_op>shrn<mode>2_vect_le,
2792 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
2793 * config/aarch64/iterators.md (srn_op): New.
2795 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
2797 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
2799 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
2800 ctx->outer instead of ctx.
2802 2021-10-20 Martin Liska <mliska@suse.cz>
2805 2021-10-19 Martin Liska <mliska@suse.cz>
2808 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
2809 * system.h (strip_whilespaces): New function.
2811 2021-10-20 Martin Liska <mliska@suse.cz>
2814 2021-10-19 Martin Liska <mliska@suse.cz>
2817 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
2820 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
2822 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
2823 Change to 8 with -Os, 11 otherwise.
2825 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
2827 * config/aarch64/aarch64.c (neoversev1_tunings):
2828 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
2829 (neoversen2_tunings): Likewise.
2831 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
2834 * config/aarch64/aarch64.md (frint_pattern): Update comment.
2835 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
2836 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
2837 * config/aarch64/arm_neon.h: Likewise.
2838 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
2840 2021-10-20 Martin Liska <mliska@suse.cz>
2842 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
2843 already declared global variable.
2844 (arm_unwind_emit_set): Use out_file as function argument.
2845 (arm_unwind_emit): Likewise.
2846 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
2847 (machopic_output_stub_indirection): Likewise.
2848 (machopic_output_indirection): Likewise.
2849 (machopic_finish): Likewise.
2850 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
2851 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
2852 * config/ia64/ia64.c (process_epilogue): Likewise.
2853 (process_cfa_adjust_cfa): Likewise.
2854 (process_cfa_register): Likewise.
2855 (process_cfa_offset): Likewise.
2856 (ia64_asm_unwind_emit): Likewise.
2857 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
2859 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
2861 * config/aarch64/aarch64-builtins.c
2862 (aarch64_general_gimple_fold_builtin):
2863 lower vld1 and vst1 variants of the neon builtins
2864 * config/aarch64/aarch64-protos.h:
2865 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
2866 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
2869 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
2871 * match.pd: Generate IFN_TRUNC.
2873 2021-10-20 Richard Biener <rguenther@suse.de>
2875 PR tree-optimization/102853
2876 * tree-data-ref.c (split_constant_offset_1): Bail out
2877 immediately if the expression traps on overflow.
2879 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
2881 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
2883 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
2885 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
2886 Remove superflous debugging message.
2887 (back_jt_path_registry::duplicate_thread_path): Same.
2889 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
2891 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
2893 (back_threader_registry::register_path): Remove m_threaded_paths.
2895 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
2897 PR tree-optimization/102814
2898 * doc/invoke.texi: Document --param=max-fsm-thread-length.
2899 * params.opt: Add --param=max-fsm-thread-length.
2900 * tree-ssa-threadbackward.c
2901 (back_threader_profitability::profitable_path_p): Fail on paths
2902 longer than max-fsm-thread-length.
2904 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
2906 PR middle-end/102764
2907 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
2908 statement to reset the current location for the outgoing edges.
2910 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
2911 Richard Biener <rguenther@suse.de>
2913 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
2914 on the same line as the threading cancellation.
2915 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
2916 Avoid threading through loop headers where the path remains in the
2919 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
2921 * tree-object-size.c (unknown): Make into a function. Adjust
2923 (unknown_object_size): Simplify implementation.
2925 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
2926 Kewen Lin <linkw@linux.ibm.com>
2928 * doc/sourcebuild.texi (Effective-Target Keywords): Document
2929 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
2930 vect_slp_v16qi_store, vect_slp_v2hi_store,
2931 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
2933 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
2935 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
2937 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
2941 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
2942 * config/rs6000/pmmintrin.h: Same.
2943 * config/rs6000/smmintrin.h: Same.
2944 * config/rs6000/tmmintrin.h: Same.
2946 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
2948 * config.gcc (extra_headers): Add nmmintrin.h.
2950 2021-10-19 Richard Biener <rguenther@suse.de>
2952 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
2953 misalignment parameter.
2954 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
2955 Do not change DR_MISALIGNMENT in place, instead pass the
2956 adjusted misalignment to vect_supportable_dr_alignment.
2957 (vect_peeling_supportable): Likewise.
2958 (vect_peeling_hash_get_lowest_cost): Adjust.
2959 (vect_enhance_data_refs_alignment): Likewise.
2960 (vect_vfa_access_size): Likewise.
2961 (vect_supportable_dr_alignment): Add misalignment
2962 parameter and simplify.
2963 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
2964 (get_group_load_store_type): Likewise.
2965 (get_load_store_type): Likewise.
2967 2021-10-19 Clément Chigot <clement.chigot@atos.net>
2969 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
2970 __tls_get_addr reference to .text csect.
2972 2021-10-19 Martin Liska <mliska@suse.cz>
2975 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
2978 2021-10-19 Richard Biener <rguenther@suse.de>
2980 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
2981 (vect_get_load_cost): Likewise.
2982 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2983 alignment support scheme and misalignment as arguments
2985 (vect_get_peeling_costs_all_drs): Compute that info here
2986 and note that we shouldn't need to.
2987 * tree-vect-stmts.c (vect_model_store_cost): Get
2988 alignment support scheme and misalignment as arguments.
2989 (vect_get_store_cost): Likewise.
2990 (vect_model_load_cost): Likewise.
2991 (vect_get_load_cost): Likewise.
2992 (vectorizable_store): Pass down alignment support scheme
2993 and misalignment to costing.
2994 (vectorizable_load): Likewise.
2996 2021-10-19 Richard Biener <rguenther@suse.de>
2998 * tree-vect-stmts.c (get_negative_load_store_type): Add
2999 offset output parameter and initialize it.
3000 (get_group_load_store_type): Likewise.
3001 (get_load_store_type): Likewise.
3002 (vectorizable_store): Use offset as computed by
3003 get_load_store_type.
3004 (vectorizable_load): Likewise.
3006 2021-10-19 Richard Biener <rguenther@suse.de>
3008 PR tree-optimization/102827
3009 * tree-if-conv.c (predicate_statements): Add pe parameter
3010 and use that edge to insert invariant stmts on.
3011 (combine_blocks): Pass through pe.
3012 (tree_if_conversion): Compute the edge to insert invariant
3013 stmts on and pass it along.
3015 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
3018 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
3019 sssubaddv2hi3): Swap the order of operators in vec_concat.
3021 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
3023 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
3024 (altivec_vmrghb_direct): New.
3025 (*altivec_vmrghh_internal): Delete.
3026 (altivec_vmrghh_direct): New.
3027 (*altivec_vmrghw_internal): Delete.
3028 (altivec_vmrghw_direct_<mode>): New.
3029 (altivec_vmrghw_direct): Delete.
3030 (*altivec_vmrglb_internal): Delete.
3031 (altivec_vmrglb_direct): New.
3032 (*altivec_vmrglh_internal): Delete.
3033 (altivec_vmrglh_direct): New.
3034 (*altivec_vmrglw_internal): Delete.
3035 (altivec_vmrglw_direct_<mode>): New.
3036 (altivec_vmrglw_direct): Delete.
3037 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
3038 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
3040 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
3041 (vsx_xxmrglw_<mode>): Adjust.
3043 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
3045 * passes.def: Change threading comment before pass_ccp pass.
3047 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
3049 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
3050 Modify the expansion for sign extension. All extensions are done
3051 within VSX registers.
3053 2021-10-19 Richard Biener <rguenther@suse.de>
3055 * tree-vect-stmts.c (get_group_load_store_type): Add
3056 misalignment output parameter and initialize it.
3057 (get_group_load_store_type): Likewise.
3058 (vectorizable_store): Remove now redundant queries.
3059 (vectorizable_load): Likewise.
3061 2021-10-19 Richard Biener <rguenther@suse.de>
3063 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
3064 check_aligned argument.
3065 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
3067 (vect_peeling_hash_insert): Add supportable_if_not_aligned
3068 argument and do not call vect_supportable_dr_alignment here.
3069 (vect_peeling_supportable): Adjust.
3070 (vect_enhance_data_refs_alignment): Compute whether the
3071 access is supported with different alignment here and
3072 pass that down to vect_peeling_hash_insert.
3073 (vect_vfa_access_size): Adjust.
3074 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
3075 (vect_get_load_cost): Likewise.
3076 (get_negative_load_store_type): Likewise.
3077 (get_group_load_store_type): Likewise.
3078 (get_load_store_type): Likewise.
3080 2021-10-19 Martin Liska <mliska@suse.cz>
3083 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
3084 * system.h (strip_whilespaces): New function.
3086 2021-10-19 dianhong xu <dianhong.xu@intel.com>
3088 * config/i386/avx512fp16intrin.h:
3089 (_mm512_set1_pch): New intrinsic.
3090 * config/i386/avx512fp16vlintrin.h:
3091 (_mm256_set1_pch): New intrinsic.
3092 (_mm_set1_pch): Ditto.
3094 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
3096 PR tree-optimization/102796
3097 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
3098 normally. Return get_tree_range for non gimple_range_ssa_p names.
3099 (gimple_ranger::range_of_stmt): Use get_tree_range for non
3100 gimple_range_ssa_p names.
3102 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
3105 * config/i386/i386.c (ix86_print_operand_address):
3106 Error out for non-address_operand asm operands.
3108 2021-10-18 Richard Biener <rguenther@suse.de>
3110 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
3111 not auto-convert dr_alignment_support to bool.
3112 (vect_peeling_supportable): Likewise.
3113 (vect_enhance_data_refs_alignment): Likewise.
3114 (vect_supportable_dr_alignment): Commonize read/write case.
3115 * tree-vect-stmts.c (vect_get_store_cost): Use
3116 dr_alignment_support, not int, for the vect_supportable_dr_alignment
3118 (vect_get_load_cost): Likewise.
3120 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
3122 * tree-object-size.c (object_sizes_execute): Consolidate LHS
3123 null check and do it early.
3125 2021-10-18 Richard Biener <rguenther@suse.de>
3127 * tree-vect-stmts.c (vectorizable_store): Use the
3128 computed alignment scheme instead of querying
3131 2021-10-18 Richard Biener <rguenther@suse.de>
3133 * tree-vect-stmts.c (vectorizable_store): Do not recompute
3134 alignment scheme already determined by get_load_store_type.
3136 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
3138 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
3139 Clone corresponding pass.
3141 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3143 * combine.c (recog_for_combine): For an unrecognized move/set of
3144 a constant, try force_const_mem to place it in the constant pool.
3145 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
3146 Delete global variables (that are no longer assigned a cost value).
3147 (cse_insn): Simplify logic for deciding whether to place a folded
3148 constant in the constant pool using force_const_mem.
3149 (cse_main): Remove zero initialization of constant_pool_entries_cost
3150 and constant_pool_entries_regcost.
3151 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
3152 fractionally more expensive, when optimizing for speed.
3154 2021-10-18 Martin Liska <mliska@suse.cz>
3156 PR gcov-profile/102746
3157 PR gcov-profile/102747
3158 * gcov.c (main): Return return_code.
3159 (output_gcov_file): Mark return_code when error happens.
3160 (generate_results): Likewise.
3161 (read_graph_file): Likewise.
3162 (read_count_file): Likewise.
3164 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3166 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
3167 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
3168 a popcount, allowing compile-time evaluation/simplification.
3169 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
3171 2021-10-18 Richard Biener <rguenther@suse.de>
3173 PR tree-optimization/102788
3174 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
3176 (vect_pattern_recog_1): Likewise.
3177 (vect_recog_bool_pattern): Continue matching the pattern
3178 even if we do not have a vector type for a conversion
3181 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3183 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
3184 Evalute SS_NEG and SS_ABS of a constant argument.
3186 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
3189 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
3190 * internal-fn.def: Add entry for COND_NEG internal function.
3191 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
3192 NEGATE_EXPR, COND_NEG mapping.
3193 * optabs.def: Add entry for cond_neg_optab.
3194 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
3195 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
3196 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
3198 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
3200 * tree-ssa-strlen.c (class strlen_pass): Rename from
3202 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
3203 (handle_assign): Same.
3204 (adjust_last_stmt): Same.
3205 (maybe_warn_overflow): Same.
3206 (handle_builtin_strlen): Same.
3207 (handle_builtin_strchr): Same.
3208 (handle_builtin_strcpy): Same.
3209 (handle_builtin_strncat): Same.
3210 (handle_builtin_stxncpy_strncat): Same.
3211 (handle_builtin_memcpy): Same.
3212 (handle_builtin_strcat): Same.
3213 (handle_alloc_call): Same.
3214 (handle_builtin_memset): Same.
3215 (handle_builtin_memcmp): Same.
3216 (get_len_or_size): Same.
3217 (strxcmp_eqz_result): Same.
3218 (handle_builtin_string_cmp): Same.
3219 (handle_pointer_plus): Same.
3220 (count_nonzero_bytes_addr): Same.
3221 (count_nonzero_bytes): Same.
3222 (handle_store): Same.
3223 (strlen_check_and_optimize_call): Same.
3224 (handle_integral_assign): Same.
3225 (check_and_optimize_stmt): Same.
3226 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
3228 2021-10-18 Richard Biener <rguenther@suse.de>
3230 PR tree-optimization/102798
3231 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
3232 Only copy points-to info to newly generated SSA names.
3234 2021-10-18 Martin Liska <mliska@suse.cz>
3236 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
3237 * gcov.c (get_cycles_count): Likewise.
3238 * lto-compress.c (lto_compression_zlib): Likewise.
3239 (lto_uncompression_zlib): Likewise.
3240 * targhooks.c (default_pch_valid_p): Likewise.
3242 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
3244 * tree-pass.h (make_pass_thread_jumps_full): New.
3245 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
3246 (try_thread_blocks): Add resolve and speed arguments.
3247 (pass_thread_jumps::execute): Inline.
3248 (do_early_thread_jumps): New.
3249 (do_thread_jumps): New.
3250 (make_pass_thread_jumps): Move.
3251 (pass_early_thread_jumps::gate): Inline.
3252 (pass_early_thread_jumps::execute): Inline.
3253 (class pass_thread_jumps_full): New.
3255 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
3257 * configure.ac: Treat powerpc64*-*-freebsd* the same as
3258 powerpc64-*-freebsd*.
3259 * configure: Regenerate.
3261 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
3263 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
3264 set_ptr_nonull to set_ptr_nonnull in comments.
3266 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
3268 PR tree-optimization/102720
3269 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
3270 of call used and clobbered sets.
3272 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3274 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
3275 skip gomp and itm when r or nodefaultlibs is given.
3276 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
3277 Update ordering of exclusions, remove duplicate 'v' addition
3278 (collect2 will add this from the main command line).
3280 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3282 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
3283 filelist and framework options in specs instead of code.
3284 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
3285 specs that are really driver ones.
3286 (DARWIN_CC1_SPEC): Likewise.
3287 (CPP_SPEC): Likewise.
3288 (SYSROOT_SPEC): Append space.
3289 (LINK_SYSROOT_SPEC): Remove most driver link specs.
3290 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
3291 (STARTFILE_SPEC): Likewise.
3292 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
3293 (ASM_SPEC): Update driver-related specs.
3294 (ASM_FINAL_SPEC): Likewise.
3295 * config/darwin.opt: Remove now unused option aliases.
3296 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
3297 (ASM_SPEC): Update driver-related specs.
3299 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
3301 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
3302 allow standard_sse_constant_p immediate constants.
3304 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3306 * config.gcc: Add tm-dwarf2.h to tm_d-file.
3308 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
3310 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
3311 that occur in abnormal phis.
3312 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
3313 abnormal and EH edges.
3314 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
3315 (rvrp_folder::value_on_edge): Ditto.
3316 (rvrp_folder::value_of_stmt): Ditto.
3317 (hybrid_folder::value_of_expr): Ditto for ranger queries.
3318 (hybrid_folder::value_on_edge): Ditto.
3319 (hybrid_folder::value_of_stmt): Ditto.
3320 * value-query.cc (gimple_range_global): Always return a range if
3321 the type is supported.
3323 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
3325 * config/pa/pa.md: Consistently use "rG" constraint for copy
3326 instruction in move patterns.
3328 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
3330 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
3333 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
3334 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
3335 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
3336 (ELFABIVERSION_AMDGPU_HSA_V4): New.
3337 (SET_SRAM_ECC_UNSUPPORTED): New.
3338 (copy_early_debug_info): Create elf flags to match the other objects.
3339 (main): Just let the attribute flags pass through.
3341 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3343 * tree-loop-distribution.c (reduction_var_overflows_first):
3344 Pass the type of reduction_var as first argument as it is also
3345 done for the load type.
3346 (loop_distribution::transform_reduction_loop): Add missing
3347 TREE_TYPE while determining precission of reduction_var.
3349 2021-10-15 Richard Biener <rguenther@suse.de>
3351 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
3353 * toplev.c (process_options): Warn when STABS debugging is
3354 enabled but not the preferred format.
3355 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
3357 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
3359 2021-10-15 Richard Biener <rguenther@suse.de>
3362 * tree-inline.c (copy_bb): Avoid underflowing nargs.
3364 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
3366 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
3367 HFmode input operand to HImode.
3368 (ix86_vectorize_vec_perm_const): Likewise.
3369 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
3371 (*avx512f_permvar_truncv8siv8hi_1_hf):
3374 2021-10-15 Richard Biener <rguenther@suse.de>
3376 PR middle-end/102682
3377 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
3378 not create a paradoxical subreg.
3380 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
3382 * config/i386/i386-expand.c (ix86_expand_vector_init):
3383 For half_vector concat for HFmode, handle them like HImode.
3385 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3387 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
3388 (back_threader::back_threader): Same.
3389 (back_threader::resolve_phi): Try to solve without looking back if
3391 (back_threader::find_paths_to_names): Same.
3392 (try_thread_blocks): Pass resolve argument to back threader.
3393 (pass_early_thread_jumps::execute): Same.
3395 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3397 * doc/invoke.texi: Remove max-fsm-thread-length,
3398 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
3400 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
3402 (back_threader_registry::back_threader_registry): Same.
3403 (back_threader_profitability::profitable_path_p): Remove
3404 param_max_fsm_thread-length.
3405 (back_threader_registry::register_path): Remove
3406 m_max_allowable_paths.
3408 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3410 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
3412 (back_threader::~back_threader): Do not release m_path.
3414 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
3416 PR tree-optimization/102738
3417 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
3419 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
3421 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
3422 (omp_mark_declare_variant): Move from c-omp.c.
3423 (omp_context_name_list_prop): Update for Fortran strings.
3424 * omp-general.h (omp_check_context_selector): New prototype.
3425 (omp_mark_declare_variant): New prototype.
3427 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3429 * config/rs6000/rs6000.c (rs6000_density_test): Move early
3430 exit test further up the function.
3432 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3434 * config/arm/arm.c (arm_add_stmt_cost): Delete.
3435 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
3437 2021-10-14 Martin Jambor <mjambor@suse.cz>
3439 * doc/invoke.texi (Optimize Options): Add entry for
3440 ipa-cp-recursive-freq-factor.
3442 2021-10-14 Tamar Christina <tamar.christina@arm.com>
3444 * match.pd: New rule.
3446 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
3449 * ipa-modref-tree.h (modref_access_node::update2):
3450 Also check that parm_offset is unchanged.
3451 (modref_ref_node::insert_access): Fix updating of
3454 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3456 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
3459 2021-10-14 Richard Biener <rguenther@suse.de>
3461 PR tree-optimization/102659
3462 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
3463 rewrite pointer typed undefined overflow operations.
3464 (predicate_statements): Likewise. Make sure to emit invariant
3465 conversions in the preheader.
3466 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
3467 non-empty preheaders.
3468 * tree-data-ref.c (dr_analyze_indices): Strip useless
3469 conversions to the MEM_REF base type.
3471 2021-10-14 Martin Liska <mliska@suse.cz>
3473 * common.opt: Stop using AUTODETECT_VALUE
3474 and use EnabledBy where possible.
3475 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
3476 * toplev.c (AUTODETECT_VALUE): Remove macro.
3477 (process_options): Simplify by using EnabledBy and
3478 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
3481 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
3483 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
3485 (vld1_s32_x3): Likewise.
3487 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3489 PR tree-optimization/102736
3490 PR tree-optimization/102736
3491 * gimple-range-path.cc (path_range_query::range_on_path_entry):
3492 Assert that the requested range is defined outside the path.
3493 (path_range_query::ssa_range_in_phi): Do not call
3494 range_on_path_entry for SSA names that are defined within the
3497 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
3499 * config/darwin-driver.c (darwin_find_version_from_kernel):
3500 Quote internal identifiers and avoid contractions in
3502 (darwin_default_min_version): Likewise.
3503 (darwin_driver_init): Likewise.
3505 2021-10-14 Martin Jambor <mjambor@suse.cz>
3507 * params.opt (ipa-cp-recursive-freq-factor): New.
3508 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
3509 scc_no, self_recursion_generated_level, same_scc and
3510 self_recursion_generated_p.
3511 (ipcp_lattice::add_value): Replaced parameter unlimited with
3512 same_lat_gen_level, usit it determine limit of values and store it to
3514 (ipcp_lattice<valtype>::print): Dump the new fileds.
3515 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
3516 parameter and store it to the new value.
3517 (self_recursively_generated_p): Removed.
3518 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
3519 instead of self_recursively_generated_p, store self generation level
3521 (value_topo_info<valtype>::add_val): Set scc_no.
3522 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
3523 recursively feeding values and self generated values by appropriate
3526 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3528 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
3529 redundant test for flag_vect_cost_model.
3531 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3533 * bitmap.c (debug): New overloaded function for auto_bitmaps.
3534 * bitmap.h (debug): Same.
3536 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
3538 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
3539 Adjust builtin call.
3540 (_mm512_mask3_fcmadd_pch): Likewise.
3541 (_mm512_mask_fmadd_pch): Likewise
3542 (_mm512_mask3_fmadd_pch): Likewise
3543 (_mm512_mask_fcmadd_round_pch): Likewise
3544 (_mm512_mask3_fcmadd_round_pch): Likewise
3545 (_mm512_mask_fmadd_round_pch): Likewise
3546 (_mm512_mask3_fmadd_round_pch): Likewise
3547 (_mm_mask_fcmadd_sch): Likewise
3548 (_mm_mask3_fcmadd_sch): Likewise
3549 (_mm_mask_fmadd_sch): Likewise
3550 (_mm_mask3_fmadd_sch): Likewise
3551 (_mm_mask_fcmadd_round_sch): Likewise
3552 (_mm_mask3_fcmadd_round_sch): Likewise
3553 (_mm_mask_fmadd_round_sch): Likewise
3554 (_mm_mask3_fmadd_round_sch): Likewise
3555 (_mm_fcmadd_round_sch): Likewise
3556 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
3557 Adjust builtin call.
3558 (_mm_mask3_fmadd_pch): Likewise
3559 (_mm256_mask_fmadd_pch): Likewise
3560 (_mm256_mask3_fmadd_pch): Likewise
3561 (_mm_mask_fcmadd_pch): Likewise
3562 (_mm_mask3_fcmadd_pch): Likewise
3563 (_mm256_mask_fcmadd_pch): Likewise
3564 (_mm256_mask3_fcmadd_pch): Likewise
3565 * config/i386/i386-builtin.def: Add mask3 builtin for complex
3566 fma, and adjust mask_builtin to corresponding expander.
3567 * config/i386/i386-expand.c (ix86_expand_round_builtin):
3568 Skip eraseing embedded rounding for expanders that emits
3570 * config/i386/sse.md (complexmove): New mode_attr.
3571 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
3572 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
3573 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
3574 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
3575 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
3576 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
3577 * config/i386/subst.md (round_embedded_complex): New subst.
3579 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
3581 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
3582 (cbranchdf4): Likewise.
3583 Add missing move patterns for TARGET_SOFT_FLOAT.
3585 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
3587 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
3588 pseudo intermediate when moving a SUBREG into a hard register,
3589 by checking ix86_hardreg_mov_ok.
3590 (ix86_expand_vector_extract): Store zero-extended SImode
3591 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
3593 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
3594 complex (SUBREG) sets of (vector) hard registers before reload, by
3595 checking ix86_hardreg_mov_ok.
3597 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
3599 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
3601 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
3603 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
3605 2021-10-13 Martin Sebor <msebor@redhat.com>
3607 PR middle-end/102630
3608 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
3610 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
3612 * collect2.c (is_lto_object_file): Release simple-object
3613 resources, close files.
3615 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
3617 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
3618 when checking mve addressing modes.
3619 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
3621 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
3622 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
3623 there is a single Ux constraint.
3625 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
3627 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
3628 multiplication on 32-bit PA 1.1 and 2.0 targets.
3630 2021-10-13 Martin Liska <mliska@suse.cz>
3632 PR gcov-profile/90364
3633 * coverage.c (build_info): Emit checksum to the global variable.
3634 (build_info_type): Add new field for checksum.
3635 (coverage_obj_finish): Pass object_checksum.
3636 (coverage_init): Use 0 as checksum for .gcno files.
3637 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
3638 * gcov.c (read_graph_file): Read also checksum.
3639 * doc/invoke.texi: Document the behaviour change.
3641 2021-10-13 Richard Biener <rguenther@suse.de>
3643 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
3644 start at integer value 2.
3645 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
3646 the iterator for GSI_LAST_NEW_STMT.
3647 (gsi_insert_seq_nodes_after): Likewise.
3648 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
3649 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
3650 arguments to gsi_replace.
3652 2021-10-13 Martin Liska <mliska@suse.cz>
3655 * common.opt: Use EnabledBy instead of detection in
3656 finish_options and process_options.
3657 * opts.c (finish_options): Remove handling of
3658 x_flag_unroll_all_loops.
3659 * toplev.c (process_options): Likewise for flag_web and
3660 flag_rename_registers.
3662 2021-10-13 Richard Biener <rguenther@suse.de>
3664 PR tree-optimization/102659
3665 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
3666 (if_convertible_gimple_assign_stmt_p): Mark the loop for
3667 rewrite when stmts with undefined behavior on integer
3669 (combine_blocks): Predicate also when we need to rewrite stmts.
3670 (predicate_statements): Rewrite affected stmts to something
3671 with well-defined behavior on overflow.
3672 (tree_if_conversion): Initialize need_to_rewrite_undefined.
3674 2021-10-13 Richard Biener <rguenther@suse.de>
3677 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
3680 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
3682 * dwarf2ctf.c: Fix typo in comment.
3684 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
3687 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
3688 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
3689 avoid generating an automatic #define of vec_cpsgn. Use the
3690 correct built-in for V4SFmode that doesn't depend on VSX.
3692 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
3696 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
3697 define_insn_and_split pattern. Add alternative 1 and split it
3698 post reload to insert operand 1 into the low part of operand 0.
3699 (*sub<mode>_1_slp): Ditto.
3700 (*and<mode>_1_slp): Ditto.
3701 (*<any_or:code><mode>_1_slp): Ditto.
3702 (*ashl<mode>3_1_slp): Ditto.
3703 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
3704 (*<any_rotate:insn><mode>3_1_slp): Ditto.
3705 (*neg<mode>_1_slp): New insn_and_split pattern.
3706 (*one_cmpl<mode>_1_slp): Ditto.
3708 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
3710 * doc/install.texi: Update MinGW and mingw-64 Binaries
3713 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
3715 * doc/extend.texi (Common Variable Attributes): Fix typos in
3716 alloc_size documentation.
3718 2021-10-12 Richard Biener <rguenther@suse.de>
3720 PR tree-optimization/102696
3721 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
3722 the tree fatally failed when we reject a BIT_FIELD_REF.
3724 2021-10-12 Richard Biener <rguenther@suse.de>
3726 PR tree-optimization/102572
3727 * tree-vect-stmts.c (vect_build_gather_load_calls): When
3728 gathering the vectorized defs for the mask pass in the
3729 desired mask vector type so invariants will be handled
3732 2021-10-12 Tamar Christina <tamar.christina@arm.com>
3734 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
3735 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
3736 *fcmuo<mode>_nor_combine): New.
3738 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
3741 * config/sparc/sparc-modes.def (OI): New integer mode.
3743 2021-10-12 Jakub Jelinek <jakub@redhat.com>
3745 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
3746 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
3749 2021-10-12 Jakub Jelinek <jakub@redhat.com>
3751 * tree-vectorizer.h (loop_cost_model): New function.
3752 (unlimited_cost_model): Use it.
3753 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
3754 call instead of flag_vect_cost_model.
3755 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
3756 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
3757 of flag_simd_cost_model.
3759 2021-10-12 liuhongt <hongtao.liu@intel.com>
3762 * config/i386/i386-expand.c (emit_reduc_half): Handle
3764 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
3765 (reduc_plus_scal_v4qi): Ditto.
3767 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
3769 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
3770 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
3771 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
3773 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
3775 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
3776 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
3777 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
3778 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
3779 _mm_cvtepu32_epi64): New.
3781 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
3783 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
3784 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
3786 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
3788 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
3789 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
3790 _mm_max_epi32, _mm_max_epu32): New.
3792 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
3794 * ipa-modref-tree.h (struct modref_access_node): Revert
3796 (struct modref_ref_node): Likewise.
3798 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
3800 * ipa-modref-tree.h (modref_tree::global_access_p): New member
3803 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
3804 ignore_stores_eaf_flags): Move to ipa-modref.h
3805 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
3806 (modref_summary::global_memory_read_p): New member function.
3807 (modref_summary::global_memory_written_p): New member function.
3808 * ipa-modref.h (modref_summary::global_memory_read_p,
3809 modref_summary::global_memory_written_p): Declare.
3810 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
3811 ignore_stores_eaf_flags): move here.
3812 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
3814 (handle_rhs_call): Rewrite.
3815 (handle_call_arg): New function.
3816 (determine_global_memory_access): New function.
3817 (handle_const_call): Remove
3818 (handle_pure_call): Remove
3819 (find_func_aliases_for_call): Update use of handle_rhs_call.
3820 (compute_points_to_sets): Handle global memory acccesses
3823 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
3825 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
3826 documentation, mention UBSAN_OPTIONS, similar to what is done
3827 for AddressSanitizer.
3829 2021-10-11 Richard Biener <rguenther@suse.de>
3831 PR middle-end/102683
3832 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
3833 availability before building an integer type for storage
3836 2021-10-11 Richard Biener <rguenther@suse.de>
3838 PR middle-end/101480
3839 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
3842 2021-10-11 Martin Liska <mliska@suse.cz>
3844 * common.opt: Remove Init(2) for some options.
3845 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
3846 use rather OPTION_SET_P.
3848 2021-10-11 Martin Liska <mliska@suse.cz>
3850 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
3851 * flag-types.h (enum ira_region): Likewise.
3852 * toplev.c (process_options): Use OPTION_SET_P instead of
3853 IRA_REGION_AUTODETECT.
3855 2021-10-11 Jakub Jelinek <jakub@redhat.com>
3857 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
3858 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
3860 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3862 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
3863 * config/s390/s390.c (s390_rawmemchr): New function.
3864 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
3865 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
3866 the pattern vfees<mode> from vx-builtins.md.
3867 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
3869 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3871 * builtins.c (get_memory_rtx): Change to external linkage.
3872 * builtins.h (get_memory_rtx): Add function prototype.
3873 * doc/md.texi (rawmemchr<mode>): Document.
3874 * internal-fn.c (expand_RAWMEMCHR): Define.
3875 * internal-fn.def (RAWMEMCHR): Add.
3876 * optabs.def (rawmemchr_optab): Add.
3877 * tree-loop-distribution.c (find_single_drs): Change return code
3878 behaviour by also returning true if no single store was found
3880 (loop_distribution::classify_partition): Respect the new return
3881 code behaviour of function find_single_drs.
3882 (loop_distribution::execute): Call new function
3883 transform_reduction_loop in order to replace rawmemchr or strlen
3884 like loops by calls into builtins.
3885 (generate_reduction_builtin_1): New function.
3886 (generate_rawmemchr_builtin): New function.
3887 (generate_strlen_builtin_1): New function.
3888 (generate_strlen_builtin): New function.
3889 (generate_strlen_builtin_using_rawmemchr): New function.
3890 (reduction_var_overflows_first): New function.
3891 (determine_reduction_stmt_1): New function.
3892 (determine_reduction_stmt): New function.
3893 (loop_distribution::transform_reduction_loop): New function.
3895 2021-10-11 Martin Liska <mliska@suse.cz>
3897 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
3898 and remove legacy hashing code.
3900 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
3903 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
3904 CONST_INT for BEGIN and END, and use gcc_assert rather than
3907 2021-10-10 Jakub Jelinek <jakub@redhat.com>
3910 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
3911 use MO_VAL_SET if loc is not sp.
3913 2021-10-10 Andrew Pinski <apinski@marvell.com>
3915 PR tree-optimization/102622
3916 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
3917 Swap the order of a?0:pow2cst and a?0:-1 transformations.
3919 2021-10-09 liuhongt <hongtao.liu@intel.com>
3922 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
3924 (ix86_use_mask_cmp_p): Ditto.
3925 (ix86_expand_sse_movcc): Ditto.
3926 * config/i386/i386.md (setcc_hf_mask): New define_insn.
3927 (movhf_mask): Ditto.
3928 (UNSPEC_MOVCC_MASK): New unspec.
3929 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
3931 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
3933 PR rtl-optimization/102627
3934 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
3936 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
3938 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
3941 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
3943 * value-range.cc (irange::debug): New.
3944 * value-range.h (irange::debug): New.
3946 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
3948 PR tree-optimization/102385
3949 * predict.h (change_edge_frequency): Declare.
3950 * predict.c (change_edge_frequency): New function.
3951 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
3953 (tree_unroll_loop): Likewise.
3954 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
3955 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
3957 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
3958 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
3959 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
3960 to retrieve the exit edges. Make all the old profile update code
3961 conditional on !single_loop_p -- the case it was written for --
3962 and use a different approach for the single-loop case.
3964 2021-10-08 Martin Liska <mliska@suse.cz>
3966 * config/alpha/alpha.c (alpha_option_override): Use new macro
3968 * config/arc/arc.c (arc_override_options): Likewise.
3969 * config/arm/arm.c (arm_option_override): Likewise.
3970 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
3971 * config/c6x/c6x.c (c6x_option_override): Likewise.
3972 * config/csky/csky.c: Likewise.
3973 * config/darwin.c (darwin_override_options): Likewise.
3974 * config/frv/frv.c (frv_option_override): Likewise.
3975 * config/i386/djgpp.h: Likewise.
3976 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
3977 (ix86_max_noce_ifcvt_seq_cost): Likewise.
3978 * config/ia64/ia64.c (ia64_option_override): Likewise.
3979 (ia64_override_options_after_change): Likewise.
3980 * config/m32c/m32c.c (m32c_option_override): Likewise.
3981 * config/m32r/m32r.c (m32r_init): Likewise.
3982 * config/m68k/m68k.c (m68k_option_override): Likewise.
3983 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
3984 * config/mips/mips.c (mips_option_override): Likewise.
3985 * config/nios2/nios2.c (nios2_option_override): Likewise.
3986 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
3987 * config/pa/pa.c (pa_option_override): Likewise.
3988 * config/riscv/riscv.c (riscv_option_override): Likewise.
3989 * config/rs6000/aix71.h: Likewise.
3990 * config/rs6000/aix72.h: Likewise.
3991 * config/rs6000/aix73.h: Likewise.
3992 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
3993 (rs6000_override_options_after_change): Likewise.
3994 (rs6000_linux64_override_options): Likewise.
3995 (glibc_supports_ieee_128bit): Likewise.
3996 (rs6000_option_override_internal): Likewise.
3997 (rs6000_file_start): Likewise.
3998 (rs6000_darwin_file_start): Likewise.
3999 * config/rs6000/rtems.h: Likewise.
4000 * config/rs6000/sysv4.h: Likewise.
4001 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
4002 * config/s390/s390.c (s390_option_override): Likewise.
4003 * config/sh/linux.h: Likewise.
4004 * config/sh/netbsd-elf.h (while): Likewise.
4005 * config/sh/sh.c (sh_option_override): Likewise.
4006 * config/sol2.c (solaris_override_options): Likewise.
4007 * config/sparc/sparc.c (sparc_option_override): Likewise.
4008 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
4009 * config/visium/visium.c (visium_option_override): Likewise.
4010 * config/vxworks.c (vxworks_override_options): Likewise.
4011 * lto-opts.c (lto_write_options): Likewise.
4012 * omp-expand.c (expand_omp_simd): Likewise.
4013 * omp-general.c (omp_max_vf): Likewise.
4014 * omp-offload.c (oacc_xform_loop): Likewise.
4015 * opts.h (OPTION_SET_P): Likewise.
4016 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
4017 * toplev.c (process_options): Likewise.
4018 * tree-predcom.c: Likewise.
4019 * tree-sra.c (analyze_all_variable_accesses): Likewise.
4021 2021-10-08 liuhongt <hongtao.liu@intel.com>
4024 * config/i386/i386.c (ix86_optab_supported_p):
4025 Return true for HFmode.
4026 * match.pd: Simplify (_Float16) ceil ((double) x) to
4027 __builtin_ceilf16 (a) when a is _Float16 type and
4028 direct_internal_fn_supported_p.
4030 2021-10-08 liuhongt <hongtao.liu@intel.com>
4033 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
4034 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
4035 (reduc_<code>_scal_v4hi): New.
4037 2021-10-08 liuhongt <hongtao.liu@intel.com>
4039 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
4040 * doc/invoke.texi (Options That Control Optimization): Update
4042 * opts.c (default_options_table): Enable auto-vectorization at
4043 O2 with very-cheap cost model.
4044 (finish_options): Use cheap cost model for
4045 explicit -ftree{,-loop}-vectorize.
4047 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
4049 * ctfc.c (ctfc_delete_container): Free hash table contents.
4051 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
4053 * toplev.c (process_options): Do not warn for GNU GIMPLE.
4055 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
4057 * tree-object-size.c (addr_object_size,
4058 compute_builtin_object_size): Drop PDECL and POFF arguments.
4059 (addr_object_size): Adjust calls.
4060 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
4063 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
4065 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
4066 representing signed and unsigned high-part multiplication resp.
4067 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
4068 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
4069 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
4070 SS_DIV, US_DIV]: Similar simplifications for saturating
4072 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
4073 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
4074 Implement compile-time evaluation for constant operands.
4075 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
4077 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
4078 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
4079 Mention the new smul_highpart and umul_highpart RTX codes.
4080 * doc/invoke.texi: Silence @xref "compilation" warnings.
4082 2021-10-07 Martin Jambor <mjambor@suse.cz>
4085 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
4086 case when the source reference description corresponds to a
4087 referance taken in a function src->caller is inlined to.
4089 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
4092 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
4094 (modref_access_node::try_merge_with): Add sanity check that there
4095 are no redundant entries in the list.
4097 2021-10-07 Richard Biener <rguenther@suse.de>
4099 PR tree-optimization/102608
4100 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
4103 2021-10-07 Martin Liska <mliska@suse.cz>
4105 * toplev.c (toplev::main): Make
4106 save_opt_decoded_options a pointer type
4107 * toplev.h: Likewise.
4109 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4111 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
4112 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
4113 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
4115 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4117 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
4118 Adjust for new -msram-ecc=any behaviour.
4119 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
4120 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
4121 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
4122 (EF_AMDGPU_XNACK_V3): ... this.
4123 (EF_AMDGPU_SRAM_ECC): Rename to ...
4124 (EF_AMDGPU_SRAM_ECC_V3): ... this.
4125 (EF_AMDGPU_FEATURE_XNACK_V4): New.
4126 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
4127 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
4128 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
4129 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
4130 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
4131 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
4132 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
4133 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
4134 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
4135 (SET_XNACK_ON): New.
4136 (SET_XNACK_OFF): New.
4138 (SET_SRAM_ECC_ON): New.
4139 (SET_SRAM_ECC_ANY): New.
4140 (SET_SRAM_ECC_OFF): New.
4141 (TEST_SRAM_ECC_ANY): New.
4142 (TEST_SRAM_ECC_ON): New.
4143 (main): Implement HSACOv4 and -msram-ecc=any.
4145 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4147 * config.in: Regenerate.
4148 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
4152 (A_FIJI): Rename to ...
4154 (A_900): Rename to ...
4156 (A_906): Rename to ...
4158 (A_908): Rename to ...
4160 (SRAMOPT): New macro.
4161 (ASM_SPEC): Adjust xnack option usage.
4162 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
4163 * configure: Regenerate.
4164 * configure.ac: Detect LLVM assembler dialect.
4166 2021-10-07 Richard Biener <rguenther@suse.de>
4168 * tree-pretty-print.c (dump_generic_node): Do not elide
4169 printing '&' when dumping with -gimple.
4171 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4173 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
4175 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
4176 (adjust_imagpart_expr): Ditto.
4177 * value-range.cc (irange::irange_intersect): Call new routine if
4178 RHS is a single pair.
4179 (irange::intersect): New wide_int version.
4180 * value-range.h (class irange): New prototype.
4182 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4184 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
4185 Add parameter to limit size when recognizing switches.
4186 (gimple_outgoing_range::edge_range_p): Check size limit.
4187 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
4188 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
4189 that exceed the size limit.
4190 (gori_compute::gori_compute): Add initializer.
4191 * params.opt (evrp-switch-limit): New.
4192 * doc/invoke.texi: Update docs.
4194 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4196 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
4197 TYPE_MAX_VALUE instead of creating new trees when possible.
4199 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4201 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
4202 zero and non-zero more efficently.
4204 2021-10-06 Richard Biener <rguenther@suse.de>
4207 * dumpfile.h (TDF_GIMPLE_VAL): New.
4208 (dump_flag): Re-order and adjust TDF_* flags. Make
4209 the enum uint32_t. Use std::underlying_type in the
4211 (optgroup_flag): Likewise for the operator overloads.
4212 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
4213 in _Literal if TDF_GIMPLE_VAL.
4214 * gimple-pretty-print.c (dump_gimple_assign): Add
4215 TDF_GIMPLE_VAL to flags when dumping operands where only
4216 is_gimple_val are allowed.
4217 (dump_gimple_cond): Likewise.
4219 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
4221 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
4224 2021-10-05 qing zhao <qing.zhao@oracle.com>
4226 PR middle-end/102359
4227 * gimplify.c (gimplify_decl_expr): Not add initialization for an
4228 auto variable when it has been initialized by frontend.
4230 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
4232 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
4235 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4237 * common/config/avr/avr-common.c (avr_handle_option): Mark
4238 argument as ATTRIBUTE_UNUSED.
4240 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4242 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
4243 Undefine before redefinition.
4245 2021-10-05 Richard Biener <rguenther@suse.de>
4247 * toplev.c (no_backend): Remove global var.
4248 (process_options): Pass in no_backend, move post_options
4249 langhook call to toplev::main.
4250 (do_compile): Pass in no_backend, move process_options call
4252 (toplev::run_self_tests): Check no_backend at the caller.
4253 (toplev::main): Call post_options and process_options
4254 split out from do_compile, do self-tests only if
4255 no_backend is initialized.
4257 2021-10-05 Richard Biener <rguenther@suse.de>
4259 * tree-cfg.c (dump_function_to_file): Dump the UID of the
4260 function as part of the name when requested.
4261 * tree-pretty-print.c (dump_function_name): Dump the UID when
4262 requested and the langhook produced the actual name.
4264 2021-10-05 Richard Biener <rguenther@suse.de>
4266 PR middle-end/102587
4267 PR middle-end/102285
4268 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
4269 zero-initialization as last resort, use the constant
4270 size as given by the DEFERRED_INIT argument to build
4273 2021-10-04 Marek Polacek <polacek@redhat.com>
4276 * doc/invoke.texi: Document -Warray-compare.
4278 2021-10-04 Richard Biener <rguenther@suse.de>
4280 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
4281 variables are not to be initialized.
4283 2021-10-04 Richard Biener <rguenther@suse.de>
4285 * expr.h (non_mem_decl_p): Declare.
4286 (mem_ref_refers_to_non_mem_p): Likewise.
4287 * expr.c (non_mem_decl_p): Export.
4288 (mem_ref_refers_to_non_mem_p): Likewise.
4289 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
4290 but check the base with mem_ref_refers_to_non_mem_p
4293 2021-10-04 Richard Biener <rguenther@suse.de>
4295 PR tree-optimization/102570
4296 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
4297 we are using clique for the internal function code.
4298 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
4299 internal function code.
4300 (print_vn_reference_ops): Print the internal function code.
4301 (vn_reference_op_compute_hash): Hash it.
4302 (copy_reference_ops_from_call): Record it.
4303 (visit_stmt): Remove the restriction around internal function
4305 (fully_constant_vn_reference_p): Use fold_const_call and handle
4307 (vn_reference_eq): Compare call return types.
4308 * tree-ssa-pre.c (create_expression_by_pieces): Handle
4309 generating calls to internal functions.
4310 (compute_avail): Remove the restriction around internal function
4313 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
4315 PR tree-optimization/102560
4316 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
4317 marker for invalid_range.
4319 2021-10-04 Richard Biener <rguenther@suse.de>
4321 PR middle-end/102587
4322 * internal-fn.c (expand_DEFERRED_INIT): Guard register
4323 initialization path an avoid initializing VLA registers
4326 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
4328 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
4330 2021-10-03 Martin Liska <mliska@suse.cz>
4332 * toplev.c (toplev::main): Check opt_index if it is a part
4335 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
4337 PR tree-optimization/102563
4338 * range-op.cc (operator_lshift::op1_range): Do not clobber
4341 2021-10-02 Martin Liska <mliska@suse.cz>
4343 * toplev.c (toplev::main): save_decoded_options[0] is program
4344 name and so it should be skipped.
4346 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
4348 PR tree-optimization/102546
4349 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
4350 X << Y is non-zero implies X is also non-zero.
4352 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4354 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4356 * config/aarch64/aarch64-tune.md: Regenerate.
4357 * doc/invoke.texi: Update docs.
4359 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4361 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4363 * config/aarch64/aarch64-tune.md: Regenerate.
4364 * doc/invoke.texi: Update docs.
4366 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4368 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4370 * config/aarch64/aarch64-tune.md: Regenerate.
4371 * doc/invoke.texi: Update docs.
4373 2021-10-01 Martin Sebor <msebor@redhat.com>
4376 * doc/invoke.texi (-Waddress): Update.
4377 * gengtype.c (write_types): Avoid -Waddress.
4378 * poly-int.h (POLY_SET_COEFF): Avoid using null.
4380 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
4383 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
4386 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4388 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
4390 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
4392 * gimple-range-path.cc (path_range_query::compute_ranges): Use
4394 * gimple-range-path.h (class path_range_query): Remove shadowed
4396 (path_range_query::get_path_oracle): New.
4398 2021-10-01 Jakub Jelinek <jakub@redhat.com>
4399 Richard Biener <rguenther@suse.de>
4402 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
4403 INT_MIN / -1 division detection from here ...
4404 (-fsanitize=signed-integer-overflow): ... and add it here.
4406 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4408 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
4410 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
4411 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
4412 (AARCH64_ISA_V9): New ISA flag.
4413 * doc/invoke.texi: Update docs.
4415 2021-10-01 Martin Liska <mliska@suse.cz>
4417 * toplev.c (toplev::main): Save decoded optimization options.
4418 * toplev.h (save_opt_decoded_options): New.
4419 * doc/extend.texi: Be more clear about optimize and target
4422 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
4424 * explow.c: Include langhooks.h.
4425 (set_stack_check_libfunc): Build a proper function type.
4427 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
4430 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
4431 not return true for external weak function symbols in medium model.
4433 2021-10-01 Jakub Jelinek <jakub@redhat.com>
4435 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
4436 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
4437 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
4438 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
4439 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
4440 OMP_CLAUSE_SCHEDULE_STATIC.
4442 2021-10-01 Richard Biener <rguenther@suse.de>
4444 PR middle-end/102518
4445 * tree-inline.c (setup_one_parameter): Avoid substituting
4446 an invariant into contexts where a GIMPLE register is not valid.
4448 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4450 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
4451 * config/arm/arm-tables.opt: Regenerate.
4452 * config/arm/arm-tune.md: Regenerate.
4453 * doc/invoke.texi: Update docs.
4455 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
4458 * config/i386/i386.md
4459 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
4462 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
4464 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
4465 omp_{c,re}alloc, fix omp_alloc/omp_free.
4467 2021-09-30 Martin Liska <mliska@suse.cz>
4469 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
4470 asm_out_file and stream directly to MYFILE.
4472 2021-09-30 Richard Biener <rguenther@suse.de>
4474 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
4475 Restore and fix condition under which we apply npeel to
4476 the DRs misalignment value.
4478 2021-09-30 Richard Biener <rguenther@suse.de>
4480 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
4481 Fix npeel check for variable amount of peeling.
4483 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
4485 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
4487 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
4489 * gimple-range.cc (gimple_ranger::debug): New.
4490 * gimple-range.h (class gimple_ranger): Add debug.
4492 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
4494 PR middle-end/102519
4495 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
4497 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
4500 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
4501 (btf_finalize): Empty the hash_map btf_var_ids.
4503 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
4505 * tree-vrp.c (thread_through_all_blocks): Return bool.
4506 (execute_vrp_threader): Return TODO_* flags.
4507 (pass_data_vrp_threader): Set todo_flags_finish to 0.
4509 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
4511 * timevar.def (TV_TREE_VRP_THREADER): New.
4512 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
4514 2021-09-29 David Faust <david.faust@oracle.com>
4516 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
4518 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
4520 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
4521 new type qualifier enum.
4522 (TYPES_TERNOP_SSSU): Likewise.
4523 (TYPES_TERNOP_PPPU): Likewise.
4524 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
4525 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
4527 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
4528 builtin and remove casts.
4529 (vqtbl1_s8): Likewise.
4530 (vqtbl1q_p8): Likewise.
4531 (vqtbl1q_s8): Likewise.
4532 (vqtbx1_s8): Likewise.
4533 (vqtbx1_p8): Likewise.
4534 (vqtbx1q_s8): Likewise.
4535 (vqtbx1q_p8): Likewise.
4536 (vtbl1_p8): Likewise.
4537 (vtbl2_p8): Likewise.
4538 (vtbx2_p8): Likewise.
4540 2021-09-29 Richard Biener <rguenther@suse.de>
4542 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
4544 (vect_update_misalignment_for_peel): Use it to update
4545 misaligned to the value necessary for an aligned access.
4546 (vect_get_peeling_costs_all_drs): Likewise.
4547 (vect_enhance_data_refs_alignment): Likewise.
4549 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4551 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
4552 emitted operations and adjust heuristic for code size.
4554 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4556 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
4557 emitted operations and adjust heuristic for code size.
4559 2021-09-29 Jakub Jelinek <jakub@redhat.com>
4561 PR middle-end/102504
4562 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
4563 in OMP_SCOPE clauses, not just on worksharing construct clauses.
4565 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
4567 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
4568 (<u>mulditi3): Call <su>muldi3_highpart.
4569 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
4570 (<u>mulsidi3): Call <su>mulsi3_highpart.
4571 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
4573 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
4575 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
4577 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
4579 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
4580 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
4583 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4585 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
4586 (AARCH64_FL_V8_7): Likewise.
4587 (AARCH64_FL_FOR_ARCH8_7): Likewise.
4588 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
4589 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
4590 * doc/invoke.texi: Document the above.
4592 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4594 * dbgcnt.c (dbg_cnt_counter): New.
4595 * dbgcnt.h (dbg_cnt_counter): New.
4596 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
4597 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
4598 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
4599 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
4602 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4604 * cfgcleanup.c (pass_jump::execute): Check
4605 flag_expensive_optimizations.
4606 (pass_jump_after_combine::gate): Same.
4607 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
4608 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
4609 * tree-ssa-threadupdate.c
4610 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
4613 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
4615 * tree-ssa-reassoc.c (biased_names): New global.
4616 (propagate_bias_p): New function.
4617 (loop_carried_phi): Remove.
4618 (propagate_rank): Propagate bias along single uses.
4619 (get_rank): Update biased_names when needed.
4621 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
4623 * passes.def (pass_reassoc): Rename parameter to early_p.
4624 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
4626 (phi_rank): Don't bias loop-carried phi ranks
4627 before vectorization pass.
4628 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
4629 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
4631 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
4633 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
4635 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
4637 2021-09-28 Jakub Jelinek <jakub@redhat.com>
4640 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
4641 special 80387 instruction XFmode constants if flag_rounding_math.
4643 2021-09-28 Richard Biener <rguenther@suse.de>
4645 PR tree-optimization/100112
4646 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
4647 referece into the hashtable twice in case last_vuse is
4648 different from the original vuse on the stmt.
4650 2021-09-28 Jakub Jelinek <jakub@redhat.com>
4652 PR middle-end/102492
4653 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
4654 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
4655 clauses on SIMD constructs.
4657 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4659 PR tree-optimization/102511
4660 * gimple-range-path.cc (path_range_query::range_on_path_entry):
4661 Return VARYING when nothing found.
4663 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
4666 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
4668 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
4669 (VALID_MMX_REG_MODE): Add V4HFmode.
4670 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
4671 vector mode condition.
4672 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
4674 (function_arg_32): Add V4HFmode.
4675 (function_arg_advance_32): Likewise.
4676 * config/i386/i386.md (mode): Add V4HF/V2HF.
4677 (MODE_SIZE): Likewise.
4678 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
4679 (V_32): Add V2HF mode.
4680 (VHF_32_64): New mode iterator.
4681 (*mov<mode>_internal): Adjust sse alternatives to support
4683 (*mov<mode>_internal): Adjust sse alternatives to support
4685 (<insn><mode>3): New define_insn for add/sub/mul/div.
4687 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
4689 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
4691 (pass_early_thread_jumps::gate): Same.
4692 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
4693 Return if !flag_thread_jumps.
4694 * tree-ssa-threadupdate.c
4695 (jt_path_registry::register_jump_thread): Assert that
4696 flag_thread_jumps is true.
4698 2021-09-28 liuhongt <hongtao.liu@intel.com>
4701 (simplify_context::simplify_binary_operation_1): Relax
4702 condition of simplifying (vec_concat:M (vec_select op0
4703 index0)(vec_select op1 index1)) to allow different modes
4704 between op0 and M, but have same inner mode.
4706 2021-09-28 liuhongt <hongtao.liu@intel.com>
4708 * config/i386/i386-expand.c (emit_reduc_half): Handle
4709 V8HF/V16HF/V32HFmode.
4710 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
4711 (REDUC_SSE_SMINMAX_MODE): Ditto.
4712 (REDUC_PLUS_MODE): Add V16HF and V32HF.
4713 (REDUC_SMINMAX_MODE): Ditto.
4715 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4717 * gimple-range-path.cc
4718 (path_range_query::precompute_ranges_in_block): Rename to...
4719 (path_range_query::compute_ranges_in_block): ...this.
4720 (path_range_query::precompute_ranges): Rename to...
4721 (path_range_query::compute_ranges): ...this.
4722 (path_range_query::precompute_relations): Rename to...
4723 (path_range_query::compute_relations): ...this.
4724 (path_range_query::precompute_phi_relations): Rename to...
4725 (path_range_query::compute_phi_relations): ...this.
4726 * gimple-range-path.h: Rename precompute* to compute*.
4727 * tree-ssa-threadbackward.c
4728 (back_threader::find_taken_edge_switch): Same.
4729 (back_threader::find_taken_edge_cond): Same.
4730 * tree-ssa-threadedge.c
4731 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
4732 (hybrid_jt_state::register_equivs_stmt): Inline...
4733 * tree-ssa-threadedge.h: ...here.
4735 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4737 * tree-vrp.c (lhs_of_dominating_assert): Remove.
4738 (class vrp_jt_state): Remove.
4739 (class vrp_jt_simplifier): Remove.
4740 (vrp_jt_simplifier::simplify): Remove.
4741 (class vrp_jump_threader): Remove.
4742 (vrp_jump_threader::vrp_jump_threader): Remove.
4743 (vrp_jump_threader::~vrp_jump_threader): Remove.
4744 (vrp_jump_threader::before_dom_children): Remove.
4745 (vrp_jump_threader::after_dom_children): Remove.
4747 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4749 * passes.def (pass_vrp_threader): New.
4750 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
4751 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
4752 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
4753 (hybrid_jt_simplifier::simplify): New.
4754 (hybrid_jt_simplifier::compute_ranges_from_state): New.
4755 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
4756 (class hybrid_jt_simplifier): New.
4757 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
4759 (class hybrid_threader): New.
4760 (hybrid_threader::hybrid_threader): New.
4761 (hybrid_threader::~hybrid_threader): New.
4762 (hybrid_threader::before_dom_children): New.
4763 (hybrid_threader::after_dom_children): New.
4764 (execute_vrp_threader): New.
4765 (class pass_vrp_threader): New.
4766 (make_pass_vrp_threader): New.
4768 2021-09-27 Martin Liska <mliska@suse.cz>
4770 * output.h (enum section_flag): New.
4771 (SECTION_FORGET): Remove.
4772 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
4773 (SECTION_STYLE_MASK): Define it based on other enum
4775 * varasm.c (switch_to_section): Remove unused handling of
4778 2021-09-27 Martin Liska <mliska@suse.cz>
4780 * common.opt: Add new variable flag_default_complex_method.
4781 * opts.c (finish_options): Handle flags related to
4782 x_flag_complex_method.
4783 * toplev.c (process_options): Remove option handling related
4784 to flag_complex_method.
4786 2021-09-27 Richard Biener <rguenther@suse.de>
4788 PR middle-end/102450
4789 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
4790 type_for_size, instead use int_mode_for_size.
4792 2021-09-27 Andrew Pinski <apinski@marvell.com>
4795 * gimplify.c (gimplify_save_expr): Return early
4796 if the type of val is error_mark_node.
4798 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4800 * tree-ssanames.c (ssa_name_has_boolean_range): Use
4803 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4805 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
4806 vrp_visit_cond_stmt.
4807 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
4808 (cprop_into_stmt): Same.
4809 (dom_opt_dom_walker::optimize_stmt): Same.
4811 2021-09-27 Richard Biener <rguenther@suse.de>
4813 PR tree-optimization/97351
4814 PR tree-optimization/97352
4815 PR tree-optimization/82426
4816 * tree-vectorizer.h (dr_misalignment): Add vector type
4818 (aligned_access_p): Likewise.
4819 (known_alignment_for_access_p): Likewise.
4820 (vect_supportable_dr_alignment): Likewise.
4821 (vect_known_alignment_in_bytes): Likewise. Refactor.
4822 (DR_MISALIGNMENT): Remove.
4823 (vect_update_shared_vectype): Likewise.
4824 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
4825 a vector type with larger alignment requirement and apply
4826 the negative step adjustment here.
4827 (vect_calculate_target_alignment): Remove.
4828 (vect_compute_data_ref_alignment): Get explicit vector type
4829 argument, do not apply a negative step alignment adjustment
4831 (vect_slp_analyze_node_alignment): Re-analyze alignment
4832 when we re-visit the DR with a bigger desired alignment but
4833 keep more precise results from smaller alignments.
4834 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
4835 (vect_slp_analyze_node_operations_1): Do not update the
4836 shared vector type on stmts.
4837 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
4838 vector type of an SLP node to the representative stmt-info.
4839 (vect_transform_stmt): Likewise.
4841 2021-09-27 liuhongt <hongtao.liu@intel.com>
4844 2021-09-09 liuhongt <hongtao.liu@intel.com>
4847 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
4848 (reduc_plus_scal_v4sf): .. this, New define_expand.
4849 (reduc_plus_scal_v2df): .. and this, New define_expand.
4851 2021-09-26 liuhongt <hongtao.liu@intel.com>
4853 * doc/extend.texi (Half-Precision): Remove storage only
4854 description for _Float16 w/o avx512fp16.
4856 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
4858 * config/pru/constraints.md (Rrio): New constraint.
4859 * config/pru/predicates.md (regio_operand): New predicate.
4860 * config/pru/pru-pragma.c (pru_register_pragmas): Register
4861 the __regio_symbol address space.
4862 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
4863 * config/pru/pru.c (pru_symref2ioregno): New helper function.
4864 (pru_legitimate_address_p): Remove.
4865 (pru_addr_space_legitimate_address_p): Use the address space
4867 (pru_nongeneric_pointer_addrspace): New helper function.
4868 (pru_insert_attributes): New function to validate __regio_symbol
4870 (TARGET_INSERT_ATTRIBUTES): New macro.
4871 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
4872 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
4873 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
4874 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
4876 (*regio_nozext_writesi): New pattern to write to I/O registers.
4877 (*regio_zext_write_r30<EQS0:mode>): Ditto.
4878 * doc/extend.texi: Document the new PRU Named Address Space.
4880 2021-09-24 Patrick Palka <ppalka@redhat.com>
4884 * real.c (encode_ieee_double): Avoid unwanted sign extension.
4885 (encode_ieee_quad): Likewise.
4887 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
4889 PR rtl-optimization/102147
4890 * ira-build.c (ira_conflict_vector_profitable_p): Make
4891 profitability calculation independent of host compiler pointer and
4894 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
4896 * gimple-range-path.cc (path_range_query::path_range_query):
4897 Move debugging header...
4898 (path_range_query::precompute_ranges): ...here.
4899 (path_range_query::internal_range_of_expr): Do not call
4900 range_on_path_entry if NAME is defined in the current block.
4902 2021-09-24 Richard Biener <rguenther@suse.de>
4904 * cfghooks.c (verify_flow_info): Verify unallocated BB and
4905 edge flags are not set.
4907 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
4909 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
4911 (jt_path_registry::register_jump_thread): Call
4912 cancel_invalid_paths.
4913 * tree-ssa-threadupdate.h (class jt_path_registry): Add
4914 cancel_invalid_paths.
4916 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
4918 PR tree-optimization/102400
4919 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
4920 result_vdef to zero value.
4922 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
4924 PR tree-optimization/102451
4925 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
4928 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
4930 * config/i386/sse.md (cond_<insn><mode>): Extend to support
4932 (cond_mul<mode>): Likewise.
4933 (cond_div<mode>): Likewise.
4934 (cond_<code><mode>): Likewise.
4935 (cond_fma<mode>): Likewise.
4936 (cond_fms<mode>): Likewise.
4937 (cond_fnma<mode>): Likewise.
4938 (cond_fnms<mode>): Likewise.
4940 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
4942 PR tree-optimization/102463
4943 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
4944 there is no range-ops handler, don't look for a relation.
4946 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
4948 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
4949 non-executable_edge flag as parameter.
4950 * gimple-range-cache.h (ranger_cache): Adjust prototype.
4951 * gimple-range-gori.cc (gori_compute::gori_compute): Take
4952 non-executable_edge flag as parameter.
4953 (gori_compute::outgoing_edge_range_p): Check new flag.
4954 * gimple-range-gori.h (gori_compute): Adjust prototype.
4955 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
4956 (gimple_ranger::range_on_edge): Check new flag.
4957 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
4958 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
4959 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
4960 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
4961 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
4962 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
4963 (simplify_using_ranges::simplify_switch_using_ranges): Clear
4964 EDGE_EXECUTABLE like it originally did.
4965 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
4966 NON_EXECUTABLE flags.
4967 (simplify_using_ranges::simplify_using_ranges): Adjust.
4968 * vr-values.h (class simplify_using_ranges): Adjust.
4969 (simplify_using_ranges::set_range_query): Add non-executable flag param.
4971 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
4974 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
4975 zero-width bit fields and return indicator.
4976 (rs6000_discover_homogeneous_aggregate): Diagnose when the
4977 presence of a zero-width bit field changes parameter passing in
4980 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
4982 * gimple-range-fold.cc (fold_using_range::range_of_phi):
4983 Remove dominator check.
4985 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
4987 * gimple-range-path.cc (path_range_query::precompute_relations):
4988 Hoist edge calculations before using EDGE_SUCC.
4990 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
4992 * configure.ac: Fix --with-multilib-list description.
4993 * configure: Regenerate.
4995 2021-09-23 Richard Biener <rguenther@suse.de>
4997 PR tree-optimization/102448
4998 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
4999 Clear alignment info copied from DR_PTR_INFO.
5001 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5003 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
5005 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
5006 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
5007 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
5008 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
5009 (vcond<mode><mode>): New expander.
5010 (vcond<mode><sseintvecmodelower>): Likewise.
5011 (vcond<sseintvecmodelower><mode>): Likewise.
5012 (vcondu<mode><sseintvecmodelower>): Likewise.
5014 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5016 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
5018 (extendv4hf<mode>2): Likewise.
5019 (extendv2hfv2df2): Likewise.
5020 (trunc<mode><ssePHmodelower>2): Likewise.
5021 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
5022 (trunc<mode>v4hf2): ... this, and drop constraints.
5023 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
5024 (truncv2dfv2hf2): ... this, and likewise.
5026 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5028 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
5030 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
5032 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
5033 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
5034 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
5036 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5038 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
5039 (fixuns_trunchfhi2): Likewise.
5040 (*fixuns_trunchfsi2zext): New define_insn.
5041 * config/i386/sse.md (ssePHmodelower): New mode_attr.
5042 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
5043 New expander for same element vector fix_truncate.
5044 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
5045 Likewise for V4HF to V4SI/V4DI fix_truncate.
5046 (fix<fixunssuffix>_truncv2hfv2di2):
5047 Likeise for V2HF to V2DI fix_truncate.
5049 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5051 * config/i386/i386.md (<code>hf3): New expander.
5053 2021-09-23 liuhongt <hongtao.liu@intel.com>
5055 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
5056 (VFH_SF_AVX512VL): Extend to handle HFmode.
5057 (VF_SF_AVX512VL): Deleted.
5059 2021-09-23 liuhongt <hongtao.liu@intel.com>
5061 * config/i386/i386.md (rinthf2): New expander.
5062 (nearbyinthf2): New expander.
5064 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5066 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
5067 (class dom_jt_state): ...this and provide virtual overrides.
5068 (dom_jt_state::register_equiv): New.
5069 (class dom_jt_simplifier): Rename from
5070 dom_jump_threader_simplifier.
5071 (dom_jump_threader_simplifier::simplify): Rename...
5072 (dom_jt_simplifier::simplify): ...to this.
5073 (pass_dominator::execute): Use dom_jt_simplifier and
5075 * tree-ssa-threadedge.c (jump_threader::jump_threader):
5077 (jt_state::register_equivs_stmt): Abstract out...
5078 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
5080 (jump_threader::thread_around_empty_blocks): Update state.
5081 (jump_threader::thread_through_normal_block): Same.
5082 (jt_state::jt_state): Remove.
5083 (jt_state::push): Remove pass specific bits. Keep block vector
5085 (jt_state::append_path): New.
5086 (jt_state::pop): Remove pass specific bits.
5087 (jt_state::register_equiv): Same.
5088 (jt_state::record_ranges_from_stmt): Same.
5089 (jt_state::register_equivs_on_edge): Same. Rename...
5090 (jt_state::register_equivs_edge): ...to this.
5091 (jt_state::dump): New.
5092 (jt_state::debug): New.
5093 (jump_threader_simplifier::simplify): Remove.
5094 (jt_state::get_path): New.
5095 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
5096 class. Expose common functionality as virtual methods.
5097 (class jump_threader_simplifier): Same. Rename...
5098 (class jt_simplifier): ...to this.
5099 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
5100 (class vrp_jt_simplifier): ...to this. Provide pass specific
5102 (class vrp_jt_state): New.
5103 (vrp_jump_threader_simplifier::simplify): Rename...
5104 (vrp_jt_simplifier::simplify): ...to this. Inline code from
5105 what used to be the base class.
5106 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
5109 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
5112 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
5115 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
5116 Richard Biener <rguenther@suse.de>
5118 * match.pd (negation simplifications): Implement some negation
5119 folding transformations from fold-const.c's fold_negate_expr.
5120 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
5121 argument, to control whether the op should be simplified prior
5122 to looking up/assigning a value number.
5123 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
5124 (vn_nary_simplify): Likewise.
5125 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
5126 now call vn_nary_build_or_lookup_1 disabling simplification.
5128 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
5130 PR tree-optimization/102087
5131 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5132 Update bound/cmp/control for niter.
5134 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
5136 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
5137 Move check for non-empty BB here.
5138 (fur_source::register_outgoing_edges): ...from here.
5140 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
5142 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
5143 Remove call to improve_range_with_equivs.
5144 (path_range_query::improve_range_with_equivs): Remove
5145 * gimple-range-path.h: Remove improve_range_with_equivs.
5147 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5149 * config/i386/avx512fp16intrin.h:
5150 (_mm512_mask_blend_ph): New intrinsic.
5151 (_mm512_permutex2var_ph): Ditto.
5152 (_mm512_permutexvar_ph): Ditto.
5153 * config/i386/avx512fp16vlintrin.h:
5154 (_mm256_mask_blend_ph): New intrinsic.
5155 (_mm256_permutex2var_ph): Ditto.
5156 (_mm256_permutexvar_ph): Ditto.
5157 (_mm_mask_blend_ph): Ditto.
5158 (_mm_permutex2var_ph): Ditto.
5159 (_mm_permutexvar_ph): Ditto.
5161 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5163 * config/i386/avx512fp16intrin.h: Add new intrinsics.
5164 (_mm512_conj_pch): New intrinsic.
5165 (_mm512_mask_conj_pch): Ditto.
5166 (_mm512_maskz_conj_pch): Ditto.
5167 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
5168 (_mm256_conj_pch): New intrinsic.
5169 (_mm256_mask_conj_pch): Ditto.
5170 (_mm256_maskz_conj_pch): Ditto.
5171 (_mm_conj_pch): Ditto.
5172 (_mm_mask_conj_pch): Ditto.
5173 (_mm_maskz_conj_pch): Ditto.
5175 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5177 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
5178 (_mm512_reduce_add_ph): New intrinsic.
5179 (_mm512_reduce_mul_ph): Ditto.
5180 (_mm512_reduce_min_ph): Ditto.
5181 (_mm512_reduce_max_ph): Ditto.
5182 * config/i386/avx512fp16vlintrin.h
5183 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
5184 (_mm256_reduce_add_ph): New intrinsic.
5185 (_mm256_reduce_mul_ph): Ditto.
5186 (_mm256_reduce_min_ph): Ditto.
5187 (_mm256_reduce_max_ph): Ditto.
5188 (_mm_reduce_add_ph): Ditto.
5189 (_mm_reduce_mul_ph): Ditto.
5190 (_mm_reduce_min_ph): Ditto.
5191 (_mm_reduce_max_ph): Ditto.
5193 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5195 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
5196 __m128h_u): New typedef.
5197 (_mm512_load_ph): New intrinsic.
5198 (_mm256_load_ph): Ditto.
5199 (_mm_load_ph): Ditto.
5200 (_mm512_loadu_ph): Ditto.
5201 (_mm256_loadu_ph): Ditto.
5202 (_mm_loadu_ph): Ditto.
5203 (_mm512_store_ph): Ditto.
5204 (_mm256_store_ph): Ditto.
5205 (_mm_store_ph): Ditto.
5206 (_mm512_storeu_ph): Ditto.
5207 (_mm256_storeu_ph): Ditto.
5208 (_mm_storeu_ph): Ditto.
5209 (_mm512_abs_ph): Ditto.
5210 * config/i386/avx512fp16vlintrin.h
5211 (_mm_abs_ph): Ditto.
5212 (_mm256_abs_ph): Ditto.
5214 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
5216 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
5218 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
5221 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
5222 is actually a full copy of the source operand into the target.
5223 Don't emit a strict low part move if source and target mode match.
5225 2021-09-22 Jakub Jelinek <jakub@redhat.com>
5227 PR middle-end/102415
5228 * omp-expand.c (expand_omp_single): If region->exit is NULL,
5229 assert region->entry is GIMPLE_OMP_SCOPE region and return.
5231 2021-09-22 Jakub Jelinek <jakub@redhat.com>
5233 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
5234 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
5235 arguments from 2 to 3.
5236 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
5237 allocate clause allocator and print align if present.
5238 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
5239 for omp_default_mem_alloc if align modifier is present. If align
5240 modifier is present, use TREE_LIST to encode both allocator and
5242 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
5243 Handle align modifier on allocator clause if present.
5245 2021-09-22 liuhongt <hongtao.liu@intel.com>
5247 * config/i386/i386.md (define_attr "isa"): Add
5249 (define_attr "enabled"): Correspond fma_or_avx512vl to
5250 TARGET_FMA || TARGET_AVX512VL.
5251 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
5256 2021-09-22 liuhongt <hongtao.liu@intel.com>
5258 * config/i386/i386.md (cstorehf3): New define_expand.
5260 2021-09-22 liuhongt <hongtao.liu@intel.com>
5262 * config/i386/i386.md (<rounding_insn>hf2): New expander.
5263 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
5264 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
5265 Extend from VF_128 to VFH_128.
5267 2021-09-22 liuhongt <hongtao.liu@intel.com>
5269 * config/i386/i386-features.c (i386-features.c): Handle
5271 * config/i386/i386.md (sqrthf2): New expander.
5272 (*sqrthf2): New define_insn.
5273 * config/i386/sse.md
5274 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
5277 2021-09-22 liuhongt <hongtao.liu@intel.com>
5279 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
5281 (_mm_mask3_fcmadd_sch): Likewise.
5282 (_mm_maskz_fcmadd_sch): Likewise.
5283 (_mm_fcmadd_sch): Likewise.
5284 (_mm_mask_fmadd_sch): Likewise.
5285 (_mm_mask3_fmadd_sch): Likewise.
5286 (_mm_maskz_fmadd_sch): Likewise.
5287 (_mm_fmadd_sch): Likewise.
5288 (_mm_mask_fcmadd_round_sch): Likewise.
5289 (_mm_mask3_fcmadd_round_sch): Likewise.
5290 (_mm_maskz_fcmadd_round_sch): Likewise.
5291 (_mm_fcmadd_round_sch): Likewise.
5292 (_mm_mask_fmadd_round_sch): Likewise.
5293 (_mm_mask3_fmadd_round_sch): Likewise.
5294 (_mm_maskz_fmadd_round_sch): Likewise.
5295 (_mm_fmadd_round_sch): Likewise.
5296 (_mm_fcmul_sch): Likewise.
5297 (_mm_mask_fcmul_sch): Likewise.
5298 (_mm_maskz_fcmul_sch): Likewise.
5299 (_mm_fmul_sch): Likewise.
5300 (_mm_mask_fmul_sch): Likewise.
5301 (_mm_maskz_fmul_sch): Likewise.
5302 (_mm_fcmul_round_sch): Likewise.
5303 (_mm_mask_fcmul_round_sch): Likewise.
5304 (_mm_maskz_fcmul_round_sch): Likewise.
5305 (_mm_fmul_round_sch): Likewise.
5306 (_mm_mask_fmul_round_sch): Likewise.
5307 (_mm_maskz_fmul_round_sch): Likewise.
5308 * config/i386/i386-builtin.def: Add corresponding new builtins.
5309 * config/i386/sse.md
5310 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
5311 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
5312 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
5314 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
5315 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
5317 * config/i386/subst.md (mask_scalarcz_name): New.
5318 (mask_scalarc_name): Ditto.
5319 (mask_scalarc_operand3): Ditto.
5320 (mask_scalarcz_operand4): Ditto.
5321 (round_scalarcz_name): Ditto.
5322 (round_scalarc_mask_operand3): Ditto.
5323 (round_scalarcz_mask_operand4): Ditto.
5324 (round_scalarc_mask_op3): Ditto.
5325 (round_scalarcz_mask_op4): Ditto.
5326 (round_scalarcz_constraint): Ditto.
5327 (round_scalarcz_nimm_predicate): Ditto.
5328 (mask_scalarcz): Ditto.
5329 (mask_scalarc): Ditto.
5330 (round_scalarcz): Ditto.
5332 2021-09-22 liuhongt <hongtao.liu@intel.com>
5334 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
5336 (_mm512_mask_fcmadd_pch): Likewise.
5337 (_mm512_mask3_fcmadd_pch): Likewise.
5338 (_mm512_maskz_fcmadd_pch): Likewise.
5339 (_mm512_fmadd_pch): Likewise.
5340 (_mm512_mask_fmadd_pch): Likewise.
5341 (_mm512_mask3_fmadd_pch): Likewise.
5342 (_mm512_maskz_fmadd_pch): Likewise.
5343 (_mm512_fcmadd_round_pch): Likewise.
5344 (_mm512_mask_fcmadd_round_pch): Likewise.
5345 (_mm512_mask3_fcmadd_round_pch): Likewise.
5346 (_mm512_maskz_fcmadd_round_pch): Likewise.
5347 (_mm512_fmadd_round_pch): Likewise.
5348 (_mm512_mask_fmadd_round_pch): Likewise.
5349 (_mm512_mask3_fmadd_round_pch): Likewise.
5350 (_mm512_maskz_fmadd_round_pch): Likewise.
5351 (_mm512_fcmul_pch): Likewise.
5352 (_mm512_mask_fcmul_pch): Likewise.
5353 (_mm512_maskz_fcmul_pch): Likewise.
5354 (_mm512_fmul_pch): Likewise.
5355 (_mm512_mask_fmul_pch): Likewise.
5356 (_mm512_maskz_fmul_pch): Likewise.
5357 (_mm512_fcmul_round_pch): Likewise.
5358 (_mm512_mask_fcmul_round_pch): Likewise.
5359 (_mm512_maskz_fcmul_round_pch): Likewise.
5360 (_mm512_fmul_round_pch): Likewise.
5361 (_mm512_mask_fmul_round_pch): Likewise.
5362 (_mm512_maskz_fmul_round_pch): Likewise.
5363 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
5365 (_mm_mask_fmadd_pch): Likewise.
5366 (_mm_mask3_fmadd_pch): Likewise.
5367 (_mm_maskz_fmadd_pch): Likewise.
5368 (_mm256_fmadd_pch): Likewise.
5369 (_mm256_mask_fmadd_pch): Likewise.
5370 (_mm256_mask3_fmadd_pch): Likewise.
5371 (_mm256_maskz_fmadd_pch): Likewise.
5372 (_mm_fcmadd_pch): Likewise.
5373 (_mm_mask_fcmadd_pch): Likewise.
5374 (_mm_mask3_fcmadd_pch): Likewise.
5375 (_mm_maskz_fcmadd_pch): Likewise.
5376 (_mm256_fcmadd_pch): Likewise.
5377 (_mm256_mask_fcmadd_pch): Likewise.
5378 (_mm256_mask3_fcmadd_pch): Likewise.
5379 (_mm256_maskz_fcmadd_pch): Likewise.
5380 (_mm_fmul_pch): Likewise.
5381 (_mm_mask_fmul_pch): Likewise.
5382 (_mm_maskz_fmul_pch): Likewise.
5383 (_mm256_fmul_pch): Likewise.
5384 (_mm256_mask_fmul_pch): Likewise.
5385 (_mm256_maskz_fmul_pch): Likewise.
5386 (_mm_fcmul_pch): Likewise.
5387 (_mm_mask_fcmul_pch): Likewise.
5388 (_mm_maskz_fcmul_pch): Likewise.
5389 (_mm256_fcmul_pch): Likewise.
5390 (_mm256_mask_fcmul_pch): Likewise.
5391 (_mm256_maskz_fcmul_pch): Likewise.
5392 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
5393 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
5394 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
5395 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
5396 * config/i386/i386-builtin.def: Add new builtins.
5397 * config/i386/i386-expand.c: Handle new builtin types.
5398 * config/i386/subst.md (SUBST_CV): New.
5399 (maskc_name): Ditto.
5400 (maskc_operand3): Ditto.
5402 (sdc_maskz_name): Ditto.
5403 (sdc_mask_op4): Ditto.
5404 (sdc_mask_op5): Ditto.
5405 (sdc_mask_mode512bit_condition): Ditto.
5407 (round_maskc_operand3): Ditto.
5408 (round_sdc_mask_operand4): Ditto.
5409 (round_maskc_op3): Ditto.
5410 (round_sdc_mask_op4): Ditto.
5411 (round_saeonly_sdc_mask_operand5): Ditto.
5412 * config/i386/sse.md (unspec): Add complex fma unspecs.
5413 (avx512fmaskcmode): New.
5414 (UNSPEC_COMPLEX_F_C_MA): Ditto.
5415 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
5416 (complexopname): Ditto.
5417 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
5418 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
5419 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
5421 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
5422 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
5424 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
5426 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
5427 rs6000-density-size-threshold, rs6000-density-penalty,
5428 rs6000-density-load-pct-threshold,
5429 rs6000-density-load-num-threshold): New parameter.
5430 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
5431 corresponding parameters.
5433 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5435 * gimple-range-path.cc (path_range_query::defined_outside_path):
5437 (path_range_query::range_on_path_entry): New.
5438 (path_range_query::internal_range_of_expr): Resolve unknowns
5440 (path_range_query::improve_range_with_equivs): New.
5441 (path_range_query::ssa_range_in_phi): Resolve unknowns with
5443 * gimple-range-path.h (class path_range_query): Add
5444 defined_outside_path, range_on_path_entry, and
5445 improve_range_with_equivs.
5447 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5449 * gimple-range-path.cc (path_range_query::add_to_imports): New.
5450 (path_range_query::add_copies_to_imports): New.
5451 (path_range_query::precompute_ranges): Call
5452 add_copies_to_imports.
5453 * gimple-range-path.h (class path_range_query): Add prototypes
5454 for add_copies_to_imports and add_to_imports.
5456 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5458 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5459 Remove useless code.
5461 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5463 * gimple-range-fold.h (class fur_source): Make oracle protected.
5464 * gimple-range-path.cc (path_range_query::path_range_query): Add
5465 resolve argument. Initialize oracle.
5466 (path_range_query::~path_range_query): Delete oracle.
5467 (path_range_query::range_of_stmt): Adapt to use relations.
5468 (path_range_query::precompute_ranges): Pre-compute relations.
5469 (class jt_fur_source): New
5470 (jt_fur_source::jt_fur_source): New.
5471 (jt_fur_source::register_relation): New.
5472 (jt_fur_source::query_relation): New.
5473 (path_range_query::precompute_relations): New.
5474 (path_range_query::precompute_phi_relations): New.
5475 * gimple-range-path.h (path_range_query): Add resolve argument.
5476 Add oracle, precompute_relations, precompute_phi_relations.
5477 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
5478 resolve argument to solver.
5480 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5482 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
5483 Rename postfold_gcond_edges to register_outgoing_edges and
5485 (fold_using_range::postfold_gcond_edges): Rename...
5486 (fur_source::register_outgoing_edges): ...to this.
5487 * gimple-range-fold.h (postfold_gcond_edges): Rename to
5488 register_outgoing_edges and move to fur_source.
5490 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5492 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
5493 dom_info_available_p.
5495 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5497 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
5498 and quick_grow_cleared instead of safe_grow_cleared.
5500 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
5503 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
5506 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
5508 * configure.ac: Detect when the assembler supports new-style
5509 architecture extensions.
5510 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
5511 the full CPU string if the assembler can grok it.
5512 (arm_rewrite_march): Likewise but for the architecture.
5513 * config.in: Regenerate.
5514 * configure: Regenerate.
5516 2021-09-21 Richard Biener <rguenther@suse.de>
5518 PR tree-optimization/102421
5519 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
5520 adjust alignment info.
5522 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
5524 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
5525 bitfield stream out.
5527 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
5529 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
5530 undefined edges, apply an equivalence if appropriate.
5531 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
5532 UNDEFINED if EDGE_EXECUTABLE is not set.
5533 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
5534 as EXECUTABLE upon startup.
5535 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
5536 EDGE_EXECUTABLE set.
5537 * vr-values.c (set_and_propagate_unexecutable): New.
5538 (simplify_using_ranges::fold_cond): Call set_and_propagate.
5539 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
5540 * vr-values.h: Add prototype.
5542 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
5544 * value-relation.cc (equiv_oracle::register_initial_def): New.
5545 (equiv_oracle::register_relation): Call register_initial_def.
5546 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
5547 (relation_oracle::register_stmt): Check def block of PHI arguments.
5548 * value-relation.h (equiv_oracle): Add new prototypes.
5550 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
5552 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
5553 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
5554 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
5555 __ROUNDING_MATH__ according to their corresponding flags.
5556 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
5557 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
5558 and __ROUNDING_MATH__.
5560 2021-09-20 Richard Biener <rguenther@suse.de>
5562 * tree-vect-stmts.c (vectorizable_load): Use the vectype
5565 2021-09-20 Richard Biener <rguenther@suse.de>
5567 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
5568 Do not compute alignment of the vectorized access here.
5570 2021-09-20 Richard Biener <rguenther@suse.de>
5572 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
5573 Store -1 for runtime alias peeling iterations.
5575 2021-09-20 Richard Biener <rguenther@suse.de>
5577 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
5579 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
5581 * input.c (string_concat_db::record_string_concatenation)
5582 (string_concat_db::get_string_concatenation): Skip for
5583 'RESERVED_LOCATION_P'.
5585 2021-09-20 Richard Biener <rguenther@suse.de>
5587 PR tree-optimization/65206
5588 * tree-data-ref.h (struct data_reference): Add alt_indices,
5590 * tree-data-ref.c (free_data_ref): Release alt_indices.
5591 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
5592 (create_data_ref): Adjust.
5593 (initialize_data_dependence_relation): Split into head
5594 and tail. When the base objects fail to match up try
5595 again with pointer-based analysis of indices.
5596 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
5597 not compare the lazily computed alternate set of indices.
5599 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
5601 * gcc.c: Test for execute OK when we find the
5602 programs for assembler linker and dsymutil and those
5603 were specified at configure-time.
5605 2021-09-19 Martin Sebor <msebor@redhat.com>
5607 PR middle-end/102403
5608 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
5609 Correct a function pre/postcondition.
5611 2021-09-19 Martin Sebor <msebor@redhat.com>
5613 PR middle-end/102243
5614 * tree-ssa-strlen.c (get_range): Handle null cfun.
5616 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
5618 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
5619 unwinder shim as a convenience library.
5621 2021-09-19 Andrew Pinski <apinski@marvell.com>
5623 * doc/install.texi: Add note about
5624 binutils 2.35 is required for LTO usage.
5626 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
5628 * tree-ssa-threadbackward.c
5629 (back_threader_registry::register_path): Use push_edge.
5630 * tree-ssa-threadedge.c
5631 (jump_threader::thread_around_empty_blocks): Same.
5632 (jump_threader::thread_through_normal_block): Same.
5633 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
5635 * tree-ssa-threadupdate.c
5636 (jt_path_registry::allocate_thread_edge): Remove.
5637 (jt_path_registry::push_edge): New.
5638 (dump_jump_thread_path): Make static.
5639 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
5642 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
5644 * gimple-range-path.cc (path_range_query::path_range_query): Add
5646 (path_range_query::dump): Remove extern declaration of dump_ranger.
5647 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
5648 * gimple-range-trace.h (dump_ranger): Add prototype.
5650 2021-09-19 John Ericson <git@JohnEricson.me>
5652 * gcc.c (find_a_program): New function, factored out of...
5653 (find_a_file): Here.
5654 (execute): Use find_a_program when looking for programs rather
5657 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
5659 * config/avr/avr-mcus.def: Add atmega324pb.
5660 * doc/avr-mmcu.texi: Corresponding changes.
5662 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
5665 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
5666 the same transformations as to @1. For comparisons against NaN,
5667 don't check HONOR_SNANS but confirm that neither operand is a
5670 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
5672 * attribs.c (make_unique_name): Delete.
5673 * attribs.h (make_unique_name): Delete.
5675 2021-09-19 Andrew Pinski <apinski@marvell.com>
5677 * lra-constraints.c (check_and_process_move): Assert
5678 that dclass and sclass are greater than or equal to NO_REGS.
5680 2021-09-18 Jakub Jelinek <jakub@redhat.com>
5682 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
5683 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
5684 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
5686 2021-09-18 liuhongt <hongtao.liu@intel.com>
5688 * config/i386/i386-features.c (remove_partial_avx_dependency):
5689 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
5690 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
5692 2021-09-18 Jakub Jelinek <jakub@redhat.com>
5694 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
5695 if file/namespace scope variable doesn't have predetermined sharing,
5696 treat it as if there was default(none).
5698 2021-09-18 liuhongt <hongtao.liu@intel.com>
5700 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
5702 (_mm_mask_fmadd_sh): Likewise.
5703 (_mm_mask3_fmadd_sh): Likewise.
5704 (_mm_maskz_fmadd_sh): Likewise.
5705 (_mm_fmadd_round_sh): Likewise.
5706 (_mm_mask_fmadd_round_sh): Likewise.
5707 (_mm_mask3_fmadd_round_sh): Likewise.
5708 (_mm_maskz_fmadd_round_sh): Likewise.
5709 (_mm_fnmadd_sh): Likewise.
5710 (_mm_mask_fnmadd_sh): Likewise.
5711 (_mm_mask3_fnmadd_sh): Likewise.
5712 (_mm_maskz_fnmadd_sh): Likewise.
5713 (_mm_fnmadd_round_sh): Likewise.
5714 (_mm_mask_fnmadd_round_sh): Likewise.
5715 (_mm_mask3_fnmadd_round_sh): Likewise.
5716 (_mm_maskz_fnmadd_round_sh): Likewise.
5717 (_mm_fmsub_sh): Likewise.
5718 (_mm_mask_fmsub_sh): Likewise.
5719 (_mm_mask3_fmsub_sh): Likewise.
5720 (_mm_maskz_fmsub_sh): Likewise.
5721 (_mm_fmsub_round_sh): Likewise.
5722 (_mm_mask_fmsub_round_sh): Likewise.
5723 (_mm_mask3_fmsub_round_sh): Likewise.
5724 (_mm_maskz_fmsub_round_sh): Likewise.
5725 (_mm_fnmsub_sh): Likewise.
5726 (_mm_mask_fnmsub_sh): Likewise.
5727 (_mm_mask3_fnmsub_sh): Likewise.
5728 (_mm_maskz_fnmsub_sh): Likewise.
5729 (_mm_fnmsub_round_sh): Likewise.
5730 (_mm_mask_fnmsub_round_sh): Likewise.
5731 (_mm_mask3_fnmsub_round_sh): Likewise.
5732 (_mm_maskz_fnmsub_round_sh): Likewise.
5733 * config/i386/i386-builtin-types.def
5734 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
5735 * config/i386/i386-builtin.def: Add new builtins.
5736 * config/i386/i386-expand.c: Handle new builtin type.
5737 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
5738 Ajdust to support FP16.
5739 (fmai_vmfmsub_<mode><round_name>): Ditto.
5740 (fmai_vmfnmadd_<mode><round_name>): Ditto.
5741 (fmai_vmfnmsub_<mode><round_name>): Ditto.
5742 (*fmai_fmadd_<mode>): Ditto.
5743 (*fmai_fmsub_<mode>): Ditto.
5744 (*fmai_fnmadd_<mode><round_name>): Ditto.
5745 (*fmai_fnmsub_<mode><round_name>): Ditto.
5746 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
5747 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
5748 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
5749 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
5750 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
5751 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
5752 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
5753 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
5754 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
5755 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
5756 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
5757 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
5758 adjust to support FP16.
5759 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
5760 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
5761 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
5764 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
5766 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
5768 (maskload<mode><avx512fmaskmodelower>): Ditto.
5769 (maskstore<mode><avx512fmaskmodelower>): Ditto.
5771 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
5773 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
5775 (ix86_expand_copysign): Ditto.
5776 (ix86_expand_xorsign): Ditto.
5777 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
5779 (ix86_build_signbit_mask): Ditto.
5780 (ix86_can_change_mode_class): Ditto.
5781 * config/i386/i386.md
5782 (SSEMODEF): Add HFmode.
5783 (ssevecmodef): Ditto.
5784 (<code>hf2): New define_expand.
5785 (*<code>hf2_1): New define_insn_and_split.
5786 (copysign<mode>): Extend to support HFmode under AVX512FP16.
5787 (xorsign<mode>): Ditto.
5788 * config/i386/sse.md (VFB): New mode iterator.
5789 (VFB_128_256): Ditto.
5791 (sseintvecmode2): Support HF vector mode.
5792 (<code><mode>2): Use new mode iterator.
5793 (*<code><mode>2): Ditto.
5794 (copysign<mode>3): Ditto.
5795 (xorsign<mode>3): Ditto.
5796 (<code><mode>3<mask_name>): Ditto.
5797 (<code><mode>3<mask_name>): Ditto.
5798 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
5799 (<sse>_andnot<mode>3<mask_name>): Ditto.
5800 (*<code><mode>3<mask_name>): Ditto.
5801 (*<code><mode>3<mask_name>): Ditto.
5803 2021-09-18 liuhongt <hongtao.liu@intel.com>
5805 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
5807 (_mm512_mask3_fmadd_ph): Likewise.
5808 (_mm512_maskz_fmadd_ph): Likewise.
5809 (_mm512_fmadd_round_ph): Likewise.
5810 (_mm512_mask_fmadd_round_ph): Likewise.
5811 (_mm512_mask3_fmadd_round_ph): Likewise.
5812 (_mm512_maskz_fmadd_round_ph): Likewise.
5813 (_mm512_fnmadd_ph): Likewise.
5814 (_mm512_mask_fnmadd_ph): Likewise.
5815 (_mm512_mask3_fnmadd_ph): Likewise.
5816 (_mm512_maskz_fnmadd_ph): Likewise.
5817 (_mm512_fnmadd_round_ph): Likewise.
5818 (_mm512_mask_fnmadd_round_ph): Likewise.
5819 (_mm512_mask3_fnmadd_round_ph): Likewise.
5820 (_mm512_maskz_fnmadd_round_ph): Likewise.
5821 (_mm512_fmsub_ph): Likewise.
5822 (_mm512_mask_fmsub_ph): Likewise.
5823 (_mm512_mask3_fmsub_ph): Likewise.
5824 (_mm512_maskz_fmsub_ph): Likewise.
5825 (_mm512_fmsub_round_ph): Likewise.
5826 (_mm512_mask_fmsub_round_ph): Likewise.
5827 (_mm512_mask3_fmsub_round_ph): Likewise.
5828 (_mm512_maskz_fmsub_round_ph): Likewise.
5829 (_mm512_fnmsub_ph): Likewise.
5830 (_mm512_mask_fnmsub_ph): Likewise.
5831 (_mm512_mask3_fnmsub_ph): Likewise.
5832 (_mm512_maskz_fnmsub_ph): Likewise.
5833 (_mm512_fnmsub_round_ph): Likewise.
5834 (_mm512_mask_fnmsub_round_ph): Likewise.
5835 (_mm512_mask3_fnmsub_round_ph): Likewise.
5836 (_mm512_maskz_fnmsub_round_ph): Likewise.
5837 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
5839 (_mm256_mask_fmadd_ph): Likewise.
5840 (_mm256_mask3_fmadd_ph): Likewise.
5841 (_mm256_maskz_fmadd_ph): Likewise.
5842 (_mm_fmadd_ph): Likewise.
5843 (_mm_mask_fmadd_ph): Likewise.
5844 (_mm_mask3_fmadd_ph): Likewise.
5845 (_mm_maskz_fmadd_ph): Likewise.
5846 (_mm256_fnmadd_ph): Likewise.
5847 (_mm256_mask_fnmadd_ph): Likewise.
5848 (_mm256_mask3_fnmadd_ph): Likewise.
5849 (_mm256_maskz_fnmadd_ph): Likewise.
5850 (_mm_fnmadd_ph): Likewise.
5851 (_mm_mask_fnmadd_ph): Likewise.
5852 (_mm_mask3_fnmadd_ph): Likewise.
5853 (_mm_maskz_fnmadd_ph): Likewise.
5854 (_mm256_fmsub_ph): Likewise.
5855 (_mm256_mask_fmsub_ph): Likewise.
5856 (_mm256_mask3_fmsub_ph): Likewise.
5857 (_mm256_maskz_fmsub_ph): Likewise.
5858 (_mm_fmsub_ph): Likewise.
5859 (_mm_mask_fmsub_ph): Likewise.
5860 (_mm_mask3_fmsub_ph): Likewise.
5861 (_mm_maskz_fmsub_ph): Likewise.
5862 (_mm256_fnmsub_ph): Likewise.
5863 (_mm256_mask_fnmsub_ph): Likewise.
5864 (_mm256_mask3_fnmsub_ph): Likewise.
5865 (_mm256_maskz_fnmsub_ph): Likewise.
5866 (_mm_fnmsub_ph): Likewise.
5867 (_mm_mask_fnmsub_ph): Likewise.
5868 (_mm_mask3_fnmsub_ph): Likewise.
5869 (_mm_maskz_fnmsub_ph): Likewise.
5870 * config/i386/i386-builtin.def: Add corresponding new builtins.
5871 * config/i386/sse.md
5872 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
5873 support HF vector modes.
5874 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
5876 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
5877 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
5878 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
5879 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
5880 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
5881 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
5882 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
5884 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
5885 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
5886 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
5887 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
5888 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
5889 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
5891 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
5892 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
5893 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
5894 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
5895 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
5896 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
5897 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
5899 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
5900 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
5901 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
5902 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
5903 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
5905 2021-09-18 liuhongt <hongtao.liu@intel.com>
5907 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
5909 (_mm512_mask_fmaddsub_ph): Likewise.
5910 (_mm512_mask3_fmaddsub_ph): Likewise.
5911 (_mm512_maskz_fmaddsub_ph): Likewise.
5912 (_mm512_fmaddsub_round_ph): Likewise.
5913 (_mm512_mask_fmaddsub_round_ph): Likewise.
5914 (_mm512_mask3_fmaddsub_round_ph): Likewise.
5915 (_mm512_maskz_fmaddsub_round_ph): Likewise.
5916 (_mm512_mask_fmsubadd_ph): Likewise.
5917 (_mm512_mask3_fmsubadd_ph): Likewise.
5918 (_mm512_maskz_fmsubadd_ph): Likewise.
5919 (_mm512_fmsubadd_round_ph): Likewise.
5920 (_mm512_mask_fmsubadd_round_ph): Likewise.
5921 (_mm512_mask3_fmsubadd_round_ph): Likewise.
5922 (_mm512_maskz_fmsubadd_round_ph): Likewise.
5923 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
5925 (_mm256_mask_fmaddsub_ph): Likewise.
5926 (_mm256_mask3_fmaddsub_ph): Likewise.
5927 (_mm256_maskz_fmaddsub_ph): Likewise.
5928 (_mm_fmaddsub_ph): Likewise.
5929 (_mm_mask_fmaddsub_ph): Likewise.
5930 (_mm_mask3_fmaddsub_ph): Likewise.
5931 (_mm_maskz_fmaddsub_ph): Likewise.
5932 (_mm256_fmsubadd_ph): Likewise.
5933 (_mm256_mask_fmsubadd_ph): Likewise.
5934 (_mm256_mask3_fmsubadd_ph): Likewise.
5935 (_mm256_maskz_fmsubadd_ph): Likewise.
5936 (_mm_fmsubadd_ph): Likewise.
5937 (_mm_mask_fmsubadd_ph): Likewise.
5938 (_mm_mask3_fmsubadd_ph): Likewise.
5939 (_mm_maskz_fmsubadd_ph): Likewise.
5940 * config/i386/i386-builtin.def: Add corresponding new builtins.
5941 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
5942 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
5943 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
5945 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
5947 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
5948 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
5949 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
5951 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
5952 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
5954 2021-09-18 liuhongt <hongtao.liu@intel.com>
5957 * config/i386/i386.c (ix86_print_operand): Handle
5958 V8HF/V16HF/V32HFmode.
5959 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
5960 * config/i386/sse.md (avx512bcst): Remove.
5962 2021-09-17 Martin Sebor <msebor@redhat.com>
5964 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
5965 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
5966 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
5967 (check_defs): Add comment.
5968 (can_skip_redundant_opnd): Update comment.
5969 (compute_uninit_opnds_pos): Adjust to namespace change.
5970 (find_pdom): Move to gimple-predicate-analysis.cc.
5972 (struct uninit_undef_val_t): New.
5973 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
5974 (find_control_equiv_block): Same.
5975 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
5976 (MAX_SWITCH_CASES): Same.
5977 (compute_control_dep_chain): Same.
5978 (find_uninit_use): Use predicate analyzer.
5979 (struct pred_info): Move to gimple-predicate-analysis.
5980 (convert_control_dep_chain_into_preds): Same.
5981 (find_predicates): Same.
5982 (collect_phi_def_edges): Same.
5983 (warn_uninitialized_phi): Use predicate analyzer.
5984 (find_def_preds): Move to gimple-predicate-analysis.
5985 (dump_pred_info): Same.
5986 (dump_pred_chain): Same.
5987 (dump_predicates): Same.
5988 (destroy_predicate_vecs): Remove.
5989 (execute_late_warn_uninitialized): New.
5990 (get_cmp_code): Move to gimple-predicate-analysis.
5991 (is_value_included_in): Same.
5992 (value_sat_pred_p): Same.
5993 (find_matching_predicate_in_rest_chains): Same.
5994 (is_use_properly_guarded): Same.
5995 (prune_uninit_phi_opnds): Same.
5996 (find_var_cmp_const): Same.
5997 (use_pred_not_overlap_with_undef_path_pred): Same.
5998 (pred_equal_p): Same.
5999 (is_neq_relop_p): Same.
6000 (is_neq_zero_form_p): Same.
6001 (pred_expr_equal_p): Same.
6002 (is_pred_expr_subset_of): Same.
6003 (is_pred_chain_subset_of): Same.
6004 (is_included_in): Same.
6005 (is_superset_of): Same.
6007 (simplify_pred): Same.
6008 (simplify_preds_2): Same.
6009 (simplify_preds_3): Same.
6010 (simplify_preds_4): Same.
6011 (simplify_preds): Same.
6013 (push_to_worklist): Same.
6014 (get_pred_info_from_cmp): Same.
6015 (is_degenerated_phi): Same.
6016 (normalize_one_pred_1): Same.
6017 (normalize_one_pred): Same.
6018 (normalize_one_pred_chain): Same.
6019 (normalize_preds): Same.
6020 (can_one_predicate_be_invalidated_p): Same.
6021 (can_chain_union_be_invalidated_p): Same.
6022 (uninit_uses_cannot_happen): Same.
6023 (pass_late_warn_uninitialized::execute): Define.
6024 * gimple-predicate-analysis.cc: New file.
6025 * gimple-predicate-analysis.h: New file.
6027 2021-09-17 Julian Brown <julian@codesourcery.com>
6029 * config/gcn/gcn.c (gimple.h): Include.
6030 (gcn_fork_join): Emit barrier for worker-level joins.
6031 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
6032 writes_gang_private bitmap parameter. Set bit for blocks
6033 containing gang-private variable writes.
6034 (worker_single_simple): Don't emit barrier after predicated block.
6035 (worker_single_copy): Don't emit barrier if we're not broadcasting
6036 anything and the block contains no gang-private writes.
6037 (neuter_worker_single): Don't predicate blocks that only contain
6038 NOPs or internal marker functions. Pass has_gang_private_write
6039 argument to worker_single_copy.
6040 (oacc_do_neutering): Add writes_gang_private bitmap handling.
6042 2021-09-17 Julian Brown <julian@codesourcery.com>
6044 * config/gcn/gcn-protos.h
6045 (gcn_goacc_create_worker_broadcast_record): Update prototype.
6046 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
6047 preallocated block of LDS memory. Do not cache/share decls for
6048 reduction temporaries between invocations.
6049 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
6050 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
6051 and return temporary LDS space at that offset. Return pointer in
6053 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
6055 (ACC_LDS_SIZE): Define as acc_lds_size.
6056 (gcn_init_machine_status): Don't initialise lds_allocated,
6057 lds_allocs, reduc_decls fields of machine function struct.
6058 (gcn_option_override): Handle default size for gang-private
6059 variables and -mgang-private-size option.
6060 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
6061 initialising M0_REG.
6062 (gcn_shared_mem_layout): New function.
6063 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
6064 gang_private_hwm variable.
6065 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
6066 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
6067 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
6068 * config/gcn/gcn.opt (gang_private_size_opt): New global.
6069 (mgang-private-size=): New option.
6070 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
6072 * doc/tm.texi: Regenerate.
6073 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
6075 (build_sender_ref): Handle sender_decl being pointer.
6076 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
6077 parameters. Pass placement argument to
6078 create_worker_broadcast_record hook invocations. Handle
6079 sender_decl being pointer and isolate_broadcasts inserting extra
6081 (blk_offset_map_t): Add typedef.
6082 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
6083 preallocated range to worker_single_copy call.
6084 (dfs_broadcast_reachable_1): New function.
6085 (idx_decl_pair_t, used_range_vec_t): New typedefs.
6086 (sort_size_descending): New function.
6087 (addr_range): New class.
6088 (splay_tree_compare_addr_range, splay_tree_free_key)
6089 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
6090 (execute_omp_oacc_neuter_broadcast): Rename to...
6091 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
6092 parameters. Arrange layout of shared memory for broadcast
6094 (execute_omp_oacc_neuter_broadcast): New function.
6095 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
6096 handling from here. Enable pass for all OpenACC routines in order
6097 to call shared memory-layout hook.
6098 * target.def (create_worker_broadcast_record): Add OFFSET
6100 (shared_mem_layout): New hook.
6102 2021-09-17 Julian Brown <julian@codesourcery.com>
6103 Thomas Schwinge <thomas@codesourcery.com>
6105 * omp-oacc-neuter-broadcast.cc
6106 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
6108 (execute_omp_oacc_neuter_broadcast): Adjust.
6110 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
6112 * value-relation.cc (class equiv_chain): Move to header file.
6113 (path_oracle::path_oracle): New.
6114 (path_oracle::~path_oracle): New.
6115 (path_oracle::register_relation): New.
6116 (path_oracle::query_relation): New.
6117 (path_oracle::reset_path): New.
6118 (path_oracle::dump): New.
6119 * value-relation.h (class equiv_chain): Move to here.
6120 (class path_oracle): New.
6122 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
6124 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
6126 * gimple-range-fold.cc (fur_depend::register_relation): Use
6127 register_stmt/edge routines.
6128 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
6129 (equiv_oracle::equiv_oracle): Create self equivalence cache.
6130 (equiv_oracle::~equiv_oracle): Release same.
6131 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
6132 are no equivalences.
6133 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
6134 (equiv_oracle::register_relation): Rename from register_equiv.
6135 (relation_chain_head::find_relation): Relocate from dom_oracle.
6136 (relation_oracle::register_stmt): New.
6137 (relation_oracle::register_edge): New.
6138 (dom_oracle::*): Rename from relation_oracle.
6139 (dom_oracle::register_relation): Adjust to call equiv_oracle.
6140 (dom_oracle::set_one_relation): Split from register_relation.
6141 (dom_oracle::register_transitives): Consolidate 2 methods.
6142 (dom_oracle::find_relation_block): Move core to relation_chain.
6143 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
6144 * value-relation.h (class relation_oracle): New pure virtual base.
6145 (class equiv_oracle): Inherit from relation_oracle and adjust.
6146 (class dom_oracle): Rename from old relation_oracle and adjust.
6148 2021-09-17 Martin Sebor <msebor@redhat.com>
6150 PR middle-end/102200
6151 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
6152 (handle_min_max_size): Change argument. Store original SSA_NAME for
6153 operands to potentially distinct (sub)objects.
6154 (compute_objsize_r): Adjust call to the above.
6156 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6158 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
6159 (rs6000_new_builtin_vectorized_function): New function.
6160 (rs6000_new_builtin_md_vectorized_function): Likewise.
6161 (rs6000_builtin_vectorized_function): Call
6162 rs6000_new_builtin_vectorized_function.
6163 (rs6000_builtin_md_vectorized_function): Call
6164 rs6000_new_builtin_md_vectorized_function.
6166 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6168 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
6169 (ASSEMBLE_PAIR): Likewise.
6170 (BUILD_ACC): Likewise.
6171 (DISASSEMBLE_ACC): Likewise.
6172 (DISASSEMBLE_PAIR): Likewise.
6173 (PMXVBF16GER2): Likewise.
6174 (PMXVBF16GER2NN): Likewise.
6175 (PMXVBF16GER2NP): Likewise.
6176 (PMXVBF16GER2PN): Likewise.
6177 (PMXVBF16GER2PP): Likewise.
6178 (PMXVF16GER2): Likewise.
6179 (PMXVF16GER2NN): Likewise.
6180 (PMXVF16GER2NP): Likewise.
6181 (PMXVF16GER2PN): Likewise.
6182 (PMXVF16GER2PP): Likewise.
6183 (PMXVF32GER): Likewise.
6184 (PMXVF32GERNN): Likewise.
6185 (PMXVF32GERNP): Likewise.
6186 (PMXVF32GERPN): Likewise.
6187 (PMXVF32GERPP): Likewise.
6188 (PMXVF64GER): Likewise.
6189 (PMXVF64GERNN): Likewise.
6190 (PMXVF64GERNP): Likewise.
6191 (PMXVF64GERPN): Likewise.
6192 (PMXVF64GERPP): Likewise.
6193 (PMXVI16GER2): Likewise.
6194 (PMXVI16GER2PP): Likewise.
6195 (PMXVI16GER2S): Likewise.
6196 (PMXVI16GER2SPP): Likewise.
6197 (PMXVI4GER8): Likewise.
6198 (PMXVI4GER8PP): Likewise.
6199 (PMXVI8GER4): Likewise.
6200 (PMXVI8GER4PP): Likewise.
6201 (PMXVI8GER4SPP): Likewise.
6202 (XVBF16GER2): Likewise.
6203 (XVBF16GER2NN): Likewise.
6204 (XVBF16GER2NP): Likewise.
6205 (XVBF16GER2PN): Likewise.
6206 (XVBF16GER2PP): Likewise.
6207 (XVF16GER2): Likewise.
6208 (XVF16GER2NN): Likewise.
6209 (XVF16GER2NP): Likewise.
6210 (XVF16GER2PN): Likewise.
6211 (XVF16GER2PP): Likewise.
6212 (XVF32GER): Likewise.
6213 (XVF32GERNN): Likewise.
6214 (XVF32GERNP): Likewise.
6215 (XVF32GERPN): Likewise.
6216 (XVF32GERPP): Likewise.
6217 (XVF64GER): Likewise.
6218 (XVF64GERNN): Likewise.
6219 (XVF64GERNP): Likewise.
6220 (XVF64GERPN): Likewise.
6221 (XVF64GERPP): Likewise.
6222 (XVI16GER2): Likewise.
6223 (XVI16GER2PP): Likewise.
6224 (XVI16GER2S): Likewise.
6225 (XVI16GER2SPP): Likewise.
6226 (XVI4GER8): Likewise.
6227 (XVI4GER8PP): Likewise.
6228 (XVI8GER4): Likewise.
6229 (XVI8GER4PP): Likewise.
6230 (XVI8GER4SPP): Likewise.
6231 (XXMFACC): Likewise.
6232 (XXMTACC): Likewise.
6233 (XXSETACCZ): Likewise.
6234 (ASSEMBLE_PAIR_V): Likewise.
6235 (BUILD_PAIR): Likewise.
6236 (DISASSEMBLE_PAIR_V): Likewise.
6239 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
6240 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
6241 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
6242 (parse_bif_attrs): Handle ismmaint.
6243 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
6244 (write_bif_static_init): Handle ismmaint.
6246 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6248 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
6250 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
6251 (rs6000_new_builtin_valid_without_lhs): New function.
6252 (rs6000_gimple_fold_new_mma_builtin): Likewise.
6253 (rs6000_gimple_fold_new_builtin): Likewise.
6255 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
6257 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
6258 Destruct stale Value objects.
6259 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
6262 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
6265 * match.pd (shift optimizations): Disable recent sign-changing
6266 optimization for shifts by zero, these will be folded later.
6268 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6270 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
6271 [power9] to [always].
6273 2021-09-17 Richard Biener <rguenther@suse.de>
6275 * tree-vect-stmts.c (vectorizable_load): Do not frob
6278 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6280 * config/i386/i386-features.c (remove_partial_avx_dependency):
6281 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
6282 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
6284 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
6286 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
6287 * config/i386/i386.md (SSE FP to FP splitters): Replace
6288 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
6289 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
6290 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
6291 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
6292 * config/i386/x86-tune.def
6293 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
6294 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
6296 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6299 * config/i386/i386-features.c (remove_partial_avx_dependency):
6300 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
6301 before generating vxorps.
6303 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6305 * config/i386/i386-options.c (processor_cost_table): Use
6306 tremont_cost for Tremont.
6307 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
6308 (tremont_memset): Likewise.
6309 (tremont_cost): Likewise.
6310 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
6313 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6315 * common/config/i386/i386-common.c: Use Haswell scheduling model
6317 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
6319 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
6321 (ix86_adjust_cost): Handle Tremont.
6322 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
6324 (X86_TUNE_USE_LEAVE): Likewise.
6325 (X86_TUNE_PUSH_MEMORY): Likewise.
6326 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
6327 (X86_TUNE_USE_CLTD): Likewise.
6328 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
6329 (X86_TUNE_AVOID_MFENCE): Likewise.
6330 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
6331 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
6332 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
6333 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
6334 (X86_TUNE_OPT_AGU): Likewise.
6335 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
6336 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
6337 (X86_TUNE_EXPAND_ABS): Likewise.
6338 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
6339 (X86_TUNE_SLOW_PSHUFB): Likewise.
6341 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
6343 PR rtl-optimization/102306
6344 * combine.c (try_combine): Abort the combination if we are about to
6345 duplicate volatile references.
6347 2021-09-17 liuhongt <hongtao.liu@intel.com>
6349 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
6351 (_mm256_undefined_ph): Likewise.
6352 (_mm512_undefined_ph): Likewise.
6353 (_mm_cvtsh_h): Likewise.
6354 (_mm256_cvtsh_h): Likewise.
6355 (_mm512_cvtsh_h): Likewise.
6356 (_mm512_castph_ps): Likewise.
6357 (_mm512_castph_pd): Likewise.
6358 (_mm512_castph_si512): Likewise.
6359 (_mm512_castph512_ph128): Likewise.
6360 (_mm512_castph512_ph256): Likewise.
6361 (_mm512_castph128_ph512): Likewise.
6362 (_mm512_castph256_ph512): Likewise.
6363 (_mm512_zextph128_ph512): Likewise.
6364 (_mm512_zextph256_ph512): Likewise.
6365 (_mm512_castps_ph): Likewise.
6366 (_mm512_castpd_ph): Likewise.
6367 (_mm512_castsi512_ph): Likewise.
6368 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
6370 (_mm256_castph_ps): Likewise.
6371 (_mm_castph_pd): Likewise.
6372 (_mm256_castph_pd): Likewise.
6373 (_mm_castph_si128): Likewise.
6374 (_mm256_castph_si256): Likewise.
6375 (_mm_castps_ph): Likewise.
6376 (_mm256_castps_ph): Likewise.
6377 (_mm_castpd_ph): Likewise.
6378 (_mm256_castpd_ph): Likewise.
6379 (_mm_castsi128_ph): Likewise.
6380 (_mm256_castsi256_ph): Likewise.
6381 (_mm256_castph256_ph128): Likewise.
6382 (_mm256_castph128_ph256): Likewise.
6383 (_mm256_zextph128_ph256): Likewise.
6385 2021-09-17 liuhongt <hongtao.liu@intel.com>
6387 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
6389 (_mm_mask_cvtsh_ss): Likewise.
6390 (_mm_maskz_cvtsh_ss): Likewise.
6391 (_mm_cvtsh_sd): Likewise.
6392 (_mm_mask_cvtsh_sd): Likewise.
6393 (_mm_maskz_cvtsh_sd): Likewise.
6394 (_mm_cvt_roundsh_ss): Likewise.
6395 (_mm_mask_cvt_roundsh_ss): Likewise.
6396 (_mm_maskz_cvt_roundsh_ss): Likewise.
6397 (_mm_cvt_roundsh_sd): Likewise.
6398 (_mm_mask_cvt_roundsh_sd): Likewise.
6399 (_mm_maskz_cvt_roundsh_sd): Likewise.
6400 (_mm_cvtss_sh): Likewise.
6401 (_mm_mask_cvtss_sh): Likewise.
6402 (_mm_maskz_cvtss_sh): Likewise.
6403 (_mm_cvtsd_sh): Likewise.
6404 (_mm_mask_cvtsd_sh): Likewise.
6405 (_mm_maskz_cvtsd_sh): Likewise.
6406 (_mm_cvt_roundss_sh): Likewise.
6407 (_mm_mask_cvt_roundss_sh): Likewise.
6408 (_mm_maskz_cvt_roundss_sh): Likewise.
6409 (_mm_cvt_roundsd_sh): Likewise.
6410 (_mm_mask_cvt_roundsd_sh): Likewise.
6411 (_mm_maskz_cvt_roundsd_sh): Likewise.
6412 * config/i386/i386-builtin-types.def
6413 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
6414 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
6415 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
6416 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
6417 * config/i386/i386-builtin.def: Add corrresponding new builtins.
6418 * config/i386/i386-expand.c: Handle new builtin types.
6419 * config/i386/sse.md (VF48_128): New mode iterator.
6420 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
6422 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
6425 2021-09-17 liuhongt <hongtao.liu@intel.com>
6427 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
6429 (_mm512_mask_cvtph_pd): Likewise.
6430 (_mm512_maskz_cvtph_pd): Likewise.
6431 (_mm512_cvt_roundph_pd): Likewise.
6432 (_mm512_mask_cvt_roundph_pd): Likewise.
6433 (_mm512_maskz_cvt_roundph_pd): Likewise.
6434 (_mm512_cvtxph_ps): Likewise.
6435 (_mm512_mask_cvtxph_ps): Likewise.
6436 (_mm512_maskz_cvtxph_ps): Likewise.
6437 (_mm512_cvtx_roundph_ps): Likewise.
6438 (_mm512_mask_cvtx_roundph_ps): Likewise.
6439 (_mm512_maskz_cvtx_roundph_ps): Likewise.
6440 (_mm512_cvtxps_ph): Likewise.
6441 (_mm512_mask_cvtxps_ph): Likewise.
6442 (_mm512_maskz_cvtxps_ph): Likewise.
6443 (_mm512_cvtx_roundps_ph): Likewise.
6444 (_mm512_mask_cvtx_roundps_ph): Likewise.
6445 (_mm512_maskz_cvtx_roundps_ph): Likewise.
6446 (_mm512_cvtpd_ph): Likewise.
6447 (_mm512_mask_cvtpd_ph): Likewise.
6448 (_mm512_maskz_cvtpd_ph): Likewise.
6449 (_mm512_cvt_roundpd_ph): Likewise.
6450 (_mm512_mask_cvt_roundpd_ph): Likewise.
6451 (_mm512_maskz_cvt_roundpd_ph): Likewise.
6452 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
6454 (_mm_mask_cvtph_pd): Likewise.
6455 (_mm_maskz_cvtph_pd): Likewise.
6456 (_mm256_cvtph_pd): Likewise.
6457 (_mm256_mask_cvtph_pd): Likewise.
6458 (_mm256_maskz_cvtph_pd): Likewise.
6459 (_mm_cvtxph_ps): Likewise.
6460 (_mm_mask_cvtxph_ps): Likewise.
6461 (_mm_maskz_cvtxph_ps): Likewise.
6462 (_mm256_cvtxph_ps): Likewise.
6463 (_mm256_mask_cvtxph_ps): Likewise.
6464 (_mm256_maskz_cvtxph_ps): Likewise.
6465 (_mm_cvtxps_ph): Likewise.
6466 (_mm_mask_cvtxps_ph): Likewise.
6467 (_mm_maskz_cvtxps_ph): Likewise.
6468 (_mm256_cvtxps_ph): Likewise.
6469 (_mm256_mask_cvtxps_ph): Likewise.
6470 (_mm256_maskz_cvtxps_ph): Likewise.
6471 (_mm_cvtpd_ph): Likewise.
6472 (_mm_mask_cvtpd_ph): Likewise.
6473 (_mm_maskz_cvtpd_ph): Likewise.
6474 (_mm256_cvtpd_ph): Likewise.
6475 (_mm256_mask_cvtpd_ph): Likewise.
6476 (_mm256_maskz_cvtpd_ph): Likewise.
6477 * config/i386/i386-builtin.def: Add corresponding new builtins.
6478 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
6479 * config/i386/i386-expand.c: Handle new builtin types.
6480 * config/i386/sse.md
6481 (VF4_128_8_256): New.
6482 (VF48H_AVX512VL): Ditto.
6483 (ssePHmode): Add HF vector modes.
6484 (castmode): Add new convertable modes.
6487 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
6488 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
6489 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
6490 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
6491 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
6492 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
6493 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
6495 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
6496 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
6497 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
6498 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
6500 2021-09-17 liuhongt <hongtao.liu@intel.com>
6502 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
6504 (_mm_cvttsh_u32): Likewise.
6505 (_mm_cvtt_roundsh_i32): Likewise.
6506 (_mm_cvtt_roundsh_u32): Likewise.
6507 (_mm_cvttsh_i64): Likewise.
6508 (_mm_cvttsh_u64): Likewise.
6509 (_mm_cvtt_roundsh_i64): Likewise.
6510 (_mm_cvtt_roundsh_u64): Likewise.
6511 * config/i386/i386-builtin.def: Add corresponding new builtins.
6512 * config/i386/sse.md
6513 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
6516 2021-09-17 liuhongt <hongtao.liu@intel.com>
6518 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
6520 (_mm512_mask_cvttph_epi32): Likewise.
6521 (_mm512_maskz_cvttph_epi32): Likewise.
6522 (_mm512_cvtt_roundph_epi32): Likewise.
6523 (_mm512_mask_cvtt_roundph_epi32): Likewise.
6524 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
6525 (_mm512_cvttph_epu32): Likewise.
6526 (_mm512_mask_cvttph_epu32): Likewise.
6527 (_mm512_maskz_cvttph_epu32): Likewise.
6528 (_mm512_cvtt_roundph_epu32): Likewise.
6529 (_mm512_mask_cvtt_roundph_epu32): Likewise.
6530 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
6531 (_mm512_cvttph_epi64): Likewise.
6532 (_mm512_mask_cvttph_epi64): Likewise.
6533 (_mm512_maskz_cvttph_epi64): Likewise.
6534 (_mm512_cvtt_roundph_epi64): Likewise.
6535 (_mm512_mask_cvtt_roundph_epi64): Likewise.
6536 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
6537 (_mm512_cvttph_epu64): Likewise.
6538 (_mm512_mask_cvttph_epu64): Likewise.
6539 (_mm512_maskz_cvttph_epu64): Likewise.
6540 (_mm512_cvtt_roundph_epu64): Likewise.
6541 (_mm512_mask_cvtt_roundph_epu64): Likewise.
6542 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
6543 (_mm512_cvttph_epi16): Likewise.
6544 (_mm512_mask_cvttph_epi16): Likewise.
6545 (_mm512_maskz_cvttph_epi16): Likewise.
6546 (_mm512_cvtt_roundph_epi16): Likewise.
6547 (_mm512_mask_cvtt_roundph_epi16): Likewise.
6548 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
6549 (_mm512_cvttph_epu16): Likewise.
6550 (_mm512_mask_cvttph_epu16): Likewise.
6551 (_mm512_maskz_cvttph_epu16): Likewise.
6552 (_mm512_cvtt_roundph_epu16): Likewise.
6553 (_mm512_mask_cvtt_roundph_epu16): Likewise.
6554 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
6555 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
6557 (_mm_mask_cvttph_epi32): Likewise.
6558 (_mm_maskz_cvttph_epi32): Likewise.
6559 (_mm256_cvttph_epi32): Likewise.
6560 (_mm256_mask_cvttph_epi32): Likewise.
6561 (_mm256_maskz_cvttph_epi32): Likewise.
6562 (_mm_cvttph_epu32): Likewise.
6563 (_mm_mask_cvttph_epu32): Likewise.
6564 (_mm_maskz_cvttph_epu32): Likewise.
6565 (_mm256_cvttph_epu32): Likewise.
6566 (_mm256_mask_cvttph_epu32): Likewise.
6567 (_mm256_maskz_cvttph_epu32): Likewise.
6568 (_mm_cvttph_epi64): Likewise.
6569 (_mm_mask_cvttph_epi64): Likewise.
6570 (_mm_maskz_cvttph_epi64): Likewise.
6571 (_mm256_cvttph_epi64): Likewise.
6572 (_mm256_mask_cvttph_epi64): Likewise.
6573 (_mm256_maskz_cvttph_epi64): Likewise.
6574 (_mm_cvttph_epu64): Likewise.
6575 (_mm_mask_cvttph_epu64): Likewise.
6576 (_mm_maskz_cvttph_epu64): Likewise.
6577 (_mm256_cvttph_epu64): Likewise.
6578 (_mm256_mask_cvttph_epu64): Likewise.
6579 (_mm256_maskz_cvttph_epu64): Likewise.
6580 (_mm_cvttph_epi16): Likewise.
6581 (_mm_mask_cvttph_epi16): Likewise.
6582 (_mm_maskz_cvttph_epi16): Likewise.
6583 (_mm256_cvttph_epi16): Likewise.
6584 (_mm256_mask_cvttph_epi16): Likewise.
6585 (_mm256_maskz_cvttph_epi16): Likewise.
6586 (_mm_cvttph_epu16): Likewise.
6587 (_mm_mask_cvttph_epu16): Likewise.
6588 (_mm_maskz_cvttph_epu16): Likewise.
6589 (_mm256_cvttph_epu16): Likewise.
6590 (_mm256_mask_cvttph_epu16): Likewise.
6591 (_mm256_maskz_cvttph_epu16): Likewise.
6592 * config/i386/i386-builtin.def: Add new builtins.
6593 * config/i386/sse.md
6594 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
6596 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
6597 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
6598 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
6599 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
6601 2021-09-17 liuhongt <hongtao.liu@intel.com>
6603 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
6604 (_mm_cvtsh_u32): Likewise.
6605 (_mm_cvt_roundsh_i32): Likewise.
6606 (_mm_cvt_roundsh_u32): Likewise.
6607 (_mm_cvtsh_i64): Likewise.
6608 (_mm_cvtsh_u64): Likewise.
6609 (_mm_cvt_roundsh_i64): Likewise.
6610 (_mm_cvt_roundsh_u64): Likewise.
6611 (_mm_cvti32_sh): Likewise.
6612 (_mm_cvtu32_sh): Likewise.
6613 (_mm_cvt_roundi32_sh): Likewise.
6614 (_mm_cvt_roundu32_sh): Likewise.
6615 (_mm_cvti64_sh): Likewise.
6616 (_mm_cvtu64_sh): Likewise.
6617 (_mm_cvt_roundi64_sh): Likewise.
6618 (_mm_cvt_roundu64_sh): Likewise.
6619 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
6620 * config/i386/i386-builtin.def: Add corresponding new builtins.
6621 * config/i386/i386-expand.c (ix86_expand_round_builtin):
6622 Handle new builtin types.
6623 * config/i386/sse.md
6624 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
6626 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
6627 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
6629 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
6631 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
6632 (altivec_resolve_new_overloaded_builtin): New forward decl.
6633 (rs6000_new_builtin_type_compatible): New function.
6634 (altivec_resolve_overloaded_builtin): Call
6635 altivec_resolve_new_overloaded_builtin.
6636 (altivec_build_new_resolved_builtin): New function.
6637 (altivec_resolve_new_overloaded_builtin): Likewise.
6638 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
6640 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
6641 name of rs6000_new_builtin_is_supported.
6643 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
6645 * config/i386/i386-protos.h (ix86_decompose_address):
6646 Change return type to bool.
6647 * config/i386/i386.c (ix86_decompose_address): Ditto.
6649 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
6652 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
6653 Added 'build/' to target, use build/%.o rule.
6654 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
6655 '$(build_exeext)' to target and 'build/' for the *.o files.
6656 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
6659 2021-09-16 Martin Jambor <mjambor@suse.cz>
6661 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
6662 the body has been removed. Remove trailing whitespace.
6664 2021-09-16 Richard Biener <rguenther@suse.de>
6666 PR middle-end/102360
6667 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
6668 of non-memory more robust.
6670 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6672 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
6673 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
6674 (leon5_adjust_cost): Increase cost of store with data dependency
6675 on ALU instruction and FPU anti-dependencies.
6676 (sparc_option_override): Add LEON5 costs
6677 (sparc_adjust_cost): Add LEON5 cost adjustments
6678 * config/sparc/sparc.h: Add LEON5
6679 * config/sparc/sparc.md: Include LEON5 scheduling information
6680 * config/sparc/sparc.opt: Add LEON5
6681 * doc/invoke.texi: Add LEON5
6682 * config/sparc/leon5.md: New file.
6684 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6686 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
6687 sensitive sequence for B2BST errata workaround.
6689 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6691 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
6692 functions with atomic instruction in the UT700 errata workaround.
6694 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6696 * config/sparc/sparc.c (next_active_non_empty_insn): New function
6697 that returns next active non empty assembly instruction.
6698 (sparc_do_work_around_errata): Use new function.
6700 2021-09-16 Daniel Cederman <cederman@gaisler.com>
6702 * config/sparc/sparc.c (store_insn_p): Add predicate for store
6704 (load_insn_p): Add predicate for load attributes.
6705 (sparc_do_work_around_errata): Use new predicates.
6707 2021-09-16 Andreas Larsson <andreas@gaisler.com>
6709 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
6712 2021-09-16 Martin Liska <mliska@suse.cz>
6714 * config/mips/netbsd.h: Fix typo in name of a macro.
6716 2021-09-16 liuhongt <hongtao.liu@intel.com>
6718 PR middle-end/102080
6719 * match.pd: Check mask type when doing cond_op related gimple
6721 * tree.c (is_truth_type_for): New function.
6722 * tree.h (is_truth_type_for): New declaration.
6724 2021-09-16 liuhongt <hongtao.liu@intel.com>
6726 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
6728 (_mm512_mask_cvtepi32_ph): Likewise.
6729 (_mm512_maskz_cvtepi32_ph): Likewise.
6730 (_mm512_cvt_roundepi32_ph): Likewise.
6731 (_mm512_mask_cvt_roundepi32_ph): Likewise.
6732 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
6733 (_mm512_cvtepu32_ph): Likewise.
6734 (_mm512_mask_cvtepu32_ph): Likewise.
6735 (_mm512_maskz_cvtepu32_ph): Likewise.
6736 (_mm512_cvt_roundepu32_ph): Likewise.
6737 (_mm512_mask_cvt_roundepu32_ph): Likewise.
6738 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
6739 (_mm512_cvtepi64_ph): Likewise.
6740 (_mm512_mask_cvtepi64_ph): Likewise.
6741 (_mm512_maskz_cvtepi64_ph): Likewise.
6742 (_mm512_cvt_roundepi64_ph): Likewise.
6743 (_mm512_mask_cvt_roundepi64_ph): Likewise.
6744 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
6745 (_mm512_cvtepu64_ph): Likewise.
6746 (_mm512_mask_cvtepu64_ph): Likewise.
6747 (_mm512_maskz_cvtepu64_ph): Likewise.
6748 (_mm512_cvt_roundepu64_ph): Likewise.
6749 (_mm512_mask_cvt_roundepu64_ph): Likewise.
6750 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
6751 (_mm512_cvtepi16_ph): Likewise.
6752 (_mm512_mask_cvtepi16_ph): Likewise.
6753 (_mm512_maskz_cvtepi16_ph): Likewise.
6754 (_mm512_cvt_roundepi16_ph): Likewise.
6755 (_mm512_mask_cvt_roundepi16_ph): Likewise.
6756 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
6757 (_mm512_cvtepu16_ph): Likewise.
6758 (_mm512_mask_cvtepu16_ph): Likewise.
6759 (_mm512_maskz_cvtepu16_ph): Likewise.
6760 (_mm512_cvt_roundepu16_ph): Likewise.
6761 (_mm512_mask_cvt_roundepu16_ph): Likewise.
6762 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
6763 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
6765 (_mm_mask_cvtepi32_ph): Likewise.
6766 (_mm_maskz_cvtepi32_ph): Likewise.
6767 (_mm256_cvtepi32_ph): Likewise.
6768 (_mm256_mask_cvtepi32_ph): Likewise.
6769 (_mm256_maskz_cvtepi32_ph): Likewise.
6770 (_mm_cvtepu32_ph): Likewise.
6771 (_mm_mask_cvtepu32_ph): Likewise.
6772 (_mm_maskz_cvtepu32_ph): Likewise.
6773 (_mm256_cvtepu32_ph): Likewise.
6774 (_mm256_mask_cvtepu32_ph): Likewise.
6775 (_mm256_maskz_cvtepu32_ph): Likewise.
6776 (_mm_cvtepi64_ph): Likewise.
6777 (_mm_mask_cvtepi64_ph): Likewise.
6778 (_mm_maskz_cvtepi64_ph): Likewise.
6779 (_mm256_cvtepi64_ph): Likewise.
6780 (_mm256_mask_cvtepi64_ph): Likewise.
6781 (_mm256_maskz_cvtepi64_ph): Likewise.
6782 (_mm_cvtepu64_ph): Likewise.
6783 (_mm_mask_cvtepu64_ph): Likewise.
6784 (_mm_maskz_cvtepu64_ph): Likewise.
6785 (_mm256_cvtepu64_ph): Likewise.
6786 (_mm256_mask_cvtepu64_ph): Likewise.
6787 (_mm256_maskz_cvtepu64_ph): Likewise.
6788 (_mm_cvtepi16_ph): Likewise.
6789 (_mm_mask_cvtepi16_ph): Likewise.
6790 (_mm_maskz_cvtepi16_ph): Likewise.
6791 (_mm256_cvtepi16_ph): Likewise.
6792 (_mm256_mask_cvtepi16_ph): Likewise.
6793 (_mm256_maskz_cvtepi16_ph): Likewise.
6794 (_mm_cvtepu16_ph): Likewise.
6795 (_mm_mask_cvtepu16_ph): Likewise.
6796 (_mm_maskz_cvtepu16_ph): Likewise.
6797 (_mm256_cvtepu16_ph): Likewise.
6798 (_mm256_mask_cvtepu16_ph): Likewise.
6799 (_mm256_maskz_cvtepu16_ph): Likewise.
6800 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
6801 * config/i386/i386-builtin.def: Add corresponding new builtins.
6802 * config/i386/i386-expand.c
6803 (ix86_expand_args_builtin): Handle new builtin types.
6804 (ix86_expand_round_builtin): Ditto.
6805 * config/i386/i386-modes.def: Declare V2HF and V6HF.
6806 * config/i386/sse.md (VI2H_AVX512VL): New.
6808 (sseintvecmode): Add HF vector modes.
6809 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
6811 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
6812 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
6813 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
6814 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
6815 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
6816 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
6817 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
6818 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
6819 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
6820 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
6821 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
6823 2021-09-16 liuhongt <hongtao.liu@intel.com>
6825 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
6827 (_mm512_mask_cvtph_epi32): Likewise.
6828 (_mm512_maskz_cvtph_epi32): Likewise.
6829 (_mm512_cvt_roundph_epi32): Likewise.
6830 (_mm512_mask_cvt_roundph_epi32): Likewise.
6831 (_mm512_maskz_cvt_roundph_epi32): Likewise.
6832 (_mm512_cvtph_epu32): Likewise.
6833 (_mm512_mask_cvtph_epu32): Likewise.
6834 (_mm512_maskz_cvtph_epu32): Likewise.
6835 (_mm512_cvt_roundph_epu32): Likewise.
6836 (_mm512_mask_cvt_roundph_epu32): Likewise.
6837 (_mm512_maskz_cvt_roundph_epu32): Likewise.
6838 (_mm512_cvtph_epi64): Likewise.
6839 (_mm512_mask_cvtph_epi64): Likewise.
6840 (_mm512_maskz_cvtph_epi64): Likewise.
6841 (_mm512_cvt_roundph_epi64): Likewise.
6842 (_mm512_mask_cvt_roundph_epi64): Likewise.
6843 (_mm512_maskz_cvt_roundph_epi64): Likewise.
6844 (_mm512_cvtph_epu64): Likewise.
6845 (_mm512_mask_cvtph_epu64): Likewise.
6846 (_mm512_maskz_cvtph_epu64): Likewise.
6847 (_mm512_cvt_roundph_epu64): Likewise.
6848 (_mm512_mask_cvt_roundph_epu64): Likewise.
6849 (_mm512_maskz_cvt_roundph_epu64): Likewise.
6850 (_mm512_cvtph_epi16): Likewise.
6851 (_mm512_mask_cvtph_epi16): Likewise.
6852 (_mm512_maskz_cvtph_epi16): Likewise.
6853 (_mm512_cvt_roundph_epi16): Likewise.
6854 (_mm512_mask_cvt_roundph_epi16): Likewise.
6855 (_mm512_maskz_cvt_roundph_epi16): Likewise.
6856 (_mm512_cvtph_epu16): Likewise.
6857 (_mm512_mask_cvtph_epu16): Likewise.
6858 (_mm512_maskz_cvtph_epu16): Likewise.
6859 (_mm512_cvt_roundph_epu16): Likewise.
6860 (_mm512_mask_cvt_roundph_epu16): Likewise.
6861 (_mm512_maskz_cvt_roundph_epu16): Likewise.
6862 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
6864 (_mm_mask_cvtph_epi32): Likewise.
6865 (_mm_maskz_cvtph_epi32): Likewise.
6866 (_mm256_cvtph_epi32): Likewise.
6867 (_mm256_mask_cvtph_epi32): Likewise.
6868 (_mm256_maskz_cvtph_epi32): Likewise.
6869 (_mm_cvtph_epu32): Likewise.
6870 (_mm_mask_cvtph_epu32): Likewise.
6871 (_mm_maskz_cvtph_epu32): Likewise.
6872 (_mm256_cvtph_epu32): Likewise.
6873 (_mm256_mask_cvtph_epu32): Likewise.
6874 (_mm256_maskz_cvtph_epu32): Likewise.
6875 (_mm_cvtph_epi64): Likewise.
6876 (_mm_mask_cvtph_epi64): Likewise.
6877 (_mm_maskz_cvtph_epi64): Likewise.
6878 (_mm256_cvtph_epi64): Likewise.
6879 (_mm256_mask_cvtph_epi64): Likewise.
6880 (_mm256_maskz_cvtph_epi64): Likewise.
6881 (_mm_cvtph_epu64): Likewise.
6882 (_mm_mask_cvtph_epu64): Likewise.
6883 (_mm_maskz_cvtph_epu64): Likewise.
6884 (_mm256_cvtph_epu64): Likewise.
6885 (_mm256_mask_cvtph_epu64): Likewise.
6886 (_mm256_maskz_cvtph_epu64): Likewise.
6887 (_mm_cvtph_epi16): Likewise.
6888 (_mm_mask_cvtph_epi16): Likewise.
6889 (_mm_maskz_cvtph_epi16): Likewise.
6890 (_mm256_cvtph_epi16): Likewise.
6891 (_mm256_mask_cvtph_epi16): Likewise.
6892 (_mm256_maskz_cvtph_epi16): Likewise.
6893 (_mm_cvtph_epu16): Likewise.
6894 (_mm_mask_cvtph_epu16): Likewise.
6895 (_mm_maskz_cvtph_epu16): Likewise.
6896 (_mm256_cvtph_epu16): Likewise.
6897 (_mm256_mask_cvtph_epu16): Likewise.
6898 (_mm256_maskz_cvtph_epu16): Likewise.
6899 * config/i386/i386-builtin-types.def: Add new builtin types.
6900 * config/i386/i386-builtin.def: Add new builtins.
6901 * config/i386/i386-expand.c
6902 (ix86_expand_args_builtin): Handle new builtin types.
6903 (ix86_expand_round_builtin): Ditto.
6904 * config/i386/sse.md (sseintconvert): New.
6906 (UNSPEC_US_FIX_NOTRUNC): Ditto.
6907 (sseintconvertsignprefix): Ditto.
6908 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
6911 2021-09-16 liuhongt <hongtao.liu@intel.com>
6913 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
6915 (_mm_cvtsi128_si16): Likewise.
6916 (_mm_mask_load_sh): Likewise.
6917 (_mm_maskz_load_sh): Likewise.
6918 (_mm_mask_store_sh): Likewise.
6919 (_mm_move_sh): Likewise.
6920 (_mm_mask_move_sh): Likewise.
6921 (_mm_maskz_move_sh): Likewise.
6922 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
6923 * config/i386/i386-builtin.def: Add corresponding new builtins.
6924 * config/i386/i386-expand.c
6925 (ix86_expand_special_args_builtin): Handle new builtin types.
6926 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
6927 * config/i386/sse.md (VI2F): New mode iterator.
6928 (vec_set<mode>_0): Use new mode iterator.
6929 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
6930 (avx512f_store<mode>_mask): Ditto.
6932 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
6934 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
6936 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
6938 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
6939 Proceed if no symbol summary or the symbol alias flag is false.
6941 2021-09-15 Jakub Jelinek <jakub@redhat.com>
6945 * varasm.c (output_constructor_regular_field): Instead of assertion
6946 that array_size_for_constructor result is equal to size of
6947 TREE_TYPE (local->val) in bytes, assert that the type size is greater
6948 or equal to array_size_for_constructor result and use type size as
6951 2021-09-15 Martin Liska <mliska@suse.cz>
6954 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
6956 2021-09-15 Martin Liska <mliska@suse.cz>
6959 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
6960 Check that we have a symbol summary for a symbol.
6962 2021-09-15 Richard Biener <rguenther@suse.de>
6965 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
6966 to inherit from elfos.h
6968 2021-09-15 liuhongt <hongtao.liu@intel.com>
6971 * config/i386/i386-expand.c
6972 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
6974 (ix86_expand_vector_set): Use blendw instead of pinsrw.
6975 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
6976 AVX512FP16 which supports 16bit vector load.
6977 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
6979 (avx512bw_interleave_high<mode><mask_name>): .. this, and
6980 extend to V32HFmode.
6981 (avx2_interleave_highv16hi<mask_name>): Rename to ..
6982 (avx2_interleave_high<mode><mask_name>): .. this, and extend
6984 (vec_interleave_highv8hi<mask_name>): Rename to ..
6985 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
6986 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
6988 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
6989 this, and extend to V32HFmode.
6990 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
6991 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
6992 (vec_interleave_lowv8hi<mask_name>): Rename to ..
6993 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
6994 (sse4_1_pblendw): Rename to ..
6995 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
6996 (avx2_pblendph): New define_expand.
6997 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
6998 sseintmodesuffix instead of ssemodesuffix.
6999 (blendsuf): New mode attr.
7001 2021-09-15 Richard Biener <rguenther@suse.de>
7003 * tree-vectorizer.h (dr_misalignment): Move out of line.
7004 (dr_target_alignment): New.
7005 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
7006 (set_dr_target_alignment): New.
7007 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
7008 * tree-vect-data-refs.c (dr_misalignment): Compute and
7009 return the group members misalignment.
7010 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
7011 (vect_analyze_data_refs_alignment): Compute alignment only
7012 for the first element of a DR group.
7013 (vect_slp_analyze_node_alignment): Likewise.
7015 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
7017 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
7018 * config/i386/avx512fp16vlintrin.h: Likewise.
7019 * config/i386/i386-builtin.def: Adjust builtin name and
7020 enumeration to match AVX512F style.
7022 2021-09-15 Richard Biener <rguenther@suse.de>
7024 PR tree-optimization/102318
7025 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
7026 previous change and do the mode conversion separately from
7027 the sign conversion.
7029 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
7030 Peter Cordes <peter@cordes.ca>
7033 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
7034 (*vec_extract<mode><ssescalarmodelower>_valign): Output
7035 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
7038 2021-09-15 Richard Biener <rguenther@suse.de>
7040 * config.gcc: Remove vax-*-openbsd* configuration.
7042 2021-09-15 Richard Biener <rguenther@suse.de>
7044 * config.gcc: Remove m68k-openbsd.
7046 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
7049 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
7051 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
7053 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
7054 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
7055 (*mma_xxsetaccz): Delete.
7056 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
7057 Use UNSPECV_MMA_XXSETACCZ. Update comment.
7058 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
7060 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
7062 * Makefile.in: Remove variables related to applying no-PIE
7063 to the exes on $build.
7064 * configure: Regenerate.
7065 * configure.ac: Remove configuration related to applying
7066 no-PIE to the exes on $build.
7068 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
7070 * config/arc/arc.md (doloop_end): Add missing mode.
7071 (loop_end): Likewise.
7073 2021-09-14 Jakub Jelinek <jakub@redhat.com>
7075 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
7076 it to recursive calls, for depth above 7 just gimplify or return.
7077 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
7078 __builtin_clear_padding and TARGET_EXPR.
7079 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
7081 2021-09-14 liuhongt <hongtao.liu@intel.com>
7083 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
7085 (_mm_mask_fpclass_sh_mask): Likewise.
7086 (_mm512_mask_fpclass_ph_mask): Likewise.
7087 (_mm512_fpclass_ph_mask): Likewise.
7088 (_mm_getexp_sh): Likewise.
7089 (_mm_mask_getexp_sh): Likewise.
7090 (_mm_maskz_getexp_sh): Likewise.
7091 (_mm512_getexp_ph): Likewise.
7092 (_mm512_mask_getexp_ph): Likewise.
7093 (_mm512_maskz_getexp_ph): Likewise.
7094 (_mm_getexp_round_sh): Likewise.
7095 (_mm_mask_getexp_round_sh): Likewise.
7096 (_mm_maskz_getexp_round_sh): Likewise.
7097 (_mm512_getexp_round_ph): Likewise.
7098 (_mm512_mask_getexp_round_ph): Likewise.
7099 (_mm512_maskz_getexp_round_ph): Likewise.
7100 (_mm_getmant_sh): Likewise.
7101 (_mm_mask_getmant_sh): Likewise.
7102 (_mm_maskz_getmant_sh): Likewise.
7103 (_mm512_getmant_ph): Likewise.
7104 (_mm512_mask_getmant_ph): Likewise.
7105 (_mm512_maskz_getmant_ph): Likewise.
7106 (_mm_getmant_round_sh): Likewise.
7107 (_mm_mask_getmant_round_sh): Likewise.
7108 (_mm_maskz_getmant_round_sh): Likewise.
7109 (_mm512_getmant_round_ph): Likewise.
7110 (_mm512_mask_getmant_round_ph): Likewise.
7111 (_mm512_maskz_getmant_round_ph): Likewise.
7112 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
7114 (_mm_fpclass_ph_mask): Likewise.
7115 (_mm256_mask_fpclass_ph_mask): Likewise.
7116 (_mm256_fpclass_ph_mask): Likewise.
7117 (_mm256_getexp_ph): Likewise.
7118 (_mm256_mask_getexp_ph): Likewise.
7119 (_mm256_maskz_getexp_ph): Likewise.
7120 (_mm_getexp_ph): Likewise.
7121 (_mm_mask_getexp_ph): Likewise.
7122 (_mm_maskz_getexp_ph): Likewise.
7123 (_mm256_getmant_ph): Likewise.
7124 (_mm256_mask_getmant_ph): Likewise.
7125 (_mm256_maskz_getmant_ph): Likewise.
7126 (_mm_getmant_ph): Likewise.
7127 (_mm_mask_getmant_ph): Likewise.
7128 (_mm_maskz_getmant_ph): Likewise.
7129 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7130 * config/i386/i386-builtin.def: Add corresponding new builtins.
7131 * config/i386/i386-expand.c
7132 (ix86_expand_args_builtin): Handle new builtin types.
7133 (ix86_expand_round_builtin): Ditto.
7134 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
7135 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
7136 to support HF vector modes.
7137 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
7139 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
7140 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
7141 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
7142 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7145 2021-09-14 liuhongt <hongtao.liu@intel.com>
7147 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
7149 (_mm512_mask_reduce_ph): Likewise.
7150 (_mm512_maskz_reduce_ph): Likewise.
7151 (_mm512_reduce_round_ph): Likewise.
7152 (_mm512_mask_reduce_round_ph): Likewise.
7153 (_mm512_maskz_reduce_round_ph): Likewise.
7154 (_mm_reduce_sh): Likewise.
7155 (_mm_mask_reduce_sh): Likewise.
7156 (_mm_maskz_reduce_sh): Likewise.
7157 (_mm_reduce_round_sh): Likewise.
7158 (_mm_mask_reduce_round_sh): Likewise.
7159 (_mm_maskz_reduce_round_sh): Likewise.
7160 (_mm512_roundscale_ph): Likewise.
7161 (_mm512_mask_roundscale_ph): Likewise.
7162 (_mm512_maskz_roundscale_ph): Likewise.
7163 (_mm512_roundscale_round_ph): Likewise.
7164 (_mm512_mask_roundscale_round_ph): Likewise.
7165 (_mm512_maskz_roundscale_round_ph): Likewise.
7166 (_mm_roundscale_sh): Likewise.
7167 (_mm_mask_roundscale_sh): Likewise.
7168 (_mm_maskz_roundscale_sh): Likewise.
7169 (_mm_roundscale_round_sh): Likewise.
7170 (_mm_mask_roundscale_round_sh): Likewise.
7171 (_mm_maskz_roundscale_round_sh): Likewise.
7172 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
7174 (_mm_mask_reduce_ph): Likewise.
7175 (_mm_maskz_reduce_ph): Likewise.
7176 (_mm256_reduce_ph): Likewise.
7177 (_mm256_mask_reduce_ph): Likewise.
7178 (_mm256_maskz_reduce_ph): Likewise.
7179 (_mm_roundscale_ph): Likewise.
7180 (_mm_mask_roundscale_ph): Likewise.
7181 (_mm_maskz_roundscale_ph): Likewise.
7182 (_mm256_roundscale_ph): Likewise.
7183 (_mm256_mask_roundscale_ph): Likewise.
7184 (_mm256_maskz_roundscale_ph): Likewise.
7185 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7186 * config/i386/i386-builtin.def: Add corresponding new builtins.
7187 * config/i386/i386-expand.c
7188 (ix86_expand_args_builtin): Handle new builtin types.
7189 (ix86_expand_round_builtin): Ditto.
7190 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
7192 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
7193 ... this, and adjust for round operands.
7194 (reduces<mode><mask_scalar_name>): Likewise, with ...
7195 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
7197 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
7198 Adjust for HF vector modes.
7199 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
7201 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
7203 2021-09-14 liuhongt <hongtao.liu@intel.com>
7205 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
7207 (_mm512_mask_rcp_ph): Likewise.
7208 (_mm512_maskz_rcp_ph): Likewise.
7209 (_mm_rcp_sh): Likewise.
7210 (_mm_mask_rcp_sh): Likewise.
7211 (_mm_maskz_rcp_sh): Likewise.
7212 (_mm512_scalef_ph): Likewise.
7213 (_mm512_mask_scalef_ph): Likewise.
7214 (_mm512_maskz_scalef_ph): Likewise.
7215 (_mm512_scalef_round_ph): Likewise.
7216 (_mm512_mask_scalef_round_ph): Likewise.
7217 (_mm512_maskz_scalef_round_ph): Likewise.
7218 (_mm_scalef_sh): Likewise.
7219 (_mm_mask_scalef_sh): Likewise.
7220 (_mm_maskz_scalef_sh): Likewise.
7221 (_mm_scalef_round_sh): Likewise.
7222 (_mm_mask_scalef_round_sh): Likewise.
7223 (_mm_maskz_scalef_round_sh): Likewise.
7224 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
7226 (_mm256_rcp_ph): Likewise.
7227 (_mm_mask_rcp_ph): Likewise.
7228 (_mm256_mask_rcp_ph): Likewise.
7229 (_mm_maskz_rcp_ph): Likewise.
7230 (_mm256_maskz_rcp_ph): Likewise.
7231 (_mm_scalef_ph): Likewise.
7232 (_mm256_scalef_ph): Likewise.
7233 (_mm_mask_scalef_ph): Likewise.
7234 (_mm256_mask_scalef_ph): Likewise.
7235 (_mm_maskz_scalef_ph): Likewise.
7236 (_mm256_maskz_scalef_ph): Likewise.
7237 * config/i386/i386-builtin.def: Add new builtins.
7238 * config/i386/sse.md (VFH_AVX512VL): New.
7239 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
7240 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
7241 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
7242 Adjust to support HF vector modes.
7243 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
7245 2021-09-14 liuhongt <hongtao.liu@intel.com>
7247 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
7249 (_mm512_mask_sqrt_ph): Likewise.
7250 (_mm512_maskz_sqrt_ph): Likewise.
7251 (_mm512_sqrt_round_ph): Likewise.
7252 (_mm512_mask_sqrt_round_ph): Likewise.
7253 (_mm512_maskz_sqrt_round_ph): Likewise.
7254 (_mm512_rsqrt_ph): Likewise.
7255 (_mm512_mask_rsqrt_ph): Likewise.
7256 (_mm512_maskz_rsqrt_ph): Likewise.
7257 (_mm_rsqrt_sh): Likewise.
7258 (_mm_mask_rsqrt_sh): Likewise.
7259 (_mm_maskz_rsqrt_sh): Likewise.
7260 (_mm_sqrt_sh): Likewise.
7261 (_mm_mask_sqrt_sh): Likewise.
7262 (_mm_maskz_sqrt_sh): Likewise.
7263 (_mm_sqrt_round_sh): Likewise.
7264 (_mm_mask_sqrt_round_sh): Likewise.
7265 (_mm_maskz_sqrt_round_sh): Likewise.
7266 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
7267 (_mm256_sqrt_ph): Likewise.
7268 (_mm_mask_sqrt_ph): Likewise.
7269 (_mm256_mask_sqrt_ph): Likewise.
7270 (_mm_maskz_sqrt_ph): Likewise.
7271 (_mm256_maskz_sqrt_ph): Likewise.
7272 (_mm_rsqrt_ph): Likewise.
7273 (_mm256_rsqrt_ph): Likewise.
7274 (_mm_mask_rsqrt_ph): Likewise.
7275 (_mm256_mask_rsqrt_ph): Likewise.
7276 (_mm_maskz_rsqrt_ph): Likewise.
7277 (_mm256_maskz_rsqrt_ph): Likewise.
7278 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7279 * config/i386/i386-builtin.def: Add corresponding new builtins.
7280 * config/i386/i386-expand.c
7281 (ix86_expand_args_builtin): Handle new builtins.
7282 (ix86_expand_round_builtin): Ditto.
7283 * config/i386/sse.md (VF_AVX512FP16VL): New.
7284 (sqrt<mode>2): Adjust for HF vector modes.
7285 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
7286 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
7288 (<sse>_rsqrt<mode>2<mask_name>): New.
7289 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
7291 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7294 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
7295 'RESERVED_LOCATION_P' locations.
7296 * warning-control.cc (get_nowarn_spec, suppress_warning)
7297 (copy_warning): Likewise.
7299 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7301 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
7302 (typedef key_type_t): ... this. Remove.
7303 (nowarn_map): Document.
7304 * diagnostic-spec.c (nowarn_map): Likewise.
7305 * warning-control.cc (convert_to_key): Evolve functions into...
7306 (get_location): ... these. Adjust all users.
7308 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7310 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
7312 2021-09-13 Jason Merrill <jason@redhat.com>
7314 * params.opt: Add destructive-interference-size and
7315 constructive-interference-size.
7316 * doc/invoke.texi: Document them.
7317 * config/aarch64/aarch64.c (aarch64_override_options_internal):
7319 * config/arm/arm.c (arm_option_override): Set them.
7320 * config/i386/i386-options.c (ix86_option_override_internal):
7323 2021-09-13 Martin Liska <mliska@suse.cz>
7324 H.J. Lu <hjl.tools@gmail.com>
7327 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
7328 for x86-64 micro levels for __builtin_cpu_supports.
7329 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
7330 Add priorities for the micro-arch levels.
7331 (enum processor_features): Add new features.
7332 * common/config/i386/i386-isas.h: Add micro-arch features.
7333 * config/i386/i386-builtins.c (get_builtin_code_for_version):
7334 Support the micro-arch levels by callsing
7335 __builtin_cpu_supports.
7336 * doc/extend.texi: Document that the levels are support by
7337 __builtin_cpu_supports.
7339 2021-09-13 Andrew Pinski <apinski@marvell.com>
7342 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
7344 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
7345 (aarch64_general_gimple_fold_builtin): Likewise.
7347 2021-09-13 Andrew Pinski <apinski@marvell.com>
7349 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
7350 to the Unsupported targets list.
7351 Remove support for m32r-*-linux* and m32rle-*-linux*.
7352 * config/m32r/linux.h: Removed.
7353 * config/m32r/t-linux: Removed.
7355 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7358 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
7359 register index for SVE predicate modes.
7361 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
7363 * tree-ssa-threadbackward.c
7364 (back_threader_profitability::profitable_path_p): Remove FSM
7366 (back_threader_registry::register_path): Same.
7367 * tree-ssa-threadedge.c
7368 (jump_threader::simplify_control_stmt_condition): Same.
7369 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
7370 Add backedge_threads argument.
7371 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
7372 backedge_threads argument.
7373 (back_jt_path_registry::back_jt_path_registry): Same.
7374 (dump_jump_thread_path): Adjust for FSM removal.
7375 (back_jt_path_registry::rewire_first_differing_edge): Same.
7376 (back_jt_path_registry::adjust_paths_after_duplication): Same.
7377 (back_jt_path_registry::update_cfg): Same.
7378 (jt_path_registry::register_jump_thread): Same.
7379 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
7381 (class back_jt_path_registry): Add backedge_threads to
7384 2021-09-13 Martin Liska <mliska@suse.cz>
7387 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
7389 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
7392 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
7393 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
7394 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
7395 TARGET_AVX256_STORE_BY_PIECES instead of
7396 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
7397 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
7398 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
7399 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
7400 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
7401 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
7403 2021-09-13 liuhongt <hongtao.liu@intel.com>
7406 * expmed.c (extract_bit_field_using_extv): Use
7407 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
7409 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
7411 * Makefile.in (OBJS): Add value-pointer-equiv.o.
7412 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
7413 value-pointer-equiv.*.
7414 (ssa_equiv_stack::ssa_equiv_stack): Same.
7415 (ssa_equiv_stack::enter): Same.
7416 (ssa_equiv_stack::leave): Same.
7417 (ssa_equiv_stack::push_replacement): Same.
7418 (ssa_equiv_stack::get_replacement): Same.
7419 (is_pointer_ssa): Same.
7420 (class pointer_equiv_analyzer): Same.
7421 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
7422 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
7423 (pointer_equiv_analyzer::set_global_equiv): Same.
7424 (pointer_equiv_analyzer::set_cond_equiv): Same.
7425 (pointer_equiv_analyzer::get_equiv): Same.
7426 (pointer_equiv_analyzer::enter): Same.
7427 (pointer_equiv_analyzer::leave): Same.
7428 (pointer_equiv_analyzer::get_equiv_expr): Same.
7429 (pta_valueize): Same.
7430 (pointer_equiv_analyzer::visit_stmt): Same.
7431 (pointer_equiv_analyzer::visit_edge): Same.
7432 (hybrid_folder::value_of_expr): Same.
7433 (hybrid_folder::value_on_edge): Same.
7434 * value-pointer-equiv.cc: New file.
7435 * value-pointer-equiv.h: New file.
7437 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
7440 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
7441 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
7443 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
7446 * config/arm/arm.md (movmisaligndi): New define_expand.
7447 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
7449 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
7452 * emit-rtl.c (gen_highpart): Use adjust_address to handle
7453 MEM rather than calling simplify_gen_subreg.
7455 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
7457 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
7458 define into a block.
7460 2021-09-13 Richard Biener <rguenther@suse.de>
7462 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
7463 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
7464 change based on DARWIN_PREFER_DWARF not being defined.
7466 2021-09-13 Richard Biener <rguenther@suse.de>
7468 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
7469 to inherit from elfos.h
7471 2021-09-13 Richard Biener <rguenther@suse.de>
7473 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
7475 2021-09-13 Richard Biener <rguenther@suse.de>
7477 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
7478 override, pick up DWARF2_DEBUG define from elfos.h
7480 2021-09-13 Richard Biener <rguenther@suse.de>
7482 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
7485 2021-09-13 Richard Biener <rguenther@suse.de>
7487 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
7490 2021-09-13 Richard Biener <rguenther@suse.de>
7492 * config/i386/cygming.h: Always default to DWARF2 debugging.
7493 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
7495 * doc/install.texi: Document binutils 2.16 as minimum
7496 requirement for mingw.
7498 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
7500 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
7501 nstmts, nloads and extra_ctor_cost.
7502 (rs6000_density_test): Add load density related heuristics. Do
7503 extra costing on vector construction statements if need.
7504 (rs6000_init_cost): Init new members.
7505 (rs6000_update_target_cost_per_stmt): New function.
7506 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
7507 rs6000_update_target_cost_per_stmt and call it.
7509 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
7511 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
7512 (rs6000_init_cost): Adjust.
7514 2021-09-13 liuhongt <hongtao.liu@intel.com>
7516 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
7517 (UNSPEC_XORSIGN): Ditto.
7519 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
7521 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
7522 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
7525 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
7527 * tree-ssa-threadbackward.c (class back_threader_registry): Use
7528 back_jt_path_registry.
7529 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
7530 fwd_jt_path_registry.
7531 * tree-ssa-threadedge.h (class jump_threader): Same..
7532 * tree-ssa-threadupdate.c
7533 (jump_thread_path_registry::jump_thread_path_registry): Rename...
7534 (jt_path_registry::jt_path_registry): ...to this.
7535 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
7536 (jt_path_registry::~jt_path_registry): ...this.
7537 (fwd_jt_path_registry::fwd_jt_path_registry): New.
7538 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
7539 (jump_thread_path_registry::allocate_thread_edge): Rename...
7540 (jt_path_registry::allocate_thread_edge): ...to this.
7541 (jump_thread_path_registry::allocate_thread_path): Rename...
7542 (jt_path_registry::allocate_thread_path): ...to this.
7543 (jump_thread_path_registry::lookup_redirection_data): Rename...
7544 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
7545 (jump_thread_path_registry::thread_block_1): Rename...
7546 (fwd_jt_path_registry::thread_block_1): ...to this.
7547 (jump_thread_path_registry::thread_block): Rename...
7548 (fwd_jt_path_registry::thread_block): ...to this.
7549 (jt_path_registry::thread_through_loop_header): Rename...
7550 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
7551 (jump_thread_path_registry::mark_threaded_blocks): Rename...
7552 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
7553 (jump_thread_path_registry::debug_path): Rename...
7554 (jt_path_registry::debug_path): ...to this.
7555 (jump_thread_path_registry::dump): Rename...
7556 (jt_path_registry::debug): ...to this.
7557 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
7558 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
7559 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
7560 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
7561 (jump_thread_path_registry::duplicate_thread_path): Rename...
7562 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
7563 drop ill-formed candidates.
7564 (jump_thread_path_registry::remove_jump_threads_including): Rename...
7565 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
7566 (jt_path_registry::thread_through_all_blocks): New.
7567 (back_jt_path_registry::update_cfg): New.
7568 (fwd_jt_path_registry::update_cfg): New.
7569 (jump_thread_path_registry::register_jump_thread): Rename...
7570 (jt_path_registry::register_jump_thread): ...to this.
7571 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
7573 (class jt_path_registry): ...here.
7574 (class fwd_jt_path_registry): New.
7575 (class back_jt_path_registry): New.
7577 2021-09-10 liuhongt <hongtao.liu@intel.com>
7580 2021-09-01 liuhongt <hongtao.liu@intel.com>
7582 * emit-rtl.c (validate_subreg): Get rid of all float-int
7585 2021-09-10 Jakub Jelinek <jakub@redhat.com>
7587 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
7588 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
7589 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
7590 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
7591 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
7592 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
7593 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
7594 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
7596 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
7597 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
7599 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
7601 * tree.h (OMP_ATOMIC_WEAK): Define.
7602 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
7603 fail memory order being encoded in the same enum and also print
7604 fail clause if present.
7605 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
7606 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
7607 handle pre_p == NULL case as a test mode that only returns value
7608 but doesn't change gimplify nor change anything otherwise, adjust
7609 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
7610 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
7611 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
7613 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
7614 COND_EXPR rhs. Set weak flag on gimple load/store for
7616 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
7617 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
7619 (expand_omp_atomic_cas): New function.
7620 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
7622 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
7623 using expand_omp_atomic_cas.
7625 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
7626 Michael Matz <matz@suse.de>
7628 * tree-pass.h (PROP_loop_opts_done): New.
7629 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
7630 Intersect with global range.
7631 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
7632 * tree-ssa-threadbackward.c
7633 (back_threader_profitability::profitable_path_p): Disable
7634 threading through latches until after loop optimizations have run.
7636 2021-09-10 David Faust <david.faust@oracle.com>
7638 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
7641 2021-09-10 David Faust <david.faust@oracle.com>
7643 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
7644 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
7645 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
7646 (bpf_expand_cbranch): New function.
7647 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
7649 (zero_extendsidi2): Only use mov32 instruction if it is available.
7650 (SIM mode iterator): Conditionalize support for SI mode.
7651 (JM mode iterator): New.
7652 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
7653 (*branch_on_di): Update name, use new JM iterator.
7654 * config/bpf/bpf.opt: (mjmpext): New option.
7658 (bpf_isa): New enum.
7660 2021-09-10 David Faust <david.faust@oracle.com>
7662 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
7663 for register-to-register extensions.
7664 (zero_extendqidi2): Likewise.
7666 2021-09-10 Richard Biener <rguenther@suse.de>
7668 PR middle-end/102273
7669 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
7671 2021-09-10 Richard Biener <rguenther@suse.de>
7673 PR middle-end/102269
7674 * gimplify.c (is_var_need_auto_init): Empty types do not need
7677 2021-09-10 Richard Biener <rguenther@suse.de>
7679 * configure.ac (--with-stabs): Remove.
7680 * configure: Regenerate.
7681 * doc/install.texi: Remove --with-stabs documentation.
7683 2021-09-10 liuhongt <hongtao.liu@intel.com>
7685 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
7687 (_mm512_mask_cmp_ph_mask): Likewise.
7688 (_mm512_cmp_round_ph_mask): Likewise.
7689 (_mm512_mask_cmp_round_ph_mask): Likewise.
7690 (_mm_cmp_sh_mask): Likewise.
7691 (_mm_mask_cmp_sh_mask): Likewise.
7692 (_mm_cmp_round_sh_mask): Likewise.
7693 (_mm_mask_cmp_round_sh_mask): Likewise.
7694 (_mm_comieq_sh): Likewise.
7695 (_mm_comilt_sh): Likewise.
7696 (_mm_comile_sh): Likewise.
7697 (_mm_comigt_sh): Likewise.
7698 (_mm_comige_sh): Likewise.
7699 (_mm_comineq_sh): Likewise.
7700 (_mm_ucomieq_sh): Likewise.
7701 (_mm_ucomilt_sh): Likewise.
7702 (_mm_ucomile_sh): Likewise.
7703 (_mm_ucomigt_sh): Likewise.
7704 (_mm_ucomige_sh): Likewise.
7705 (_mm_ucomineq_sh): Likewise.
7706 (_mm_comi_round_sh): Likewise.
7707 (_mm_comi_sh): Likewise.
7708 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
7709 (_mm_mask_cmp_ph_mask): Likewise.
7710 (_mm256_cmp_ph_mask): Likewise.
7711 (_mm256_mask_cmp_ph_mask): Likewise.
7712 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7713 * config/i386/i386-builtin.def: Add corresponding new builtins.
7714 * config/i386/i386-expand.c
7715 (ix86_expand_args_builtin): Handle new builtin types.
7716 (ix86_expand_round_builtin): Ditto.
7717 * config/i386/i386.md (ssevecmode): Add HF mode.
7718 (MODEFH): New mode iterator.
7719 * config/i386/sse.md
7720 (V48H_AVX512VL): New mode iterator to support HF vector modes.
7721 Ajdust corresponding description.
7722 (ssecmpintprefix): New.
7723 (VI12_AVX512VL): Adjust to support HF vector modes.
7724 (cmp_imm_predicate): Likewise.
7725 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
7727 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
7728 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
7729 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
7731 2021-09-10 liuhongt <hongtao.liu@intel.com>
7733 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
7734 (_mm512_mask_max_ph): Likewise.
7735 (_mm512_maskz_max_ph): Likewise.
7736 (_mm512_min_ph): Likewise.
7737 (_mm512_mask_min_ph): Likewise.
7738 (_mm512_maskz_min_ph): Likewise.
7739 (_mm512_max_round_ph): Likewise.
7740 (_mm512_mask_max_round_ph): Likewise.
7741 (_mm512_maskz_max_round_ph): Likewise.
7742 (_mm512_min_round_ph): Likewise.
7743 (_mm512_mask_min_round_ph): Likewise.
7744 (_mm512_maskz_min_round_ph): Likewise.
7745 (_mm_max_sh): Likewise.
7746 (_mm_mask_max_sh): Likewise.
7747 (_mm_maskz_max_sh): Likewise.
7748 (_mm_min_sh): Likewise.
7749 (_mm_mask_min_sh): Likewise.
7750 (_mm_maskz_min_sh): Likewise.
7751 (_mm_max_round_sh): Likewise.
7752 (_mm_mask_max_round_sh): Likewise.
7753 (_mm_maskz_max_round_sh): Likewise.
7754 (_mm_min_round_sh): Likewise.
7755 (_mm_mask_min_round_sh): Likewise.
7756 (_mm_maskz_min_round_sh): Likewise.
7757 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
7758 (_mm256_max_ph): Likewise.
7759 (_mm_mask_max_ph): Likewise.
7760 (_mm256_mask_max_ph): Likewise.
7761 (_mm_maskz_max_ph): Likewise.
7762 (_mm256_maskz_max_ph): Likewise.
7763 (_mm_min_ph): Likewise.
7764 (_mm256_min_ph): Likewise.
7765 (_mm_mask_min_ph): Likewise.
7766 (_mm256_mask_min_ph): Likewise.
7767 (_mm_maskz_min_ph): Likewise.
7768 (_mm256_maskz_min_ph): Likewise.
7769 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7770 * config/i386/i386-builtin.def: Add corresponding new builtins.
7771 * config/i386/i386-expand.c
7772 (ix86_expand_args_builtin): Handle new builtin types.
7773 * config/i386/sse.md
7774 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
7775 support HF vector modes.
7776 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
7777 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
7779 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
7781 * config/i386/subst.md (round_saeonly_mode512bit_condition):
7782 Adjust for HF vector modes.
7784 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
7786 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
7787 (_mm_mask_add_sh): Likewise.
7788 (_mm_maskz_add_sh): Likewise.
7789 (_mm_sub_sh): Likewise.
7790 (_mm_mask_sub_sh): Likewise.
7791 (_mm_maskz_sub_sh): Likewise.
7792 (_mm_mul_sh): Likewise.
7793 (_mm_mask_mul_sh): Likewise.
7794 (_mm_maskz_mul_sh): Likewise.
7795 (_mm_div_sh): Likewise.
7796 (_mm_mask_div_sh): Likewise.
7797 (_mm_maskz_div_sh): Likewise.
7798 (_mm_add_round_sh): Likewise.
7799 (_mm_mask_add_round_sh): Likewise.
7800 (_mm_maskz_add_round_sh): Likewise.
7801 (_mm_sub_round_sh): Likewise.
7802 (_mm_mask_sub_round_sh): Likewise.
7803 (_mm_maskz_sub_round_sh): Likewise.
7804 (_mm_mul_round_sh): Likewise.
7805 (_mm_mask_mul_round_sh): Likewise.
7806 (_mm_maskz_mul_round_sh): Likewise.
7807 (_mm_div_round_sh): Likewise.
7808 (_mm_mask_div_round_sh): Likewise.
7809 (_mm_maskz_div_round_sh): Likewise.
7810 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7811 * config/i386/i386-builtin.def: Add corresponding new builtins.
7812 * config/i386/i386-expand.c
7813 (ix86_expand_round_builtin): Handle new builtins.
7814 * config/i386/sse.md (VF_128): Change description.
7815 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
7816 Adjust to support HF vector modes.
7817 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
7820 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
7822 * config/i386/i386-expand.c
7823 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
7824 * config/i386/i386.c
7825 (ix86_preferred_simd_mode): Handle HF mode.
7826 * config/i386/sse.md (V_256H): New mode iterator.
7827 (avx_vextractf128<mode>): Use it.
7828 (VEC_INIT_MODE): Align vector HFmode condition to vector
7829 HImodes since there're no real HF instruction used.
7830 (VEC_INIT_HALF_MODE): Ditto.
7832 (VIHF_AVX512BW): Ditto.
7833 (*vec_extracthf): Ditto.
7834 (VEC_EXTRACT_MODE): Ditto.
7836 2021-09-10 Richard Biener <rguenther@suse.de>
7839 * config/dbx.h: Remove.
7840 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
7841 * config/lynx.h: Likewise.
7843 2021-09-10 liuhongt <hongtao.liu@intel.com>
7845 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
7846 right into ANDNOT + AND + IOR, using paradoxical subregs.
7847 (ix86_split_copysign_const): Remove.
7848 (ix86_split_copysign_var): Ditto.
7849 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
7850 (ix86_split_copysign_var): Ditto.
7851 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
7852 (@copysign<mode>3_var): Ditto.
7854 2021-09-09 qing zhao <qing.zhao@oracle.com>
7856 * builtins.c (expand_builtin_memset): Make external visible.
7857 * builtins.h (expand_builtin_memset): Declare extern.
7858 * common.opt (ftrivial-auto-var-init=): New option.
7859 * doc/extend.texi: Document the uninitialized attribute.
7860 * doc/invoke.texi: Document -ftrivial-auto-var-init.
7861 * flag-types.h (enum auto_init_type): New enumerated type
7863 * gimple-fold.c (clear_padding_type): Add one new parameter.
7864 (clear_padding_union): Likewise.
7865 (clear_padding_emit_loop): Likewise.
7866 (clear_type_padding_in_mask): Likewise.
7867 (gimple_fold_builtin_clear_padding): Handle this new parameter.
7868 * gimplify.c (gimple_add_init_for_auto_var): New function.
7869 (gimple_add_padding_init_for_auto_var): New function.
7870 (is_var_need_auto_init): New function.
7871 (gimplify_decl_expr): Add initialization to automatic variables per
7873 (gimplify_call_expr): Add one new parameter for call to
7874 __builtin_clear_padding.
7875 (gimplify_init_constructor): Add padding initialization in the end.
7876 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
7877 (expand_DEFERRED_INIT): New function.
7878 * internal-fn.def (DEFERRED_INIT): New internal function.
7879 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
7880 * tree-sra.c (generate_subtree_deferred_init): New function.
7881 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
7882 calls to .DEFERRED_INIT.
7883 (sra_modify_deferred_init): New function.
7884 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
7885 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
7886 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
7888 (check_defs): Likewise.
7889 (warn_uninitialized_vars): Likewise.
7890 * tree-ssa.c (ssa_undefined_value_p): Likewise.
7891 * tree.c (build_common_builtin_nodes): Build tree node for
7892 BUILT_IN_CLEAR_PADDING when needed.
7894 2021-09-09 Richard Biener <rguenther@suse.de>
7896 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
7899 2021-09-09 Richard Biener <rguenther@suse.de>
7901 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
7902 DOM walk from get_loop_body_in_dom_order using a worklist
7905 2021-09-09 liuhongt <hongtao.liu@intel.com>
7907 * config.gcc: Add avx512fp16vlintrin.h.
7908 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
7909 (_mm512_mask_add_ph): Likewise.
7910 (_mm512_maskz_add_ph): Likewise.
7911 (_mm512_sub_ph): Likewise.
7912 (_mm512_mask_sub_ph): Likewise.
7913 (_mm512_maskz_sub_ph): Likewise.
7914 (_mm512_mul_ph): Likewise.
7915 (_mm512_mask_mul_ph): Likewise.
7916 (_mm512_maskz_mul_ph): Likewise.
7917 (_mm512_div_ph): Likewise.
7918 (_mm512_mask_div_ph): Likewise.
7919 (_mm512_maskz_div_ph): Likewise.
7920 (_mm512_add_round_ph): Likewise.
7921 (_mm512_mask_add_round_ph): Likewise.
7922 (_mm512_maskz_add_round_ph): Likewise.
7923 (_mm512_sub_round_ph): Likewise.
7924 (_mm512_mask_sub_round_ph): Likewise.
7925 (_mm512_maskz_sub_round_ph): Likewise.
7926 (_mm512_mul_round_ph): Likewise.
7927 (_mm512_mask_mul_round_ph): Likewise.
7928 (_mm512_maskz_mul_round_ph): Likewise.
7929 (_mm512_div_round_ph): Likewise.
7930 (_mm512_mask_div_round_ph): Likewise.
7931 (_mm512_maskz_div_round_ph): Likewise.
7932 * config/i386/avx512fp16vlintrin.h: New header.
7933 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
7934 Add new builtin types.
7935 * config/i386/i386-builtin.def: Add corresponding builtins.
7936 * config/i386/i386-expand.c
7937 (ix86_expand_args_builtin): Handle new builtin types.
7938 (ix86_expand_round_builtin): Likewise.
7939 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
7940 * config/i386/sse.md (VFH): New mode_iterator.
7942 (avx512fmaskmode): Add HF vector modes.
7943 (avx512fmaskhalfmode): Likewise.
7944 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
7946 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
7947 (mul<mode>3<mask_name><round_name>): Likewise.
7948 (*mul<mode>3<mask_name><round_name>): Likewise.
7949 (div<mode>3): Likewise.
7950 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
7951 * config/i386/subst.md (SUBST_V): Add HF vector modes.
7952 (SUBST_A): Likewise.
7953 (round_mode512bit_condition): Adjust for V32HFmode.
7955 2021-09-09 liuhongt <hongtao.liu@intel.com>
7958 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
7959 (reduc_plus_scal_v4sf): .. this, New define_expand.
7960 (reduc_plus_scal_v2df): .. and this, New define_expand.
7962 2021-09-09 liuhongt <hongtao.liu@intel.com>
7965 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
7968 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
7971 * doc/trouble.texi (Copy Assignment): Fix description of
7972 behaviour and fix code in example.
7974 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
7977 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
7978 r11 instead of r12 for restoring CR.
7980 2021-09-08 Jakub Jelinek <jakub@redhat.com>
7981 liuhongt <hongtao.liu@intel.com>
7984 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
7985 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
7986 into AND with mask and XOR, using paradoxical subregs.
7987 (ix86_split_xorsign): Remove.
7988 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
7990 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
7992 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
7994 2021-09-08 Jakub Jelinek <jakub@redhat.com>
7997 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
7998 operands[2], emit abs<mode>2 instead.
7999 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
8000 first alternative even for avx, add another alternative with
8001 =&Yv <- 0, Yv, Yvm constraints.
8002 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
8003 to op1, emit vpandn instead.
8005 2021-09-08 liuhongt <hongtao.liu@intel.com>
8007 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
8008 (_mm256_set_ph): Likewise.
8009 (_mm512_set_ph): Likewise.
8010 (_mm_setr_ph): Likewise.
8011 (_mm256_setr_ph): Likewise.
8012 (_mm512_setr_ph): Likewise.
8013 (_mm_set1_ph): Likewise.
8014 (_mm256_set1_ph): Likewise.
8015 (_mm512_set1_ph): Likewise.
8016 (_mm_setzero_ph): Likewise.
8017 (_mm256_setzero_ph): Likewise.
8018 (_mm512_setzero_ph): Likewise.
8019 (_mm_set_sh): Likewise.
8020 (_mm_load_sh): Likewise.
8021 (_mm_store_sh): Likewise.
8022 * config/i386/i386-builtin-types.def (V8HF): New type.
8023 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
8024 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
8025 Support vector HFmodes.
8026 (ix86_expand_vector_init_one_nonzero): Likewise.
8027 (ix86_expand_vector_init_one_var): Likewise.
8028 (ix86_expand_vector_init_interleave): Likewise.
8029 (ix86_expand_vector_init_general): Likewise.
8030 (ix86_expand_vector_set): Likewise.
8031 (ix86_expand_vector_extract): Likewise.
8032 (ix86_expand_vector_init_concat): Likewise.
8033 (ix86_expand_sse_movcc): Handle vector HFmodes.
8034 (ix86_expand_vector_set_var): Ditto.
8035 * config/i386/i386-modes.def: Add HF vector modes in comment.
8036 * config/i386/i386.c (classify_argument): Add HF vector modes.
8037 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
8038 (ix86_vector_mode_supported_p): Likewise.
8039 (ix86_set_reg_reg_cost): Handle vector HFmode.
8040 (ix86_get_ssemov): Handle vector HFmode.
8041 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
8043 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
8045 (function_arg_advance_32): Ditto.
8046 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
8047 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
8048 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
8049 (VALID_SSE2_REG_VHF_MODE): New.
8050 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
8051 (SSE_REG_MODE_P): Add vector HFmode.
8052 * config/i386/i386.md (mode): Add HF vector modes.
8053 (MODE_SIZE): Likewise.
8054 (ssemodesuffix): Add ph suffix for HF vector modes.
8055 * config/i386/sse.md (VFH_128): New mode iterator.
8056 (VMOVE): Adjust for HF vector modes.
8058 (V_256_512): Likewise.
8060 (avx512fmaskmode): Likewise.
8061 (shuffletype): Likewise.
8062 (sseinsnmode): Likewise.
8063 (ssedoublevecmode): Likewise.
8064 (ssehalfvecmode): Likewise.
8065 (ssehalfvecmodelower): Likewise.
8066 (ssePScmode): Likewise.
8067 (ssescalarmode): Likewise.
8068 (ssescalarmodelower): Likewise.
8069 (sseintprefix): Likewise.
8071 (bcstscalarsuff): Likewise.
8072 (xtg_mode): Likewise.
8073 (VI12HF_AVX512VL): New mode_iterator.
8074 (VF_AVX512FP16): Likewise.
8076 (VIHF_256): Likewise.
8077 (VIHF_AVX512BW): Likewise.
8078 (V16_256): Likewise.
8079 (V32_512): Likewise.
8080 (sseintmodesuffix): New mode_attr.
8081 (sse): Add scalar and vector HFmodes.
8082 (ssescalarmode): Add vector HFmode mapping.
8083 (ssescalarmodesuffix): Add sh suffix for HFmode.
8084 (*<sse>_vm<insn><mode>3): Use VFH_128.
8085 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
8086 (*ieee_<ieee_maxmin><mode>3): Likewise.
8087 (<avx512>_blendm<mode>): New define_insn.
8088 (vec_setv8hf): New define_expand.
8089 (vec_set<mode>_0): New define_insn for HF vector set.
8090 (*avx512fp16_movsh): Likewise.
8091 (avx512fp16_movsh): Likewise.
8092 (vec_extract_lo_v32hi): Rename to ...
8093 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
8095 (vec_extract_hi_v32hi): Likewise.
8096 (vec_extract_hi_<mode>): Likewise.
8097 (vec_extract_lo_v16hi): Likewise.
8098 (vec_extract_lo_<mode>): Likewise.
8099 (vec_extract_hi_v16hi): Likewise.
8100 (vec_extract_hi_<mode>): Likewise.
8101 (vec_set_hi_v16hi): Likewise.
8102 (vec_set_hi_<mode>): Likewise.
8103 (vec_set_lo_v16hi): Likewise.
8104 (vec_set_lo_<mode>): Likewise.
8105 (*vec_extract<mode>_0): New define_insn_and_split for HF
8107 (*vec_extracthf): New define_insn.
8108 (VEC_EXTRACT_MODE): Add HF vector modes.
8109 (PINSR_MODE): Add V8HF.
8110 (sse2p4_1): Likewise.
8111 (pinsr_evex_isa): Likewise.
8112 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
8113 insert for V8HFmode.
8114 (pbroadcast_evex_isa): Add HF vector modes.
8115 (AVX2_VEC_DUP_MODE): Likewise.
8116 (VEC_INIT_MODE): Likewise.
8117 (VEC_INIT_HALF_MODE): Likewise.
8118 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
8120 (avx2_pbroadcast<mode>_1): Likewise.
8121 (<avx512>_vec_dup<mode>_1): Likewise.
8122 (<avx512>_vec_dup<mode><mask_name>): Likewise.
8123 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
8126 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
8127 H.J. Lu <hongjiu.lu@intel.com>
8128 Liu Hongtao <hongtao.liu@intel.com>
8129 Wang Hongyu <hongyu.wang@intel.com>
8130 Xu Dianhong <dianhong.xu@intel.com>
8132 * common/config/i386/cpuinfo.h (get_available_features):
8133 Detect FEATURE_AVX512FP16.
8134 * common/config/i386/i386-common.c
8135 (OPTION_MASK_ISA_AVX512FP16_SET,
8136 OPTION_MASK_ISA_AVX512FP16_UNSET,
8137 OPTION_MASK_ISA2_AVX512FP16_SET,
8138 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
8139 (OPTION_MASK_ISA2_AVX512BW_UNSET,
8140 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
8141 (ix86_handle_option): Handle -mavx512fp16.
8142 * common/config/i386/i386-cpuinfo.h (enum processor_features):
8143 Add FEATURE_AVX512FP16.
8144 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
8145 * config.gcc: Add avx512fp16intrin.h.
8146 * config/i386/avx512fp16intrin.h: New intrinsic header.
8147 * config/i386/cpuid.h: Add bit_AVX512FP16.
8148 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
8149 * config/i386/i386-builtins.c: Support _Float16 type for i386
8151 (ix86_register_float16_builtin_type): New function.
8152 (ix86_float16_type_node): New.
8153 * config/i386/i386-c.c (ix86_target_macros_internal): Define
8155 * config/i386/i386-expand.c (ix86_expand_branch): Support
8157 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
8158 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
8159 (ix86_expand_fp_movcc): Ditto.
8160 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
8161 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
8162 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
8163 * config/i386/i386.c (ix86_get_ssemov): Use
8164 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
8165 (ix86_get_excess_precision): Use
8166 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
8168 (sse_store_index): Use SFmode cost for HFmode cost.
8169 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
8170 GPR cost for HFmode.
8171 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
8172 (ix86_mangle_type): Add manlging for _Float16 type.
8173 (inline_secondary_memory_needed): No memory is needed for
8174 16bit movement between gpr and sse reg under
8176 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
8177 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
8178 (ix86_division_cost): Ditto.
8179 (ix86_rtx_costs): Ditto.
8180 (ix86_add_stmt_cost): Ditto.
8181 (ix86_optab_supported_p): Ditto.
8182 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
8183 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
8184 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
8185 * config/i386/i386.md (mode): Add HFmode.
8186 (MODE_SIZE): Add HFmode.
8187 (isa): Add avx512fp16.
8188 (enabled): Handle avx512fp16.
8189 (ssemodesuffix): Add sh suffix for HFmode.
8190 (comm): Add mult, div.
8191 (plusminusmultdiv): New code iterator.
8192 (insn): Add mult, div.
8193 (*movhf_internal): Adjust for avx512fp16 instruction.
8194 (*movhi_internal): Ditto.
8195 (*cmpi<unord>hf): New define_insn for HFmode.
8196 (*ieee_s<ieee_maxmin>hf3): Likewise.
8197 (extendhf<mode>2): Likewise.
8198 (trunc<mode>hf2): Likewise.
8199 (float<floatunssuffix><mode>hf2): Likewise.
8200 (*<insn>hf): Likewise.
8201 (cbranchhf4): New expander.
8202 (movhfcc): Likewise.
8203 (<insn>hf3): Likewise.
8206 * config/i386/i386.opt: Add mavx512fp16.
8207 * config/i386/immintrin.h: Include avx512fp16intrin.h.
8208 * doc/invoke.texi: Add mavx512fp16.
8209 * doc/extend.texi: Add avx512fp16 Usage Notes.
8211 2021-09-08 liuhongt <hongtao.liu@intel.com>
8213 * common.opt: Support -fexcess-precision=16.
8214 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
8215 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
8216 EXCESS_PRECISION_TYPE_FLOAT16.
8217 * config/arm/arm.c (arm_excess_precision): Ditto.
8218 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
8219 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
8220 when EXCESS_PRECISION_TYPE_FLOAT16.
8221 * config/s390/s390.c (s390_excess_precision): Ditto.
8222 * coretypes.h (enum excess_precision_type): Add
8223 EXCESS_PRECISION_TYPE_FLOAT16.
8224 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
8225 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
8226 * doc/extend.texi (Half-Precision): Document
8227 -fexcess-precision=16.
8228 * flag-types.h (enum excess_precision): Add
8229 EXCESS_PRECISION_FLOAT16.
8230 * target.def (excess_precision): Update document.
8231 * tree.c (excess_precision_type): Set excess_precision_type to
8232 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
8234 2021-09-08 liuhongt <hongtao.liu@intel.com>
8236 * doc/extend.texi: (@node Floating Types): Adjust the wording.
8237 (@node Half-Precision): Ditto.
8239 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
8242 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
8243 'CONST_INT_P (src)' to the condition of the block that tries to
8244 eliminate literal when loading integer contant.
8246 2021-09-07 David Faust <david.faust@oracle.com>
8248 * doc/extend.texi (BPF Type Attributes) New node.
8249 Document new preserve_access_index attribute.
8250 Document new preserve_access_index builtin.
8251 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
8253 2021-09-07 David Faust <david.faust@oracle.com>
8255 * config/bpf/bpf.c: Adjust includes.
8256 (bpf_handle_preserve_access_index_attribute): New function.
8257 (bpf_attribute_table): Use it here.
8258 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
8259 (bpf_option_override): Handle "-mco-re" option.
8260 (bpf_asm_init_sections): New.
8261 (TARGET_ASM_INIT_SECTIONS): Redefine.
8262 (bpf_file_end): New.
8263 (TARGET_ASM_FILE_END): Redefine.
8264 (bpf_init_builtins): Add "__builtin_preserve_access_index".
8265 (bpf_core_compute, bpf_core_get_index): New.
8266 (is_attr_preserve_access): New.
8267 (bpf_expand_builtin): Handle new builtins.
8268 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
8269 (bpf_core_walk): New.
8270 (bpf_resolve_overloaded_builtin): New.
8271 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
8273 (pass_bpf_core_attr): New RTL pass.
8274 * config/bpf/bpf-passes.def: New file.
8275 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
8276 * config/bpf/coreout.c: New file.
8277 * config/bpf/coreout.h: Likewise.
8278 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
8279 (coreout.o): New rule.
8280 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
8281 * config.gcc (bpf): Add coreout.h to extra_headers.
8282 Add coreout.o to extra_objs.
8283 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
8285 2021-09-07 David Faust <david.faust@oracle.com>
8287 * btfout.c (get_btf_id): Function is no longer static.
8288 * ctfc.h: Expose it here.
8290 2021-09-07 David Faust <david.faust@oracle.com>
8292 * ctfc.c (ctf_lookup_tree_type): New function.
8295 2021-09-07 David Faust <david.faust@oracle.com>
8297 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
8298 * ctfc.h: Analogous change.
8300 2021-09-07 David Faust <david.faust@oracle.com>
8302 * dwarf2out.c (lookup_type_die): Function is no longer static.
8303 * dwarf2out.h: Expose it here.
8305 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8307 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
8308 (ctf_debug_early_finish): New definition.
8309 (ctf_debug_finish): Likewise.
8310 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
8311 (ctf_debug_early_finish): New declaration.
8312 (ctf_debug_finish): Likewise.
8313 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
8314 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
8316 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8318 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
8319 support when compiling for CO-RE.
8320 * config/bpf/bpf.opt: Add new command line option -mco-re.
8322 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8324 * flag-types.h (enum debug_info_type): Add new enum
8325 DINFO_TYPE_BTF_WITH_CORE.
8326 (BTF_WITH_CORE_DEBUG): New bitmask.
8327 * flags.h (btf_with_core_debuginfo_p): New declaration.
8328 * opts.c (btf_with_core_debuginfo_p): New definition.
8330 2021-09-07 Jason Merrill <jason@redhat.com>
8332 * tree.h (error_operand_p): Change to inline function.
8334 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
8336 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
8337 (empty_block_with_phis_p): ...this.
8338 (potentially_threadable_block): Same.
8339 (jump_threader::thread_through_normal_block): Same.
8341 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
8344 * dwarf2out.c (mark_base_types): New overloaded function.
8345 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
8346 as the compilation unit, and call move_marked_base_types afterward.
8348 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
8351 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
8353 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
8355 2021-09-07 Richard Biener <rguenther@suse.de>
8357 PR tree-optimization/102226
8358 * tree-vect-loop.c (vect_transform_cycle_phi): Record
8359 the converted value for the epilogue PHI use.
8361 2021-09-07 Martin Liska <mliska@suse.cz>
8363 PR gcov-profile/80223
8364 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
8365 options, do not inline when no_profile_instrument_function
8366 attributes are different in early inliner. It's fine to inline
8367 it after PGO instrumentation.
8369 2021-09-07 Richard Biener <rguenther@suse.de>
8371 PR tree-optimization/101555
8372 * tree-ssa-pre.c (translate_vuse_through_block): Do not
8373 perform an alias walk to determine the validity of the
8374 mem at the start of the block which is already guaranteed
8375 by means of prune_clobbered_mems.
8376 (phi_translate_1): Pass edge to translate_vuse_through_block.
8378 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
8381 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
8382 (remainder<mode>3): Likewise.
8384 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
8386 * config/mips/mips.c (mips_file_start): add .module for
8389 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
8391 * wide-int.cc (wi::clz): Reorder tests to ensure the result
8392 is zero for all negative values.
8394 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
8396 * doc/invoke.texi (-foffload-options): Fix @opindex.
8398 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
8401 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
8402 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
8403 source alternative for AVX.
8405 2021-09-06 liuhongt <hongtao.liu@intel.com>
8407 PR middle-end/102182
8408 * optabs.c (expand_fix): Add from1 to avoid from being
8411 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
8413 * dwarf2out.c (modified_type_die): Deal with all array types earlier
8414 and use local variable consistently throughout the function.
8416 2021-09-06 Jakub Jelinek <jakub@redhat.com>
8418 PR tree-optimization/102207
8419 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
8420 were promoted from signed to wider unsigned type.
8422 2021-09-06 Andrew Pinski <apinski@marvell.com>
8424 PR tree-optimization/63184
8425 * match.pd: Add simplification of pointer_diff of two pointer_plus
8426 with addr_expr in the first operand of each pointer_plus.
8427 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
8428 in the first operand of each pointer_plus.
8430 2021-09-06 Richard Biener <rguenther@suse.de>
8432 PR tree-optimization/102176
8433 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
8435 (vect_bb_slp_scalar_cost): Use the computed set of
8436 vectorized scalar stmts instead of relying on the out-of-date
8437 and not accurate PURE_SLP_STMT.
8438 (vect_bb_vectorization_profitable_p): Compute the set
8439 of vectorized scalar stmts.
8441 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
8443 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
8444 GIMPLE_COND special casing.
8445 (path_range_query::range_defined_in_block): Use range_of_stmt
8446 instead of calling fold_range directly.
8448 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
8450 * gimple-range-path.cc (path_range_query::range_of_expr): Set
8451 m_undefined_path when appropriate.
8452 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
8453 (path_range_query::unreachable_path_p): New.
8454 (path_range_query::precompute_ranges): Set m_undefined_path.
8455 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
8456 (path_range_query::internal_range_of_expr): New.
8457 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
8458 Use unreachable_path_p.
8460 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
8462 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8463 Remove argument and call find_taken_edge.
8464 (back_threader::resolve_phi): Do not calculate taken edge before
8465 calling maybe_register_path.
8466 (back_threader::find_paths_to_names): Same.
8468 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
8470 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
8471 * config/h8300/testcompare.md (store_c): Update name, use new
8473 (store_neg_c, store_shifted_c): New patterns.
8475 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
8478 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
8479 instead of r12 for CR save, in all cases.
8481 2021-09-03 Andrew Pinski <apinski@marvell.com>
8483 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
8484 Handle error_mark_node as the type of the type_decl.
8486 2021-09-03 Andrew Pinski <apinski@marvell.com>
8488 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
8490 (aarch64_simd_types): Likewise.
8491 (aarch64_simd_intOI_type_node): Likewise.
8492 (aarch64_simd_intCI_type_node): Likewise.
8493 (aarch64_simd_intXI_type_node): Likewise.
8494 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
8495 (aarch64_fp16_ptr_type_node): Likewise.
8496 (aarch64_bf16_type_node): Likewise.
8497 (aarch64_bf16_ptr_type_node): Likewise.
8499 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8501 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
8503 (minus_op1_op2_relation_effect): ...here.
8504 (class operator_pointer_diff): New.
8505 (operator_pointer_diff::op1_op2_relation_effect): Call
8506 minus_op1_op2_relation_effect.
8507 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
8509 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8511 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
8512 Add may_peel_loop_headers.
8513 (back_threader_registry::thread_through_all_blocks): Same.
8514 (try_thread_blocks): Pass may_peel_loop_headers argument.
8515 (pass_early_thread_jumps::execute): Same.
8517 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8519 * tree-ssa-threadedge.c (has_phis_p): New.
8520 (forwarder_block_p): New.
8521 (potentially_threadable_block): Call forwarder_block_p.
8522 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
8523 (jump_threader::thread_through_normal_block): Call
8526 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8528 * tree-ssa-threadbackward.c (back_threader::dump): New.
8529 (back_threader::debug): New.
8530 (back_threader_profitability::profitable_path_p): Dump blocks
8531 even if we are bailing early.
8533 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8535 * tree-ssa-threadupdate.c (cancel_thread): New.
8536 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
8537 (jump_thread_path_registry::mark_threaded_blocks): Same.
8538 (jump_thread_path_registry::register_jump_thread): Same.
8540 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8542 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
8543 which objects are available.
8544 (jt_state::pop): Same.
8545 (jt_state::register_equiv): Same.
8546 (jt_state::register_equivs_on_edge): Same.
8548 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8550 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
8551 Move pop until after a thread is registered.
8553 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8555 * tree-ssa-threadupdate.c (debug): New.
8557 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8559 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
8560 (push_dump_file::~push_dump_file): New.
8561 (dump_ranger): Change dump_file temporarily while dumping
8563 * gimple-range-trace.h (class push_dump_file): New.
8565 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8567 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
8568 (dump_ranger): Dump function name.
8570 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8572 * gimple-range-path.cc (path_range_query::range_defined_in_block):
8573 Adjust for non-null.
8574 (path_range_query::adjust_for_non_null_uses): New.
8575 (path_range_query::precompute_ranges): Call
8576 adjust_for_non_null_uses.
8577 * gimple-range-path.h: Add m_non_null and
8578 adjust_for_non_null_uses.
8580 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8582 * gimple-range-path.cc (path_range_query::dump): Dump path
8584 (path_range_query::precompute_ranges): Dump entire path.
8586 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8588 * value-relation.cc (relation_oracle::debug): New.
8589 * value-relation.h (relation_oracle::debug): New.
8591 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8593 * tree-ssa-loop-ch.c: Remove unnecessary include file.
8595 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8597 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
8598 Skip statements with no defining BB.
8599 * gimple-range-path.cc (path_range_query::range_defined_in_block):
8600 Do not get confused by statements with no defining BB.
8602 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
8604 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
8605 gimple_range_adjustment. Add support for constants.
8606 (adjust_realpart_expr): New.
8607 (gimple_range_adjustment): Move IMAGPART_EXPR code to
8608 adjust_imagpart_expr.
8609 * range-op.cc (integral_table::integral_table): Add entry for
8612 2021-09-03 Jakub Jelinek <jakub@redhat.com>
8614 * omp-expand.c (expand_omp_atomic_pipeline): Use
8615 IFN_ATOMIC_COMPARE_EXCHANGE instead of
8616 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
8619 2021-09-03 Jakub Jelinek <jakub@redhat.com>
8622 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
8623 that is false if DECL_BIT_FIELD.
8624 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
8625 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
8626 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
8627 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
8628 of DECL_FIELD_ABI_IGNORED.
8629 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
8630 SET_DECL_FIELD_ABI_IGNORED instead of writing to
8631 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
8632 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
8633 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
8634 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
8636 2021-09-03 liuhongt <hongtao.liu@intel.com>
8639 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
8640 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
8641 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
8643 2021-09-02 Martin Sebor <msebor@redhat.com>
8645 PR tree-optimization/17506
8647 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
8649 2021-09-02 Richard Biener <rguenther@suse.de>
8651 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
8652 fix for PR78185 and continue processing when leaving
8655 2021-09-02 Jakub Jelinek <jakub@redhat.com>
8657 PR tree-optimization/99591
8658 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
8661 2021-09-02 Richard Biener <rguenther@suse.de>
8664 2021-09-02 Richard Biener <rguenther@suse.de>
8666 PR tree-optimization/102155
8667 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
8668 over a part of the RPO array and do not recurse here.
8669 Dump blocks marked as always executed.
8670 (fill_always_executed_in): Walk over the RPO array and
8671 process loops whose header we run into.
8672 (loop_invariant_motion_in_fun): Compute the first RPO
8673 using rev_post_order_and_mark_dfs_back_seme in iteration
8674 order and pass that to fill_always_executed_in.
8676 2021-09-02 liuhongt <hongtao.liu@intel.com>
8678 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
8679 * config/i386/i386.c (enum x86_64_reg_class): Add
8681 (merge_classes): Handle X86_64_SSEHF_CLASS.
8682 (examine_argument): Ditto.
8683 (construct_container): Ditto.
8684 (classify_argument): Ditto, and set HFmode/HCmode to
8686 (function_value_32): Return _FLoat16/Complex Float16 by
8688 (function_value_64): Return _Float16/Complex Float16 by SSE
8690 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
8691 (ix86_secondary_reload): Require gpr as intermediate register
8692 to store _Float16 from sse register when sse4 is not
8694 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
8696 (ix86_scalar_mode_supported_p): Ditto.
8697 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
8698 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
8699 (VALID_INT_MODE_P): Add HFmode and HCmode.
8700 * config/i386/i386.md (*pushhf_rex64): New define_insn.
8702 (*movhf_internal): Ditto.
8703 * doc/extend.texi (Half-Precision Floating Point): Documemt
8706 2021-09-02 Richard Biener <rguenther@suse.de>
8708 PR tree-optimization/102155
8709 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
8710 over a part of the RPO array and do not recurse here.
8711 Dump blocks marked as always executed.
8712 (fill_always_executed_in): Walk over the RPO array and
8713 process loops whose header we run into.
8714 (loop_invariant_motion_in_fun): Compute the first RPO
8715 using rev_post_order_and_mark_dfs_back_seme in iteration
8716 order and pass that to fill_always_executed_in.
8718 2021-09-02 YunQiang Su <syq@debian.org>
8721 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
8723 * config/mips/mips.c (mips_module_isa_name): New.
8724 mips_file_start: add .module mipsREV to all asm output
8726 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
8728 PR tree-optimization/102152
8729 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
8730 comparison to a scalar comparison before calling
8731 update_stmt_if_modified.
8733 2021-09-01 Andrew Pinski <apinski@marvell.com>
8736 * config/aarch64/aarch64.c (aarch64_expand_setmem):
8737 Check STRICT_ALIGNMENT before creating an overlapping
8740 2021-09-01 Martin Sebor <msebor@redhat.com>
8742 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
8743 (check_access): Pass additional argument.
8744 (check_memop_access): Remove template and make a member function.
8745 (maybe_check_dealloc_call): Make a pass_waccess member function.
8746 (class pass_waccess): Add, rename, and remove members.
8747 (pass_waccess::pass_waccess): Adjust to name change.
8748 (pass_waccess::~pass_waccess): Same.
8749 (check_alloca): Make a member function.
8750 (check_alloc_size_call): Same.
8751 (check_strcat): Same.
8752 (check_strncat): Same.
8753 (check_stxcpy): Same.
8754 (check_stxncpy): Same.
8755 (check_strncmp): Same.
8756 (maybe_warn_rdwr_sizes): Rename...
8757 (pass_waccess::maybe_check_access_sizes): ...to this.
8758 (pass_waccess::check_call): Adjust to name changes.
8759 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
8761 (pass_waccess::execute): Adjust to name changes.
8762 * gimple-ssa-warn-access.h (check_memop_access): Remove.
8763 * pointer-query.cc (access_ref::phi): Handle null pointer.
8764 (access_ref::inform_access): Same.
8765 (pointer_query::put_ref): Modify a cached value, not a copy of it.
8766 (pointer_query::dump): New function.
8767 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
8769 * pointer-query.h (pointer_query::dump): Declare.
8770 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
8771 (dump_strlen_info): Use function query.
8772 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
8774 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
8776 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
8778 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
8780 * doc/extend.texi: Document unavailable attribute.
8781 * print-tree.c (print_node): Handle unavailable attribute.
8782 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
8783 * tree.c (error_unavailable_use): New.
8784 * tree.h (TREE_UNAVAILABLE): New.
8785 (error_unavailable_use): New.
8787 2021-09-01 Jakub Jelinek <jakub@redhat.com>
8789 PR tree-optimization/102124
8790 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
8791 MINUS_EXPR, if itype is unsigned with smaller precision than type,
8792 add an extra cast to signed variant of itype to ensure sign-extension.
8794 2021-09-01 Martin Liska <mliska@suse.cz>
8796 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
8797 edges and rather use colors for TRUE and FALSE edges.
8799 2021-09-01 Richard Biener <rguenther@suse.de>
8801 PR tree-optimization/93491
8802 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
8803 after processing the stmt itself. Do not consider
8804 pure functions possibly not returning. Properly avoid
8805 adding possibly trapping calls to EXP_GEN when there's
8806 a preceeding possibly not returning call.
8807 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
8810 2021-09-01 Richard Biener <rguenther@suse.de>
8812 PR tree-optimization/102139
8813 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
8814 type to record a std::pair of the stmt-info and the innermost
8816 (dr_vec_info::group): New member.
8817 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
8818 (vect_compute_data_ref_alignment): Verify the recorded
8819 base alignment can be used.
8820 (data_ref_pair): Remove.
8821 (dr_group_sort_cmp): Adjust.
8822 (vect_analyze_data_ref_accesses): Store the group-ID in the
8823 dr_vec_info and operate on a vector of dr_vec_infos.
8825 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
8827 * read-md.c (md_reader::handle_enum): support value assignation.
8828 * doc/md.texi: record define_c_enum value assignation support.
8830 2021-09-01 Jakub Jelinek <jakub@redhat.com>
8832 PR tree-optimization/102141
8833 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
8834 argument. If false, emit stmts after gsi instead of before, and
8836 (bswap_replace): Adjust callers. When converting output of bswap,
8837 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
8840 2021-09-01 liuhongt <hongtao.liu@intel.com>
8842 * emit-rtl.c (validate_subreg): Get rid of all float-int
8845 2021-09-01 liuhongt <hongtao.liu@intel.com>
8848 2021-08-30 liuhongt <hongtao.liu@intel.com>
8850 * expmed.c (extract_bit_field_1): Make sure we're playing with
8851 integral modes before call extract_integral_bit_field.
8852 (extract_integral_bit_field): Add a parameter of type
8853 scalar_int_mode which corresponds to of tmode.
8854 And call extract_and_convert_fixed_bit_field instead of
8855 extract_fixed_bit_field and convert_extracted_bit_field.
8856 (extract_and_convert_fixed_bit_field): New function, it's a
8857 combination of extract_fixed_bit_field and
8858 convert_extracted_bit_field.
8860 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
8862 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
8864 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
8866 * omp-general.h (omp_is_reference): Rename to...
8867 (omp_privatize_by_reference): ... this. Adjust all users...
8868 * omp-general.c: ... here, ...
8869 * gimplify.c: ... here, ...
8870 * omp-expand.c: ... here, ...
8871 * omp-low.c: ... here.
8873 2021-08-31 Martin Sebor <msebor@redhat.com>
8875 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
8876 pointer element for equality to zero, not that of the cotaining
8879 2021-08-31 Martin Sebor <msebor@redhat.com>
8881 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
8883 2021-08-31 Martin Sebor <msebor@redhat.com>
8885 * function.h (function): Add comments.
8886 (get_range_query): Same. Add attribute returns nonnull.
8888 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
8890 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
8891 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
8892 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
8894 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
8896 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
8898 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
8899 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
8900 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
8901 (m32c-*-elf*, m32c-*-rtems*): Likewise.
8902 * config/cris/cris.h (DBX_NO_XREFS): Remove.
8903 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
8904 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
8905 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
8906 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
8907 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
8909 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
8911 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
8912 allowed on target constructs and only with particular other clauses.
8913 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
8915 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
8916 structs when 'ancestor' is used.
8917 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
8919 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
8920 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
8922 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
8924 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
8925 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
8927 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
8928 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
8929 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
8930 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
8931 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
8932 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
8933 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
8934 would be paradoxical.
8936 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
8938 * combine.c (combine_simplify_rtx): Avoid converting an explicit
8939 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
8940 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
8942 2021-08-31 Richard Biener <rguenther@suse.de>
8944 PR tree-optimization/102142
8945 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
8946 condition under which to unset the visited flag.
8948 2021-08-31 Richard Biener <rguenther@suse.de>
8950 PR middle-end/102129
8951 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
8952 possibly trapping expressions across calls.
8954 2021-08-31 Jakub Jelinek <jakub@redhat.com>
8956 PR tree-optimization/102134
8957 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
8958 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
8961 2021-08-31 Andrew Pinski <apinski@marvell.com>
8964 * collect-utils.c (setup_signals): New declaration.
8965 * collect-utils.h (setup_signals): New function.
8966 * collect2.c (handler): Delete.
8967 (main): Instead of manually setting up the signals,
8968 just call setup_signals.
8969 * lto-wrapper.c (main): Likewise.
8971 2021-08-31 Andrew Pinski <apinski@marvell.com>
8974 * config/i386/i386-protos.h (x86_output_aligned_bss):
8975 Change align argument to unsigned type.
8976 (x86_elf_aligned_decl_common): Likewise.
8977 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
8978 (x86_output_aligned_bss): Likewise.
8980 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
8982 * config/mips/mips.c (mips_module_isa_name): New.
8983 mips_file_start: add .module mipsREV to all asm output
8985 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
8987 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
8988 use enum instead of int for 'isa' member.
8989 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
8990 config/mips/netbsd.h: replace hardcoded numbers with enum.
8992 2021-08-31 liuhongt <hongtao.liu@intel.com>
8994 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
8995 define_split to define_insn_and_split.
8996 (*avx2_eq<mode>3): Removed.
8997 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
8998 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
8999 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
9001 (*avx2_gt<mode>3): Removed.
9002 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
9003 define_insn to define_expand, and adjust pattern.
9004 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
9006 2021-08-30 David Malcolm <dmalcolm@redhat.com>
9009 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
9011 2021-08-30 Jason Merrill <jason@redhat.com>
9013 * doc/invoke.texi: Document -Wmissing-requires.
9015 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9017 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
9018 TARGET_EXTRA_BUILTINS guard.
9020 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9022 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
9023 initialization of V2DI_type_node and unsigned_V2DI_type_node.
9025 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9027 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
9028 decl when new_builtins_are_live.
9029 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
9032 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
9034 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
9035 OPTION_MASK_P10_FUSION_2STORE.
9036 (POWERPC_MASKS): Likewise.
9037 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
9038 store fusion for Power10.
9039 (is_fusable_store): New.
9040 (power10_sched_reorder): Likewise.
9041 (rs6000_sched_reorder): Do Power10 specific reordering.
9042 (rs6000_sched_reorder2): Likewise.
9043 * config/rs6000/rs6000.opt: Add new option.
9045 2021-08-30 Richard Biener <rguenther@suse.de>
9047 PR tree-optimization/102128
9048 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
9049 Move scanning for if-converted scalar code to the caller
9050 and instead delay clearing the visited flag for profitable
9052 (vect_slp_region): Cost all subgraphs before scheduling.
9053 For if-converted BB vectorization scan for scalar COND_EXPRs
9054 and do not vectorize if any found and the cost model is
9057 2021-08-30 Richard Biener <rguenther@suse.de>
9059 * common.opt (fexceptions): Mark
9060 EnabledBy(fnon-call-exceptions).
9061 * doc/invoke.texi (fnon-call-exceptions): Document this
9062 enables -fexceptions.
9064 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
9066 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
9067 is not already defined.
9069 2021-08-30 liuhongt <hongtao.liu@intel.com>
9071 * expmed.c (extract_bit_field_1): Make sure we're playing with
9072 integral modes before call extract_integral_bit_field.
9073 (extract_integral_bit_field): Add a parameter of type
9074 scalar_int_mode which corresponds to of tmode.
9075 And call extract_and_convert_fixed_bit_field instead of
9076 extract_fixed_bit_field and convert_extracted_bit_field.
9077 (extract_and_convert_fixed_bit_field): New function, it's a
9078 combination of extract_fixed_bit_field and
9079 convert_extracted_bit_field.
9081 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
9083 * config/darwin.c (darwin_libc_has_function): Do not run
9084 the checks for x86 or modern Darwin. Make sure that there
9085 is a value set for darwin_macosx_version_min before testing.
9087 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
9089 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
9091 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
9093 * ipa-modref-tree.h (modref_access_node::merge): Break out
9094 logic combining offsets and logic merging ranges to ...
9095 (modref_access_node::combined_offsets): ... here
9096 (modref_access_node::update2): ... here
9097 (modref_access_node::closer_pair_p): New member function.
9098 (modref_access_node::forced_merge): New member function.
9099 (modre_ref_node::insert): Do merging when table is full.
9101 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
9104 * config.gcc: MIPS: use N64 ABI by default if the triple end
9105 with -gnuabi64, which is used by Debian since 2013.
9107 2021-08-28 Alexandre Oliva <oliva@adacore.com>
9109 * ipa-modref.c (analyze_function): Skip debug stmts.
9110 * tree-inline.c (estimate_num_insn): Consider builtins even
9111 without a cgraph_node.
9113 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
9115 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
9116 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
9117 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
9118 * config/h8300/h8300.md (eqne): New code iterator.
9119 (geultu, geultu_to_c): Similarly.
9120 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
9121 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
9122 (cmp<mode>_c): New pattern
9124 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
9126 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
9128 (dom_opt_dom_walker::optimize_stmt): Use it.
9130 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
9132 * config/darwin.c (finalize_ctors): Add a section-start linker-
9134 (finalize_dtors): Likewise.
9135 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
9137 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
9139 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
9140 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
9141 old initialization logic when new builtins are enabled.
9142 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
9143 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
9144 (write_init_file): Likewise.
9146 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
9148 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
9149 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
9150 (amdgcn-* | gcn-*) Likewise.
9152 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
9154 * symbol-summary.h: Added missing template keyword.
9156 2021-08-27 Richard Biener <rguenther@suse.de>
9158 PR tree-optimization/45178
9159 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
9160 infinite loops without exit do not mark control dependent
9161 edges of the latch necessary.
9163 2021-08-27 konglin1 <lingling.kong@intel.com>
9166 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
9168 (<avx512>scattersi<mode>): Likewise.
9169 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
9170 (*avx512f_scatterdi<VI48F:mode>): Likewise
9172 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
9174 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
9175 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
9176 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
9177 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
9178 P10_BUILTIN_PEXTD on Power10.
9180 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
9182 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
9183 signedness for some Power10 bifs.
9185 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
9188 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
9189 computed alignment if the entire struct has attribute packed.
9191 2021-08-27 liuhongt <hongtao.liu@intel.com>
9195 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
9196 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
9197 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
9198 IX86_BUILTIN_SHUFPS256.
9199 (ix86_masked_all_ones): New function.
9201 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
9203 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
9204 (conditional moves with memory inputs splitters): Ditto.
9205 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
9207 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9209 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
9210 search after merging.
9212 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9214 * config/rs6000/rs6000-overload.def: Add remaining overloads.
9216 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9218 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
9220 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9222 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
9223 crypto, and htm stanzas.
9225 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9227 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
9229 2021-08-26 Martin Sebor <msebor@redhat.com>
9231 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
9232 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
9233 (warn_uninitialized_vars): Same. Reduce visibility of locals.
9234 (warn_uninitialized_phi): Same.
9236 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
9238 * tree-ssa-ccp.c (get_individual_bits): Helper function to
9239 extract the individual bits from a widest_int constant (mask).
9240 (gray_code_bit_flips): New read-only table for effiently
9241 enumerating permutations/combinations of bits.
9242 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
9243 by unknown counts that are guaranteed less than the target
9244 precision and four or fewer unknown bits by enumeration.
9245 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
9246 enumeration under the same conditions. Handle remaining
9247 shifts as a mask based upon the minimum possible shift value.
9249 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
9250 Richard Biener <rguenther@suse.de>
9252 * match.pd (shift transformations): Remove a redundant
9253 !POINTER_TYPE_P check.
9255 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
9258 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
9259 Set all_regs to true in the call to replace_rtx.
9261 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9263 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
9264 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
9265 hiting --param modref-max-refs limit.
9266 (modref_tree:insert_base): Be smarter when hitting
9267 --param modref-max-bases limit. Add new parameter REF.
9268 (modref_tree:insert): Update.
9269 (modref_tree:merge): Update.
9270 * ipa-modref.c (read_modref_records): Update.
9272 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9274 * params.opt: (modref-max-adjustments): Add full stop.
9276 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9278 * ipa-modref-tree.h (modref_ref_node::verify): New member
9280 (modref_ref_node::insert): Use it.
9281 (modref_ref_node::try_mere_with): Fix off by one error.
9283 2021-08-26 Martin Liska <mliska@suse.cz>
9284 Stefan Kneifel <stefan.kneifel@bluewin.ch>
9286 * cgraph.h (create_version_clone_with_body): Add new parameter.
9287 * cgraphclones.c: Likewise.
9288 * multiple_target.c (create_dispatcher_calls): Do not use
9290 (create_target_clone): Likewise here.
9292 2021-08-26 Jonathan Yong <10walls@gmail.com>
9294 * doc/extend.texi: Add note about reserved priorities
9295 to the constructor attribute.
9297 2021-08-25 Martin Sebor <msebor@redhat.com>
9299 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
9300 range table header alone.
9301 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
9303 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
9305 * doc/invoke.texi: Document --param modref-max-adjustments.
9306 * ipa-modref-tree.c (test_insert_search_collapse): Update.
9307 (test_merge): Update.
9308 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
9309 (modref_access_node::operator==): Fix handling of access ranges.
9310 (modref_access_node::contains): Constify parameter; handle also
9311 mismatched parm offsets.
9312 (modref_access_node::update): New function.
9313 (modref_access_node::merge): New function.
9314 (unspecified_modref_access_node): Update constructor.
9315 (modref_ref_node::insert_access): Add record_adjustments parameter;
9317 (modref_ref_node::try_merge_with): New private function.
9318 (modref_tree::insert): New record_adjustments parameter.
9319 (modref_tree::merge): New record_adjustments parameter.
9320 (modref_tree::copy_from): Update.
9321 * ipa-modref.c (dump_access): Dump adjustments field.
9322 (get_access): Update constructor.
9323 (record_access): Update call of insert.
9324 (record_access_lto): Update call of insert.
9325 (merge_call_side_effects): Add record_adjustments parameter.
9326 (get_access_for_fnspec): Update.
9327 (process_fnspec): Update.
9328 (analyze_call): Update.
9329 (analyze_function): Update.
9330 (read_modref_records): Update.
9331 (ipa_merge_modref_summary_after_inlining): Update.
9332 (propagate_unknown_call): Update.
9333 (modref_propagate_in_scc): Update.
9334 * params.opt (param-max-modref-adjustments=): New.
9336 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
9338 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
9340 (xxspltiw_v4si): Use vecperm type attribute.
9341 (xxspltiw_v4si_inst): Use vecperm type attribute.
9342 (xxspltiw_v4sf_inst): Likewise.
9343 (xxspltidp_v2df): Use vecperm type attribute. Use
9344 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
9345 (xxspltidp_v2df_inst): Likewise.
9346 (xxsplti32dx_v4si): Use vecperm type attribute.
9347 (xxsplti32dx_v4si_inst): Likewise.
9348 (xxsplti32dx_v4sf_inst): Likewise.
9349 (xxblend_<mode>): Likewise.
9350 (xxpermx): Likewise.
9351 (xxpermx_inst): Likewise.
9354 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
9357 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
9358 * diagnostic.c (diagnostic_initialize_input_context): New function.
9359 * diagnostic.h (diagnostic_initialize_input_context): Declare.
9360 * input.c (default_charset_callback): New function.
9361 (file_cache::initialize_input_context): New function.
9362 (file_cache_slot::create): Added ability to convert the input
9363 according to the input context.
9364 (file_cache::file_cache): Initialize the new input context.
9365 (class file_cache_slot): Added new m_alloc_offset member.
9366 (file_cache_slot::file_cache_slot): Initialize the new member.
9367 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
9368 (file_cache_slot::maybe_grow): Likewise.
9369 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
9370 (file_cache_slot::get_next_line): Likewise.
9371 * input.h (class file_cache): Added input context member.
9373 2021-08-25 Richard Biener <rguenther@suse.de>
9375 PR tree-optimization/102046
9376 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
9377 update ->any_pattern when swapping operands.
9379 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
9382 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
9383 (ix86_decompose_address): Remove retval check for ASHIFT,
9384 allow non-canonical zero extend if AND mask covers ASHIFT
9386 (ix86_legitimate_address_p): Adjust condition for decompose.
9387 (ix86_rtx_costs): Adjust cost for lea with non-canonical
9389 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
9391 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
9393 PR tree-optimization/101145
9394 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
9396 (number_of_iterations_lt): Invoke above function.
9397 (adjust_cond_for_loop_until_wrap):
9398 Merge to number_of_iterations_until_wrap.
9399 (number_of_iterations_cond): Update invokes for
9400 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
9402 2021-08-25 konglin1 <lingling.kong@intel.com>
9405 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
9407 (_mm512_mask_fpclass_ps_mask): Ditto.
9409 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
9411 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
9412 (vec_unpacku_hi_v8hi): Likewise.
9413 (vec_unpacku_lo_v16qi): Likewise.
9414 (vec_unpacku_lo_v8hi): Likewise.
9415 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
9416 (vec_unpacku_lo_<VP_small_lc>): Likewise.
9418 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
9420 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
9421 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
9422 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
9423 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
9425 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9427 PR middle-end/102031
9428 * simplify-rtx.c (simplify_truncation): When comparing precisions
9429 use "subreg_prec" variable, not "subreg_mode".
9431 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9433 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
9436 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9438 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
9439 various pointer type nodes.
9440 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
9441 values for various pointer types.
9442 (ptr_V16QI_type_node): New macro.
9443 (ptr_V1TI_type_node): New macro.
9444 (ptr_V2DI_type_node): New macro.
9445 (ptr_V2DF_type_node): New macro.
9446 (ptr_V4SI_type_node): New macro.
9447 (ptr_V4SF_type_node): New macro.
9448 (ptr_V8HI_type_node): New macro.
9449 (ptr_unsigned_V16QI_type_node): New macro.
9450 (ptr_unsigned_V1TI_type_node): New macro.
9451 (ptr_unsigned_V8HI_type_node): New macro.
9452 (ptr_unsigned_V4SI_type_node): New macro.
9453 (ptr_unsigned_V2DI_type_node): New macro.
9454 (ptr_bool_V16QI_type_node): New macro.
9455 (ptr_bool_V8HI_type_node): New macro.
9456 (ptr_bool_V4SI_type_node): New macro.
9457 (ptr_bool_V2DI_type_node): New macro.
9458 (ptr_bool_V1TI_type_node): New macro.
9459 (ptr_pixel_type_node): New macro.
9460 (ptr_intQI_type_node): New macro.
9461 (ptr_uintQI_type_node): New macro.
9462 (ptr_intHI_type_node): New macro.
9463 (ptr_uintHI_type_node): New macro.
9464 (ptr_intSI_type_node): New macro.
9465 (ptr_uintSI_type_node): New macro.
9466 (ptr_intDI_type_node): New macro.
9467 (ptr_uintDI_type_node): New macro.
9468 (ptr_intTI_type_node): New macro.
9469 (ptr_uintTI_type_node): New macro.
9470 (ptr_long_integer_type_node): New macro.
9471 (ptr_long_unsigned_type_node): New macro.
9472 (ptr_float_type_node): New macro.
9473 (ptr_double_type_node): New macro.
9474 (ptr_long_double_type_node): New macro.
9475 (ptr_dfloat64_type_node): New macro.
9476 (ptr_dfloat128_type_node): New macro.
9477 (ptr_ieee128_type_node): New macro.
9478 (ptr_ibm128_type_node): New macro.
9479 (ptr_vector_pair_type_node): New macro.
9480 (ptr_vector_quad_type_node): New macro.
9481 (ptr_long_long_integer_type_node): New macro.
9482 (ptr_long_long_unsigned_type_node): New macro.
9484 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9486 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
9487 and power9-64 stanzas.
9489 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9490 Tom de Vries <tdevries@suse.de>
9492 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
9493 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
9494 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
9495 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
9496 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
9497 * config/nvptx/nvptx-c.c: New source file.
9498 (nvptx_cpu_cpp_builtins): Move implementation here.
9500 2021-08-24 Martin Sebor <msebor@redhat.com>
9502 PR middle-end/101600
9503 PR middle-end/101977
9504 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
9505 the phrasing of a warning.
9506 (check_access): Use the remaining size after subtracting any offset
9507 rather than the whole object size.
9508 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
9509 clear for any nonnull PHI argument.
9510 (compute_objsize): Clear argument.
9512 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9514 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
9516 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9518 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
9521 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
9523 * value-relation.cc (rr_transitive_table): New.
9524 (relation_transitive): New.
9525 (value_relation::swap): Remove.
9526 (value_relation::apply_transitive): New.
9527 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
9528 (relation_oracle::register_relation): Call register_transitives.
9529 (relation_oracle::register_transitives): New.
9530 * value-relation.h (relation_oracle): Add new temporary bitmap and
9533 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
9536 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
9537 from integer to a pseudo vector register.
9539 2021-08-24 Richard Biener <rguenther@suse.de>
9541 PR tree-optimization/100089
9542 * tree-vectorizer.h (vect_slp_bb): Rename to ...
9543 (vect_slp_if_converted_bb): ... this and get the original
9544 loop as new argument.
9545 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
9546 pass original loop to vect_slp_if_converted_bb.
9547 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
9548 If orig_loop was passed scan the not vectorized stmts
9549 for COND_EXPRs and force not profitable if found.
9550 (vect_slp_region): Pass down all SLP instances to costing
9551 if orig_loop was specified.
9552 (vect_slp_bbs): Pass through orig_loop.
9553 (vect_slp_bb): Rename to ...
9554 (vect_slp_if_converted_bb): ... this and get the original
9555 loop as new argument.
9556 (vect_slp_function): Adjust.
9558 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
9561 * config/arm/arm.md (attribute arch): Add fix_vlldm.
9562 (arch_enabled): Use it.
9563 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
9564 use when erratum mitigation is needed.
9566 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
9569 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
9570 * doc/invoke.texi (Arm Options): Document it.
9571 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
9572 (ALL_QUIRKS): Add quirk_vlldm.
9573 (cortex-m33): Add quirk_vlldm.
9574 (cortex-m35p, cortex-m55): Likewise.
9575 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
9576 targetting an affected CPU and not explicitly controlled on
9579 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
9581 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
9582 (lazy_load_multiple_insn): Likewise.
9584 2021-08-24 liuhongt <hongtao.liu@intel.com>
9587 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
9588 Enable avx512 embedded broadcast.
9589 (*<avx512>_vternlog<mode>_all): Ditto.
9590 (<avx512>_vternlog<mode>_mask): Ditto.
9592 2021-08-24 liuhongt <hongtao.liu@intel.com>
9595 * config/i386/i386.c (ix86_rtx_costs): Define cost for
9597 * config/i386/i386.h (STRIP_UNARY): New macro.
9598 * config/i386/predicates.md (reg_or_notreg_operand): New
9600 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
9601 (*<avx512>_vternlog<mode>_1): New pre_reload
9602 define_insn_and_split.
9603 (*<avx512>_vternlog<mode>_2): Ditto.
9604 (*<avx512>_vternlog<mode>_3): Ditto.
9605 (any_logic1,any_logic2): New code iterator.
9606 (logic_op): New code attribute.
9607 (ternlogsuffix): Extend to VNxDF and VNxSF.
9609 2021-08-24 Richard Biener <rguenther@suse.de>
9611 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
9612 * params.opt (--param vect-inner-loop-cost-factor): Adjust
9614 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
9615 inner_loop_cost_factor to the minimum of the estimated number
9616 of iterations of the inner loop and vect-inner-loop-cost-factor.
9618 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9619 Richard Biener <rguenther@suse.de>
9621 * config/i386/i386-features.c (compute_convert_gain): Provide
9622 more accurate values for CONST_INT, when optimizing for size.
9623 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
9624 * config/i386/i386.h (COSTS_N_BYTES): to here.
9626 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9627 Jakub Jelinek <jakub@redhat.com>
9629 PR middle-end/102029
9630 * match.pd (shift transformations): Add an additional check for
9631 !POINTER_TYPE_P in the recently added left shift transformation.
9633 2021-08-24 liuhongt <hongtao.liu@intel.com>
9635 PR tree-optimization/100089
9636 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
9637 loop vectorizer when cost model is very-cheap.
9639 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9641 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
9642 asprintf, which is not available on AIX.
9644 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9646 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
9647 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
9649 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9651 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
9652 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
9654 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
9655 (rbtree.o): Likewise.
9656 (rs6000-gen-builtins): Likewise.
9657 (rs6000-builtins.c): Likewise.
9658 (rs6000-builtins.h): Likewise.
9659 (rs6000.o): Add dependency.
9660 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
9661 (rs6000-vecdefines.h): New target.
9662 (rs6000-builtins.o): Likewise.
9663 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
9664 (rs6000-c.o): Likewise.
9666 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
9669 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
9670 Diagnose buffer overrun.
9671 (safe_inc_pos): Fix overrun detection.
9672 (match_identifier): Diagnose buffer overrun.
9673 (match_integer): Likewise.
9674 (match_to_right_bracket): Likewise.
9676 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
9678 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
9679 Improve range compare.
9680 (modref_access_node::contains): New member function.
9681 (modref_access_node::search): Remove.
9682 (modref_access_node::insert): Be smarter about subaccesses.
9684 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
9686 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
9687 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
9688 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
9691 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
9693 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
9695 * config/h8300/jumpcall.md (call, call_value): Restrict to
9696 !SIBLING_CALL_P cases.
9697 (subcall, sibcall_value): New patterns & expanders.
9698 * config/h8300/proepi.md (epilogue): Pass new argument to
9699 h8300_expand_epilogue.
9700 (sibcall_epilogue): New expander.
9701 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
9703 (h8300_ok_for_sibcall_p): New function.
9704 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
9706 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9708 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
9709 Handle case where the operand is already the desired mode.
9711 2021-08-23 Richard Biener <rguenther@suse.de>
9714 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
9715 in addition to has_gimple_body.
9717 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
9719 PR middle-end/101949
9720 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
9723 2021-08-23 Martin Liska <mliska@suse.cz>
9725 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
9728 2021-08-23 Richard Biener <rguenther@suse.de>
9730 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9731 Properly scale the inner loop cost only once.
9733 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9735 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
9736 Provide bounds for unsigned (and signed with non-negative operands)
9737 division and modulus.
9739 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9741 * simplify-rtx.c (simplify_truncation): Generalize simplification
9742 of (truncate:A (subreg:B X)).
9743 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
9744 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
9745 already has the desired machine mode.
9746 (test_scalar_int_ops): Add tests that useless extensions and
9747 truncations are optimized away.
9748 (test_scalar_int_ext_ops): New self-test function to confirm
9749 that truncations of extensions are correctly simplified.
9750 (test_scalar_int_ext_ops2): New self-test function to check
9751 truncations of truncations, extensions of extensions, and
9752 truncations of extensions.
9753 (test_scalar_ops): Call the above two functions with a
9754 representative sampling of integer machine modes.
9756 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
9758 * match.pd (shift transformations): Change the sign of an
9759 LSHIFT_EXPR if it reduces the number of explicit conversions.
9761 2021-08-23 Jakub Jelinek <jakub@redhat.com>
9763 PR tree-optimization/86723
9764 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
9765 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
9766 non-memory permutation of 64-bit src which only has bytes of
9767 0 or [5..8] and n->range is 4.
9768 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
9769 find_bswap_or_nop_finalize caller, support bswap with some bytes
9770 zeroed, as long as at least two bytes are not zeroed.
9771 (bswap_replace): Add mask argument and handle masking of bswap
9773 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
9774 caller, punt if cast64_to_32 or mask is not all ones.
9775 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
9776 caller, for now punt if cast64_to_32.
9778 2021-08-23 Richard Biener <rguenther@suse.de>
9780 PR tree-optimization/79334
9781 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
9782 a type also for COMPONENT_REFs.
9783 (vn_reference_may_trap): Check ARRAY_REF with constant index
9784 against the array domain.
9786 2021-08-23 liuhongt <hongtao.liu@intel.com>
9789 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
9790 TARGET_AVX512BW to condition.
9792 2021-08-23 Jakub Jelinek <jakub@redhat.com>
9795 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
9796 register variables already during early_dwarf if possible.
9798 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
9800 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
9802 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
9805 * config/arm/arm.opt: Fix typo.
9806 * config/arm/t-rmprofile: Fix typo.
9808 2021-08-23 Jakub Jelinek <jakub@redhat.com>
9810 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
9811 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
9812 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
9813 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
9814 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
9815 if either grainsize or num_tasks clause has the strict modifier.
9817 2021-08-23 Martin Liska <mliska@suse.cz>
9819 * dbgcnt.def (DEBUG_COUNTER): New counter.
9820 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
9822 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
9824 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
9826 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
9829 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
9831 * config/mips/mips.c (mips_msa_output_shift_immediate): New
9833 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
9834 vlshr<mode>3): Call it.
9836 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
9837 Martin Liska <mliska@suse.cz>
9839 PR middle-end/101949
9840 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
9843 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
9845 * config/mips/mips.c (mips_function_rodata_section,
9846 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
9848 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
9850 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
9852 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
9854 * configure.ac (thread-local storage support): Remove tls_first_major
9855 and tls_first_minor. Use "$conftest_s" to check support.
9856 * configure: Regenerate.
9858 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
9860 * configure.ac: Fixup formatting.
9862 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
9864 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
9865 * configure.ac: ... update all callers.
9867 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
9870 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
9871 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
9872 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
9873 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
9874 Remove remaining checks for in-tree assembler.
9875 * configure: Regenerate.
9877 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
9879 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
9880 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
9882 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
9883 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
9884 Improve logical shifts by 27 bits for H8/S.
9885 (get_shift_alg): Corresponding changes.
9886 (h8300_option_override): Revert to loops for -Os when profitable.
9888 2021-08-20 Richard Biener <rguenther@suse.de>
9890 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
9892 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
9893 index as group_id when dataref_groups were not computed.
9894 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
9895 we advace to the next BB.
9897 2021-08-20 Jakub Jelinek <jakub@redhat.com>
9899 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
9902 2021-08-20 Martin Liska <mliska@suse.cz>
9904 PR gcov-profile/89961
9905 * gcov.c (make_gcov_file_name): Rewrite using std::string.
9906 (mangle_name): Simplify, do not used the second argument.
9907 (strip_extention): New function.
9908 (get_md5sum): Likewise.
9909 (get_gcov_intermediate_filename): Handle properly -p and -x
9911 (output_gcov_file): Use string type.
9912 (generate_results): Likewise.
9913 (md5sum_to_hex): Remove.
9915 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
9917 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
9918 (UNSPEC_XXSPLTIW): Move to vsx.md.
9919 (UNSPEC_XXSPLTID): Move to vsx.md.
9920 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
9921 (UNSPEC_XXBLEND): Move to vsx.md.
9922 (UNSPEC_XXPERMX): Move to vsx.md.
9923 (VM3): Move to vsx.md.
9924 (VM3_char): Move to vsx.md.
9925 (xxspltiw_v4si): Move to vsx.md.
9926 (xxspltiw_v4sf): Move to vsx.md.
9927 (xxspltiw_v4sf_inst): Move to vsx.md.
9928 (xxspltidp_v2df): Move to vsx.md.
9929 (xxspltidp_v2df_inst): Move to vsx.md.
9930 (xxsplti32dx_v4si_inst): Move to vsx.md.
9931 (xxsplti32dx_v4sf): Move to vsx.md.
9932 (xxsplti32dx_v4sf_inst): Move to vsx.md.
9933 (xxblend_<mode>): Move to vsx.md.
9934 (xxpermx): Move to vsx.md.
9935 (xxpermx_inst): Move to vsx.md.
9936 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
9937 (UNSPEC_XXSPLTIW): Move from altivec.md.
9938 (UNSPEC_XXSPLTID): Move from altivec.md.
9939 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
9940 (UNSPEC_XXBLEND): Move from altivec.md.
9941 (UNSPEC_XXPERMX): Move from altivec.md.
9942 (VM3): Move from altivec.md.
9943 (VM3_char): Move from altivec.md.
9944 (xxspltiw_v4si): Move from altivec.md.
9945 (xxspltiw_v4sf): Move from altivec.md.
9946 (xxspltiw_v4sf_inst): Move from altivec.md.
9947 (xxspltidp_v2df): Move from altivec.md.
9948 (xxspltidp_v2df_inst): Move from altivec.md.
9949 (xxsplti32dx_v4si_inst): Move from altivec.md.
9950 (xxsplti32dx_v4sf): Move from altivec.md.
9951 (xxsplti32dx_v4sf_inst): Move from altivec.md.
9952 (xxblend_<mode>): Move from altivec.md.
9953 (xxpermx): Move from altivec.md.
9954 (xxpermx_inst): Move from altivec.md.
9956 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
9958 * tree-vect-generic.c (expand_vector_operations_1): Use either
9959 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
9960 when constructing scalar splat expressions.
9962 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
9965 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
9966 pointer to __vector_pair *.
9968 2021-08-19 Martin Sebor <msebor@redhat.com>
9970 * gimple-range.cc: Add comments.
9971 * gimple-range.h: Same.
9973 2021-08-19 Martin Sebor <msebor@redhat.com>
9975 PR middle-end/101984
9976 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
9979 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
9981 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
9982 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
9983 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
9985 2021-08-19 Jakub Jelinek <jakub@redhat.com>
9987 PR middle-end/101950
9988 * optabs.c (expand_clrsb_using_clz): New function.
9989 (expand_unop): Use it as another clrsb expansion fallback.
9991 2021-08-19 liuhongt <hongtao.liu@intel.com>
9994 2021-07-28 liuhongt <hongtao.liu@intel.com>
9997 * config/i386/i386.h (processor_costs): Add new member
9999 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
10000 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
10001 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
10002 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
10003 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
10004 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
10005 generic_cost, core_cost): Initialize integer_to_sse same value
10007 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
10008 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10009 Use integer_to_sse instead of sse_op to calculate the cost of
10012 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
10014 * config.gcc: Include rpath.opt for Darwin.
10015 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
10017 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10019 PR bootstrap/101959
10020 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
10023 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
10025 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
10026 pattern and type cast.
10027 (vld4_lane_f32): Use float RTL pattern.
10028 (vld4q_lane_f64): Use float type cast.
10030 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
10032 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
10035 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10037 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
10038 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
10039 (hash_map_tests_c_tests): Call it.
10041 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10043 * ggc.h (enum ggc_collect): New.
10044 (ggc_collect): Use it.
10045 * ggc-page.c: Adjust.
10046 * ggc-common.c: Likewise.
10047 * ggc-tests.c: Likewise.
10048 * read-rtl-function.c: Likewise.
10049 * selftest-run-tests.c: Likewise.
10050 * doc/gty.texi (Invoking the garbage collector): Likewise.
10052 2021-08-18 liuhongt <hongtao.liu@intel.com>
10055 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
10057 * config/i386/sse.md (*sse3_haddv2df3_low): Add
10058 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
10059 (*sse3_hsubv2df3_low): Ditto.
10060 * config/i386/x86-tune.def
10061 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
10063 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10065 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
10066 (gori_compute::compute_operand_range): Add tracing.
10067 (gori_compute::logical_combine): Ditto.
10068 (gori_compute::compute_logical_operands): Ditto.
10069 (gori_compute::compute_operand1_range): Ditto.
10070 (gori_compute::compute_operand2_range): Ditto.
10071 (gori_compute::outgoing_edge_range_p): Ditto.
10072 * gimple-range-gori.h (class gori_compute): Add range_tracer.
10074 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10076 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
10077 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
10078 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
10079 * params.opt (--param=evrp-mode): Adjust options.
10081 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10083 * Makefile.in (OBJS): Add gimple-range-trace.o.
10084 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
10085 * gimple-range-fold.cc: Adjust headers.
10086 * gimple-range-trace.cc: New.
10087 * gimple-range-trace.h: New.
10088 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
10089 (gimple_ranger::range_of_expr): Add tracing.
10090 (gimple_ranger::range_on_entry): Ditto.
10091 (gimple_ranger::range_on_exit): Ditto.
10092 (gimple_ranger::range_on_edge): Ditto.
10093 (gimple_ranger::fold_range_internal): Ditto.
10094 (gimple_ranger::dump_bb): Do not calculate edge range twice.
10095 (trace_ranger::*): Remove.
10096 (enable_ranger): Never create a trace_ranger.
10097 (debug_seed_ranger): Move to gimple-range-trace.cc.
10098 (dump_ranger): Ditto.
10099 (debug_ranger): Ditto.
10100 * gimple-range.h: Include gimple-range-trace.h.
10101 (range_on_entry, range_on_exit): No longer virtual.
10102 (class trace_ranger): Remove.
10103 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
10105 2021-08-17 Martin Sebor <msebor@redhat.com>
10107 PR middle-end/101854
10108 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
10109 in gimple-ssa-warn-access.cc.
10110 * calls.c (alloc_max_size): Move code to check_alloca.
10111 (get_size_range): Move to pointer-query.cc.
10112 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
10113 (get_attr_nonstring_decl): Move to tree.c.
10114 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
10115 (append_attrname): Same.
10116 (maybe_warn_rdwr_sizes): Same.
10117 (initialize_argument_information): Move code to
10118 gimple-ssa-warn-access.cc.
10119 * calls.h (maybe_warn_alloc_args_overflow): Move to
10120 gimple-ssa-warn-access.h.
10121 (get_attr_nonstring_decl): Move to tree.h.
10122 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
10123 (enum size_range_flags): Move to pointer-query.h.
10124 (get_size_range): Same.
10125 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
10126 to avoid Clang -Wunused-function.
10127 (get_size_range): Declare static.
10128 (maybe_emit_free_warning): Rename...
10129 (maybe_check_dealloc_call): ...to this for consistency.
10130 (class pass_waccess): Add members.
10131 (pass_waccess::~pass_waccess): Defined.
10132 (alloc_max_size): Move here from calls.c.
10133 (maybe_warn_alloc_args_overflow): Same.
10134 (check_alloca): New function.
10135 (check_alloc_size_call): New function.
10136 (check_strncat): Handle another warning flag.
10137 (pass_waccess::check_builtin): Handle alloca.
10138 (fntype_argno_type): Move here from calls.c.
10139 (append_attrname): Same.
10140 (maybe_warn_rdwr_sizes): Same.
10141 (pass_waccess::check_call): Define.
10142 (check_nonstring_args): New function.
10143 (pass_waccess::check): Call new member functions.
10144 (pass_waccess::execute): Enable ranger.
10145 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
10146 (maybe_warn_nonstring_arg): Same.
10147 * gimple-ssa-warn-restrict.c: Remove #include.
10148 * pointer-query.cc (get_size_range): Move here from calls.c.
10149 * pointer-query.h (enum size_range_flags): Same.
10150 (get_size_range): Same.
10151 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
10152 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
10154 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
10156 * ggc.h (ggc_collect): Add 'force_collect' parameter.
10157 * ggc-page.c (ggc_collect): Use that one instead of global
10158 'ggc_force_collect'. Adjust all users.
10159 * doc/gty.texi (Invoking the garbage collector): Update.
10160 * ggc-internal.h (ggc_force_collect): Remove.
10161 * ggc-common.c (ggc_force_collect): Likewise.
10162 * selftest.h (forcibly_ggc_collect): Remove.
10163 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
10164 * read-rtl-function.c (test_loading_labels): Adjust.
10165 * selftest-run-tests.c (run_tests): Likewise.
10167 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
10169 * config/darwin.c (darwin_file_end): Reset and reclaim the
10170 section names table at the end of compile.
10172 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
10175 * config.in: Regenerate.
10176 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
10177 (ASM_SPEC): Pass options to disable branch shortening where
10179 * configure: Regenerate.
10180 * configure.ac: Detect versions of 'as' that support the
10181 optimisation which has the bug.
10183 2021-08-17 Richard Biener <rguenther@suse.de>
10185 * optabs-query.c (supports_vec_gather_load_p): Also check
10187 (supports_vec_scatter_store_p): Likewise.
10188 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
10189 back to masked variants if non-masked are not supported.
10190 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
10191 When we need to use masked gather/scatter but do not have
10192 a mask set up a constant true one.
10193 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
10194 non-SSA_NAME masks.
10196 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
10198 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
10199 algorithm as PLUS_EXPR to improve subtraction bit bounds.
10200 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
10202 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
10204 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
10205 calculate the mask-value pair result of a multiplication by an
10207 (bit_value_binop) [MULT_EXPR]: Call it from here for
10208 multiplications by (sparse) non-negative constants.
10210 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
10213 * config.gcc (gcc_cv_initfini_array): Leave undefined for
10214 uclinuxfdpiceabi targets.
10216 2021-08-17 Alexandre Oliva <oliva@adacore.com>
10218 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
10219 reverse debug stmts.
10221 2021-08-17 Alexandre Oliva <oliva@adacore.com>
10223 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
10225 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10227 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
10228 (__LD4Q_LANE_FUNC): Likewise.
10229 (vld4_lane_u8): Define without macro.
10230 (vld4_lane_u16): Likewise.
10231 (vld4_lane_u32): Likewise.
10232 (vld4_lane_u64): Likewise.
10233 (vld4_lane_s8): Likewise.
10234 (vld4_lane_s16): Likewise.
10235 (vld4_lane_s32): Likewise.
10236 (vld4_lane_s64): Likewise.
10237 (vld4_lane_f16): Likewise.
10238 (vld4_lane_f32): Likewise.
10239 (vld4_lane_f64): Likewise.
10240 (vld4_lane_p8): Likewise.
10241 (vld4_lane_p16): Likewise.
10242 (vld4_lane_p64): Likewise.
10243 (vld4q_lane_u8): Likewise.
10244 (vld4q_lane_u16): Likewise.
10245 (vld4q_lane_u32): Likewise.
10246 (vld4q_lane_u64): Likewise.
10247 (vld4q_lane_s8): Likewise.
10248 (vld4q_lane_s16): Likewise.
10249 (vld4q_lane_s32): Likewise.
10250 (vld4q_lane_s64): Likewise.
10251 (vld4q_lane_f16): Likewise.
10252 (vld4q_lane_f32): Likewise.
10253 (vld4q_lane_f64): Likewise.
10254 (vld4q_lane_p8): Likewise.
10255 (vld4q_lane_p16): Likewise.
10256 (vld4q_lane_p64): Likewise.
10257 (vld4_lane_bf16): Likewise.
10258 (vld4q_lane_bf16): Likewise.
10260 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10262 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
10263 (__LD3Q_LANE_FUNC): Delete.
10264 (vld3_lane_u8): Define without macro.
10265 (vld3_lane_u16): Likewise.
10266 (vld3_lane_u32): Likewise.
10267 (vld3_lane_u64): Likewise.
10268 (vld3_lane_s8): Likewise.
10269 (vld3_lane_s16): Likewise.
10270 (vld3_lane_s32): Likewise.
10271 (vld3_lane_s64): Likewise.
10272 (vld3_lane_f16): Likewise.
10273 (vld3_lane_f32): Likewise.
10274 (vld3_lane_f64): Likewise.
10275 (vld3_lane_p8): Likewise.
10276 (vld3_lane_p16): Likewise.
10277 (vld3_lane_p64): Likewise.
10278 (vld3q_lane_u8): Likewise.
10279 (vld3q_lane_u16): Likewise.
10280 (vld3q_lane_u32): Likewise.
10281 (vld3q_lane_u64): Likewise.
10282 (vld3q_lane_s8): Likewise.
10283 (vld3q_lane_s16): Likewise.
10284 (vld3q_lane_s32): Likewise.
10285 (vld3q_lane_s64): Likewise.
10286 (vld3q_lane_f16): Likewise.
10287 (vld3q_lane_f32): Likewise.
10288 (vld3q_lane_f64): Likewise.
10289 (vld3q_lane_p8): Likewise.
10290 (vld3q_lane_p16): Likewise.
10291 (vld3q_lane_p64): Likewise.
10292 (vld3_lane_bf16): Likewise.
10293 (vld3q_lane_bf16): Likewise.
10295 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10297 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
10298 (__LD2Q_LANE_FUNC): Likewise.
10299 (vld2_lane_u8): Define without macro.
10300 (vld2_lane_u16): Likewise.
10301 (vld2_lane_u32): Likewise.
10302 (vld2_lane_u64): Likewise.
10303 (vld2_lane_s8): Likewise.
10304 (vld2_lane_s16): Likewise.
10305 (vld2_lane_s32): Likewise.
10306 (vld2_lane_s64): Likewise.
10307 (vld2_lane_f16): Likewise.
10308 (vld2_lane_f32): Likewise.
10309 (vld2_lane_f64): Likewise.
10310 (vld2_lane_p8): Likewise.
10311 (vld2_lane_p16): Likewise.
10312 (vld2_lane_p64): Likewise.
10313 (vld2q_lane_u8): Likewise.
10314 (vld2q_lane_u16): Likewise.
10315 (vld2q_lane_u32): Likewise.
10316 (vld2q_lane_u64): Likewise.
10317 (vld2q_lane_s8): Likewise.
10318 (vld2q_lane_s16): Likewise.
10319 (vld2q_lane_s32): Likewise.
10320 (vld2q_lane_s64): Likewise.
10321 (vld2q_lane_f16): Likewise.
10322 (vld2q_lane_f32): Likewise.
10323 (vld2q_lane_f64): Likewise.
10324 (vld2q_lane_p8): Likewise.
10325 (vld2q_lane_p16): Likewise.
10326 (vld2q_lane_p64): Likewise.
10327 (vld2_lane_bf16): Likewise.
10328 (vld2q_lane_bf16): Likewise.
10330 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10332 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
10335 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10337 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
10338 (rank_for_schedule): Use it.
10340 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10342 PR rtl-optimization/91598
10343 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
10344 insns after memory reads and before memory writes.
10346 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
10348 * rtl.h (CONST_VECTOR_P): New macro.
10349 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
10350 code testing macros.
10351 (aarch64_ptrue_all_mode): Likewise.
10352 (aarch64_expand_mov_immediate): Likewise.
10353 (aarch64_const_vec_all_in_range_p): Likewise.
10354 (aarch64_rtx_costs): Likewise.
10355 (aarch64_legitimate_constant_p): Likewise.
10356 (aarch64_simd_valid_immediate): Likewise.
10357 (aarch64_simd_make_constant): Likewise.
10358 (aarch64_convert_mult_to_shift): Likewise.
10359 (aarch64_expand_sve_vec_perm): Likewise.
10360 (aarch64_vec_fpconst_pow_of_2): Likewise.
10362 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10364 PR tree-optimization/101938
10365 * range-op.cc (operator_abs::op1_range): Special case
10366 -TYPE_MIN_VALUE for flag_wrapv.
10368 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
10370 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
10373 2021-08-17 Jakub Jelinek <jakub@redhat.com>
10375 * tree.def (OMP_SCOPE): New tree code.
10376 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
10377 * tree-nested.c (convert_nonlocal_reference_stmt,
10378 convert_local_reference_stmt, convert_gimple_call): Handle
10380 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
10381 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
10382 * gimple.c (gimple_build_omp_scope): New function.
10383 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
10384 * gimple.h (gimple_build_omp_scope): Declare.
10385 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
10386 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
10387 gimple_omp_scope_set_clauses): New inline functions.
10388 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
10389 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
10390 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
10391 * gimple-walk.c (walk_gimple_stmt): Likewise.
10392 * gimple-low.c (lower_stmt): Likewise.
10393 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
10394 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
10395 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
10396 allowing task reductions. Reject inscan reductions on scope.
10397 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
10398 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
10399 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
10400 (estimate_num_insns): Likewise.
10401 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
10402 contexts if var isn't privatized there.
10403 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
10404 (scan_omp_1_stmt): Likewise.
10405 (maybe_add_implicit_barrier_cancel): Look through outer
10407 (lower_omp_scope): New function.
10408 (lower_omp_task_reductions): Handle OMP_SCOPE.
10409 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
10410 (diagnose_sb_1, diagnose_sb_2): Likewise.
10411 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
10412 (expand_omp): Handle GIMPLE_OMP_SCOPE.
10413 (omp_make_gimple_edges): Likewise.
10414 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
10416 2021-08-17 Richard Biener <rguenther@suse.de>
10418 PR tree-optimization/101925
10419 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
10420 reverse on COMPONENT_REF and ARRAY_REF according to
10421 what reverse_storage_order_for_component_p does.
10422 (vn_reference_eq): Compare reversed on reference ops.
10423 (reverse_storage_order_for_component_p): New overload.
10424 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
10425 on the reference looked up.
10427 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
10429 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
10430 shifts on the H8/S.
10431 (h8300_option_override): Use loops on H8/S more often when optimizing
10433 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
10434 accordingly. Handle various arithmetic right shifts with special
10435 sequences that we couldn't handle before.
10437 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
10439 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
10441 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
10443 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
10444 * config/sparc/sparc.c (sparc_gcov_type_size): New.
10445 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
10446 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
10447 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
10448 * doc/tm.texi.in: Regenerate.
10449 * target.def (gcov_type_size): New target hook.
10450 * targhooks.c (default_gcov_type_size): New.
10451 * targhooks.h (default_gcov_type_size): Declare.
10452 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
10454 (gimple_gen_time_profiler): Likewise.
10456 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
10458 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
10460 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
10462 PR middle-end/101931
10463 * omp-low.c (omp_runtime_api_call): Update for routines
10464 added in the meanwhile.
10466 2021-08-16 Martin Liska <mliska@suse.cz>
10468 PR tree-optimization/100393
10469 * tree-switch-conversion.c (group_cluster::dump): Use
10470 get_comparison_count.
10471 (jump_table_cluster::find_jump_tables): Pre-compute number of
10472 comparisons and then decrement it. Cache also max_ratio.
10473 (jump_table_cluster::can_be_handled): Change signature.
10474 * tree-switch-conversion.h (get_comparison_count): New.
10476 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
10478 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
10479 encodings only when -fgnat-encodings=all is specified.
10480 (add_bound_info): Likewise.
10481 (add_byte_size_attribute): Likewise.
10482 (gen_member_die): Likewise.
10484 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
10486 * omp-oacc-neuter-broadcast.cc
10487 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
10489 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
10491 * omp-oacc-neuter-broadcast.cc
10492 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
10495 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
10497 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
10498 (execute_omp_oacc_neuter_broadcast): ... here.
10499 (install_var_field, build_receiver_ref, build_sender_ref): Take
10500 'field_map_t *' parameter. Adjust all users.
10501 (worker_single_copy, neuter_worker_single): Take a
10502 'record_field_map_t *' parameter. Adjust all users.
10504 2021-08-16 liuhongt <hongtao.liu@intel.com>
10507 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
10510 2021-08-16 Martin Liska <mliska@suse.cz>
10513 * multiple_target.c (create_dispatcher_calls): Make default
10514 function local only if it is a definition.
10516 2021-08-16 Martin Liska <mliska@suse.cz>
10519 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
10520 consider equal SSA_NAMEs when one is a param.
10522 2021-08-16 liuhongt <hongtao.liu@intel.com>
10525 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
10526 Support vpermi2b for V32QI/V16QImode.
10527 (ix86_extract_perm_from_pool_constant): New function.
10528 (ix86_expand_vec_one_operand_perm_avx512): Support
10529 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
10530 (expand_vec_perm_1): Adjust comments for upper.
10531 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
10533 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
10534 (pshufb_truncv4siv4hi_operand): Ditto.
10535 (pshufb_truncv8hiv8qi_operand): Ditto.
10536 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
10537 New pre_reload define_insn_and_split.
10538 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
10539 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
10540 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
10541 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
10542 (*avx512f_permvar_truncv4div4si_1): Ditto.
10543 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
10544 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
10545 (*avx512f_pshufd_truncv2div2si_1): Ditto.
10547 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
10549 * config/riscv/multilib-generator: Support code model option for
10551 * doc/install.texi: Add document of new option for
10552 --with-multilib-generator.
10554 2021-08-15 Clément Chigot <clement.chigot@atos.net>
10556 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
10557 (rs6000_legitimize_tls_address_aix): Use it.
10558 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
10559 xcoff_tls_exec_model_detected is true.
10561 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
10563 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
10564 to allow a bit more code growth, saving many dozens of cycles.
10565 (h8300_option_override): Adjus shift_alg_si if optimizing for
10567 (get_shift_alg): Use special + inline shifts for residuals
10570 2021-08-14 Stafford Horne <shorne@gmail.com>
10573 * config/or1k/or1k-opts.h: New file.
10574 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
10575 Support generating gotha relocations if -mcmodel=large is
10577 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
10579 * config/or1k/or1k.opt (mcmodel=): New option.
10580 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
10582 2021-08-14 Martin Sebor <msebor@redhat.com>
10584 PR middle-end/101791
10585 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
10586 to valid_new_delete_pair_p.
10587 * tree.c (valid_new_delete_pair_p): Add argument.
10588 * tree.h (valid_new_delete_pair_p): Same.
10590 2021-08-14 Jakub Jelinek <jakub@redhat.com>
10593 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
10594 <case E_V64QImode>: For this mode assert
10595 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
10597 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
10600 * config/rs6000/altivec.md (xxeval): Use register_predicate
10601 instead of altivec_register_predicate.
10603 2021-08-13 Martin Sebor <msebor@redhat.com>
10605 PR middle-end/101734
10606 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
10607 (maybe_warn_operand): Call it.
10609 2021-08-13 Martin Liska <mliska@suse.cz>
10612 * attribs.c (decl_attributes): Make naked functions "noipa"
10615 2021-08-13 Martin Liska <mliska@suse.cz>
10618 * symtab.c (symtab_node::noninterposable_alias): Do not create
10619 local aliases for target_clone functions as the clonning pass
10622 2021-08-13 Martin Liska <mliska@suse.cz>
10624 * opts.c (LIVE_PATCHING_OPTION): Define.
10625 (control_options_for_live_patching): Use it in error messages.
10627 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
10629 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
10630 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
10631 ignore_stores_eaf_flags): New constants.
10632 (remove_useless_eaf_flags): New function.
10633 (eaf_flags_useful_p): Use it.
10634 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
10636 (modref_lattice::init): Add EAF_NOREAD.
10637 (modref_lattice::add_escape_point): Do not reacord escape point if
10639 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
10640 use remove_useless_eaf_flags.
10641 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
10642 (modref_lattice::merge_direct_load): Add EAF_NOREAD
10643 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
10644 (analyze_parms): Use remove_useless_eaf_flags.
10645 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
10646 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
10647 use remove_useless_eaf_flags.
10648 (modref_propagate_flags_in_scc): Update.
10649 * ipa-modref.h: Turn eaf_flags_t back to char.
10650 * tree-core.h (EAF_NOT_RETURNED): Fix.
10651 (EAF_NOREAD): New constant
10652 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
10654 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
10655 (handle_pure_call): Likewise.
10657 2021-08-12 Jakub Jelinek <jakub@redhat.com>
10659 * tree.def (OMP_MASKED): New tree code.
10660 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
10661 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
10662 OMP_CLAUSE_FILTER_EXPR): Define.
10663 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
10664 (omp_clause_code_name): Likewise.
10665 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
10666 * tree-nested.c (convert_nonlocal_omp_clauses,
10667 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
10668 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
10669 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
10670 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
10671 (dump_generic_node): Handle OMP_MASTER.
10672 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
10673 * gimple.c (gimple_build_omp_masked): New function.
10674 (gimple_copy): Handle GIMPLE_OMP_MASKED.
10675 * gimple.h (gimple_build_omp_masked): Declare.
10676 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
10677 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
10678 gimple_omp_masked_set_clauses): New inline functions.
10679 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
10680 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
10681 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
10682 * gimple-walk.c (walk_gimple_stmt): Likewise.
10683 * gimple-low.c (lower_stmt): Likewise.
10684 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
10685 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
10686 that take one expression rather than decl or constant, force
10687 gimplification of that into a SSA_NAME or temporary unless min
10689 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
10690 (gimplify_expr): Handle OMP_MASKED.
10691 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
10692 (estimate_num_insns): Likewise.
10693 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
10694 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
10695 diagnostics for existence of masked construct.
10696 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
10697 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
10698 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
10701 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
10704 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
10705 (ldexp<mode>3): Use avx512f_scalef<mode>2.
10706 (UNSPEC_SCALEF): Move from ...
10707 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
10709 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
10711 * ipa-split.c (consider_split): Fix condition testing void functions.
10713 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
10715 * doc/invoke.texi: Remove docs for threader-mode param.
10716 * flag-types.h (enum threader_mode): Remove.
10717 * params.opt: Remove threader-mode param.
10718 * tree-ssa-threadbackward.c (class back_threader): Remove
10719 path_is_unreachable_p.
10720 Make find_paths private.
10721 Add maybe_thread and thread_through_all_blocks.
10722 Remove reference marker for m_registry.
10723 Remove reference marker for m_profit.
10724 (back_threader::back_threader): Adjust for registry and profit not
10726 (dump_path): Move down.
10727 (debug): Move down.
10728 (class thread_jumps): Remove.
10729 (class back_threader_registry): Remove m_all_paths.
10731 (thread_jumps::thread_through_all_blocks): Move to back_threader
10733 (fsm_find_thread_path): Remove
10734 (back_threader::maybe_thread): New.
10735 (back_threader::thread_through_all_blocks): Move from
10737 (back_threader_registry::back_threader_registry): Remove
10739 (back_threader_registry::~back_threader_registry): Remove.
10740 (thread_jumps::find_taken_edge): Remove.
10741 (thread_jumps::check_subpath_and_update_thread_path): Remove.
10742 (thread_jumps::maybe_register_path): Remove.
10743 (thread_jumps::handle_phi): Remove.
10744 (handle_assignment_p): Remove.
10745 (thread_jumps::handle_assignment): Remove.
10746 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
10747 (thread_jumps::find_jump_threads_backwards): Remove.
10748 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
10749 (try_thread_blocks): Rename find_jump_threads_backwards to
10751 (pass_early_thread_jumps::execute): Same.
10753 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
10755 * tree-core.h (omp_clause_proc_bind_kind): Add
10756 OMP_CLAUSE_PROC_BIND_PRIMARY.
10757 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
10758 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
10760 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
10762 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
10763 fno-common reference.
10764 * config/arc/arc.c (arc_override_options): Remove overriding of
10767 2021-08-12 Jakub Jelinek <jakub@redhat.com>
10770 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
10771 If d->testing_p, return true after performing checks instead of
10772 actually expanding the insn.
10773 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
10774 !TARGET_AVX512BW and return false.
10776 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
10778 * configure.ac (PE linker --disable-dynamicbase support): New check.
10779 * configure: Regenerate.
10780 * config.in: Likewise.
10781 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
10782 (LINK_SPEC): Use it.
10783 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
10784 (LINK_SPEC): Likewise.
10786 2021-08-12 liuhongt <hongtao.liu@intel.com>
10789 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
10790 post_reload define_insn_and_split.
10791 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
10792 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
10793 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
10794 (*avx2_zero_extendv8hiv8si2_2): Ditto.
10795 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
10796 (*avx512f_zero_extendv8siv8di2_2): Ditto.
10797 (*avx2_zero_extendv4siv4di2_2): Ditto.
10798 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
10799 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
10802 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
10804 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
10807 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
10809 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
10811 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
10813 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
10814 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
10815 initialization of pcvoid_type_node here...
10816 (altivec_init_builtins): ...from here.
10817 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
10818 RS6000_BTI_const_ptr_void.
10819 (pcvoid_type_node): New macro.
10821 2021-08-11 Richard Biener <rguenther@suse.de>
10824 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
10825 hard-register accesses.
10827 2021-08-11 Richard Biener <rguenther@suse.de>
10829 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
10830 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
10831 to determine has_volatile_ops.
10833 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
10835 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
10837 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
10838 Alexandre Oliva <oliva@adacore.com>
10840 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
10842 (modref_lattice::merge_deref): Fix handling of indirect scape points.
10843 (update_escape_summary_1): Likewise.
10844 (update_escape_summary): Likewise.
10845 (ipa_merge_modref_summary_after_inlining): Likewise.
10847 2021-08-11 Richard Biener <rguenther@suse.de>
10849 PR middle-end/101858
10850 * fold-const.c (fold_binary_loc): Guard simplification
10851 of X < (cast) (1 << Y) to integer types.
10853 2021-08-11 Richard Biener <rguenther@suse.de>
10855 PR tree-optimization/101861
10856 * tree-vect-stmts.c (vectorizable_load): Fix error in
10857 previous change with regard to gather vectorization.
10859 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10862 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
10864 (vdup_n_s16): Likewise.
10865 (vdup_n_s32): Likewise.
10866 (vdup_n_s64): Likewise.
10867 (vdup_n_u8): Likewise.
10868 (vdup_n_u16): Likewise.
10869 (vdup_n_u32): Likewise.
10870 (vdup_n_u64): Likewise.
10871 (vdup_n_p8): Likewise.
10872 (vdup_n_p16): Likewise.
10873 (vdup_n_p64): Likewise.
10874 (vdup_n_f16): Likewise.
10875 (vdup_n_f32): Likewise.
10876 (vdupq_n_s8): Likewise.
10877 (vdupq_n_s16): Likewise.
10878 (vdupq_n_s32): Likewise.
10879 (vdupq_n_s64): Likewise.
10880 (vdupq_n_u8): Likewise.
10881 (vdupq_n_u16): Likewise.
10882 (vdupq_n_u32): Likewise.
10883 (vdupq_n_u64): Likewise.
10884 (vdupq_n_p8): Likewise.
10885 (vdupq_n_p16): Likewise.
10886 (vdupq_n_p64): Likewise.
10887 (vdupq_n_f16): Likewise.
10888 (vdupq_n_f32): Likewise.
10889 (vmov_n_s8): Replace call to builtin with call to corresponding
10891 (vmov_n_s16): Likewise.
10892 (vmov_n_s32): Likewise.
10893 (vmov_n_s64): Likewise.
10894 (vmov_n_u8): Likewise.
10895 (vmov_n_u16): Likewise.
10896 (vmov_n_u32): Likewise.
10897 (vmov_n_u64): Likewise.
10898 (vmov_n_p8): Likewise.
10899 (vmov_n_p16): Likewise.
10900 (vmov_n_f16): Likewise.
10901 (vmov_n_f32): Likewise.
10902 (vmovq_n_s8): Likewise.
10903 (vmovq_n_s16): Likewise.
10904 (vmovq_n_s32): Likewise.
10905 (vmovq_n_s64): Likewise.
10906 (vmovq_n_u8): Likewise.
10907 (vmovq_n_u16): Likewise.
10908 (vmovq_n_u32): Likewise.
10909 (vmovq_n_u64): Likewise.
10910 (vmovq_n_p8): Likewise.
10911 (vmovq_n_p16): Likewise.
10912 (vmovq_n_f16): Likewise.
10913 (vmovq_n_f32): Likewise.
10914 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
10916 2021-08-11 liuhongt <hongtao.liu@intel.com>
10919 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
10920 when TARGET_AVX512F and TARGET_SSE_MATH.
10922 2021-08-10 Jakub Jelinek <jakub@redhat.com>
10925 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
10926 for V32HImode if !TARGET_AVX512BW.
10927 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
10928 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
10929 early, but actually check the permutation.
10931 2021-08-10 Richard Biener <rguenther@suse.de>
10933 PR tree-optimization/101809
10934 * tree-vect-stmts.c (get_load_store_type): Allow emulated
10935 gathers with offset vector nunits being a constant multiple
10936 of the data vector nunits.
10937 (vect_get_gather_scatter_ops): Use the appropriate nunits
10938 for the offset vector defs.
10939 (vectorizable_store): Adjust call to
10940 vect_get_gather_scatter_ops.
10941 (vectorizable_load): Likewise. Handle the case of less
10942 offset vectors than data vectors.
10944 2021-08-10 Jakub Jelinek <jakub@redhat.com>
10947 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
10948 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
10951 2021-08-10 Richard Biener <rguenther@suse.de>
10953 PR tree-optimization/101801
10954 PR tree-optimization/101819
10955 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
10956 * tree-vect-loop.c (vect_emulated_vector_p): New function.
10957 (vectorizable_reduction): Re-instantiate a check for emulated
10959 * tree-vect-stmts.c (vectorizable_shift): Likewise.
10960 (vectorizable_operation): Likewise. Cost emulated vector
10961 operations according to the scalar sequence synthesized by
10964 2021-08-10 Richard Biener <rguenther@suse.de>
10966 PR middle-end/101824
10967 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
10968 volatile in case the variable was.
10970 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
10973 * config/i386/constraints.md (BC): Document for integer SSE
10974 constant all bits set operand.
10975 (BF): New constraint for const floating-point all bits set
10977 * config/i386/i386.c (standard_sse_constant_p): Likewise.
10978 (standard_sse_constant_opcode): Likewise.
10979 * config/i386/sse.md (sseconstm1): New mode attribute.
10980 (mov<mode>_internal): Replace BC with <sseconstm1>.
10982 2021-08-10 liuhongt <hongtao.liu@intel.com>
10984 * config/i386/sse.md (cond_<insn><mode>): New expander.
10985 (VI248_AVX512VLBW): New mode iterator.
10986 * config/i386/predicates.md
10987 (nonimmediate_or_const_vec_dup_operand): New predicate.
10989 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
10991 PR tree-optimization/101741
10992 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
10993 type of parameter for toupper/tolower.
10995 2021-08-09 Martin Jambor <mjambor@suse.cz>
10997 PR testsuite/101654
10998 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
11000 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
11002 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
11004 (is_store_insn1): Verify source is a register.
11006 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
11009 * config/i386/mmx.md (<any_logic:code>v2sf3):
11010 Rename from *mmx_<any_logic:code>v2sf3
11012 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11014 * config/nvptx/nvptx.c: Cross-reference parts adapted in
11015 'gcc/omp-oacc-neuter-broadcast.cc'.
11016 * omp-low.c: Likewise.
11017 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
11020 2021-08-09 Julian Brown <julian@codesourcery.com>
11021 Kwok Cheung Yeung <kcy@codesourcery.com>
11022 Thomas Schwinge <thomas@codesourcery.com>
11024 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
11025 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
11026 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
11027 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
11028 (gcn_fork_join): Update comment.
11029 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
11030 (macc_experimental_workers): Remove unused option.
11032 2021-08-09 Julian Brown <julian@codesourcery.com>
11033 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
11034 Kwok Cheung Yeung <kcy@codesourcery.com>
11035 Thomas Schwinge <thomas@codesourcery.com>
11037 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
11038 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
11039 Add documentation hook.
11040 * doc/tm.texi: Regenerate.
11041 * omp-oacc-neuter-broadcast.cc: New file.
11042 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
11043 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
11044 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
11045 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
11046 * target.def (goacc.create_worker_broadcast_record): Add target
11048 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
11050 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
11051 Rename prototype to...
11052 (gcn_goacc_create_worker_broadcast_record): ... this.
11053 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
11055 (gcn_goacc_create_worker_broadcast_record): ... this.
11056 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
11058 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
11060 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
11063 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
11064 the right iterator.
11066 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11068 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
11070 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11072 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
11073 consider '-foffload-abi'.
11074 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
11075 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
11078 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11080 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
11083 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11085 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
11087 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11089 * doc/gty.texi (Files): Update.
11091 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11093 * doc/gty.texi (Files): Fix GTY header file example.
11095 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
11097 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
11098 determine the upper and lower bounds from a mask-value pair.
11099 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
11100 absolute value and unsigned absolute value expressions.
11101 (bit_value_binop): Initialize *VAL's precision.
11102 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
11103 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
11104 support when the operands are unknown but potentially equal.
11105 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
11107 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
11109 * config/aarch64/aarch64.md
11110 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
11112 2021-08-08 Sergei Trofimovich <siarheit@google.com>
11114 * lra-constraints.c: Fix s/otput/output/ typo.
11116 2021-08-06 Martin Sebor <msebor@redhat.com>
11118 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
11119 (expand_builtin_strcat): Same.
11120 (expand_builtin_stpncpy): Same.
11121 (expand_builtin_strncat): Same.
11122 (check_read_access): Same.
11123 (check_memop_access): Same.
11124 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
11125 (expand_builtin_strnlen): Same.
11126 (expand_builtin_memcpy): Same.
11127 (expand_builtin_memmove): Same.
11128 (expand_builtin_mempcpy): Same.
11129 (expand_builtin_strcpy): Same.
11130 (expand_builtin_strcpy_args): Same.
11131 (expand_builtin_stpcpy_1): Same.
11132 (expand_builtin_strncpy): Same.
11133 (expand_builtin_memset): Same.
11134 (expand_builtin_bzero): Same.
11135 (expand_builtin_strcmp): Same.
11136 (expand_builtin_strncmp): Same.
11137 (expand_builtin): Remove handlers.
11138 (fold_builtin_strlen): Add a comment.
11139 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
11140 * calls.c (maybe_warn_nonstring_arg): Same.
11141 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
11142 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
11143 (gimple_fold_builtin_stpcpy): Same.
11144 * gimple-ssa-warn-access.cc (has_location): New function.
11145 (get_location): Same.
11146 (get_callee_fndecl): Same.
11147 (call_nargs): Same.
11149 (warn_string_no_nul): Define.
11150 (unterminated_array): Same.
11151 (check_nul_terminated_array): Same.
11152 (maybe_warn_nonstring_arg): Same.
11153 (maybe_warn_for_bound): Same.
11154 (warn_for_access): Same.
11155 (check_access): Same.
11156 (check_memop_access): Same.
11157 (check_read_access): Same.
11158 (warn_dealloc_offset): Use helper functions.
11159 (maybe_emit_free_warning): Same.
11160 (class pass_waccess): Add members.
11161 (check_strcat): New function.
11162 (check_strncat): New function.
11163 (check_stxcpy): New function.
11164 (check_stxncpy): New function.
11165 (check_strncmp): New function.
11166 (pass_waccess::check_builtin): New function.
11167 (pass_waccess::check): Call it.
11168 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
11170 (maybe_warn_for_bound): Same.
11171 (check_access): Same.
11172 (check_memop_access): Same.
11173 (check_read_access): Same.
11174 * pointer-query.h (struct access_data): Define a ctor overload.
11176 2021-08-06 Richard Biener <rguenther@suse.de>
11178 PR tree-optimization/101801
11179 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
11180 (vect_can_vectorize_without_simd_p): ... to this.
11181 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
11182 (vect_can_vectorize_without_simd_p): ... to this and fold
11183 in vect_min_worthwhile_factor.
11184 (vect_min_worthwhile_factor): Remove.
11185 (vectorizable_reduction): Adjust and remove the cost part.
11186 * tree-vect-stmts.c (vectorizable_shift): Likewise.
11187 (vectorizable_operation): Likewise.
11189 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
11192 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
11193 Add general_gr_operand predicate to operand 3.
11195 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
11197 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
11198 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
11200 2021-08-06 Tamar Christina <tamar.christina@arm.com>
11202 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
11203 register_svprfop): Pass vec<> by pointer.
11204 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
11205 * langhooks.c (lhd_simulate_enum_decl): Likewise.
11206 * langhooks.h (struct lang_hooks_for_types): Likewise.
11208 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11210 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
11211 __builtin_memcpy instead of constructing an additional
11212 __builtin_aarch64_simd_oi one vector at a time.
11213 (vst1q_bf16_x2): Likewise.
11214 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
11215 an additional __builtin_aarch64_simd_ci one vector at a time.
11216 (vst1q_bf16_x3): Likewise.
11217 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
11218 (vst1q_bf16_x4): Likewise.
11219 (vst2_bf16): Use __builtin_memcpy instead of constructing an
11220 additional __builtin_aarch64_simd_oi one vector at a time.
11221 (vst2q_bf16): Likewise.
11222 (vst3_bf16): Use __builtin_memcpy instead of constructing an
11223 additional __builtin_aarch64_simd_ci mode one vector at a
11225 (vst3q_bf16): Likewise.
11226 (vst4_bf16): Use __builtin_memcpy instead of constructing an
11227 additional __builtin_aarch64_simd_xi one vector at a time.
11228 (vst4q_bf16): Likewise.
11230 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11232 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
11233 (__ST2Q_LANE_FUNC): Delete.
11234 (vst2_lane_f16): Use __builtin_memcpy to copy vector
11235 structure instead of constructing __builtin_aarch64_simd_oi
11236 one vector at a time.
11237 (vst2_lane_f32): Likewise.
11238 (vst2_lane_f64): Likewise.
11239 (vst2_lane_p8): Likewise.
11240 (vst2_lane_p16): Likewise.
11241 (vst2_lane_p64): Likewise.
11242 (vst2_lane_s8): Likewise.
11243 (vst2_lane_s16): Likewise.
11244 (vst2_lane_s32): Likewise.
11245 (vst2_lane_s64): Likewise.
11246 (vst2_lane_u8): Likewise.
11247 (vst2_lane_u16): Likewise.
11248 (vst2_lane_u32): Likewise.
11249 (vst2_lane_u64): Likewise.
11250 (vst2_lane_bf16): Likewise.
11251 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
11252 structure instead of using a union.
11253 (vst2q_lane_f32): Likewise.
11254 (vst2q_lane_f64): Likewise.
11255 (vst2q_lane_p8): Likewise.
11256 (vst2q_lane_p16): Likewise.
11257 (vst2q_lane_p64): Likewise.
11258 (vst2q_lane_s8): Likewise.
11259 (vst2q_lane_s16): Likewise.
11260 (vst2q_lane_s32): Likewise.
11261 (vst2q_lane_s64): Likewise.
11262 (vst2q_lane_u8): Likewise.
11263 (vst2q_lane_u16): Likewise.
11264 (vst2q_lane_u32): Likewise.
11265 (vst2q_lane_u64): Likewise.
11266 (vst2q_lane_bf16): Likewise.
11268 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11270 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
11271 (__ST3Q_LANE_FUNC): Delete.
11272 (vst3_lane_f16): Use __builtin_memcpy to copy vector
11273 structure instead of constructing __builtin_aarch64_simd_ci
11274 one vector at a time.
11275 (vst3_lane_f32): Likewise.
11276 (vst3_lane_f64): Likewise.
11277 (vst3_lane_p8): Likewise.
11278 (vst3_lane_p16): Likewise.
11279 (vst3_lane_p64): Likewise.
11280 (vst3_lane_s8): Likewise.
11281 (vst3_lane_s16): Likewise.
11282 (vst3_lane_s32): Likewise.
11283 (vst3_lane_s64): Likewise.
11284 (vst3_lane_u8): Likewise.
11285 (vst3_lane_u16): Likewise.
11286 (vst3_lane_u32): Likewise.
11287 (vst3_lane_u64): Likewise.
11288 (vst3_lane_bf16): Likewise.
11289 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
11290 structure instead of using a union.
11291 (vst3q_lane_f32): Likewise.
11292 (vst3q_lane_f64): Likewise.
11293 (vst3q_lane_p8): Likewise.
11294 (vst3q_lane_p16): Likewise.
11295 (vst3q_lane_p64): Likewise.
11296 (vst3q_lane_s8): Likewise.
11297 (vst3q_lane_s16): Likewise.
11298 (vst3q_lane_s32): Likewise.
11299 (vst3q_lane_s64): Likewise.
11300 (vst3q_lane_u8): Likewise.
11301 (vst3q_lane_u16): Likewise.
11302 (vst3q_lane_u32): Likewise.
11303 (vst3q_lane_u64): Likewise.
11304 (vst3q_lane_bf16): Likewise.
11306 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11308 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
11309 (__ST4Q_LANE_FUNC): Delete.
11310 (vst4_lane_f16): Use __builtin_memcpy to copy vector
11311 structure instead of constructing __builtin_aarch64_simd_xi
11312 one vector at a time.
11313 (vst4_lane_f32): Likewise.
11314 (vst4_lane_f64): Likewise.
11315 (vst4_lane_p8): Likewise.
11316 (vst4_lane_p16): Likewise.
11317 (vst4_lane_p64): Likewise.
11318 (vst4_lane_s8): Likewise.
11319 (vst4_lane_s16): Likewise.
11320 (vst4_lane_s32): Likewise.
11321 (vst4_lane_s64): Likewise.
11322 (vst4_lane_u8): Likewise.
11323 (vst4_lane_u16): Likewise.
11324 (vst4_lane_u32): Likewise.
11325 (vst4_lane_u64): Likewise.
11326 (vst4_lane_bf16): Likewise.
11327 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
11328 structure instead of using a union.
11329 (vst4q_lane_f32): Likewise.
11330 (vst4q_lane_f64): Likewise.
11331 (vst4q_lane_p8): Likewise.
11332 (vst4q_lane_p16): Likewise.
11333 (vst4q_lane_p64): Likewise.
11334 (vst4q_lane_s8): Likewise.
11335 (vst4q_lane_s16): Likewise.
11336 (vst4q_lane_s32): Likewise.
11337 (vst4q_lane_s64): Likewise.
11338 (vst4q_lane_u8): Likewise.
11339 (vst4q_lane_u16): Likewise.
11340 (vst4q_lane_u32): Likewise.
11341 (vst4q_lane_u64): Likewise.
11342 (vst4q_lane_bf16): Likewise.
11344 2021-08-06 Martin Liska <mliska@suse.cz>
11346 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
11347 a target option is restored, it can have
11348 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
11349 and error should not be emitted.
11351 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
11353 * gcov-io.h (gcov_write): Declare.
11354 * gcov-io.c (gcov_write): New.
11355 (gcov_write_counter): Remove.
11356 (gcov_write_tag_length): Likewise.
11357 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
11358 gcov_write_unsigned().
11359 * doc/invoke.texi (fprofile-info-section): Mention
11360 __gcov_info_to_gdca().
11362 2021-08-06 Martin Sebor <msebor@redhat.com>
11364 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
11365 arguments to by-reference.
11366 (iterate_fix_dominators): Same.
11367 * dominance.h (iterate_fix_dominators): Same.
11368 * ipa-prop.h: Call auto_vec::to_vec_legacy.
11369 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
11370 arguments to by-reference.
11371 (debug_data_dependence_relation): Same.
11372 (dump_data_dependence_relations): Same.
11373 * tree-data-ref.h (debug_data_dependence_relation): Same.
11374 (dump_data_dependence_relations): Same.
11375 * tree-predcom.c (dump_chains): Same.
11376 (initialize_root_vars_lm): Same.
11377 (determine_unroll_factor): Same.
11378 (replace_phis_by_defined_names): Same.
11379 (insert_init_seqs): Same.
11380 (pcom_worker::tree_predictive_commoning_loop): Call
11381 auto_vec::to_vec_legacy.
11382 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
11383 arguments to by-reference.
11384 * tree-ssa-threadbackward.c (populate_worklist): Same.
11385 (back_threader::resolve_def): Same.
11386 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
11387 (vect_enhance_data_refs_alignment): Same.
11388 (vect_check_lower_bound): Same.
11389 (vect_prune_runtime_alias_test_list): Same.
11390 (vect_permute_store_chain): Same.
11391 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
11392 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
11393 * tree-vectorizer.h (vect_permute_store_chain): Same.
11394 * vec.c (test_init): New function.
11395 (vec_c_tests): Call new function.
11396 * vec.h (vec): Declare ctors, dtor, and assignment.
11397 (auto_vec::vec_to_legacy): New function.
11398 (vec::copy): Adjust initialization.
11400 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
11403 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
11404 callee only uses GPRs.
11405 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
11406 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
11407 * config/i386/x86gprintrin.h: Add
11408 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
11409 to disable non-GPR ISAs.
11411 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
11413 PR middle-end/101787
11414 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
11416 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
11418 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
11419 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
11420 (vect_is_extending_load, vect_is_integer_truncation): New functions,
11421 moved from aarch64.c but given different names.
11422 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
11423 (aarch64_is_reduction, aarch64_reduc_type)
11424 (aarch64_embedded_comparison_type, aarch64_comparison_type)
11425 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
11426 in favor of the above. Update callers accordingly.
11428 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11431 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
11432 writing .cpu directive in asm output.
11433 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
11434 (arm_last_printed_arch_string): Delete.
11435 (arm_last-printed_fpu_string): Delete.
11436 (arm_configure_build_target): If use of floating-point/SIMD is
11437 disabled, remove all fp/simd related features from the target ISA.
11438 (last_arm_targ_options): New variable.
11439 (arm_print_asm_arch_directives): Add new parameters. Change order
11440 of emitted directives and handle all cases here.
11441 (arm_file_start): Always call arm_print_asm_arch_directives, move
11442 all generation of .arch/.arch_extension here.
11443 (arm_file_end): Call arm_print_asm_arch.
11444 (arm_declare_function_name): Call arm_print_asm_arch_directives
11445 instead of printing .arch/.fpu directives directly.
11447 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11449 * config/arm/arm.c (arm_configure_build_target): Don't call
11450 arm_option_reconfigure_globals.
11451 (arm_option_restore): Call arm_option_reconfigure_globals after
11452 reconfiguring the target.
11453 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
11455 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11457 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
11458 arch_name is always set.
11460 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
11462 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
11463 of vec_select high-half from being added into Neon subtract
11466 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
11468 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
11469 of vec_select high-half from being added into Neon add cost.
11471 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
11473 * cfgloop.h (loops_list::loops_list): Add one optional argument
11474 root and adjust accordingly, update loop tree walking and factor
11476 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
11478 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
11480 PR tree-optimization/101626
11481 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
11482 reverse scalar storage order on a pointer or vector component.
11484 2021-08-05 liuhongt <hongtao.liu@intel.com>
11486 * config/i386/sse.md (cond_<code><mode>): New expander.
11488 2021-08-05 liuhongt <hongtao.liu@intel.com>
11490 * config/i386/sse.md (cond_<code><mode>): New expander.
11492 2021-08-05 liuhongt <hongtao.liu@intel.com>
11494 * config/i386/sse.md (cond_<code><mode>): New expander.
11496 2021-08-04 David Malcolm <dmalcolm@redhat.com>
11499 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
11501 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
11504 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
11505 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
11507 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
11510 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
11511 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
11512 data with SSE register from one memory location to another.
11514 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11516 * config/s390/s390.c (expand_perm_with_vpdi): New function.
11517 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
11518 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
11519 parameterized expander.
11520 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
11522 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11524 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
11525 (struct expand_vec_perm_d): Define struct.
11526 (expand_perm_with_merge): New function.
11527 (vectorize_vec_perm_const_1): New function.
11528 (s390_vectorize_vec_perm_const): New function.
11529 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
11531 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11533 * config/s390/vector.md (V_HW_64): Remove mode iterator.
11534 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
11535 * config/s390/vx-builtins.md
11536 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
11539 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11541 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
11543 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
11545 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
11546 instead of an unspec.
11548 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
11550 * config/s390/s390-modes.def: Add more vector modes to support
11551 concatenation of two vectors.
11552 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
11554 (s390_expand_merge): Likewise.
11555 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
11556 (s390_expand_merge): New function.
11557 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
11558 Remove constant definitions.
11559 * config/s390/vector.md (V_HW_2): Add mode iterators.
11560 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
11561 (vec_2x_nelts, vec_2x_wide): New mode attributes.
11562 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
11563 New pattern definitions.
11564 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
11565 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
11566 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
11567 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
11569 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
11571 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
11572 emit vec merge pattern.
11574 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
11576 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
11578 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
11579 vec_select high-half from being added into Neon multiply
11581 * rtlanal.c (vec_series_highpart_p): Define.
11582 * rtlanal.h (vec_series_highpart_p): Declare.
11584 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
11586 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
11588 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
11589 vec_select cost from being added into Neon multiply cost.
11591 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
11593 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
11594 which old_loop_vinfo is an epilogue loop that handles a constant
11595 number of iterations.
11597 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
11599 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
11600 when a reanalyzed loop fails to be cheaper than the current
11603 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
11605 * config/aarch64/aarch64.c: Fix a typo.
11607 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
11609 PR gcov-profile/101773
11610 * gcov-io.c (gcov_close): Check return code of a fclose.
11612 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
11615 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
11616 .file statement when needed.
11618 2021-08-04 Richard Biener <rguenther@suse.de>
11620 * tree-vect-data-refs.c (vect_check_gather_scatter):
11621 Include widening conversions only when the result is
11622 still handed by native gather or the current offset
11623 size not already matches the data size.
11624 Also succeed analysis in case there's no native support,
11625 noted by a IFN_LAST ifn and a NULL decl.
11626 (vect_analyze_data_refs): Always consider gathers.
11627 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
11628 Test for no IFN gather rather than decl gather.
11629 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
11630 gather-scatter info and cost emulated gathers accordingly.
11631 (vect_truncate_gather_scatter_offset): Properly test for
11633 (vect_use_strided_gather_scatters_p): Likewise.
11634 (get_load_store_type): Handle emulated gathers and its
11636 (vectorizable_load): Likewise. Emulate them by extracting
11637 scalar offsets, doing scalar loads and a vector construct.
11639 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
11642 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
11643 argument to set m_max_size.
11644 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
11645 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
11646 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
11648 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
11649 Marc Glisse <marc.glisse@inria.fr>
11651 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
11652 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
11653 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
11655 2021-08-04 Richard Biener <rguenther@suse.de>
11657 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
11658 out code to decompose vector loads ...
11659 (optimize_vector_load): ... here. Generalize it to
11660 handle intermediate widening and TARGET_MEM_REF loads
11661 and apply it to loads with a supported vector mode as well.
11663 2021-08-04 Richard Biener <rguenther@suse.de>
11665 PR tree-optimization/101756
11666 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
11667 the result of the reduction epilogue is compatible to the original
11670 2021-08-04 liuhongt <hongtao.liu@intel.com>
11673 * config/i386/i386.md (peephole2): Refine predicate from
11674 register_operand to general_reg_operand.
11676 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
11678 * gimple-range-path.h (path_range_query::dump): Mark override.
11680 2021-08-04 Richard Biener <rguenther@suse.de>
11682 PR tree-optimization/101769
11683 * tree-tailcall.c (eliminate_tail_call): Add the created loop
11684 for the first recursion and return it via the new output parameter.
11685 (optimize_tail_call): Pass through new output param.
11686 (tree_optimize_tail_calls_1): After creating all latches,
11687 add the created loop to the loop tree. Do not mark loops for fixup.
11689 2021-08-04 Martin Liska <mliska@suse.cz>
11691 * doc/invoke.texi: Document threader-mode param.
11693 2021-08-04 liuhongt <hongtao.liu@intel.com>
11695 * config/i386/sse.md (cond_fma<mode>): New expander.
11696 (cond_fms<mode>): Ditto.
11697 (cond_fnma<mode>): Ditto.
11698 (cond_fnms<mode>): Ditto.
11700 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
11702 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
11704 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
11706 * config/rs6000/constraints.md: Remove "e" from the list of available
11707 constraint characters.
11709 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
11711 PR gcov-profile/71672
11712 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
11714 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
11716 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
11718 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
11720 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
11721 try XMM31 to avoid vzeroupper.
11723 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11725 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
11726 -mcpu=neoverse-512tvb.
11727 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
11728 * config/aarch64/aarch64-tune.md: Regenerate.
11729 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
11730 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
11731 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
11732 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
11733 (aarch64_adjust_body_cost): Likewise.
11735 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11737 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
11738 record issue information for operations that occur in the
11741 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11743 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
11744 parameter. Detect cases in which an Advanced SIMD MLA would almost
11745 certainly require a MOV.
11746 (aarch64_count_ops): Update accordingly.
11748 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11750 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
11751 function, split out from...
11752 (aarch64_detect_vector_stmt_subtype): ...here.
11753 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
11755 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11757 * config/aarch64/aarch64-protos.h (sve_vec_cost):
11758 Add gather_load_x32_cost and gather_load_x64_cost.
11759 * config/aarch64/aarch64.c (generic_sve_vector_cost)
11760 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
11761 accordingly, using the values given by the scalar_load * number
11762 of elements calculation that we used previously.
11763 (aarch64_detect_vector_stmt_subtype): Use the new fields.
11765 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11767 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
11768 function, split out from...
11769 (aarch64_adjust_body_cost): ...here.
11771 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11773 * config/aarch64/fractional-cost.h: New file.
11774 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
11775 and cost_fraction.h.
11776 (vec_cost_fraction): New typedef.
11777 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
11778 (aarch64_detect_vector_stmt_subtype): Likewise.
11779 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
11780 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
11782 (aarch64_adjust_body_cost): Likewise.
11783 (aarch64_test_cost_fraction): New function.
11784 (aarch64_run_selftests): Call it.
11786 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
11788 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
11790 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
11792 (aarch64_estimated_poly_value): Likewise. Use the least significant
11793 set bit for the minimum and likely values. Use the most significant
11794 set bit for the maximum value.
11796 2021-08-03 liuhongt <hongtao.liu@intel.com>
11798 * config/i386/sse.md (cond_<insn><mode>): New expander.
11799 (cond_mul<mode>): Ditto.
11801 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
11803 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
11805 2021-08-03 liuhongt <hongtao.liu@intel.com>
11807 * config/i386/sse.md (cond_<insn><mode>):New expander.
11808 (cond_mul<mode>): Ditto.
11809 (cond_div<mode>): Ditto.
11811 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
11813 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
11814 check stack_realign_needed for stack realignment.
11815 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
11816 than the largest integer supported by vector register.
11817 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
11818 (MOVE_MAX): Set to bytes of the largest integer supported by
11820 (STORE_MAX_PIECES): New.
11822 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
11824 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
11825 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
11826 data from one memory location to another.
11828 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
11830 PR middle-end/90773
11831 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
11833 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
11835 PR tree-optimization/101724
11836 * params.opt: Remove --param=threader-iterative.
11837 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
11840 2021-08-02 Tom de Vries <tdevries@suse.de>
11842 PR middle-end/101665
11843 * doc/extend.texi (nonnull attribute): Improve documentation.
11845 2021-08-02 Andrew Pinski <apinski@marvell.com>
11847 PR rtl-optimization/101683
11848 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
11850 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
11852 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
11853 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
11854 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
11855 (tree_ssa_phiop_worker): Update call to function above.
11857 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
11860 * config/i386/i386.md (bsr_rex64_1_zext): New.
11861 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
11862 with gen_bsr_rex64_1_zext.
11864 2021-07-31 Jakub Jelinek <jakub@redhat.com>
11867 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
11868 define_insn patterns.
11869 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
11870 Add combine splitters for constant - clz.
11871 (clz<mode>2): Use a temporary pseudo for bsr result.
11873 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
11875 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
11876 _mm_floor_sd, _mm_floor_ss): New.
11878 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
11880 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
11881 _mm_ceil_sd, _mm_ceil_ss): New.
11883 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
11885 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
11886 _mm_blend_ps, _mm_blendv_ps): New.
11888 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
11889 Uroš Bizjak <ubizjak@gmail.com>
11891 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
11892 to generate a conditional move using the carry flag after sub $1.
11893 (peephole2): Eliminate a register-to-register move by inverting
11894 the condition of a conditional move.
11896 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
11898 * config/mmix/mmix.md ("call", "call_value", "*call_real")
11899 ("*call_value_real"): Don't generate rtx mentioning the generic
11900 operands 1 and 2 to "call", and similarly for "call_value".
11901 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
11902 (mmix_print_operand): Use '!' instead of 'p'.
11904 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
11906 * doc/md.texi (call): Correct information about operand 2.
11907 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
11909 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
11911 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
11913 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
11915 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
11917 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
11919 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
11920 const_basic_block..
11921 (*::get_bb_range): Ditto.
11922 (*::bb_range_p): Ditto.
11923 * gimple-range-cache.h: Change prototypes.
11925 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
11927 PR middle-end/90773
11928 * builtins.c (builtin_memcpy_read_str): Change the mode argument
11929 from scalar_int_mode to fixed_size_mode.
11930 (builtin_strncpy_read_str): Likewise.
11931 (gen_memset_value_from_prev): New function.
11932 (builtin_memset_read_str): Change the mode argument from
11933 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
11934 and support CONST_VECTOR.
11935 (builtin_memset_gen_str): Likewise.
11936 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
11938 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
11939 with fixed_size_mode.
11940 (builtin_memset_read_str): Likewise.
11941 * expr.c (widest_int_mode_for_size): Renamed to ...
11942 (widest_fixed_size_mode_for_size): Add a bool argument to
11943 indicate if QI vector mode can be used.
11944 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
11945 instead of widest_int_mode_for_size.
11946 (pieces_addr::adjust): Change the mode argument from
11947 scalar_int_mode to fixed_size_mode.
11948 (op_by_pieces_d): Make m_len read-only. Add a bool member,
11949 m_qi_vector_mode, to indicate that QI vector mode can be used.
11950 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
11951 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
11952 instead of widest_int_mode_for_size.
11953 (op_by_pieces_d::get_usable_mode): Change the mode argument from
11954 scalar_int_mode to fixed_size_mode. Call
11955 widest_fixed_size_mode_for_size instead of
11956 widest_int_mode_for_size.
11957 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
11958 function to return the smallest integer or QI vector mode.
11959 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
11960 instead of widest_int_mode_for_size. Call
11961 smallest_fixed_size_mode_for_size instead of
11962 smallest_int_mode_for_size.
11963 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
11964 indicate that QI vector mode can be used and pass it to
11965 op_by_pieces_d::op_by_pieces_d.
11966 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
11967 instead of widest_int_mode_for_size. Pass memsetp to
11968 widest_fixed_size_mode_for_size to support QI vector mode.
11969 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
11970 (store_by_pieces): Pass memsetp to
11971 store_by_pieces_d::store_by_pieces_d.
11972 (clear_by_pieces_1): Removed.
11973 (clear_by_pieces): Replace clear_by_pieces_1 with
11974 builtin_memset_read_str and pass true to store_by_pieces_d to
11975 support vector mode broadcast.
11976 (string_cst_read_str): Change the mode argument from
11977 scalar_int_mode to fixed_size_mode.
11978 * expr.h (by_pieces_constfn): Change scalar_int_mode to
11980 (by_pieces_prev): Likewise.
11981 * rtl.h (lowpart_subreg_regno): New.
11982 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
11983 simplify_subreg_regno.
11984 * target.def (gen_memset_scratch_rtx): New hook.
11985 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
11986 * doc/tm.texi: Regenerated.
11988 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
11991 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
11992 TARGET_EXPR instead of MODIFY_EXPR.
11994 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
11997 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
11998 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
11999 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
12001 (vec_cmpu<IMSA:mode><mode_i>): New expander.
12003 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
12006 * config/i386/i386-options.c (ix86_option_override_internal):
12007 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
12009 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12012 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
12013 explicitly dereferencing __a.
12014 (vld1_s64): Likewise.
12015 (vld1_u64): Likewise.
12016 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
12017 and change to VAR13.
12019 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
12021 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
12022 use of m_range_analyzer.
12023 (loop_versioning::lv_dom_walker::before_dom_children): Same.
12024 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
12025 (loop_versioning::prune_loop_conditions): Replace vr_values use
12026 with range_query interface.
12027 (pass_loop_versioning::execute): Use ranger.
12029 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12032 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
12033 enum values, and emit a warning if they mismatch.
12035 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
12037 * cfgloop.h (as_const): New function.
12038 (class loop_iterator): Rename to ...
12039 (class loops_list): ... this.
12040 (loop_iterator::next): Rename to ...
12041 (loops_list::Iter::fill_curr_loop): ... this and adjust.
12042 (loop_iterator::loop_iterator): Rename to ...
12043 (loops_list::loops_list): ... this and adjust.
12044 (loops_list::Iter): New class.
12045 (loops_list::iterator): New type.
12046 (loops_list::const_iterator): New type.
12047 (loops_list::begin): New function.
12048 (loops_list::end): Likewise.
12049 (loops_list::begin const): Likewise.
12050 (loops_list::end const): Likewise.
12051 (FOR_EACH_LOOP): Remove.
12052 (FOR_EACH_LOOP_FN): Remove.
12053 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
12054 for loop with loops_list instance.
12055 (sort_sibling_loops): Likewise.
12056 (disambiguate_loops_with_multiple_latches): Likewise.
12057 (verify_loop_structure): Likewise.
12058 * cfgloopmanip.c (create_preheaders): Likewise.
12059 (force_single_succ_latches): Likewise.
12060 * config/aarch64/falkor-tag-collision-avoidance.c
12061 (execute_tag_collision_avoidance): Likewise.
12062 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
12063 * config/s390/s390.c (s390_adjust_loops): Likewise.
12064 * doc/loop.texi: Likewise.
12065 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
12066 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
12067 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
12068 (loop_versioning::make_versioning_decisions): Likewise.
12069 * gimple-ssa-split-paths.c (split_paths): Likewise.
12070 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
12071 * graphite.c (canonicalize_loop_form): Likewise.
12072 (graphite_transform_loops): Likewise.
12073 * ipa-fnsummary.c (analyze_function_body): Likewise.
12074 * ipa-pure-const.c (analyze_function): Likewise.
12075 * loop-doloop.c (doloop_optimize_loops): Likewise.
12076 * loop-init.c (loop_optimizer_finalize): Likewise.
12077 (fix_loop_structure): Likewise.
12078 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
12079 (move_loop_invariants): Likewise.
12080 * loop-unroll.c (decide_unrolling): Likewise.
12081 (unroll_loops): Likewise.
12082 * modulo-sched.c (sms_schedule): Likewise.
12083 * predict.c (predict_loops): Likewise.
12084 (pass_profile::execute): Likewise.
12085 * profile.c (branch_prob): Likewise.
12086 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
12087 (sel_find_rgns): Likewise.
12088 * tree-cfg.c (replace_loop_annotate): Likewise.
12089 (replace_uses_by): Likewise.
12090 (move_sese_region_to_fn): Likewise.
12091 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
12092 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
12093 * tree-parloops.c (parallelize_loops): Likewise.
12094 * tree-predcom.c (tree_predictive_commoning): Likewise.
12095 * tree-scalar-evolution.c (scev_initialize): Likewise.
12096 (scev_reset): Likewise.
12097 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
12098 * tree-ssa-live.c (remove_unused_locals): Likewise.
12099 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
12100 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
12101 (tree_ssa_lim_initialize): Likewise.
12102 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
12103 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
12104 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
12105 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
12106 (free_numbers_of_iterations_estimates): Likewise.
12107 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
12108 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
12109 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
12110 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
12111 (pass_scev_cprop::execute): Likewise.
12112 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12113 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
12114 * tree-ssa-threadupdate.c
12115 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
12116 * tree-vectorizer.c (vectorize_loops): Likewise.
12117 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
12119 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
12121 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
12122 generating a VOIDmode register for e.g the
12123 function_arg_info::end_marker.
12125 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
12127 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
12128 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
12129 (cc mode_attr): Similarly.
12130 (ccz subst_attr): Similarly.
12131 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
12132 * config/h8300/testcompare.md: Remove various cc0 based patterns
12133 that had been commented out. Add pattern to set CCZ from a bit
12136 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
12137 Julian Brown <julian@codesourcery.com>
12138 Kwok Cheung Yeung <kcy@codesourcery.com>
12140 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
12141 'update_stmt' after modification.
12142 (pass_oacc_loop_designation): New function, extracted out of...
12143 (pass_oacc_device_lower): ... this.
12144 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
12145 (make_pass_oacc_loop_designation): New
12146 * passes.def: Add it.
12147 * tree-parloops.c (create_parallel_loop): Adjust.
12148 * tree-pass.h (make_pass_oacc_loop_designation): New.
12150 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
12152 * flag-types.h (enum threader_mode): New.
12153 * params.opt: Add entry for --param=threader-mode.
12154 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
12155 (class back_threader): New.
12156 (back_threader::back_threader): New.
12157 (back_threader::~back_threader): New.
12158 (back_threader::maybe_register_path): New.
12159 (back_threader::find_taken_edge): New.
12160 (back_threader::find_taken_edge_switch): New.
12161 (back_threader::find_taken_edge_cond): New.
12162 (back_threader::resolve_def): New.
12163 (back_threader::resolve_phi): New.
12164 (back_threader::find_paths_to_names): New.
12165 (back_threader::find_paths): New.
12168 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
12169 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
12170 (pass_thread_jumps::execute): Abstract out code...
12171 (try_thread_blocks): ...here.
12172 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
12173 Abstract out threading candidate code to...
12174 (single_succ_to_potentially_threadable_block): ...here.
12175 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
12177 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
12178 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
12179 Return bool from register_jump_thread.
12181 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
12183 * target.def: in0 and in1 do not need to be registers.
12184 * doc/tm.texi: Regenerate.
12186 2021-07-29 liuhongt <hongtao.liu@intel.com>
12189 * config/i386/i386.c (ix86_widen_mult_cost): New function.
12190 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
12193 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
12196 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
12197 (rs6000_preferred_doloop_mode): New hook.
12198 * doc/tm.texi: Regenerate.
12199 * doc/tm.texi.in: Add hook preferred_doloop_mode.
12200 * target.def (preferred_doloop_mode): New hook.
12201 * targhooks.c (default_preferred_doloop_mode): New hook.
12202 * targhooks.h (default_preferred_doloop_mode): New hook.
12203 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
12204 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
12205 and compute_doloop_base_on_mode.
12207 2021-07-28 Martin Sebor <msebor@redhat.com>
12209 PR middle-end/101494
12210 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
12211 and size computation.
12213 2021-07-28 Martin Sebor <msebor@redhat.com>
12215 PR middle-end/101601
12216 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
12218 Handle pointers to functions.
12220 2021-07-28 Martin Sebor <msebor@redhat.com>
12222 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
12223 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
12224 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
12225 (access_ref::access_ref): Same.
12226 (access_ref::phi): Same.
12227 (access_ref::get_ref): Same.
12228 (access_ref::size_remaining): Same.
12229 (access_ref::offset_in_range): Same.
12230 (access_ref::add_offset): Same.
12231 (access_ref::inform_access): Same.
12232 (ssa_name_limit_t::visit_phi): Same.
12233 (ssa_name_limit_t::leave_phi): Same.
12234 (ssa_name_limit_t::next): Same.
12235 (ssa_name_limit_t::next_phi): Same.
12236 (ssa_name_limit_t::~ssa_name_limit_t): Same.
12237 (pointer_query::pointer_query): Same.
12238 (pointer_query::get_ref): Same.
12239 (pointer_query::put_ref): Same.
12240 (pointer_query::flush_cache): Same.
12241 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
12242 (check_nul_terminated_array): Same.
12243 (unterminated_array): Same.
12244 (maybe_warn_for_bound): Same.
12245 (check_read_access): Same.
12246 (warn_for_access): Same.
12247 (get_size_range): Same.
12248 (check_access): Same.
12249 (gimple_call_alloc_size): Move to tree.c.
12250 (gimple_parm_array_size): Move to pointer-query.cc.
12251 (get_offset_range): Same.
12252 (gimple_call_return_array): Same.
12253 (handle_min_max_size): Same.
12254 (handle_array_ref): Same.
12255 (handle_mem_ref): Same.
12256 (compute_objsize): Same.
12257 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
12258 (call_dealloc_argno): Same.
12259 (fndecl_dealloc_argno): Same.
12260 (new_delete_mismatch_p): Same.
12261 (matching_alloc_calls_p): Same.
12262 (warn_dealloc_offset): Same.
12263 (maybe_emit_free_warning): Same.
12264 * builtins.h (check_nul_terminated_array): Move to
12265 gimple-ssa-warn-access.h.
12266 (check_nul_terminated_array): Same.
12267 (warn_string_no_nul): Same.
12268 (unterminated_array): Same.
12269 (class ssa_name_limit_t): Same.
12270 (class pointer_query): Same.
12271 (struct access_ref): Same.
12272 (class range_query): Same.
12273 (struct access_data): Same.
12274 (gimple_call_alloc_size): Same.
12275 (gimple_parm_array_size): Same.
12276 (compute_objsize): Same.
12277 (class access_data): Same.
12278 (maybe_emit_free_warning): Same.
12279 * calls.c (initialize_argument_information): Remove call to
12280 maybe_emit_free_warning.
12281 * gimple-array-bounds.cc: Include new header..
12282 * gimple-fold.c: Same.
12283 * gimple-ssa-sprintf.c: Same.
12284 * gimple-ssa-warn-restrict.c: Same.
12285 * passes.def: Add pass_warn_access.
12286 * tree-pass.h (make_pass_warn_access): Declare.
12287 * tree-ssa-strlen.c: Include new headers.
12288 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
12289 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
12290 * gimple-ssa-warn-access.cc: New file.
12291 * gimple-ssa-warn-access.h: New file.
12292 * pointer-query.cc: New file.
12293 * pointer-query.h: New file.
12295 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12297 PR middle-end/101624
12298 * ubsan.c (maybe_instrument_pointer_overflow,
12299 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
12300 PARM_DECLs or RESULT_DECLs.
12301 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
12303 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12305 PR middle-end/101642
12306 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
12307 to type of bswap16 for comparison.
12308 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
12310 2021-07-28 Richard Biener <rguenther@suse.de>
12312 PR tree-optimization/101615
12313 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
12314 at CTOR SLP graph entries.
12316 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12318 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
12319 Add "r,w" alternative.
12321 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
12324 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
12325 AVX_U128_DIRTY when all bits are zero.
12327 2021-07-28 Richard Biener <rguenther@suse.de>
12329 PR tree-optimization/101615
12330 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
12331 external nodes cannot be permuted so make them perm_out 0.
12333 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
12336 * config.in: Regenerate.
12337 * config/gcn/gcn-hsa.h (A_FIJI): New define.
12338 (A_900): New define.
12339 (A_906): New define.
12340 (A_908): New define.
12341 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
12342 * config/gcn/gcn.c (output_file_start): Adjust attributes according
12343 to the assembler capabilities.
12344 * config/gcn/mkoffload.c (main): Likewise.
12345 * configure: Regenerate.
12346 * configure.ac: Add tests for LLVM assembler attribute features.
12348 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
12350 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
12351 cond_false and cond_true on branches.
12353 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
12355 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
12358 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
12360 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
12362 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12365 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
12366 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
12367 and has special !TARGET_AVX512VL expansion.
12368 (vashrv2di3<mask_name>): Rename to ...
12369 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
12370 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
12372 2021-07-28 Martin Uecker <muecker@gwdg.de>
12374 * calls.c (maybe_warn_rdwr_sizes): Correct argument
12375 numbers in warning that were switched.
12377 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
12379 PR tree-optimization/101596
12380 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
12381 by using new_type's precision instead.
12383 2021-07-28 liuhongt <hongtao.liu@intel.com>
12386 * config/i386/i386.h (processor_costs): Add new member
12388 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
12389 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
12390 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
12391 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
12392 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
12393 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
12394 generic_cost, core_cost): Initialize integer_to_sse same value
12396 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
12397 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12398 Use integer_to_sse instead of sse_op to calculate the cost of
12401 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12403 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
12405 (write_init_file): Call write_ovld_static_init.
12407 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12409 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
12411 (write_init_file): Call write_bif_static_init.
12413 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12415 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
12416 (TYPE_MAP_SIZE): New macro.
12417 (type_map): New initialized variable.
12418 (typemap_cmp): New function.
12419 (write_type_node): Likewise.
12420 (write_fntype_init): Implement.
12422 2021-07-27 Martin Sebor <msebor@redhat.com>
12424 PR tree-optimization/101584
12425 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
12426 (check_defs): Call it.
12428 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
12430 * tree-ssa-dom.c (dom_jump_threader_simplifier):
12431 Put avail_exprs_stack in the class, instead of passing it to
12432 jump_threader_simplifier.
12433 (dom_jump_threader_simplifier::simplify): Add state argument.
12434 (dom_opt_dom_walker): Add state.
12435 (pass_dominator::execute): Pass state to threader.
12436 (dom_opt_dom_walker::before_dom_children): Use state.
12437 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
12438 arguments by state.
12439 (jump_threader::record_temporary_equivalences_from_phis):
12440 Register equivalences through the state variable.
12441 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
12442 Record ranges in a statement through the state variable.
12443 (jump_threader::simplify_control_stmt_condition): Pass state to
12445 (jump_threader::simplify_control_stmt_condition_1): Same.
12446 (jump_threader::thread_around_empty_blocks): Remove obsolete
12448 (jump_threader::thread_through_normal_block): Record equivalences
12449 on edge through the state variable.
12450 (jump_threader::thread_across_edge): Abstract state pushing.
12451 (jt_state::jt_state): New.
12452 (jt_state::push): New.
12453 (jt_state::pop): New.
12454 (jt_state::register_equiv): New.
12455 (jt_state::record_ranges_from_stmt): New.
12456 (jt_state::register_equivs_on_edge): New.
12457 (jump_threader_simplifier::jump_threader_simplifier): Move from
12459 (jump_threader_simplifier::simplify): Add state argument.
12460 * tree-ssa-threadedge.h (class jt_state): New.
12461 (class jump_threader): Add state to constructor.
12462 (class jump_threader_simplifier): Add state to simplify. Remove
12463 avail_exprs_stack from class.
12464 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
12466 (vrp_jump_threader::vrp_jump_threader): Add state.
12467 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
12469 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
12471 * Makefile.in (OBJS): Add gimple-range-path.o.
12472 * gimple-range-path.cc: New file.
12473 * gimple-range-path.h: New file.
12475 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
12477 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
12478 inside vec_duplicate for all patterns.
12479 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
12480 Push sign/zero-extension inside vec_duplicate.
12482 2021-07-27 Richard Biener <rguenther@suse.de>
12484 PR tree-optimization/101573
12485 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
12486 looking at uninitialized PHI arg defs in some constrained cases.
12487 (warn_uninitialized_vars): Call it.
12488 (execute_early_warn_uninitialized): Calculate dominators.
12490 2021-07-27 Richard Biener <rguenther@suse.de>
12492 PR tree-optimization/39821
12493 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
12494 vector_stmt for widening arithmetic.
12495 (vectorizable_conversion): Adjust.
12497 2021-07-27 Martin Jambor <mjambor@suse.cz>
12499 * cgraph.h (ipa_replace_map): New field force_load_ref.
12500 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
12501 aded new flag load_dereferenced, adjusted comments.
12502 (ipa_get_param_dereferenced): New function.
12503 (ipa_set_param_dereferenced): Likewise.
12504 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
12505 * ipa-cp.c: Include gimple.h.
12506 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
12507 (get_replacement_map): New parameter force_load_ref, set the
12508 appropriate flag in ipa_replace_map if set.
12509 (struct symbol_and_index_together): New type.
12510 (adjust_refs_in_act_callers): New function.
12511 (adjust_references_in_caller): Likewise.
12512 (create_specialized_node): When appropriate, call
12513 adjust_references_in_caller and force only load references.
12514 * ipa-prop.c (load_from_dereferenced_name): New function.
12515 (ipa_analyze_controlled_uses): Also detect loads from a
12516 dereference, harden testing of call statements.
12517 (ipa_write_node_info): Stream the dereferenced flag.
12518 (ipa_read_node_info): Likewise.
12519 (ipa_set_jf_constant): Also create refdesc when jump function
12520 references a variable.
12521 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
12522 also on references of variables and return a symtab_node. Adjust
12524 (propagate_controlled_uses): Also remove references to VAR_DECLs.
12526 2021-07-27 Jakub Jelinek <jakub@redhat.com>
12528 PR middle-end/101586
12529 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
12530 positions above or equal to sz except for diagnostics of flexible
12533 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
12535 PR tree-optimization/78888
12536 * gimple-range-fold.cc (get_letter_range): New.
12537 (fold_using_range::range_of_builtin_call): Call get_letter_range.
12539 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
12541 PR tree-optimization/78888
12542 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
12543 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
12545 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
12546 Marc Glisse <marc.glisse@inria.fr>
12548 * match.pd (rotate): Simplify equality/inequality of rotations.
12549 (bswap): Simplify equality/inequality tests of byte swapping.
12551 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12553 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
12556 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12558 * range-op.cc (operator_lshift::fold_range): Pass rel to
12559 base class fold_range.
12560 (operator_rshift::fold_range): Same.
12562 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
12565 * toplev.h (min_align_loops_log): Remove declaration.
12566 (min_align_jumps_log, min_align_labels_log): Likewise.
12567 (min_align_functions_log): Likewise.
12569 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12571 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
12573 (execute_vrp): Abstract out simplification of conditionals...
12574 (simplify_casted_conds): ...here.
12576 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
12578 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
12579 Add gimple argument.
12580 (array_bounds_checker::check_array_ref): Same.
12581 (array_bounds_checker::check_addr_expr): Same.
12582 (array_bounds_checker::check_array_bounds): Pass statement to
12583 check_array_bounds and check_addr_expr.
12584 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
12585 (check_addr_expr): Same.
12586 (get_value_range): Same.
12588 2021-07-26 Tamar Christina <tamar.christina@arm.com>
12590 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
12591 (sdot_prod, udot_prod): ... This.
12592 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
12594 (<sur>dot_prod<vsi2qi>): ... this.
12595 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
12596 Change operands order.
12597 (<sur>sadv16qi): Use new operands order.
12598 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
12599 vdotq_s32): Use new RTL ordering.
12601 2021-07-26 Tamar Christina <tamar.christina@arm.com>
12603 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
12604 aarch64_types_ternop_suss_qualifiers): New.
12605 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
12606 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
12607 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
12609 2021-07-23 Jakub Jelinek <jakub@redhat.com>
12611 PR rtl-optimization/101562
12612 * expmed.c (store_integral_bit_field): Only use movstrict_optab
12613 if the operand isn't paradoxical.
12615 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
12617 * gimple-array-bounds.h (class array_bounds_checker): Change
12618 ranges type to range_query.
12620 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12622 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
12623 __builtin_memcpy instead of constructing
12624 __builtin_aarch64_simd_oi one vector at a time.
12625 (vst1_u64_x2): Likewise.
12626 (vst1_f64_x2): Likewise.
12627 (vst1_s8_x2): Likewise.
12628 (vst1_p8_x2): Likewise.
12629 (vst1_s16_x2): Likewise.
12630 (vst1_p16_x2): Likewise.
12631 (vst1_s32_x2): Likewise.
12632 (vst1_u8_x2): Likewise.
12633 (vst1_u16_x2): Likewise.
12634 (vst1_u32_x2): Likewise.
12635 (vst1_f16_x2): Likewise.
12636 (vst1_f32_x2): Likewise.
12637 (vst1_p64_x2): Likewise.
12638 (vst1q_s8_x2): Likewise.
12639 (vst1q_p8_x2): Likewise.
12640 (vst1q_s16_x2): Likewise.
12641 (vst1q_p16_x2): Likewise.
12642 (vst1q_s32_x2): Likewise.
12643 (vst1q_s64_x2): Likewise.
12644 (vst1q_u8_x2): Likewise.
12645 (vst1q_u16_x2): Likewise.
12646 (vst1q_u32_x2): Likewise.
12647 (vst1q_u64_x2): Likewise.
12648 (vst1q_f16_x2): Likewise.
12649 (vst1q_f32_x2): Likewise.
12650 (vst1q_f64_x2): Likewise.
12651 (vst1q_p64_x2): Likewise.
12653 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12655 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
12656 __builtin_memcpy instead of constructing
12657 __builtin_aarch64_simd_ci one vector at a time.
12658 (vst1_u64_x3): Likewise.
12659 (vst1_f64_x3): Likewise.
12660 (vst1_s8_x3): Likewise.
12661 (vst1_p8_x3): Likewise.
12662 (vst1_s16_x3): Likewise.
12663 (vst1_p16_x3): Likewise.
12664 (vst1_s32_x3): Likewise.
12665 (vst1_u8_x3): Likewise.
12666 (vst1_u16_x3): Likewise.
12667 (vst1_u32_x3): Likewise.
12668 (vst1_f16_x3): Likewise.
12669 (vst1_f32_x3): Likewise.
12670 (vst1_p64_x3): Likewise.
12671 (vst1q_s8_x3): Likewise.
12672 (vst1q_p8_x3): Likewise.
12673 (vst1q_s16_x3): Likewise.
12674 (vst1q_p16_x3): Likewise.
12675 (vst1q_s32_x3): Likewise.
12676 (vst1q_s64_x3): Likewise.
12677 (vst1q_u8_x3): Likewise.
12678 (vst1q_u16_x3): Likewise.
12679 (vst1q_u32_x3): Likewise.
12680 (vst1q_u64_x3): Likewise.
12681 (vst1q_f16_x3): Likewise.
12682 (vst1q_f32_x3): Likewise.
12683 (vst1q_f64_x3): Likewise.
12684 (vst1q_p64_x3): Likewise.
12686 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
12689 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
12690 hard register when LRA is in progress.
12692 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12694 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
12695 __builtin_memcpy instead of using a union.
12696 (vst1q_s8_x4): Likewise.
12697 (vst1_s16_x4): Likewise.
12698 (vst1q_s16_x4): Likewise.
12699 (vst1_s32_x4): Likewise.
12700 (vst1q_s32_x4): Likewise.
12701 (vst1_u8_x4): Likewise.
12702 (vst1q_u8_x4): Likewise.
12703 (vst1_u16_x4): Likewise.
12704 (vst1q_u16_x4): Likewise.
12705 (vst1_u32_x4): Likewise.
12706 (vst1q_u32_x4): Likewise.
12707 (vst1_f16_x4): Likewise.
12708 (vst1q_f16_x4): Likewise.
12709 (vst1_f32_x4): Likewise.
12710 (vst1q_f32_x4): Likewise.
12711 (vst1_p8_x4): Likewise.
12712 (vst1q_p8_x4): Likewise.
12713 (vst1_p16_x4): Likewise.
12714 (vst1q_p16_x4): Likewise.
12715 (vst1_s64_x4): Likewise.
12716 (vst1_u64_x4): Likewise.
12717 (vst1_p64_x4): Likewise.
12718 (vst1q_s64_x4): Likewise.
12719 (vst1q_u64_x4): Likewise.
12720 (vst1q_p64_x4): Likewise.
12721 (vst1_f64_x4): Likewise.
12722 (vst1q_f64_x4): Likewise.
12724 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
12726 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
12727 instead of constructing __builtin_aarch64_simd_oi one vector
12729 (vst2_u64): Likewise.
12730 (vst2_f64): Likewise.
12731 (vst2_s8): Likewise.
12732 (vst2_p8): Likewise.
12733 (vst2_s16): Likewise.
12734 (vst2_p16): Likewise.
12735 (vst2_s32): Likewise.
12736 (vst2_u8): Likewise.
12737 (vst2_u16): Likewise.
12738 (vst2_u32): Likewise.
12739 (vst2_f16): Likewise.
12740 (vst2_f32): Likewise.
12741 (vst2_p64): Likewise.
12742 (vst2q_s8): Likewise.
12743 (vst2q_p8): Likewise.
12744 (vst2q_s16): Likewise.
12745 (vst2q_p16): Likewise.
12746 (vst2q_s32): Likewise.
12747 (vst2q_s64): Likewise.
12748 (vst2q_u8): Likewise.
12749 (vst2q_u16): Likewise.
12750 (vst2q_u32): Likewise.
12751 (vst2q_u64): Likewise.
12752 (vst2q_f16): Likewise.
12753 (vst2q_f32): Likewise.
12754 (vst2q_f64): Likewise.
12755 (vst2q_p64): Likewise.
12757 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12759 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
12760 instead of constructing __builtin_aarch64_simd_ci one vector
12762 (vst3_u64): Likewise.
12763 (vst3_f64): Likewise.
12764 (vst3_s8): Likewise.
12765 (vst3_p8): Likewise.
12766 (vst3_s16): Likewise.
12767 (vst3_p16): Likewise.
12768 (vst3_s32): Likewise.
12769 (vst3_u8): Likewise.
12770 (vst3_u16): Likewise.
12771 (vst3_u32): Likewise.
12772 (vst3_f16): Likewise.
12773 (vst3_f32): Likewise.
12774 (vst3_p64): Likewise.
12775 (vst3q_s8): Likewise.
12776 (vst3q_p8): Likewise.
12777 (vst3q_s16): Likewise.
12778 (vst3q_p16): Likewise.
12779 (vst3q_s32): Likewise.
12780 (vst3q_s64): Likewise.
12781 (vst3q_u8): Likewise.
12782 (vst3q_u16): Likewise.
12783 (vst3q_u32): Likewise.
12784 (vst3q_u64): Likewise.
12785 (vst3q_f16): Likewise.
12786 (vst3q_f32): Likewise.
12787 (vst3q_f64): Likewise.
12788 (vst3q_p64): Likewise.
12790 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12792 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
12793 instead of constructing __builtin_aarch64_simd_xi one vector
12795 (vst4_u64): Likewise.
12796 (vst4_f64): Likewise.
12797 (vst4_s8): Likewise.
12798 (vst4_p8): Likewise.
12799 (vst4_s16): Likewise.
12800 (vst4_p16): Likewise.
12801 (vst4_s32): Likewise.
12802 (vst4_u8): Likewise.
12803 (vst4_u16): Likewise.
12804 (vst4_u32): Likewise.
12805 (vst4_f16): Likewise.
12806 (vst4_f32): Likewise.
12807 (vst4_p64): Likewise.
12808 (vst4q_s8): Likewise.
12809 (vst4q_p8): Likewise.
12810 (vst4q_s16): Likewise.
12811 (vst4q_p16): Likewise.
12812 (vst4q_s32): Likewise.
12813 (vst4q_s64): Likewise.
12814 (vst4q_u8): Likewise.
12815 (vst4q_u16): Likewise.
12816 (vst4q_u32): Likewise.
12817 (vst4q_u64): Likewise.
12818 (vst4q_f16): Likewise.
12819 (vst4q_f32): Likewise.
12820 (vst4q_f64): Likewise.
12821 (vst4q_p64): Likewise.
12823 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12825 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
12826 instead of constructing __builtin_aarch64_simd_oi one vector
12828 (vtbx4_u8): Likewise.
12829 (vtbx4_p8): Likewise.
12831 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12833 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
12834 instead of constructing __builtin_aarch64_simd_oi one vector
12836 (vtbl3_u8): Likewise.
12837 (vtbl3_p8): Likewise.
12838 (vtbl4_s8): Likewise.
12839 (vtbl4_u8): Likewise.
12840 (vtbl4_p8): Likewise.
12842 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12844 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
12845 instead of constructing __builtin_aarch64_simd_oi one vector
12847 (vqtbx2_u8): Likewise.
12848 (vqtbx2_p8): Likewise.
12849 (vqtbx2q_s8): Likewise.
12850 (vqtbx2q_u8): Likewise.
12851 (vqtbx2q_p8): Likewise.
12852 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
12853 __builtin_aarch64_simd_ci one vector at a time.
12854 (vqtbx3_u8): Likewise.
12855 (vqtbx3_p8): Likewise.
12856 (vqtbx3q_s8): Likewise.
12857 (vqtbx3q_u8): Likewise.
12858 (vqtbx3q_p8): Likewise.
12859 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
12860 __builtin_aarch64_simd_xi one vector at a time.
12861 (vqtbx4_u8): Likewise.
12862 (vqtbx4_p8): Likewise.
12863 (vqtbx4q_s8): Likewise.
12864 (vqtbx4q_u8): Likewise.
12865 (vqtbx4q_p8): Likewise.
12867 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
12869 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
12870 instead of constructing __builtin_aarch64_simd_oi one vector
12872 (vqtbl2_u8): Likewise.
12873 (vqtbl2_p8): Likewise.
12874 (vqtbl2q_s8): Likewise.
12875 (vqtbl2q_u8): Likewise.
12876 (vqtbl2q_p8): Likewise.
12877 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
12878 __builtin_aarch64_simd_ci one vector at a time.
12879 (vqtbl3_u8): Likewise.
12880 (vqtbl3_p8): Likewise.
12881 (vqtbl3q_s8): Likewise.
12882 (vqtbl3q_u8): Likewise.
12883 (vqtbl3q_p8): Likewise.
12884 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
12885 __builtin_aarch64_simd_xi one vector at a time.
12886 (vqtbl4_u8): Likewise.
12887 (vqtbl4_p8): Likewise.
12888 (vqtbl4q_s8): Likewise.
12889 (vqtbl4q_u8): Likewise.
12890 (vqtbl4q_p8): Likewise.
12892 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
12895 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
12897 2021-07-22 Andrew Pinski <apinski@marvell.com>
12899 PR tree-optimization/10153
12900 * tree-tailcall.c (create_tailcall_accumulator):
12901 Don't call fold_convert as the type should be correct already.
12902 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
12903 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
12905 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
12907 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
12908 varying_p check for null/non-null check.
12910 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
12912 PR tree-optimization/101511
12913 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
12914 is in ssa2's equiv set, and don't trap if so.
12916 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
12918 PR tree-optimization/101497
12919 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
12922 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
12924 PR tree-optimization/101496
12925 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
12926 first, then vrp_visit_cond_Stmt.
12928 2021-07-22 liuhongt <hongtao.liu@intel.com>
12930 * config/i386/i386-expand.c
12931 (ix86_broadcast_from_integer_constant): Rename to ..
12932 (ix86_broadcast_from_constant): .. this, and extend it to
12934 (ix86_expand_vector_move): Extend to float mode.
12935 * config/i386/i386-features.c
12936 (replace_constant_pool_with_broadcast): Remove.
12937 (remove_partial_avx_dependency_gate): Ditto.
12938 (constant_pool_broadcast): Ditto.
12939 (class pass_constant_pool_broadcast): Ditto.
12940 (make_pass_constant_pool_broadcast): Ditto.
12941 (remove_partial_avx_dependency): Adjust gate.
12942 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
12943 * config/i386/i386-protos.h
12944 (make_pass_constant_pool_broadcast): Remove.
12946 2021-07-22 liuhongt <hongtao.liu@intel.com>
12948 * config/i386/constraints.md (Wb): New constraint.
12950 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
12952 (*ashlqi3_1): Ditto.
12953 (*<insn><mode>3_1): Split to ..
12954 (*ashr<mode>3_1): this, ...
12955 (*lshr<mode>3_1): and this, also extend this pattern to avx512
12957 (*<insn><mode>3_1): Split to ..
12958 (*ashr<mode>3_1): this, ...
12959 (*lshrqi3_1): and this, also extend this pattern to avx512
12961 (*lshrhi3_1): And this, also extend this pattern to avx512
12963 * config/i386/sse.md (k<code><mode>): New define_split after
12964 it to convert generic shift pattern to mask shift ones.
12966 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
12967 Joseph Myers <joseph@codesourcery.com>
12968 Cesar Philippidis <cesar@codesourcery.com>
12970 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
12971 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
12973 * tree-pretty-print.c (dump_omp_clause): Likewise.
12974 * omp-general.c (oacc_verify_routine_clauses): Likewise.
12975 * gimplify.c (gimplify_scan_omp_clauses)
12976 (gimplify_adjust_omp_clauses): Likewise.
12977 * tree-nested.c (convert_nonlocal_omp_clauses)
12978 (convert_local_omp_clauses): Likewise.
12979 * omp-low.c (scan_sharing_clauses): Likewise.
12980 * omp-offload.c (execute_oacc_device_lower): Update.
12982 2021-07-21 Martin Sebor <msebor@redhat.com>
12984 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
12986 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
12988 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
12991 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
12993 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
12995 (write_fntype_init): New stub function.
12996 (write_init_bif_table): Likewise.
12997 (write_init_ovld_table): New function.
12998 (write_init_file): Implement.
13000 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13002 * config/rs6000/rs6000-gen-builtins.c
13003 (write_autogenerated_header): New function.
13004 (write_decls): Likewise.
13005 (write_extern_fntype): New callback function.
13006 (write_header_file): Implement.
13008 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13010 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
13013 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13015 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
13017 (complete_base_type): Likewise.
13018 (construct_fntype_id): Likewise.
13019 (parse_bif_entry): Call contruct_fntype_id.
13020 (parse_ovld_entry): Likewise.
13022 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13024 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
13025 (MAXOVLDSTANZAS): New macro.
13026 (ovld_stanzas): New variable.
13027 (curr_ovld_stanza): Likewise.
13028 (MAXOVLDS): New macro.
13029 (ovlddata): New struct.
13030 (ovlds): New variable.
13031 (curr_ovld): Likewise.
13032 (max_ovld_args): Likewise.
13033 (parse_ovld_entry): New function.
13034 (parse_ovld_stanza): Likewise.
13035 (parse_ovld): Implement.
13037 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13039 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
13042 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13044 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
13045 (parse_prototype): Implement.
13047 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13049 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
13050 (curr_bif_stanza): New variable.
13051 (stanza_entry): New struct.
13052 (stanza_map): New initialized variable.
13053 (enable_string): Likewise.
13054 (fnkinds): New enum.
13055 (typelist): New struct.
13056 (attrinfo): Likewise.
13057 (MAXRESTROPNDS): New macro.
13058 (prototype): New struct.
13059 (MAXBIFS): New macro.
13060 (bifdata): New struct.
13061 (bifs): New variable.
13062 (curr_bif): Likewise.
13063 (bif_order): Likewise.
13064 (bif_index): Likewise.
13065 (fatal): New function.
13066 (stanza_name_to_stanza): Likewise.
13067 (parse_bif_attrs): New stub function.
13068 (parse_prototype): Likewise.
13069 (parse_bif_entry): New function.
13070 (parse_bif_stanza): Likewise.
13071 (parse_bif): Implement.
13072 (set_bif_order): New function.
13073 (create_bif_order): Implement.
13075 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13077 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
13078 (num_bifs): New variable.
13079 (num_ovld_stanzas): Likewise.
13080 (num_ovlds): Likewise.
13081 (parse_codes): New enum.
13082 (bif_rbt): New variable.
13083 (ovld_rbt): Likewise.
13084 (fntype_rbt): Likewise.
13085 (bifo_rbt): Likewise.
13086 (parse_bif): New stub function.
13087 (create_bif_order): Likewise.
13088 (parse_ovld): Likewise.
13089 (write_header_file): Likewise.
13090 (write_init_file): Likewise.
13091 (write_defines_file): Likewise.
13092 (delete_output_files): New function.
13095 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
13098 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
13099 from CRC32 _builtin functions.
13101 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
13103 * coverage.c (build_gcov_info_var_registration): Mark the object placed
13104 in the linker set as referenced so that it does not get optimized away.
13106 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
13109 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
13111 * config.gcc (riscv*-*-*): Detect which python is available.
13113 2021-07-21 Jakub Jelinek <jakub@redhat.com>
13115 PR middle-end/101535
13116 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
13117 contexts in which decl isn't privatized and for ORT_TARGET return
13118 false if decl is mapped.
13120 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
13122 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
13123 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
13124 has been successfully unrolled.
13126 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
13128 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
13129 exit condition if no tail loop is needed, and if the original exit
13130 condition should therefore be kept as-is.
13131 (tree_transform_and_unroll_loop): Handle that case here too.
13133 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
13135 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
13137 (free_data_refs): Likewise.
13138 * tree-data-ref.h (free_dependence_relations): Likewise.
13139 (free_data_refs): Likewise.
13140 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
13142 (struct component): Likewise.
13143 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
13144 (pcom_worker::~pcom_worker): Likewise.
13145 (pcom_worker::release_chain): Adjust as auto_vec changes.
13146 (pcom_worker::loop): Rename to ...
13147 (pcom_worker::m_loop): ... this.
13148 (pcom_worker::datarefs): Rename to ...
13149 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
13150 (pcom_worker::dependences): Rename to ...
13151 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
13152 (pcom_worker::chains): Rename to ...
13153 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
13154 (pcom_worker::looparound_phis): Rename to ...
13155 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
13157 (pcom_worker::cache): Rename to ...
13158 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
13159 (pcom_worker::release_chain): Adjust for auto_vec changes.
13160 (pcom_worker::release_chains): Adjust for auto_vec and renaming
13162 (release_component): Remove.
13163 (release_components): Adjust for release_component removal.
13164 (component_of): Adjust to use vec.
13165 (merge_comps): Likewise.
13166 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
13167 (pcom_worker::determine_offset): Likewise.
13168 (class comp_ptrs): Remove.
13169 (pcom_worker::split_data_refs_to_components): Adjust for renaming
13170 changes, for comp_ptrs removal with auto_vec.
13171 (pcom_worker::suitable_component_p): Adjust for renaming changes.
13172 (pcom_worker::filter_suitable_components): Adjust for release_component
13174 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
13175 (pcom_worker::find_looparound_phi): Likewise.
13176 (pcom_worker::add_looparound_copies): Likewise.
13177 (pcom_worker::determine_roots_comp): Likewise.
13178 (pcom_worker::single_nonlooparound_use): Likewise.
13179 (pcom_worker::execute_pred_commoning_chain): Likewise.
13180 (pcom_worker::execute_pred_commoning): Likewise.
13181 (pcom_worker::try_combine_chains): Likewise.
13182 (pcom_worker::prepare_initializers_chain): Likewise.
13183 (pcom_worker::prepare_initializers): Likewise.
13184 (pcom_worker::prepare_finalizers_chain): Likewise.
13185 (pcom_worker::prepare_finalizers): Likewise.
13186 (pcom_worker::tree_predictive_commoning_loop): Likewise.
13188 2021-07-20 Martin Sebor <msebor@redhat.com>
13190 PR middle-end/101397
13191 * builtins.c (gimple_call_return_array): Add argument. Correct
13192 offsets for memchr, mempcpy, stpcpy, and stpncpy.
13193 (compute_objsize_r): Adjust offset computation for argument returning
13196 2021-07-20 Martin Sebor <msebor@redhat.com>
13198 PR middle-end/101300
13199 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
13201 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
13203 * function.c (assign_parm_setup_block): Use adjust_address instead
13204 of change_address to preserve MEM_EXPR and friends.
13206 2021-07-20 Martin Sebor <msebor@redhat.com>
13208 * cfgloop.h (single_likely_exit): Adjust by-value argument to
13209 by-const-reference.
13210 * cfgloopanal.c (single_likely_exit): Same.
13211 * cgraph.h (struct cgraph_node): Same.
13212 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
13213 * genautomata.c (merge_states): Same.
13214 * genextract.c (VEC_char_to_string): Same.
13215 * genmatch.c (dt_node::gen_kids_1): Same.
13216 (walk_captures): Adjust by-value argument to by-reference.
13217 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
13218 to by-const-reference.
13219 * gimple.c (gimple_build_call_vec): Same.
13220 (gimple_build_call_internal_vec): Same.
13221 (gimple_build_switch): Same.
13222 (sort_case_labels): Same.
13223 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
13225 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
13226 by-const-reference.
13227 (gimple_build_call_internal_vec): Same.
13228 (gimple_build_switch): Same.
13229 (sort_case_labels): Same.
13230 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
13232 * haifa-sched.c (calc_priorities): Adjust by-value argument to
13233 by-const-reference.
13234 (sched_init_luids): Same.
13235 (haifa_init_h_i_d): Same.
13236 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
13237 (adjust_callers_for_value_intersection): Adjust by-value argument to
13239 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
13240 by-const-reference.
13241 (find_more_contexts_for_caller_subset): Same.
13242 (find_aggregate_values_for_callers_subset): Same.
13243 (copy_useful_known_contexts): Same.
13244 * ipa-fnsummary.c (remap_edge_summaries): Same.
13245 (remap_freqcounting_predicate): Same.
13246 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
13248 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
13249 to by-const-reference.
13250 * ipa-predicate.h (predicate::remap_after_inlining): Same.
13251 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
13252 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
13253 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
13254 * read-rtl.c (add_overload_instance): Same.
13255 * rtl.h (native_decode_rtx): Same.
13256 (native_decode_vector_rtx): Same.
13257 * sched-int.h (sched_init_luids): Same.
13258 (haifa_init_h_i_d): Same.
13259 * simplify-rtx.c (native_decode_vector_rtx): Same.
13260 (native_decode_rtx): Same.
13261 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
13262 (shrink_wrap_one_built_in_call_with_conds): Same.
13263 (shrink_wrap_conditional_dead_built_in_calls): Same.
13264 * tree-data-ref.c (create_runtime_alias_checks): Same.
13265 (compute_all_dependences): Same.
13266 * tree-data-ref.h (compute_all_dependences): Same.
13267 (create_runtime_alias_checks): Same.
13268 (index_in_loop_nest): Same.
13269 * tree-if-conv.c (mask_exists): Same.
13270 * tree-loop-distribution.c (class loop_distribution): Same.
13271 (loop_distribution::create_rdg_vertices): Same.
13272 (dump_rdg_partitions): Same.
13273 (debug_rdg_partitions): Same.
13274 (partition_contains_all_rw): Same.
13275 (loop_distribution::distribute_loop): Same.
13276 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
13277 (oacc_entry_exit_single_gang): Same.
13278 * tree-ssa-loop-im.c (hoist_memory_references): Same.
13279 (loop_suitable_for_sm): Same.
13280 * tree-ssa-loop-niter.c (bound_index): Same.
13281 * tree-ssa-reassoc.c (update_ops): Same.
13282 (swap_ops_for_binary_stmt): Same.
13283 (rewrite_expr_tree): Same.
13284 (rewrite_expr_tree_parallel): Same.
13285 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
13286 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
13287 * tree-ssa-structalias.c (process_all_all_constraints): Same.
13288 (make_constraints_to): Same.
13289 (handle_lhs_call): Same.
13290 (find_func_aliases_for_builtin_call): Same.
13291 (sort_fieldstack): Same.
13292 (check_for_overlaps): Same.
13293 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
13294 (vect_create_cond_for_unequal_addrs): Same.
13295 (vect_create_cond_for_lower_bounds): Same.
13296 (vect_create_cond_for_alias_checks): Same.
13297 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
13298 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
13299 (vect_make_slp_decision): Same.
13300 (vect_slp_bbs): Same.
13301 (duplicate_and_interleave): Same.
13302 (vect_transform_slp_perm_load): Same.
13303 (vect_schedule_slp): Same.
13304 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
13305 (vect_schedule_slp): Same.
13306 (duplicate_and_interleave): Same.
13307 * tree.c (build_vector_from_ctor): Same.
13308 (build_vector): Same.
13309 (check_vector_cst): Same.
13310 (check_vector_cst_duplicate): Same.
13311 (check_vector_cst_fill): Same.
13312 (check_vector_cst_stepped): Same.
13313 * tree.h (build_vector_from_ctor): Same.
13315 2021-07-20 Jakub Jelinek <jakub@redhat.com>
13318 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
13319 type from bool to int.
13320 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
13321 EASY_VECTOR_MSB case if either step or copies is not 1.
13322 (vspltis_shifted): Fix comment typo.
13323 (easy_altivec_constant): Change return type from bool to int, instead
13324 of returning true return byte size of the element mode that should be
13325 used to synthetize the constant.
13326 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
13327 that vspltis_shifted is 0, handle the case where easy_altivec_constant
13328 assumes using different vector mode from CONST_VECTOR's mode.
13329 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
13330 easy_altivec_constant to determine mode in which -1 >> -1 should be
13331 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
13333 2021-07-20 Richard Biener <rguenther@suse.de>
13336 * dwarf2out.h (dwarf_file_data): Add key member.
13337 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
13338 (dwarf_file_hasher::hash): Hash key.
13339 (lookup_filename): Remap the filename and store it in the
13340 filename member of dwarf_file_data when creating a new
13342 (file_name_acquire): Do not remap the filename again.
13343 (maybe_emit_file): Likewise.
13345 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
13347 * config/aarch64/aarch64-simd-builtins.def: Use two variant
13348 generators for all TBL/TBX intrinsics and rename to
13349 consistent forms: qtbl[1234] or qtbx[1234].
13350 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
13352 (aarch64_qtbl1<mode>): This.
13353 (aarch64_tbx1<mode>): Rename to...
13354 (aarch64_qtbx1<mode>): This.
13355 (aarch64_tbl2v16qi): Delete.
13356 (aarch64_tbl3<mode>): Rename to...
13357 (aarch64_qtbl2<mode>): This.
13358 (aarch64_tbx4<mode>): Rename to...
13359 (aarch64_qtbx2<mode>): This.
13360 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
13361 renamed qtbl1 and qtbl2 RTL patterns.
13362 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
13364 (vqtbl1_s8): Likewise.
13365 (vqtbl1_u8): Likewise.
13366 (vqtbl1q_p8): Likewise.
13367 (vqtbl1q_s8): Likewise.
13368 (vqtbl1q_u8): Likewise.
13369 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
13370 (vqtbx1_u8): Likewise.
13371 (vqtbx1_p8): Likewise.
13372 (vqtbx1q_s8): Likewise.
13373 (vqtbx1q_u8): Likewise.
13374 (vqtbx1q_p8): Likewise.
13375 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
13376 (vtbl1_u8): Likewise.
13377 (vtbl1_p8): Likewise.
13378 (vtbl2_s8): Likewise
13379 (vtbl2_u8): Likewise.
13380 (vtbl2_p8): Likewise.
13381 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
13382 (vtbl3_u8): Likewise.
13383 (vtbl3_p8): Likewise.
13384 (vtbl4_s8): Likewise.
13385 (vtbl4_u8): Likewise.
13386 (vtbl4_p8): Likewise.
13387 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
13388 (vtbx2_u8): Likewise.
13389 (vtbx2_p8): Likewise.
13390 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
13391 (vqtbl2_u8): Likewise.
13392 (vqtbl2_p8): Likewise.
13393 (vqtbl2q_s8): Likewise.
13394 (vqtbl2q_u8): Likewise.
13395 (vqtbl2q_p8): Likewise.
13396 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
13397 (vqtbx2_u8): Likewise.
13398 (vqtbx2_p8): Likewise.
13399 (vqtbx2q_s8): Likewise.
13400 (vqtbx2q_u8): Likewise.
13401 (vqtbx2q_p8): Likewise.
13402 (vtbx4_s8): Likewise.
13403 (vtbx4_u8): Likewise.
13404 (vtbx4_p8): Likewise.
13406 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
13409 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
13411 (define_peephole2 atomic_loaddi_fpu): Ditto.
13413 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
13415 * config.gcc (riscv*-*-*): Detect which python is available.
13417 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
13419 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
13420 (smul<mode>3_highpart): ... this.
13421 (mulhu_<mode>): Rename to...
13422 (umul<mode>3_highpart): ... this.
13423 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
13424 MULHU_V2DI, MULHU_V4SI): Adjust.
13426 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
13428 PR tree-optimization/100696
13429 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
13430 * internal-fn.def (IFN_MULH): New internal function.
13431 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
13432 recog normal multiply highpart as IFN_MULH.
13433 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
13436 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
13438 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
13439 (BTF_DEBUGGING_INFO): Likewise.
13440 * doc/tm.texi.in: Document the new macros.
13441 * doc/tm.texi: Regenerated.
13442 * toplev.c: Guard initialization of debug hooks.
13444 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
13446 * flags.h (ctf_debuginfo_p): New function declaration.
13447 * opts.c (ctf_debuginfo_p): New function definition.
13449 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
13452 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
13453 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
13454 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
13455 * config/gcn/gcn-valu.md: Add a warning comment.
13456 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
13457 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
13458 * config/gcn/gcn.md: Add a warning comment.
13459 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
13460 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
13462 (EF_AMDGPU_XNACK): New.
13463 (EF_AMDGPU_SRAM_ECC): New.
13465 (copy_early_debug_info): Use elf_flags.
13466 (main): Handle -mxnack and -msram-ecc options.
13467 * doc/invoke.texi: Document -mxnack and -msram-ecc.
13469 2021-07-19 Andrew Pinski <apinski@marvell.com>
13472 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
13473 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
13475 2021-07-19 Richard Biener <rguenther@suse.de>
13477 PR tree-optimization/101505
13478 * tree-vect-patterns.c (vect_determine_precisions): Walk
13479 PHIs also for loop vectorization.
13481 2021-07-19 Richard Biener <rguenther@suse.de>
13483 * gimple.h (gimple_expr_type): Remove.
13484 * doc/gimple.texi: Remove gimple_expr_type documentation.
13486 2021-07-19 Richard Biener <rguenther@suse.de>
13488 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
13489 (ao_ref_init_from_vn_reference): Likewise.
13490 (fully_constant_reference): Likewise.
13491 (vn_reference_lookup_call): Do not set vr->type to random
13493 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
13495 * tree-vect-generic.c (expand_vector_piecewise): Pass in
13496 whether we expanded parallel.
13497 (expand_vector_parallel): Adjust.
13498 (expand_vector_addition): Likewise.
13499 (expand_vector_comparison): Likewise.
13500 (expand_vector_operation): Likewise.
13501 (expand_vector_scalar_condition): Likewise.
13502 (expand_vector_conversion): Likewise.
13504 2021-07-19 Richard Biener <rguenther@suse.de>
13506 * tree-vrp.c (register_edge_assert_for_2): Use the
13508 (vrp_folder::fold_predicate_in): Likewise.
13509 * vr-values.c (gimple_assign_nonzero_p): Likewise.
13510 (vr_values::extract_range_from_comparison): Likewise.
13511 (vr_values::extract_range_from_ubsan_builtin): Use the
13512 type of the first operand.
13513 (vr_values::extract_range_basic): Push down type
13514 computation, use the appropriate LHS.
13515 (vr_values::extract_range_from_assignment): Use the
13518 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
13521 * common/config/i386/i386-common.c (ix86_handle_option): For
13522 -mgeneral-regs-only, enable the GPR only instructions which are
13523 enabled implicitly by SSE ISAs unless they have been disabled
13526 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
13529 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
13530 ix86_avx_u128_mode_needed.
13531 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
13532 returns AVX register.
13534 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
13536 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
13537 (handle_const_call): Liekise
13538 (handle_pure_call): Liekise
13540 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
13542 PR tree-optimization/96542
13543 * range-op.cc (range_operator::wi_fold_in_parts): New.
13544 (range_operator::fold_range): Call wi_fold_in_parts.
13545 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
13546 * range-op.h (wi_fold_in_parts): Add prototype.
13548 2021-07-16 David Malcolm <dmalcolm@redhat.com>
13550 * doc/analyzer.texi: Add __analyzer_dump_state.
13552 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13554 * config/rs6000/rbtree.c: New file.
13555 * config/rs6000/rbtree.h: New file.
13557 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13559 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
13560 (typeinfo): Add restr field.
13561 (match_bracketed_pair): New function.
13562 (match_const_restriction): Implement.
13564 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13566 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
13568 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13570 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
13571 (basetype): Likewise.
13572 (typeinfo): Likewise.
13573 (handle_pointer): New function.
13574 (match_basetype): New stub function.
13575 (match_const_restriction): Likewise.
13576 (match_type): New function.
13578 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13580 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
13582 (advance_line): Likewise.
13583 (safe_inc_pos): Likewise.
13584 (match_identifier): Likewise.
13585 (match_integer): Likewise.
13586 (match_to_right_bracket): Likewise.
13588 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13590 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
13591 (ovld_file): Likewise.
13592 (header_file): Likewise.
13593 (init_file): Likewise.
13594 (defines_file): Likewise.
13595 (pgm_path): Likewise.
13596 (bif_path): Likewise.
13597 (ovld_path): Likewise.
13598 (header_path): Likewise.
13599 (init_path): Likewise.
13600 (defines_path): Likewise.
13601 (LINELEN): New macro.
13602 (linebuf): New variable.
13606 (bif_diag): New function.
13607 (ovld_diag): Likewise.
13609 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13611 * config/rs6000/rs6000-builtin-new.def: New.
13612 * config/rs6000/rs6000-overload.def: New.
13614 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13616 * config/rs6000/rs6000-gen-builtins.c: New.
13618 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
13620 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
13621 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
13622 * gengtype-state.c (state_writer::write_state_file_list): Add a
13623 parameter to the fileslist expression for the number of build
13625 (read_state_files_list): Detect build headers and strip the
13626 initial "./" or ".\" from their names.
13627 * gengtype.c (build_headers): New global variable.
13628 (num_build_headers): Likewise.
13629 (open_base_files): Emit #include for each build header.
13630 (main): Detect and count build headers.
13631 * gengtype.h (build_headers): New extern variable.
13632 (num_build_headers): Likewise.
13634 2021-07-16 Richard Biener <rguenther@suse.de>
13636 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
13637 the type of the LHS.
13638 (find_bswap_or_nop_1): Likewise.
13639 (find_bswap_or_nop): Likewise.
13640 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
13642 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
13643 Remove unused parameters, pass in the scalar type. Fix
13644 internal store function handling.
13645 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
13646 (vect_get_vector_types_for_stmt): Move down check for
13647 existing vector stmt after we've determined a scalar type.
13648 Pass down the used scalar type to vect_get_smallest_scalar_type.
13649 * tree-vect-generic.c (expand_vector_condition): Use
13650 the type of the LHS.
13651 (expand_vector_scalar_condition): Likewise.
13652 (expand_vector_operations_1): Likewise.
13653 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
13654 (vect_recog_dot_prod_pattern): Likewise.
13655 (vect_recog_sad_pattern): Likewise.
13656 (vect_recog_widen_op_pattern): Likewise.
13657 (vect_recog_widen_sum_pattern): Likewise.
13658 (vect_recog_mixed_size_cond_pattern): Likewise.
13660 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
13662 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
13663 (dump_eaf_flags): Dump EAF_NOT_RETURNED
13664 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
13665 and EAF_NOT_RETURNED.
13666 (modref_summary::useful_p): Likewise.
13667 (modref_summary_lto::useful_p): Likewise.
13668 (struct) modref_summary_lto: Use eaf_fleags_t.
13669 (deref_flags): Handle EAF_NOT_RETURNED.
13670 (struct escape_point): Use min_flags.
13671 (modref_lattice::init): Add EAF_NOT_RETURNED.
13672 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
13673 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
13675 (analyze_parms): Also analyze const functions; update conition on
13677 (modref_write): Update streaming.
13678 (read_section): Update streaming.
13679 (remap_arg_flags): Use eaf_flags_t.
13680 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
13681 * ipa-modref.h: (eaf_flags_t): New typedef.
13682 (struct modref_summary): Use eaf_flags_t.
13683 * tree-core.h (EAF_NOT_RETURNED): New constant.
13685 2021-07-16 Richard Biener <rguenther@suse.de>
13687 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
13688 the type of the LHS.
13689 (gimple_assign_nonnegative_warnv_p): Likewise.
13690 (gimple_call_nonnegative_warnv_p): Likewise. Return false
13691 if the call has no LHS.
13692 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
13693 * tree-eh.c (stmt_could_throw_1_p): Likewise.
13694 * tree-inline.c (insert_init_stmt): Likewise.
13695 * tree-ssa-loop-niter.c (get_val_for): Likewise.
13696 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
13698 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
13699 gassign *. Use the type of the lhs.
13700 (vn_nary_op_lookup_stmt): Adjust.
13701 (vn_nary_op_insert_stmt): Likewise.
13703 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
13705 * config/s390/predicates.md (bras_sym_operand): Accept all
13706 functions in 64-bit mode, use UNSPEC_PLT31.
13707 (larl_operand): Use UNSPEC_PLT31.
13708 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
13709 (legitimize_pic_address): Likewise.
13710 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
13712 (s390_delegitimize_address): Use UNSPEC_PLT31.
13713 (s390_output_addr_const_extra): Likewise.
13714 (print_operand): Add @PLT to TLS calls, handle %K.
13715 (s390_function_profiler): Mark __fentry__/_mcount as function,
13716 use %K, use UNSPEC_PLT31.
13717 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
13718 (s390_emit_call): Use UNSPEC_PLT31.
13719 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
13720 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
13721 (*movdi_64): Use %K.
13722 (reload_base_64): Likewise.
13723 (*sibcall_brc): Likewise.
13724 (*sibcall_brcl): Likewise.
13725 (*sibcall_value_brc): Likewise.
13726 (*sibcall_value_brcl): Likewise.
13728 (*brasl): Likewise.
13729 (*bras_r): Likewise.
13730 (*brasl_r): Likewise.
13731 (*bras_tls): Likewise.
13732 (*brasl_tls): Likewise.
13733 (main_base_64): Likewise.
13734 (reload_base_64): Likewise.
13735 (@split_stack_call<mode>): Likewise.
13737 2021-07-16 Richard Biener <rguenther@suse.de>
13739 PR tree-optimization/101467
13740 * tree-vect-stmts.c (vect_gen_while): Properly guard
13741 make_temp_ssa_name usage.
13743 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
13745 * config.gcc: Don't use forked print-sysroot-suffix.sh and
13746 t-sysroot-suffix for C-SKY.
13747 * config/csky/print-sysroot-suffix.sh: Delete.
13748 * config/csky/t-csky-linux: Delete.
13749 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
13750 instead of CSKY_MULTILIB_DIRNAMES.
13752 2021-07-16 Richard Biener <rguenther@suse.de>
13754 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
13755 conversion issues with the partial reduction of the reused
13756 vector accumulator.
13758 2021-07-16 Richard Biener <rguenther@suse.de>
13760 * config/i386/i386-options.c (ix86_option_override_internal): Set
13761 param_vect_partial_vector_usage to zero if not set.
13763 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
13766 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
13767 (VALID_INT_MODE_P): Add SDmode and DDmode.
13768 Add TDmode for TARGET_64BIT.
13769 (VALID_DFP_MODE_P): Remove.
13770 * config/i386/i386.c (ix86_hard_regno_mode_ok):
13771 Do not use VALID_DFP_MODE_P.
13773 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
13775 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
13777 (fold_using_range::fold_stmt): Ditto.
13778 (fold_using_range::range_of_range_op): Ditto.
13779 (fold_using_range::range_of_phi): Ditto.
13780 (fold_using_range::range_of_call): Ditto.
13781 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
13782 (fold_using_range::range_of_builtin_call): Ditto.
13783 (fold_using_range::range_of_cond_expr): Ditto.
13784 * gimple-range-fold.h (gimple_range_type): New.
13786 2021-07-15 Martin Sebor <msebor@redhat.com>
13788 PR middle-end/97027
13789 * tree-ssa-strlen.c (handle_assign): New function.
13790 (maybe_warn_overflow): Add argument.
13791 (nonzero_bytes_for_type): New function.
13792 (count_nonzero_bytes): Handle more tree types. Call
13793 nonzero_bytes_for_tye.
13794 (count_nonzero_bytes): Handle types.
13795 (handle_store): Handle stores from function calls.
13796 (strlen_check_and_optimize_call): Move code to handle_assign. Call
13797 it for assignments from function calls.
13799 2021-07-15 David Malcolm <dmalcolm@redhat.com>
13804 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
13806 2021-07-15 David Malcolm <dmalcolm@redhat.com>
13808 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
13810 2021-07-15 Martin Sebor <msebor@redhat.com>
13814 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
13815 (operand_compare::verify_hash_value): Same.
13816 * tree-core.h (OEP_DECL_NAME): New.
13818 2021-07-15 Martin Jambor <mjambor@suse.cz>
13820 * profile-count.h (profile_count::value): Change the return type to
13822 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
13824 * tree-cfg.c (dump_function_to_file): Likewise.
13826 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
13829 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
13830 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
13833 2021-07-15 Richard Biener <rguenther@suse.de>
13835 * tree-vectorizer.h (vect_gen_while): Match up with
13836 vect_gen_while_not.
13837 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
13838 of vect_gen_while_not.
13839 (vect_gen_while_not): Adjust.
13840 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
13842 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
13844 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
13845 (ranger_cache::range_of_def): Call adjust_range.
13846 (ranger_cache::entry_range): Same.
13847 * gimple-range-cache.h (non_null_ref::adjust_range): New.
13848 * gimple-range.cc (gimple_ranger::range_of_expr): Call
13850 (gimple_ranger::range_on_entry): Same.
13852 2021-07-15 Tamar Christina <tamar.christina@arm.com>
13855 2021-07-14 Tamar Christina <tamar.christina@arm.com>
13857 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
13859 2021-07-15 Tamar Christina <tamar.christina@arm.com>
13862 2021-07-14 Tamar Christina <tamar.christina@arm.com>
13864 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
13865 (sdot_prod, udot_prod): ...These.
13866 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
13867 (aarch64_<sur>dot<vsi2qi>): Rename to...
13868 (<sur>dot_prod<vsi2qi>): ...This.
13869 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
13872 2021-07-15 Jakub Jelinek <jakub@redhat.com>
13874 PR middle-end/101437
13875 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
13876 types even if they have non-BLKmode TYPE_MODE.
13878 2021-07-15 Richard Biener <rguenther@suse.de>
13881 * gcc.c (process_command): Process -gtoggle like process_options
13882 would after parsing options.
13884 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
13886 * cfgexpand.c (expand_asm_loc): Adjust.
13887 (expand_asm_stmt): Likewise.
13888 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
13889 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
13890 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
13891 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
13892 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
13893 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
13894 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
13895 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
13896 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
13897 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
13898 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
13899 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
13900 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
13901 * doc/tm.texi: Regenerate.
13902 * target.def: Add location argument to md_asm_adjust.
13904 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
13906 * tree-diagnostic.c (diagnostic_report_current_function): Use the
13907 diagnostic's location, not input_location.
13909 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
13911 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
13913 (expand_asm_stmt): Likewise.
13915 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
13917 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
13918 addressed memory rtx, if any.
13919 (rs6000_split_multireg_move): Fix code formatting.
13920 Handle MMA build built-ins with operands in adjacent memory locations.
13922 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
13924 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
13927 2021-07-14 Jason Merrill <jason@redhat.com>
13929 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
13931 2021-07-14 Jason Merrill <jason@redhat.com>
13933 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
13936 2021-07-14 Tamar Christina <tamar.christina@arm.com>
13938 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13939 Remove erroneous line.
13941 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
13943 * params.opt (param_evrp_mode): Change default.
13945 2021-07-14 Tamar Christina <tamar.christina@arm.com>
13947 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
13948 (sdot_prod, udot_prod): ...These.
13949 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
13950 (aarch64_<sur>dot<vsi2qi>): Rename to...
13951 (<sur>dot_prod<vsi2qi>): ...This.
13952 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
13955 2021-07-14 Tamar Christina <tamar.christina@arm.com>
13957 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
13959 2021-07-14 Tamar Christina <tamar.christina@arm.com>
13961 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
13963 2021-07-14 Tamar Christina <tamar.christina@arm.com>
13965 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
13967 2021-07-14 Tamar Christina <tamar.christina@arm.com>
13969 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
13970 (usdot_prod<vsi2qi>): ... This.
13971 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
13972 (usdot_prod): ...This.
13973 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
13974 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
13976 (@<sur>dot_prod<vsi2qi>): ...This.
13977 * config/aarch64/aarch64-sve-builtins-base.cc
13978 (svusdot_impl::expand): Use it.
13980 2021-07-14 Tamar Christina <tamar.christina@arm.com>
13982 * optabs.def (usdot_prod_optab): New.
13983 * doc/md.texi: Document it and clarify other dot prod optabs.
13984 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
13985 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
13986 * optabs.c (expand_widen_pattern_expr): Likewise.
13987 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
13988 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
13989 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
13991 (vect_widened_op_tree): Optionally ignore
13993 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
13995 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
13998 * config/i386/driver-i386.c (host_detect_local_cpu): Check
13999 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
14000 Enable UINTR only for 64-bit codegen.
14001 * config/i386/i386-options.c
14002 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
14004 * config/i386/i386.h (ARCH_ARG): New.
14005 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
14006 "[arch|tune] 64" for 64-bit codegen.
14008 2021-07-14 Richard Biener <rguenther@suse.de>
14010 PR tree-optimization/101445
14011 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
14012 of the IV in the correct direction for negative stride
14015 2021-07-14 Jakub Jelinek <jakub@redhat.com>
14018 * godump.c (godump_str_hash): New type.
14019 (godump_container::pot_dummy_types): Use string_hash instead of
14020 ptr_hash in the hash_set.
14022 2021-07-14 Richard Biener <rguenther@suse.de>
14024 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
14025 vector types where the old vector type has a multiple of
14026 the new vector type elements.
14027 (vect_create_partial_epilog): New function, split out from...
14028 (vect_create_epilog_for_reduction): ... here.
14029 (vect_transform_cycle_phi): Reduce the re-used accumulator
14030 to the new vector type.
14032 2021-07-14 Alexandre Oliva <oliva@adacore.com>
14034 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
14035 non-'t'-sized arg check.
14037 2021-07-14 Alexandre Oliva <oliva@adacore.com>
14039 * tree-cfg.c (cleanup_dead_labels_eh): Update
14040 post_landing_pad label upon change of landing pad block's
14042 (cleanup_dead_labels): Check that a removed label is not that
14045 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
14047 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
14049 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
14050 Add Neon to general purpose register case for zero-extend
14052 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
14053 case to prevent some cases opting to go through memory.
14054 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
14055 * rtl.c (rtvec_series_p): Define predicate to determine
14056 whether a vector contains a linear series of integers.
14057 * rtl.h (rtvec_series_p): Define.
14058 * rtlanal.c (vec_series_lowpart_p): Define predicate to
14059 determine if a vector selection is equivalent to the low part
14061 * rtlanal.h (vec_series_lowpart_p): Define.
14062 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
14063 Add vec_select -> subreg simplification.
14065 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
14067 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
14068 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
14069 _mm_test_mix_ones_zeros): New.
14071 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
14072 Richard Biener <rguenther@suse.de>
14074 * gimple.c (gimple_could_trap_p_1): Make S argument a
14075 "const gimple*". Preserve constness in call to
14076 gimple_asm_volatile_p.
14077 (gimple_could_trap_p): Make S argument a "const gimple*".
14078 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
14079 Update function prototypes.
14081 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14083 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
14084 (_loop_vec_info::main_loop_edge): New field.
14085 (_loop_vec_info::skip_main_loop_edge): Likewise.
14086 (_loop_vec_info::skip_this_loop_edge): Likewise.
14087 (_loop_vec_info::reusable_accumulators): Likewise.
14088 (_stmt_vec_info::reduc_scalar_results): Likewise.
14089 (_stmt_vec_info::reused_accumulator): Likewise.
14090 (vect_get_main_loop_result): Declare.
14091 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
14092 reduc_scalar_inputs.
14093 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
14094 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
14095 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
14096 skip_main_loop_edge and skip_this_loop_edge fields.
14097 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
14098 (vect_emit_reduction_init_stmts): New function.
14099 (get_initial_def_for_reduction): Use it.
14100 (get_initial_defs_for_reduction): Likewise. Change the vinfo
14101 parameter to a loop_vec_info.
14102 (vect_create_epilog_for_reduction): Store the scalar results
14103 in the reduc_info. If an epilogue loop is reusing an accumulator
14104 from the main loop, and if the epilogue loop can also be skipped,
14105 try to place the reduction code in the join block. Record
14106 accumulators that could potentially be reused by epilogue loops.
14107 (vect_transform_cycle_phi): When vectorizing epilogue loops,
14108 try to reuse accumulators from the main loop. Record the initial
14109 value in reduc_info for non-SLP reductions too.
14111 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14113 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
14114 adjustment handling. Take the neutral value as an argument,
14115 in place of the code argument.
14116 (vect_transform_cycle_phi): Update accordingly. Handle the
14117 initial values of cond reductions separately from code reductions.
14118 Choose the adjustment here rather than in
14119 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
14121 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14123 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
14124 (neutral_op_for_reduction): ...this, providing a more general
14126 (vect_create_epilog_for_reduction): Update accordingly.
14127 (vectorizable_reduction): Likewise.
14128 (vect_transform_cycle_phi): Likewise.
14130 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14132 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
14133 reduc_info instead of the original stmt_vec_info.
14134 (vect_transform_cycle_phi): Update accordingly.
14136 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14138 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
14139 reduc_info as an additional parameter.
14140 (vect_transform_cycle_phi): Update accordingly.
14142 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14144 * tree-vectorizer.h: Include tree-ssa-operands.h.
14145 (vect_phi_initial_value): New function.
14146 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
14147 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
14148 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
14149 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
14151 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14153 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
14154 the phi results to vectype after creating them. Remove later
14155 conversion code that thus becomes redundant.
14157 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14159 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
14160 the new_phis vector with a reduc_inputs vector. Combine handling
14161 of reduction chains and ncopies > 1.
14163 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14165 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
14166 scalar_results to group_size elements after reducing down from
14167 N*group_size elements. Construct an array_slice of the live-out
14168 stmts and assert that there is one stmt per scalar result.
14170 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14172 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
14173 nested_in_vect_loop and use double_reduc everywhere. Remove dead
14174 assignment to "loop".
14176 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14178 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
14179 vector types first. For scalar types, consider both the preferred
14180 vector mode and the alternative vector modes.
14181 * optabs-query.c (can_vec_mask_load_store_p): Use the same
14182 structure as above, in particular using related_vector_mode
14183 for modes provided by autovectorize_vector_modes.
14185 2021-07-13 Jakub Jelinek <jakub@redhat.com>
14186 Richard Biener <rguenther@suse.de>
14188 PR tree-optimization/101419
14189 * tree-pass.h (PROP_objsz): Define.
14190 (make_pass_early_object_sizes): Declare.
14191 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
14192 there to pass_early_object_sizes, drop parameter.
14193 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
14194 drop parameter, move pass_post_ipa_warn right after that.
14195 * tree-object-size.c (pass_object_sizes::execute): Rename to...
14196 (object_sizes_execute): ... this. Add insert_min_max_p argument.
14197 (pass_data_object_sizes): Move after object_sizes_execute.
14198 (pass_object_sizes): Likewise. In execute method call
14199 object_sizes_execute, drop set_pass_param method and insert_min_max_p
14200 non-static data member and its initializer in the ctor.
14201 (pass_data_early_object_sizes, pass_early_object_sizes,
14202 make_pass_early_object_sizes): New.
14203 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
14204 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
14206 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
14209 * config/riscv/constraints.md ("S"): Update description and remove
14211 * doc/md.texi (Machine Constraints): Document the 'S' constraints
14214 2021-07-13 Richard Biener <rguenther@suse.de>
14217 2021-07-12 Richard Biener <rguenther@suse.de>
14219 * tree-vect-slp.c (vect_slp_region): Show the number of
14220 SLP graph entries in the optimization message.
14222 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
14224 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
14226 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
14227 return type to long.
14228 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
14231 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
14233 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
14234 Query relation between the 2 operands and use it.
14236 2021-07-12 Sergei Trofimovich <siarheit@google.com>
14238 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
14240 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
14243 * config/i386/predicates.md (vec_setm_sse41_operand):
14244 Rename from vec_setm_operand.
14245 (vec_setm_avx2_operand): New predicate.
14246 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
14247 Use vec_setm_sse41_operand as operand 2 predicate.
14248 (vec_set<V_256_512:mode): New expander.
14249 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
14250 as operand 2 predicate.
14252 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
14254 PR tree-optimization/101335
14255 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
14257 2021-07-12 Andrew Pinski <apinski@marvell.com>
14259 * tree-ssa-phiopt.c (match_simplify_replacement): Move
14260 insert of the sequence before the movement of the
14261 statement. Check if to see if the statement is used
14262 outside of the original phi to see if we should move it.
14264 2021-07-12 Richard Biener <rguenther@suse.de>
14266 * dump-context.h (debug_dump_context::debug_dump_context):
14267 Add FILE * parameter defaulted to stderr.
14268 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
14269 * tree-vect-slp.c (dot_slp_tree): New functions.
14271 2021-07-12 Richard Biener <rguenther@suse.de>
14273 PR tree-optimization/101373
14274 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
14275 references when the BB may not return.
14276 (compute_avail): Pass in the function we're working on and
14277 replace cfun references with it. Externally throwing
14278 const calls also possibly terminate the function.
14279 (pass_pre::execute): Pass down the function we're working on.
14280 * gcse.c (compute_hash_table_work): Externally throwing
14281 const/pure calls also need record_last_mem_set_info.
14282 * postreload-gcse.c (record_opr_changes): Looping or externally
14283 throwing const/pure calls also need record_last_mem_set_info.
14285 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
14287 * recog.c (memory_address_addr_space_p): Change the type to bool.
14288 Return true/false instead of 1/0.
14289 (offsettable_memref_p): Ditto.
14290 (offsettable_nonstrict_memref_p): Ditto.
14291 (offsettable_address_addr_space_p): Ditto.
14292 Change the type of addressp indirect function to bool.
14293 * recog.h (memory_address_addr_space_p): Change the type to bool.
14294 (strict_memory_address_addr_space_p): Ditto.
14295 (offsettable_memref_p): Ditto.
14296 (offsettable_nonstrict_memref_p): Ditto.
14297 (offsettable_address_addr_space_p): Ditto.
14298 * reload.c (maybe_memory_address_addr_space_p): Ditto.
14299 (strict_memory_address_addr_space_p): Change the type to bool.
14300 Return true/false instead of 1/0.
14301 (maybe_memory_address_addr_space_p): Change the type to bool.
14303 2021-07-12 Richard Biener <rguenther@suse.de>
14305 * tree-vect-slp.c (vect_slp_region): Show the number of
14306 SLP graph entries in the optimization message.
14308 2021-07-12 Richard Biener <rguenther@suse.de>
14310 PR tree-optimization/101394
14311 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
14312 copies from abnormals for a full redundancy.
14314 2021-07-12 Richard Biener <rguenther@suse.de>
14316 PR middle-end/101423
14317 * gimple.c (gimple_could_trap_p_1): Internal function calls
14319 * tree-eh.c (tree_could_trap_p): Likewise.
14321 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14324 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
14326 (vmulq_n_u32): Likewise.
14327 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
14328 (vmulq_n_f32): Likewise.
14329 (vmul_n_f16): Likewise.
14330 (vmulq_n_f16): Likewise.
14332 2021-07-12 Martin Liska <mliska@suse.cz>
14334 PR sanitizer/101425
14335 * gcc.c (check_offload_target_name): Call
14336 candidates_list_and_hint only if we have a candidate.
14338 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14341 * config/arm/neon.md (vec_init): Move to ...
14342 * config/arm/vec-common.md (vec_init): ... here.
14343 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
14345 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
14347 PR tree-optimization/101403
14348 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
14349 signedness of the shift is not the same as the signedness of
14350 the type extension.
14352 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
14353 Uroš Bizjak <ubizjak@gmail.com>
14355 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
14356 divmod of a constant numerator with new define_insn_and_split.
14358 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
14361 * config/i386/i386-expand.c (ix86_expand_call): If a call is
14362 to a non-local-binding, or local but to a public symbol, then
14363 assume that it might be indirected via the lazy symbol binder.
14364 Mark R10 and R10 as clobbered in that case.
14366 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
14369 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
14370 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
14371 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
14373 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
14375 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
14376 log2 range of operands[3] to [1,31].
14377 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
14379 2021-07-09 Sergei Trofimovich <siarheit@google.com>
14381 * doc/md.texi: Don't split @smallexample in multiple @groups.
14383 2021-07-09 Sergei Trofimovich <siarheit@google.com>
14385 * doc/md.texi: Add missing 'see' word.
14387 2021-07-09 Andrew Pinski <apinski@marvell.com>
14389 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
14390 to take sequence and gimple_match_op. Accept the case where
14391 op is a SSA_NAME and one statement in the sequence.
14392 Also allow constants.
14393 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
14394 Update call to phiopt_early_allow. Discard the sequence if not
14397 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
14402 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
14403 instead of ISA_HAS_MSA.
14404 (mips_expand_vec_unpack): Likewise.
14405 (mips_expand_vector_init): Likewise.
14407 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
14409 * config/rs6000/vsx.md (mods_<mode>): Rename to...
14410 (mod<mode>3): ... this.
14411 (modu_<mode>): Rename to...
14412 (umod<mode>3): ... this.
14413 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
14414 MODU_V4SI): Adjust.
14416 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
14418 * config/h8300/shiftrotate.md (variable shifts): Expose condition
14419 code handling for the test before the loop.
14421 2021-07-08 Martin Jambor <mjambor@suse.cz>
14424 * ipa-sra.c (class isra_call_summary): New member
14425 m_before_any_store, initialize it in the constructor.
14426 (isra_call_summary::dump): Dump the new field.
14427 (ipa_sra_call_summaries::duplicate): Copy it.
14428 (process_scan_results): Set it.
14429 (isra_write_edge_summary): Stream it.
14430 (isra_read_edge_summary): Likewise.
14431 (param_splitting_across_edge): Only override
14432 safe_to_import_accesses if m_before_any_store is set.
14434 2021-07-08 Martin Sebor <msebor@redhat.com>
14436 PR bootstrap/101374
14437 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
14438 Use Object Size Type 0 instead of 1.
14440 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
14442 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
14445 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
14447 * match.pd: Simplify an extend-operate-truncate sequence involving
14450 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
14451 Richard Biener <rguenther@suse.de>
14453 PR tree-optimization/40210
14454 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
14455 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
14456 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
14458 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
14461 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
14463 * config/i386/mmx.md (V_32): New mode iterator.
14464 (mov<V_32:mode>): Use V_32 mode iterator.
14465 (*mov<V_32:mode>_internal): Ditto.
14466 (*push<V_32:mode>2_rex64): Ditto.
14467 (*push<V_32:mode>2): Ditto.
14468 (movmisalign<V_32:mode>): Ditto.
14469 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
14470 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
14471 (vec_unpacks_lo_v4qi): New expander.
14472 (vec_unpacks_hi_v4qi): Ditto.
14473 (vec_unpacku_lo_v4qi): Ditto.
14474 (vec_unpacku_hi_v4qi): Ditto.
14475 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
14476 (VALID_INT_MODE_P): Ditto.
14478 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
14481 * config/rs6000/rs6000.md (udivti3): New insn.
14482 (divti3): New insn.
14483 (umodti3): New insn.
14484 (modti3): New insn.
14486 2021-07-07 Martin Sebor <msebor@redhat.com>
14488 PR tree-optimization/100137
14489 PR tree-optimization/99121
14490 PR tree-optimization/97027
14491 * builtins.c (access_ref::access_ref): Also set offmax.
14492 (access_ref::offset_in_range): Define new function.
14493 (access_ref::add_offset): Set offmax.
14494 (access_ref::inform_access): Handle access_none.
14495 (handle_mem_ref): Clear ostype.
14496 (compute_objsize_r): Handle ASSERT_EXPR.
14497 * builtins.h (struct access_ref): Add offmax member.
14498 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
14499 compute_objsize() and simplify.
14501 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
14503 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
14504 and VSX_BUILTIN_STXVP.
14506 2021-07-07 Martin Sebor <msebor@redhat.com>
14509 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
14510 a stray %K from error_at() missed in r12-2088.
14512 2021-07-07 Richard Biener <rguenther@suse.de>
14514 PR tree-optimization/99728
14515 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
14517 (mem_refs_may_alias_p): Add assert we handled aggregate
14519 (sm_seq_valid_bb): Give up when running into aggregate copies.
14520 (ref_indep_loop_p): Handle aggregate copies as never
14521 being invariant themselves but allow other refs to be
14522 disambiguated against them.
14523 (can_sm_ref_p): Do not try to apply store-motion to aggregate
14526 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
14529 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
14532 2021-07-06 Martin Sebor <msebor@redhat.com>
14534 * gimple-pretty-print.c (percent_G_format): Remove.
14535 * tree-diagnostic.c (default_tree_printer): Remove calls.
14536 * tree-pretty-print.c (percent_K_format): Remove.
14537 * tree-pretty-print.h (percent_K_format): Remove.
14539 2021-07-06 Martin Sebor <msebor@redhat.com>
14541 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
14542 Remove %K and use error_at.
14543 (aarch64_expand_fcmla_builtin): Same.
14544 (aarch64_expand_builtin_tme): Same.
14545 (aarch64_expand_builtin_memtag): Same.
14546 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
14547 (arm_expand_builtin): Same.
14548 * config/arm/arm.c (bounds_check): Same.
14550 2021-07-06 Martin Sebor <msebor@redhat.com>
14552 * builtins.c (warn_string_no_nul): Remove %G.
14553 (maybe_warn_for_bound): Same.
14554 (warn_for_access): Same.
14555 (check_access): Same.
14556 (check_strncat_sizes): Same.
14557 (expand_builtin_strncat): Same.
14558 (expand_builtin_strncmp): Same.
14559 (expand_builtin): Same.
14560 (expand_builtin_object_size): Same.
14561 (warn_dealloc_offset): Same.
14562 (maybe_emit_free_warning): Same.
14563 * calls.c (maybe_warn_alloc_args_overflow): Same.
14564 (maybe_warn_nonstring_arg): Same.
14565 (maybe_warn_rdwr_sizes): Same.
14566 * expr.c (expand_expr_real_1): Remove %K.
14567 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
14568 (gimple_fold_builtin_strncat): Same.
14569 * gimple-ssa-sprintf.c (format_directive): Same.
14570 (handle_printf_call): Same.
14571 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
14572 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
14573 (maybe_diag_access_bounds): Same. Call gimple_location.
14574 (check_bounds_or_overlap): Same.
14575 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
14576 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
14577 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
14578 (maybe_diag_stxncpy_trunc): Same.
14579 (handle_builtin_stxncpy_strncat): Same.
14580 (maybe_warn_pointless_strcmp): Same.
14581 * tree-ssa-uninit.c (maybe_warn_operand): Same.
14583 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
14586 * config/i386/predicates.md (vec_setm_operand): Enable
14587 register_operand for TARGET_SSE4_1.
14588 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
14589 as operand 2 predicate. Call ix86_expand_vector_set_var
14590 for non-constant index operand.
14591 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
14592 Call ix86_expand_vector_set_var for non-constant index operand.
14594 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
14596 * config/h8300/jumpcall.md (*branch): When possible, generate
14597 the comparison in CCZN mode.
14598 * config/h8300/predicates.md (simple_memory_operand): Reject all
14599 auto-increment addressing modes.
14601 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
14603 PR bootstrap/100246
14604 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
14607 2021-07-06 Richard Biener <rguenther@suse.de>
14609 * doc/md.texi (vec_fmaddsub<mode>4): Document.
14610 (vec_fmsubadd<mode>4): Likewise.
14611 * optabs.def (vec_fmaddsub$a4): Add.
14612 (vec_fmsubadd$a4): Likewise.
14613 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
14614 (IFN_VEC_FMSUBADD): Likewise.
14615 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
14616 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
14617 (addsub_pattern::build): Likewise.
14618 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
14619 and CFN_VEC_FMSUBADD are not transparent for permutes.
14620 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
14621 (vec_fmsubadd<mode>4): Likewise.
14623 2021-07-06 Richard Biener <rguenther@suse.de>
14625 * doc/invoke.texi (fmove-loop-stores): Document.
14626 * common.opt (fmove-loop-stores): New option.
14627 * opts.c (default_options_table): Enable -fmove-loop-stores
14628 at -O1 but not -Og.
14629 * tree-ssa-loop-im.c (pass_lim::execute): Pass
14630 flag_move_loop_stores instead of true to
14631 loop_invariant_motion_in_fun.
14633 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
14635 * doc/install.texi: Document --with-dsymutil.
14637 2021-07-06 Andrew Pinski <apinski@marvell.com>
14639 PR tree-optimization/101256
14640 * dbgcnt.def (phiopt_edge_range): New counter.
14641 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
14642 Check to make sure the new name is defined in the same
14643 bb as the conditional before duplicating range info.
14644 Also add debug counter.
14646 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
14648 PR rtl-optimization/100328
14649 * config/i386/i386-options.c (ix86_option_override_internal):
14650 Set param_ira_consider_dup_in_all_alts to 0.
14652 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
14654 PR rtl-optimization/100328
14655 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
14657 * ira.c (ira_get_dup_out_num): Adjust as parameter
14658 param_ira_consider_dup_in_all_alts.
14659 * params.opt (ira-consider-dup-in-all-alts): New.
14660 * ira-conflicts.c (process_regs_for_copy): Add one parameter
14661 single_input_op_has_cstr_p.
14662 (get_freq_for_shuffle_copy): New function.
14663 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
14664 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
14666 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
14668 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
14669 generate condition code aware RTL directly.
14671 2021-07-05 Andrew Pinski <apinski@marvell.com>
14673 PR tree-optimization/101039
14674 * match.pd (A CMP 0 ? A : -A): New patterns.
14675 * tree-ssa-phiopt.c (abs_replacement): Delete function.
14676 (tree_ssa_phiopt_worker): Don't call abs_replacement.
14677 Update comment about abs_replacement.
14679 2021-07-05 Andrew Pinski <apinski@marvell.com>
14681 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
14682 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
14684 2021-07-05 Andrew Pinski <apinski@marvell.com>
14686 * tree-ssa-phiopt.c (match_simplify_replacement):
14687 Add early_p argument. Call gimple_simplify_phiopt
14688 instead of gimple_simplify.
14689 (tree_ssa_phiopt_worker): Update call to
14690 match_simplify_replacement and allow unconditionally.
14691 (phiopt_early_allow): New function.
14692 (gimple_simplify_phiopt): New function.
14694 2021-07-05 Andrew Pinski <apinski@marvell.com>
14696 PR middle-end/101237
14697 * fold-const.c (negate_expr_p): Remove call to element_mode
14698 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
14699 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
14700 (fold_negate_expr_1): Likewise.
14701 (const_unop): Likewise.
14702 (fold_cond_expr_with_comparison): Likewise.
14703 (fold_binary_loc): Likewise.
14704 (fold_ternary_loc): Likewise.
14705 (tree_call_nonnegative_warnv_p): Likewise.
14706 * match.pd (-(A + B) -> (-B) - A): Likewise.
14708 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
14710 * configure.ac: Handle --with-dsymutil in the same way as we
14711 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
14712 Extract the type and version for the dsymutil configured or
14713 found by the default searches.
14714 * config.in: Regenerated.
14715 * configure: Regenerated.
14716 * collect2.c (do_dsymutil): Handle locating dsymutil in the
14717 same way as for the assembler and linker.
14718 * config/darwin.h (DSYMUTIL): Delete.
14719 * gcc.c: Report a configured dsymutil correctly.
14720 * exec-tool.in: Allow for dsymutil.
14722 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
14724 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
14725 Handle V4QI and V2HI modes.
14726 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
14727 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
14728 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
14729 Handle 4-byte dual operands with TARGET_XOP and single operands
14730 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
14731 mmx_pshufbv4qi3 for TARGET_SSSE3.
14732 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
14733 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
14734 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
14735 (expand_vec_perm_even_odd_1): Handle V4QI mode.
14736 (expand_vec_perm_broadcast_1): Handle V4QI mode.
14737 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
14738 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
14739 (mmx_pshufbv4qi3): Ditto.
14740 (*mmx_pblendw32): Ditto.
14741 (*mmx_pblendw64): Rename from *mmx_pblendw.
14742 (mmx_punpckhbw_low): New insn_and_split pattern.
14743 (mmx_punpcklbw_low): Ditto.
14745 2021-07-05 Richard Biener <rguenther@suse.de>
14747 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
14748 set LOOP_C_INFINITE on the vectorized loop.
14750 2021-07-05 Richard Biener <rguenther@suse.de>
14752 PR middle-end/101291
14753 * cfgloopmanip.c (loop_version): Set the loop copy of the
14754 versioned loop to the new loop.
14756 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
14759 * config.gcc: Ensure that Darwin biarch definitions are
14760 added before i386.h.
14761 * config/i386/darwin.h (TARGET_64BIT): Remove.
14762 (PR80556_WORKAROUND): New.
14763 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
14764 (DARWIN_SUBARCH_SPEC): New.
14765 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
14766 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
14767 (REAL_LIBGCC_SPEC): Remove.
14768 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
14769 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
14770 (REAL_LIBGCC_SPEC): Remove.
14772 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
14774 PR middle-end/101294
14775 * expr.c (store_constructor): Don't use vec_duplicate on vector.
14777 2021-07-02 Martin Sebor <msebor@redhat.com>
14779 PR middle-end/98871
14780 PR middle-end/98512
14781 * diagnostic.c (get_any_inlining_info): New.
14782 (update_effective_level_from_pragmas): Handle inlining context.
14783 (diagnostic_enabled): Same.
14784 (diagnostic_report_diagnostic): Same.
14785 * diagnostic.h (struct diagnostic_info): Add ctor.
14786 (struct diagnostic_context): Add new member.
14787 * tree-diagnostic.c (set_inlining_locations): New.
14788 (tree_diagnostics_defaults): Set new callback pointer.
14790 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
14792 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
14794 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
14795 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
14796 lxvp and stxvp built-ins.
14797 (mma_init_builtins): Handle lxvp and stxvp built-ins.
14798 (builtin_function_type): Likewise.
14799 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
14801 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
14803 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
14804 additional argument for the code.
14805 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
14806 argument for the code. Just return if the ZN bits are useful or
14807 not rather than the old style CC_* enums.
14808 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
14809 more generic shiftqi patterns.
14810 (shifthi_noscratch, shiftsi_noscratch): Similarly.
14811 (shiftqi_noscratch_set_flags): New pattern.
14812 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
14814 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
14816 PR tree-optimization/101223
14817 * range-op.cc (build_lt): Add -1 for signed values.
14818 (built_gt): Subtract -1 for signed values.
14820 2021-07-02 David Faust <david.faust@oracle.com>
14822 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
14823 (btf_asm_type): Likewise.
14825 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
14827 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
14828 an array of rtx rather than a pointer to rtx. Add code argument.
14829 (compute_a_shift_length): Similarly.
14830 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
14831 shift itself isn't an operand. Create dummy operand[0] to carry
14832 a mode and pass a suitable rtx code to compute_a_shift_length.
14833 (get_shift_alg): Adjust operand number of clobber in output templates.
14834 (output_a_shift): Make first argument an array of rtx rather than
14835 a pointer to rtx. Add code argument for the type of shift.
14836 Adjust now that the shift itself is no longer an operand.
14837 (compute_a_shift_length): Similarly.
14838 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
14839 iterator rather than nshift_operator.
14840 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
14841 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
14842 and compute_a_shift_length.
14843 (shiftqi_noscratch_clobber_flags): Likewise.
14844 (shifthi_noscratch_clobber_flags): Likewise.
14845 (shiftsi_noscratch_clobber_flags): Likewise.
14847 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
14850 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
14851 dsymutil for BTF/CTF.
14853 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
14856 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
14857 segment to include BTF.
14858 (BTF_INFO_SECTION_NAME): New.
14860 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
14862 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
14863 (small_data_operand, memreg_operand, small_insn_p): Likewise.
14864 * config/m32r/m32r.c (call_operand): Adjust return type.
14865 (small_data_operand, memreg_operand): Likewise.
14867 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
14869 * config/frv/frv-protos.h (integer_register_operand): Adjust return
14871 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
14872 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
14873 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
14874 (move_source_operand, move_destination_operand): Likewise.
14875 (condexec_source_operand, condexec_dest_operand): Likewise.
14876 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
14877 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
14878 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
14879 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
14880 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
14881 (even_fpr_operand, odd_fpr_operand): Likewise.
14882 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
14883 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
14884 (uint4_operand, uint1_operand, int_2word_operand): Likewise
14885 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
14886 (relational_operator, float_relational_operator): Likewise.
14887 (ccr_eqne_operator, minmax_operator): Likewise.
14888 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
14889 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
14890 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
14891 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
14892 (quad_acc_operand, accg_operand): Likewise.
14894 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
14896 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
14897 return type to a bool.
14898 (nonimmediate_nonstack_operand): Likewise.
14899 (xstormy16_splittable_below100_operand): Likewise.
14900 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
14902 (xstormy16_splittable_below100_operand): Likewise.
14904 2021-07-02 Richard Biener <rguenther@suse.de>
14906 PR tree-optimization/101293
14907 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
14908 with combined offsets.
14909 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
14910 combined with the rest of the offset.
14912 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
14914 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
14915 PIC encodings for PE-COFF targets.
14917 2021-07-02 Jakub Jelinek <jakub@redhat.com>
14920 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
14921 Return nullptr for TImode inner mode.
14923 2021-07-02 Richard Biener <rguenther@suse.de>
14925 PR tree-optimization/101280
14926 PR tree-optimization/101173
14927 * gimple-loop-interchange.cc
14928 (tree_loop_interchange::valid_data_dependences): Properly
14929 guard all dependence checks with DDR_REVERSED_P or its
14932 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
14934 * config/i386/i386-expand.c (ix86_expand_builtin):
14935 Add branch to clear odata when ZF is set for asedecenc_expand
14936 and wideaesdecenc_expand.
14938 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
14940 * config/i386/gcc-auto-profile: regenerate
14942 2021-07-02 liuhongt <hongtao.liu@intel.com>
14944 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
14945 (trunc<mode><pmov_dst_4_lower>2): this.
14947 2021-07-01 David Malcolm <dmalcolm@redhat.com>
14949 * diagnostic.h (diagnostic_context::m_file_cache): New field.
14950 * input.c (class fcache): Rename to...
14951 (class file_cache_slot): ...this, making most members private and
14952 prefixing fields with "m_".
14953 (file_cache_slot::get_file_path): New accessor.
14954 (file_cache_slot::get_use_count): New accessor.
14955 (file_cache_slot::missing_trailing_newline_p): New accessor.
14956 (file_cache_slot::inc_use_count): New.
14957 (fcache_buffer_size): Move to...
14958 (file_cache_slot::buffer_size): ...here.
14959 (fcache_line_record_size): Move to...
14960 (file_cache_slot::line_record_size): ...here.
14961 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
14962 (fcache_tab_size): Move to file_cache::num_file_slots.
14963 (diagnostic_file_cache_init): Update for move of fcache_tab
14964 to global_dc->m_file_cache.
14965 (diagnostic_file_cache_fini): Likewise.
14966 (lookup_file_in_cache_tab): Convert to...
14967 (file_cache::lookup_file): ...this.
14968 (diagnostics_file_cache_forcibly_evict_file): Update for move of
14969 fcache_tab to global_dc->m_file_cache, moving most of
14970 implementation to...
14971 (file_cache::forcibly_evict_file): ...this new function and...
14972 (file_cache_slot::evict): ...this new function.
14973 (evicted_cache_tab_entry): Convert to...
14974 (file_cache::evicted_cache_tab_entry): ...this.
14975 (add_file_to_cache_tab): Convert to...
14976 (file_cache::add_file): ...this, moving bulk of implementation
14978 (file_cache_slot::create): ..this new function.
14979 (file_cache::file_cache): New.
14980 (file_cache::~file_cache): New.
14981 (lookup_or_add_file_to_cache_tab): Convert to...
14982 (file_cache::lookup_or_add_file): ..this new function.
14983 (fcache::fcache): Rename to...
14984 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
14986 (fcache::~fcache): Rename to...
14987 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
14989 (needs_read): Convert to...
14990 (file_cache_slot::needs_read_p): ...this.
14991 (needs_grow): Convert to...
14992 (file_cache_slot::needs_grow_p): ...this.
14993 (maybe_grow): Convert to...
14994 (file_cache_slot::maybe_grow): ...this.
14995 (read_data): Convert to...
14996 (file_cache_slot::read_data): ...this.
14997 (maybe_read_data): Convert to...
14998 (file_cache_slot::maybe_read_data): ...this.
14999 (get_next_line): Convert to...
15000 (file_cache_slot::get_next_line): ...this.
15001 (goto_next_line): Convert to...
15002 (file_cache_slot::goto_next_line): ...this.
15003 (read_line_num): Convert to...
15004 (file_cache_slot::read_line_num): ...this.
15005 (location_get_source_line): Update for moving of globals to
15006 global_dc->m_file_cache.
15007 (location_missing_trailing_newline): Likewise.
15008 * input.h (class file_cache_slot): New forward decl.
15009 (class file_cache): New.
15011 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
15013 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
15014 128-bit floating point conditional move support.
15015 (have_compare_and_set_mask): Add IEEE 128-bit floating point
15017 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
15018 (mov<mode>cc_p10, IEEE128 iterator): New insn.
15019 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
15020 (fpmask<mode>, IEEE128 iterator): New insn.
15021 (xxsel<mode>, IEEE128 iterator): New insn.
15023 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
15026 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
15028 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
15030 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15032 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
15034 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
15035 (vec_duplicate<mode>): New expander.
15037 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
15040 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15042 (ix86_byte_broadcast): New function.
15043 (ix86_convert_const_wide_int_to_broadcast): Likewise.
15044 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
15045 size is 16 bytes or bigger.
15046 (ix86_broadcast_from_integer_constant): New function.
15047 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
15048 to broadcast if mode size is 16 bytes or bigger.
15049 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
15051 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
15053 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15055 * config/i386/predicates.md (ix86_endbr_immediate_operand):
15056 Return true/false instead of 1/0.
15057 (movq_parallel): Ditto.
15059 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15061 * recog.c (general_operand): Return true/false instead of 1/0.
15062 (register_operand): Ditto.
15063 (immediate_operand): Ditto.
15064 (const_int_operand): Ditto.
15065 (const_scalar_int_operand): Ditto.
15066 (const_double_operand): Ditto.
15067 (push_operand): Ditto.
15068 (pop_operand): Ditto.
15069 (memory_operand): Ditto.
15070 (indirect_operand): Ditto.
15072 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15074 * genpreds.c (write_predicate_subfunction):
15075 Change the type of written subfunction to bool.
15076 (write_one_predicate_function):
15077 Change the type of written function to bool.
15078 (write_tm_preds_h): Ditto.
15079 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
15080 * recog.c (general_operand): Change the type to bool.
15081 (address_operand): Ditto.
15082 (register_operand): Ditto.
15083 (pmode_register_operand): Ditto.
15084 (scratch_operand): Ditto.
15085 (immediate_operand): Ditto.
15086 (const_int_operand): Ditto.
15087 (const_scalar_int_operand): Ditto.
15088 (const_double_operand): Ditto.
15089 (nonimmediate_operand): Ditto.
15090 (nonmemory_operand): Ditto.
15091 (push_operand): Ditto.
15092 (pop_operand): Ditto.
15093 (memory_operand): Ditto.
15094 (indirect_operand): Ditto.
15095 (ordered_comparison_operator): Ditto.
15096 (comparison_operator): Ditto.
15097 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
15098 Change the type of indirect predicate function to bool.
15099 * config/rs6000/rs6000.c (easy_vector_constant):
15100 Change the type to bool.
15101 * config/mips/mips-protos.h (m16_based_address_p):
15102 Change the type of operand 3 to bool.
15104 2021-07-01 Richard Biener <rguenther@suse.de>
15106 PR tree-optimization/101280
15107 PR tree-optimization/101173
15108 * gimple-loop-interchange.cc
15109 (tree_loop_interchange::valid_data_dependences): Revert
15110 previous change and instead correctly handle DDR_REVERSED_P
15113 2021-07-01 Richard Biener <rguenther@suse.de>
15115 PR tree-optimization/101278
15116 * tree-ssa-dse.c (dse_classify_store): First check for
15117 uses, then ignore stmt for chaining purposes.
15119 2021-07-01 Richard Biener <rguenther@suse.de>
15121 PR tree-optimization/100778
15122 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
15123 vectorized ops ahead of their scalar BB.
15125 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15128 * config/i386/i386.md (*nabs<dwi>2_doubleword):
15129 New insn_and_split pattern.
15130 (*nabs<dwi>2_1): Ditto.
15131 * config/i386/i386-features.c
15132 (general_scalar_chain::compute_convert_gain):
15133 Handle (NEG (ABS (...))) RTX. Rewrite src code
15134 scanner as switch statement.
15135 (general_scalar_chain::convert_insn):
15136 Handle (NEG (ABS (...))) RTX.
15137 (general_scalar_to_vector_candidate_p):
15138 Detect (NEG (ABS (...))) RTX. Reorder case statements
15139 for (AND (NOT (...) ...)) fallthrough.
15141 2021-07-01 Richard Biener <rguenther@suse.de>
15143 PR tree-optimization/101178
15144 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
15145 (slpg::perm_in): Add.
15146 (slpg::get_perm_in): Remove.
15147 (slpg::get_perm_materialized): Add.
15148 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
15149 during permute propagation and materialization.
15151 2021-07-01 Jakub Jelinek <jakub@redhat.com>
15154 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
15156 2021-07-01 Jakub Jelinek <jakub@redhat.com>
15158 PR middle-end/94366
15159 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
15160 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
15161 use boolean_type_node instead of integer_type_node as NE_EXPR type.
15162 (lower_reduction_clauses): Likewise.
15164 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15166 * config/gcn/gcn.c: Include dwarf2.h.
15167 (gcn_addr_space_debug): New function.
15168 (TARGET_ADDR_SPACE_DEBUG): New hook.
15170 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15172 * common/config/gcn/gcn-common.c
15173 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
15174 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
15176 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
15177 (gcn_frame_pointer_rqd): New function.
15178 (TARGET_FRAME_POINTER_REQUIRED): New hook.
15180 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15182 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
15183 prologue register saves.
15184 (gcn_debug_unwind_info): Use UI_DWARF2.
15185 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
15186 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
15187 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
15188 (DWARF_LINK_REGISTER): New define.
15189 (FIRST_PSEUDO_REGISTER): Increment.
15190 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
15191 (CALL_USED_REGISTERS): Likewise.
15192 (REGISTER_NAMES): Likewise.
15194 2021-06-30 Richard Biener <rguenther@suse.de>
15196 PR tree-optimization/101267
15197 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
15198 API and use SLP compatible interface of vect_is_simple_use.
15199 Reject not vectorized SLP defs for callers that do not support
15201 (vect_check_store_rhs): Handle masked stores and pass down
15202 the appropriate operator index.
15203 (vectorizable_call): Adjust.
15204 (vectorizable_store): Likewise.
15205 (vectorizable_load): Likewise. Handle SLP pecularity of
15207 (vect_is_simple_use): Remove special-casing of masked stores.
15209 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
15211 * common.opt (foffload): Remove help as Driver only.
15212 * gcc.c (display_help): Add -foffload.
15214 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
15216 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
15217 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
15218 a candidate list; better inform no offload target is configured
15219 and fix hint extraction when passed target is not '\0' at [len].
15220 * common.opt (foffload): Add tailing '.'.
15221 (foffload-options): Likewise; fix flag name in the help string.
15223 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
15226 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
15227 (vcage_f32): Gate comparison on __FAST_MATH__.
15228 (vcageq_f32): Likewise.
15229 (vcale_f32): Likewise.
15230 (vcaleq_f32): Likewise.
15231 (vcagt_f32): Likewise.
15232 (vcagtq_f32): Likewise.
15233 (vcalt_f32): Likewise.
15234 (vcaltq_f32): Likewise.
15235 (vcage_f16): Likewise.
15236 (vcageq_f16): Likewise.
15237 (vcale_f16): Likewise.
15238 (vcaleq_f16): Likewise.
15239 (vcagt_f16): Likewise.
15240 (vcagtq_f16): Likewise.
15241 (vcalt_f16): Likewise.
15242 (vcaltq_f16): Likewise.
15244 2021-06-30 Richard Biener <rguenther@suse.de>
15246 PR tree-optimization/101264
15247 * tree-vect-slp.c (vect_optimize_slp): Propagate the
15248 computed perm_in to all "any" permute successors
15249 we cannot de-duplicate immediately.
15251 2021-06-30 liuhongt <hongtao.liu@intel.com>
15254 * config/i386/sse.md
15255 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
15257 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
15259 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
15260 * config/i386/subst.md (maskz_scalar): New define_subst.
15261 (maskz_scalar_name): New subst_attr.
15262 (maskz_scalar_op5): Ditto.
15263 (round_saeonly_maskz_scalar_op5): Ditto.
15264 (round_saeonly_maskz_scalar_operand5): Ditto.
15266 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
15268 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
15269 Increase code CSECT alignment to at least 32 bytes.
15270 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
15271 alignment designation.
15273 2021-06-29 Sergei Trofimovich <siarheit@google.com>
15275 * doc/generic.texi: Fix s/net yet/not yet/ typo.
15277 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15279 PR tree-optimization/101254
15280 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
15281 wrapping/non-wrapping when setting the result range.
15283 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15285 * value-query.cc (gimple_range_global): Allow phis.
15287 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15289 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
15290 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
15291 statement for location context.
15292 (check_for_binary_op_overflow): Ditto.
15293 (simplify_using_ranges::get_vr_for_comparison): Ditto.
15294 (simplify_using_ranges::compare_name_with_value): Ditto.
15295 (simplify_using_ranges::compare_names): Ditto.
15296 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
15297 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
15298 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
15299 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
15300 (simplify_using_ranges::two_valued_val_range_p): Ditto.
15301 (simplify_using_ranges::simplify): Ditto.
15302 * vr-values.h: Adjust prototypes.
15304 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
15307 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
15309 2021-06-29 Julian Brown <julian@codesourcery.com>
15311 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
15312 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
15313 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
15315 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
15316 (MAX_FIXED_MODE_SIZE): Change to 128.
15318 2021-06-29 Julian Brown <julian@codesourcery.com>
15320 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
15321 (s_mnemonic): Add clrsb.
15322 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
15323 (clrsb<mode>2): Add expander for SImode/DImode.
15325 2021-06-29 Julian Brown <julian@codesourcery.com>
15327 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
15328 muldi3): Add patterns.
15330 2021-06-29 Julian Brown <julian@codesourcery.com>
15332 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
15333 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
15335 2021-06-29 Julian Brown <julian@codesourcery.com>
15337 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
15339 2021-06-29 Joseph Myers <joseph@codesourcery.com>
15341 * btfout.c, ctfout.c: Include "memmodel.h".
15343 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15345 * gcc.c (check_offload_target_name): Cast len argument to
15346 %q.*s to 'int'; avoid -Wstringop-truncation warning.
15348 2021-06-29 Richard Biener <rguenther@suse.de>
15350 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
15351 to "any" permute nodes and relax "any" permute proapgation
15352 during iterative backward propagation.
15354 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15357 * common.opt (-foffload=): Update description.
15358 (-foffload-options=): New.
15359 * doc/invoke.texi (C Language Options): Document
15360 -foffload and -foffload-options.
15361 * gcc.c (check_offload_target_name): New, split off from
15362 handle_foffload_option.
15363 (check_foffload_target_names): New.
15364 (handle_foffload_option): Handle -foffload=default.
15365 (driver_handle_option): Update for -foffload-options.
15366 * lto-opts.c (lto_write_options): Use -foffload-options
15367 instead of -foffload.
15368 * lto-wrapper.c (merge_and_complain, append_offload_options):
15370 * opts.c (common_handle_option): Likewise.
15372 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15374 * doc/invoke.texi (C Language Options): Sort options
15375 alphabetically in optlist and also the description itself.
15376 Remove leftover -fallow-single-precision from and add missing
15377 -fgnu-tm to the optlist.
15379 2021-06-29 Richard Biener <rguenther@suse.de>
15381 * tree-vect-slp.c (slpg_vertex::visited): Remove.
15382 (vect_slp_perms_eq): Handle -1 permutes.
15383 (vect_optimize_slp): Rewrite permute propagation.
15385 2021-06-29 Jakub Jelinek <jakub@redhat.com>
15388 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
15389 perform the optimization in GENERIC when sanitizing and x has a
15392 2021-06-29 Richard Biener <rguenther@suse.de>
15394 PR tree-optimization/101242
15395 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
15396 PHIs with not represented initial values as leafs.
15398 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
15400 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
15401 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
15402 "register" keyword.
15403 (pdp11_initial_elimination_offset) Remove unused variable.
15404 (pdp11_cmp_length) Ditto.
15405 (pdp11_insn_cost): Ditto, and fix signedness warning.
15407 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
15409 * btfout.c: Include tm_p.h.
15412 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15414 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
15416 (bpf_expand_epilogue): Likewise.
15417 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
15418 Do not define DBX_DEBUGGING_INFO.
15420 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15422 * doc/invoke.texi: Document the CTF and BTF debug info options.
15424 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15425 David Faust <david.faust@oracle.com>
15426 Jose E. Marchesi <jose.marchesi@oracle.com>
15427 Weimin Pan <weimin.pan@oracle.com>
15429 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
15430 GTFILES. Add new object files.
15431 * common.opt: Add CTF and BTF debug info options.
15432 * btfout.c: New file.
15433 * ctfc.c: Likewise.
15434 * ctfc.h: Likewise.
15435 * ctfout.c: Likewise.
15436 * dwarf2ctf.c: Likewise.
15437 * dwarf2ctf.h: Likewise.
15438 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
15440 * dwarf2out.c (dwarf2out_source_line): Likewise.
15441 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
15443 (debug_format_do_cu): New function.
15444 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
15446 Include dwarf2ctf.c.
15447 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
15449 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
15450 (CTF_DEBUG): New bitmask.
15451 (BTF_DEBUG): Likewise.
15452 (enum ctf_debug_info_levels): New enum.
15453 * gengtype.c (open_base_files): Handle ctfc.h.
15454 (main): Handle uint32_t type.
15455 * flags.h (btf_debuginfo_p): New definition.
15456 (dwarf_based_debuginfo_p): Likewise.
15457 * opts.c (debug_type_names): Add entries for CTF and BTF.
15458 (btf_debuginfo_p): New function.
15459 (dwarf_based_debuginfo_p): Likewise.
15460 (common_handle_option): Handle -gctfN and -gbtf options.
15461 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
15462 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
15465 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
15467 * dwarf2out.c (AT_class): Function is no longer static.
15468 (AT_int): Likewise.
15469 (AT_unsigned): Likewise.
15470 (AT_loc): Likewise.
15471 (get_AT): Likewise.
15472 (get_AT_string): Likewise.
15473 (get_AT_flag): Likewise.
15474 (get_AT_unsigned): Likewise.
15475 (get_AT_ref): Likewise.
15476 (new_die_raw): Likewise.
15477 (lookup_decl_die): Likewise.
15478 (base_type_die): Likewise.
15479 (add_name_attribute): Likewise.
15480 (add_AT_int): Likewise.
15481 (add_AT_unsigned): Likewise.
15482 (add_AT_loc): Likewise.
15483 (dw_get_die_tag): New function.
15484 (dw_get_die_child): Likewise.
15485 (dw_get_die_sib): Likewise.
15486 (struct dwarf_file_data): Move from here to dwarf2out.h
15487 (struct dw_attr_struct): Likewise.
15488 * dwarf2out.h: Analogous changes.
15490 2021-06-28 Martin Jambor <mjambor@suse.cz>
15493 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
15494 members m_dead_stmts and m_dead_ssas.
15495 * ipa-param-manipulation.c
15496 (ipa_param_body_adjustments::mark_dead_statements): New function.
15497 (ipa_param_body_adjustments::common_initialization): Call it on
15498 all removed but not split parameters.
15499 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
15501 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
15503 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
15504 dead debug statements.
15505 (copy_phis_for_bb): Do not copy dead PHI nodes.
15507 2021-06-28 Martin Jambor <mjambor@suse.cz>
15510 * symtab-clones.h (clone_info): Removed member param_adjustments.
15511 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
15512 deal with pass-through splits now.
15513 (ipa_param_performed_split): Removed.
15514 (ipa_param_adjustments::modify_call): Adjusted parameters.
15515 (class ipa_param_body_adjustments): Adjusted parameters of
15516 register_replacement, modify_gimple_stmt and modify_call_stmt.
15517 (ipa_verify_edge_has_no_modifications): Declare.
15518 (ipa_edge_modifications_finalize): Declare.
15519 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
15520 performed_splits processing, pas only edge to padjs->modify_call,
15521 check that call arguments were not modified if they should not have
15523 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
15525 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
15526 (ipa_edge_modification_info): Likewise.
15527 (ipa_edge_modification_sum): Likewise.
15528 (ipa_edge_modifications): New edge summary.
15529 (ipa_verify_edge_has_no_modifications): New function.
15530 (transitive_split_p): Removed.
15531 (transitive_split_map): Likewise.
15532 (init_transitive_splits): Likewise.
15533 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
15534 summary instead of performed_splits.
15535 (ipa_param_body_adjustments::register_replacement): Drop dummy
15536 parameter, set base_index of the created ipa_param_body_replacement.
15537 (phi_arg_will_live_p): New function.
15538 (ipa_param_body_adjustments::common_initialization): Do not create
15539 IPA_SRA dummy decls.
15540 (simple_tree_swap_info): Removed.
15541 (remap_split_decl_to_dummy): Likewise.
15542 (record_argument_state_1): New function.
15543 (record_argument_state): Likewise.
15544 (ipa_param_body_adjustments::modify_call_stmt): New parameter
15545 orig_stmt. Do not work with dummy decls, save necessary info about
15546 changes to ipa_edge_modifications.
15547 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
15548 orig_stmt, pass it to modify_call_stmt.
15549 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
15550 modify_gimple_stmt.
15551 (ipa_edge_modifications_finalize): New function.
15552 * tree-inline.c (remap_gimple_stmt): Pass original statement to
15553 modify_gimple_stmt.
15554 (copy_phis_for_bb): Do not copy dead PHI nodes.
15555 (expand_call_inline): Do not remap performed_splits.
15556 (update_clone_info): Likewise.
15557 * toplev.c: Include ipa-param-manipulation.h.
15558 (toplev::finalize): Call ipa_edge_modifications_finalize.
15560 2021-06-28 Andrew Pinski <apinski@marvell.com>
15562 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
15563 info if we're the only things setting the target PHI.
15564 (value_replacement): Don't duplicate range here.
15565 (minmax_replacement): Likewise.
15567 2021-06-28 Richard Biener <rguenther@suse.de>
15569 PR tree-optimization/101229
15570 * gimple-walk.c (gimple_walk_op): Handle PHIs.
15572 2021-06-28 Martin Liska <mliska@suse.cz>
15574 * config/v850/v850.c (construct_dispose_instruction): Allocate
15576 (construct_prepare_instruction): Likewise.
15578 2021-06-28 Martin Liska <mliska@suse.cz>
15580 * config/v850/v850.c (v850_option_override): Build default
15582 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
15583 ignored for inlining.
15584 (TARGET_CAN_INLINE_P): New.
15586 2021-06-28 Richard Biener <rguenther@suse.de>
15588 PR tree-optimization/101207
15589 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
15590 permute eliding for load permutations properly.
15592 2021-06-28 Richard Biener <rguenther@suse.de>
15594 PR tree-optimization/101173
15595 * gimple-loop-interchange.cc
15596 (tree_loop_interchange::valid_data_dependences): Disallow outer
15597 loop dependence distance of zero.
15599 2021-06-28 liuhongt <hongtao.liu@intel.com>
15602 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
15603 define_insn_and_split.
15604 (*avx_cmp<mode>3_ltint): Ditto.
15605 (*avx2_pcmp<mode>3_3): Ditto.
15606 (*avx2_pcmp<mode>3_4): Ditto.
15607 (*avx2_pcmp<mode>3_5): Ditto.
15609 2021-06-28 liuhongt <hongtao.liu@intel.com>
15611 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
15612 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
15613 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
15614 IX86_BUILTIN_PBLENDVB128): Replace icode with
15616 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
15618 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
15619 New pre_reload splitter.
15621 2021-06-27 Andrew Pinski <apinski@marvell.com>
15623 PR middle-end/101230
15624 * fold-const.c (fold_ternary_loc): Check
15625 the return value of invert_tree_comparison.
15627 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
15629 * config.gcc: Add SPDX License Identifier.
15630 (powerpc-ibm-aix789): Default to aix73.h.
15631 (powerpc-ibm-aix7.2.*.*): New stanza.
15632 * config/rs6000/aix72.h: Add SPDX License Identifier.
15633 * config/rs6000/aix73.h: New file.
15635 2021-06-26 Jason Merrill <jason@redhat.com>
15637 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
15639 2021-06-26 Andrew Pinski <apinski@marvell.com>
15641 * genmatch.c (lower_cond): Copy for_subst_vec
15642 for the simplify also.
15643 (lower): Swap the order for lower_for and lower_cond.
15645 2021-06-26 Andrew Pinski <apinski@marvell.com>
15647 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
15648 flow senatitive info on the moved ssa set.
15650 2021-06-26 Andrew Pinski <apinski@marvell.com>
15652 * fold-const.c (fold_cond_expr_with_comparison):
15653 Exand arg0 into comp_code, arg00, and arg01.
15654 (fold_ternary_loc): Use invert_tree_comparison
15655 instead of fold_invert_truthvalue for the case
15656 where we have A CMP B ? C : A.
15658 2021-06-25 Martin Sebor <msebor@redhat.com>
15660 PR middle-end/101216
15661 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
15663 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
15665 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
15667 2021-06-25 Richard Biener <rguenther@suse.de>
15669 PR tree-optimization/101202
15670 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
15673 2021-06-25 Richard Biener <rguenther@suse.de>
15675 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
15676 STMT_VINFO_REDUC_DEF from the original representative.
15678 2021-06-25 Martin Sebor <msebor@redhat.com>
15680 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
15681 gimple_no_warning_p and gimple_set_no_warning with
15682 warning_suppressed_p, and suppress_warning.
15684 (maybe_warn_for_bound): Same.
15685 (warn_for_access): Same.
15686 (check_access): Same.
15687 (expand_builtin_strncmp): Same.
15688 (fold_builtin_varargs): Same.
15689 * calls.c (maybe_warn_nonstring_arg): Same.
15690 (maybe_warn_rdwr_sizes): Same.
15691 * cfgexpand.c (expand_call_stmt): Same.
15692 * cgraphunit.c (check_global_declaration): Same.
15693 * fold-const.c (fold_undefer_overflow_warnings): Same.
15694 (fold_truth_not_expr): Same.
15695 (fold_unary_loc): Same.
15696 (fold_checksum_tree): Same.
15697 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
15698 (array_bounds_checker::check_mem_ref): Same.
15699 (array_bounds_checker::check_addr_expr): Same.
15700 (array_bounds_checker::check_array_bounds): Same.
15701 * gimple-expr.c (copy_var_decl): Same.
15702 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
15703 (gimple_fold_builtin_strncat): Same.
15704 (gimple_fold_builtin_stxcpy_chk): Same.
15705 (gimple_fold_builtin_stpcpy): Same.
15706 (gimple_fold_builtin_sprintf): Same.
15707 (fold_stmt_1): Same.
15708 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
15709 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
15710 * gimple-ssa-sprintf.c (handle_printf_call): Same.
15711 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
15712 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
15713 * gimple-ssa-warn-restrict.h: Adjust declarations.
15714 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
15715 gimple_no_warning_p and gimple_set_no_warning with
15716 warning_suppressed_p, and suppress_warning.
15717 (check_call): Same.
15718 (check_bounds_or_overlap): Same.
15719 * gimple.c (gimple_build_call_from_tree): Same.
15720 * gimplify.c (gimplify_return_expr): Same.
15721 (gimplify_cond_expr): Same.
15722 (gimplify_modify_expr_complex_part): Same.
15723 (gimplify_modify_expr): Same.
15724 (gimple_push_cleanup): Same.
15725 (gimplify_expr): Same.
15726 * omp-expand.c (expand_omp_for_generic): Same.
15727 (expand_omp_taskloop_for_outer): Same.
15728 * omp-low.c (lower_rec_input_clauses): Same.
15729 (lower_lastprivate_clauses): Same.
15730 (lower_send_clauses): Same.
15731 (lower_omp_target): Same.
15732 * tree-cfg.c (pass_warn_function_return::execute): Same.
15733 * tree-complex.c (create_one_component_var): Same.
15734 * tree-inline.c (remap_gimple_op_r): Same.
15735 (copy_tree_body_r): Same.
15736 (declare_return_variable): Same.
15737 (expand_call_inline): Same.
15738 * tree-nested.c (lookup_field_for_decl): Same.
15739 * tree-sra.c (create_access_replacement): Same.
15740 (generate_subtree_copies): Same.
15741 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
15742 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
15743 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
15744 * tree-ssa-loop-im.c (execute_sm): Same.
15745 * tree-ssa-phiopt.c (cond_store_replacement): Same.
15746 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
15747 (handle_builtin_strcpy): Same.
15748 (maybe_diag_stxncpy_trunc): Same.
15749 (handle_builtin_stxncpy_strncat): Same.
15750 (handle_builtin_strcat): Same.
15751 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
15752 (set_no_uninit_warning): Same.
15753 (uninit_undefined_value_p): Same.
15754 (warn_uninit): Same.
15755 (maybe_warn_operand): Same.
15756 * tree-vrp.c (compare_values_warnv): Same.
15757 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
15758 (test_for_singularity): Same.
15759 * gimple.h (warning_suppressed_p): New function.
15760 (suppress_warning): Same.
15761 (copy_no_warning): Same.
15762 (gimple_set_block): Call gimple_set_location.
15763 (gimple_set_location): Call copy_warning.
15765 2021-06-25 Martin Sebor <msebor@redhat.com>
15767 * tree.h (warning_suppressed_at, copy_warning,
15768 warning_suppressed_p, suppress_warning): New functions.
15770 2021-06-25 Martin Sebor <msebor@redhat.com>
15772 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
15773 * gengtype.c (open_base_files): Add diagnostic-spec.h.
15774 * diagnostic-spec.c: New file.
15775 * diagnostic-spec.h: New file.
15776 * tree.h (no_warning, all_warnings, suppress_warning_at): New
15778 * warning-control.cc: New file.
15780 2021-06-25 liuhongt <hongtao.liu@intel.com>
15783 * config/i386/i386.c (x86_order_regs_for_local_alloc):
15786 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
15788 PR tree-optimization/101189
15789 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
15790 LHS range of condition to postfold routine.
15791 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
15792 FALSE edge if the LHS range supports it being taken.
15793 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
15795 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
15797 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
15798 (relation_oracle::find_relation_block): Check correct bitmap.
15799 (relation_oracle::dump): Do not dump NULL blocks.
15801 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
15803 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
15804 range_on_edge instead of manually calculating.
15806 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
15808 * range-op.cc: Fix comment.
15810 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
15813 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
15814 Handle V8QI and V4HI modes.
15815 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
15817 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
15818 (mmxpackmode): New mode attribute.
15819 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
15820 (mmxunpackmode): New mode attribute.
15821 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
15822 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
15823 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
15824 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
15825 * config/i386/i386.md (extsuffix): Move from ...
15826 * config/i386/sse.md: ... here.
15828 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
15830 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
15831 (dwarf2out_finish): ...instead of here.
15833 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
15835 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
15836 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
15837 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
15838 * configure: Regenerate.
15840 2021-06-24 Richard Biener <rguenther@suse.de>
15842 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
15843 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
15844 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
15846 2021-06-24 Richard Biener <rguenther@suse.de>
15848 * config/i386/sse.md (avx_addsubv4df3): Rename to
15850 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
15851 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
15852 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
15853 * config/i386/i386-builtin.def: Adjust.
15854 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
15855 * optabs.def (vec_addsub_optab): New optab.
15856 * tree-vect-slp-patterns.c (class addsub_pattern): New.
15857 (slp_patterns): Add addsub_pattern.
15858 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
15859 across CFN_VEC_ADDSUB.
15860 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
15862 * doc/md.texi (vec_addsub<mode>3): Document.
15864 2021-06-24 Jakub Jelinek <jakub@redhat.com>
15866 PR middle-end/101170
15867 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
15868 where regno + subreg_regno_offset wraps around use 0 as starting
15871 2021-06-24 Jakub Jelinek <jakub@redhat.com>
15873 PR middle-end/101172
15874 * stor-layout.c (finish_bitfield_representative): If nextf has
15875 error_mark_node type, set repr type to error_mark_node too.
15877 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
15879 * config/s390/s390.c (s390_function_profiler): Ignore labelno
15881 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
15883 2021-06-24 Richard Biener <rguenther@suse.de>
15885 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
15886 across operations that have different semantics on different
15889 2021-06-24 Jakub Jelinek <jakub@redhat.com>
15891 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
15892 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
15893 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
15894 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
15895 OMP_TARGET user outer_ctx instead of ctx for placeholders and
15896 initializer/combiner gimplification.
15897 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
15898 on target constructs.
15899 (lower_rec_input_clauses): Likewise.
15900 (lower_omp_target): Likewise.
15901 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
15902 on target if in_reduction is present.
15904 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
15906 * tree-predcom.c (class pcom_worker): New class.
15907 (release_chain): Renamed to...
15908 (pcom_worker::release_chain): ...this.
15909 (release_chains): Renamed to...
15910 (pcom_worker::release_chains): ...this.
15911 (aff_combination_dr_offset): Renamed to...
15912 (pcom_worker::aff_combination_dr_offset): ...this.
15913 (determine_offset): Renamed to...
15914 (pcom_worker::determine_offset): ...this.
15915 (class comp_ptrs): New class.
15916 (split_data_refs_to_components): Renamed to...
15917 (pcom_worker::split_data_refs_to_components): ...this,
15918 and update with class comp_ptrs.
15919 (suitable_component_p): Renamed to...
15920 (pcom_worker::suitable_component_p): ...this.
15921 (filter_suitable_components): Renamed to...
15922 (pcom_worker::filter_suitable_components): ...this.
15923 (valid_initializer_p): Renamed to...
15924 (pcom_worker::valid_initializer_p): ...this.
15925 (find_looparound_phi): Renamed to...
15926 (pcom_worker::find_looparound_phi): ...this.
15927 (add_looparound_copies): Renamed to...
15928 (pcom_worker::add_looparound_copies): ...this.
15929 (determine_roots_comp): Renamed to...
15930 (pcom_worker::determine_roots_comp): ...this.
15931 (determine_roots): Renamed to...
15932 (pcom_worker::determine_roots): ...this.
15933 (single_nonlooparound_use): Renamed to...
15934 (pcom_worker::single_nonlooparound_use): ...this.
15935 (remove_stmt): Renamed to...
15936 (pcom_worker::remove_stmt): ...this.
15937 (execute_pred_commoning_chain): Renamed to...
15938 (pcom_worker::execute_pred_commoning_chain): ...this.
15939 (execute_pred_commoning): Renamed to...
15940 (pcom_worker::execute_pred_commoning): ...this.
15941 (struct epcc_data): New member worker.
15942 (execute_pred_commoning_cbck): Call execute_pred_commoning
15943 with pcom_worker pointer.
15944 (find_use_stmt): Renamed to...
15945 (pcom_worker::find_use_stmt): ...this.
15946 (find_associative_operation_root): Renamed to...
15947 (pcom_worker::find_associative_operation_root): ...this.
15948 (find_common_use_stmt): Renamed to...
15949 (pcom_worker::find_common_use_stmt): ...this.
15950 (combinable_refs_p): Renamed to...
15951 (pcom_worker::combinable_refs_p): ...this.
15952 (reassociate_to_the_same_stmt): Renamed to...
15953 (pcom_worker::reassociate_to_the_same_stmt): ...this.
15954 (stmt_combining_refs): Renamed to...
15955 (pcom_worker::stmt_combining_refs): ...this.
15956 (combine_chains): Renamed to...
15957 (pcom_worker::combine_chains): ...this.
15958 (try_combine_chains): Renamed to...
15959 (pcom_worker::try_combine_chains): ...this.
15960 (prepare_initializers_chain): Renamed to...
15961 (pcom_worker::prepare_initializers_chain): ...this.
15962 (prepare_initializers): Renamed to...
15963 (pcom_worker::prepare_initializers): ...this.
15964 (prepare_finalizers_chain): Renamed to...
15965 (pcom_worker::prepare_finalizers_chain): ...this.
15966 (prepare_finalizers): Renamed to...
15967 (pcom_worker::prepare_finalizers): ...this.
15968 (tree_predictive_commoning_loop): Renamed to...
15969 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
15970 some calls and remove some cleanup code.
15971 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
15972 (static variable looparound_phis): Remove.
15973 (static variable name_expansions): Remove.
15975 2021-06-24 Richard Biener <rguenther@suse.de>
15977 * tree-vect-slp.c (slpg_vertex): New struct.
15978 (vect_slp_build_vertices): Adjust.
15979 (vect_optimize_slp): Likewise. Maintain an outgoing permute
15980 and a materialized one.
15982 2021-06-24 Richard Biener <rguenther@suse.de>
15984 PR tree-optimization/101105
15985 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
15986 Only ignore steps when they are equal or scalar order is preserved.
15988 2021-06-24 liuhongt <hongtao.liu@intel.com>
15991 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
15992 Adjust comments for ix86_expand_vecop_qihi2.
15993 (ix86_expand_vecmul_qihi): Renamed to ..
15994 (ix86_expand_vecop_qihi2): Adjust function prototype to
15995 support shift operation, add static to definition.
15996 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
15997 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
15998 ix86_expand_vec_shift_qihi_constant.
15999 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
16000 (ix86_expand_vec_shift_qihi_constant): Deleted.
16001 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
16003 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
16004 condition TARGET_64BIT.
16005 (mul<mode>3): Ditto.
16006 (<insn><mode>3): Ditto.
16007 (vlshr<mode>3): Extend to support avx512 vlshr.
16008 (v<insn><mode>3): New expander for
16010 (v<insn>v8qi3): Ditto.
16011 (vashrv8hi3<mask_name>): Renamed to ..
16012 (vashr<mode>3): And extend to support V16QImode for avx512.
16013 (vashrv16qi3): Deleted.
16014 (vashrv2di3<mask_name>): Extend expander to support avx512
16017 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
16019 * doc/lto.texi (Design Overview): Update that slim objects are
16022 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
16024 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
16025 of OTHER_POWER10_MASKS so it will not be enabled by default.
16027 2021-06-23 Richard Biener <rguenther@suse.de>
16028 Martin Jambor <mjambor@suse.cz>
16030 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
16031 param replacement unconditionally. Adjust comment.
16033 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16035 * Makefile.in (OBJS): Add gimple-range-fold.o
16036 * gimple-range-fold.cc: New.
16037 * gimple-range-fold.h: New.
16038 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
16039 (gimple_range_calc_op2): Ditto.
16040 * gimple-range-gori.h: Move prototypes to here.
16041 * gimple-range.cc: Adjust include files.
16042 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
16043 (fur_source::get_operand): Ditto.
16044 (fur_source::get_phi_operand): Ditto.
16045 (fur_source::query_relation): Ditto.
16046 (fur_source::register_relation): Ditto.
16047 (class fur_edge): Ditto.
16048 (fur_edge::fur_edge): Ditto.
16049 (fur_edge::get_operand): Ditto.
16050 (fur_edge::get_phi_operand): Ditto.
16051 (fur_stmt::fur_stmt): Ditto.
16052 (fur_stmt::get_operand): Ditto.
16053 (fur_stmt::get_phi_operand): Ditto.
16054 (fur_stmt::query_relation): Ditto.
16055 (class fur_depend): Relocate to gimple-range-fold.h.
16056 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
16057 (fur_depend::register_relation): Ditto.
16058 (fur_depend::register_relation): Ditto.
16059 (class fur_list): Ditto.
16060 (fur_list::fur_list): Ditto.
16061 (fur_list::get_operand): Ditto.
16062 (fur_list::get_phi_operand): Ditto.
16063 (fold_range): Ditto.
16064 (adjust_pointer_diff_expr): Ditto.
16065 (gimple_range_adjustment): Ditto.
16066 (gimple_range_base_of_assignment): Ditto.
16067 (gimple_range_operand1): Ditto.
16068 (gimple_range_operand2): Ditto.
16069 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
16070 (gimple_range_calc_op2): Ditto.
16071 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
16072 (fold_using_range::range_of_range_op): Ditto.
16073 (fold_using_range::range_of_address): Ditto.
16074 (fold_using_range::range_of_phi): Ditto.
16075 (fold_using_range::range_of_call): Ditto.
16076 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
16077 (fold_using_range::range_of_builtin_call): Ditto.
16078 (fold_using_range::range_of_cond_expr): Ditto.
16079 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
16080 (fold_using_range::relation_fold_and_or): Ditto.
16081 (fold_using_range::postfold_gcond_edges): Ditto.
16082 * gimple-range.h: Add gimple-range-fold.h to include files. Change
16083 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
16084 (gimple_range_handler): Relocate to gimple-range-fold.h.
16085 (gimple_range_ssa_p): Ditto.
16086 (range_compatible_p): Ditto.
16087 (class fur_source): Ditto.
16088 (class fur_stmt): Ditto.
16089 (class fold_using_range): Ditto.
16090 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
16091 (gimple_range_calc_op2): Ditto.
16093 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16095 PR tree-optimization/101148
16096 PR tree-optimization/101014
16097 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
16098 (ranger_cache::~ranger_cache): Adjust.
16099 (ranger_cache::block_range): Check if propagation disallowed.
16100 (ranger_cache::propagate_cache): Disallow propagation if new value
16101 can't be stored properly.
16102 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
16104 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16106 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
16107 (sbr_vector::set_bb_range): Return true.
16108 (class sbr_sparse_bitmap): Adjust.
16109 (sbr_sparse_bitmap::set_bb_range): Return value.
16110 (block_range_cache::set_bb_range): Return value.
16111 (ranger_cache::propagate_cache): Use return value to print msg.
16112 * gimple-range-cache.h (class block_range_cache): Adjust.
16114 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16116 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
16118 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
16120 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
16121 so this pattern can be used for test/compare removal. Pass
16122 current insn to compute_logical_op_length and output_logical_op.
16123 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
16124 (h8300_and_costs): Add argument to compute_logical_op_length.
16125 (output_logical_op): Add new argument. Use it to determine if the
16126 condition codes are used and adjust the output accordingly.
16127 (compute_logical_op_length): Add new argument and update length
16128 computations when condition codes are used.
16129 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
16131 (output_logical_op): Likewise.
16133 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
16136 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
16137 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
16138 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
16139 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
16140 * config/i386/sse.md (unspec): ... here.
16142 2021-06-23 Martin Liska <mliska@suse.cz>
16145 * optc-save-gen.awk: Put back arm_fp16_format to
16148 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
16151 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
16154 (clz<mode>2): Update RTX pattern for additions.
16156 2021-06-23 Jakub Jelinek <jakub@redhat.com>
16158 PR middle-end/101167
16159 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
16160 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
16162 2021-06-22 Sergei Trofimovich <siarheit@google.com>
16164 * doc/rtl.texi: drop unbalanced parenthesis.
16166 2021-06-22 Richard Biener <rguenther@suse.de>
16168 PR middle-end/101156
16169 * gimplify.c (gimplify_expr): Remove premature incorrect
16172 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16174 PR tree-optimization/101159
16175 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
16178 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16180 PR middle-end/101160
16181 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
16182 clear crtl->return_rtx instead of keeping it referencing a pseudo.
16184 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16185 Andrew Pinski <apinski@marvell.com>
16187 PR tree-optimization/101162
16188 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
16191 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16193 * range-op.cc (range_relational_tests): New.
16194 (range_op_tests): Call range_relational_tests.
16196 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16198 * range-op.cc (operator_cast::lhs_op1_relation): New.
16199 (operator_identity::lhs_op1_relation): Mew.
16201 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16203 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
16205 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16207 * range-op.cc (operator_plus::lhs_op1_relation): New.
16208 (operator_plus::lhs_op2_relation): New.
16210 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16212 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
16213 relation_oracle if dominators exist.
16214 (ranger_cache::~ranger_cache): Dispose of oracle.
16215 (ranger_cache::dump_bb): Dump oracle.
16216 * gimple-range.cc (fur_source::fur_source): New.
16217 (fur_source::get_operand): Use mmeber query.
16218 (fur_source::get_phi_operand): Use member_query.
16219 (fur_source::query_relation): New.
16220 (fur_source::register_dependency): Delete.
16221 (fur_source::register_relation): New.
16222 (fur_edge::fur_edge): Adjust.
16223 (fur_edge::get_phi_operand): Fix comment.
16224 (fur_edge::query): Delete.
16225 (fur_stmt::fur_stmt): Adjust.
16226 (fur_stmt::query): Delete.
16227 (fur_depend::fur_depend): Adjust.
16228 (fur_depend::register_relation): New.
16229 (fur_depend::register_relation): New.
16230 (fur_list::fur_list): Adjust.
16231 (fur_list::get_operand): Use member query.
16232 (fold_using_range::range_of_range_op): Process and query relations.
16233 (fold_using_range::range_of_address): Adjust dependency call.
16234 (fold_using_range::range_of_phi): Ditto.
16235 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
16236 (fold_using_range::relation_fold_and_or): New.
16237 (fold_using_range::postfold_gcond_edges): New.
16238 * gimple-range.h (class gimple_ranger): Adjust.
16239 (class fur_source): Adjust members.
16240 (class fur_stmt): Ditto.
16241 (class fold_using_range): Ditto.
16243 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16245 * range-op.cc (range_operator::wi_fold): Apply relation effect.
16246 (range_operator::fold_range): Adjust and apply relation effect.
16247 (*::fold_range): Add relation parameters.
16248 (*::op1_range): Ditto.
16249 (*::op2_range): Ditto.
16250 (range_operator::lhs_op1_relation): New.
16251 (range_operator::lhs_op2_relation): New.
16252 (range_operator::op1_op2_relation): New.
16253 (range_operator::op1_op2_relation_effect): New.
16254 (relop_early_resolve): New.
16255 (operator_equal::op1_op2_relation): New.
16256 (operator_equal::fold_range): Call relop_early_resolve.
16257 (operator_not_equal::op1_op2_relation): New.
16258 (operator_not_equal::fold_range): Call relop_early_resolve.
16259 (operator_lt::op1_op2_relation): New.
16260 (operator_lt::fold_range): Call relop_early_resolve.
16261 (operator_le::op1_op2_relation): New.
16262 (operator_le::fold_range): Call relop_early_resolve.
16263 (operator_gt::op1_op2_relation): New.
16264 (operator_gt::fold_range): Call relop_early_resolve.
16265 (operator_ge::op1_op2_relation): New.
16266 (operator_ge::fold_range): Call relop_early_resolve.
16267 * range-op.h (class range_operator): Adjust parameters and methods.
16269 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16271 * Makefile.in (OBJS): Add value-relation.o.
16272 * gimple-range.h: Adjust include files.
16273 * tree-data-ref.c: Adjust include file order.
16274 * value-query.cc (range_query::get_value_range): Default to no oracle.
16275 (range_query::query_relation): New.
16276 (range_query::query_relation): New.
16277 * value-query.h (class range_query): Adjust.
16278 * value-relation.cc: New.
16279 * value-relation.h: New.
16281 2021-06-22 Richard Biener <rguenther@suse.de>
16283 PR tree-optimization/101151
16284 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
16287 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
16289 * config/riscv/riscv.c (thead_c906_tune_info): New.
16290 (riscv_tune_info_table): Use new tune.
16292 2021-06-22 Richard Biener <rguenther@suse.de>
16294 PR tree-optimization/101158
16295 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
16296 checking after checking for matching operation.
16298 2021-06-22 Richard Biener <rguenther@suse.de>
16300 PR tree-optimization/101159
16301 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
16302 missing NULL vectype check.
16304 2021-06-22 Richard Biener <rguenther@suse.de>
16306 PR tree-optimization/101154
16307 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
16309 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16312 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
16313 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
16314 * config/i386/i386.c (ix86_expand_prologue): Clear it.
16315 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
16316 Remove "" from match_operand. Emit new insns using emit_move_insn and
16317 set ix86_last_zero_store_uid to INSN_UID of the last store.
16318 Add peephole2s for 1/2/4 stores of const0_rtx following previous
16321 2021-06-22 Martin Liska <mliska@suse.cz>
16323 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
16326 2021-06-22 Martin Liska <mliska@suse.cz>
16328 * gcov-io.h: Remove padding entries.
16330 2021-06-22 liuhongt <hongtao.liu@intel.com>
16332 PR tree-optimization/97770
16333 * tree-vect-patterns.c (vect_recog_popcount_pattern):
16335 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
16337 2021-06-22 liuhongt <hongtao.liu@intel.com>
16340 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
16341 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
16342 (expand<mode>_mask): this ..
16343 (*expand<mode>_mask): New pre_reload splitter to transform
16344 v{,p}expand* to vmov* when mask is zero, all ones, or has all
16345 ones in it's lower part, otherwise still generate
16348 2021-06-22 liuhongt <hongtao.liu@intel.com>
16351 * config/i386/i386-expand.c
16352 (ix86_expand_special_args_builtin): Keep constm1_operand only
16353 if it satisfies insn's operand predicate.
16355 2021-06-21 Jason Merrill <jason@redhat.com>
16358 * df-scan.c (df_ref_record): Check that regno < endregno.
16359 * function.c (assign_parms, expand_function_end): Do nothing with a
16360 TYPE_EMPTY_P result.
16362 2021-06-21 Richard Biener <rguenther@suse.de>
16364 PR tree-optimization/101120
16365 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
16367 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
16368 DR chain DCE capability.
16369 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
16370 * tree-vect-stmts.c (vectorizable_load): Remove unused
16371 loads in the DR chain for SLP.
16373 2021-06-21 Jakub Jelinek <jakub@redhat.com>
16375 PR inline-asm/100785
16376 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
16377 output or input operands were already error_mark_node.
16378 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
16379 remove all inputs, outputs and clobbers from the asm and
16380 set template to "".
16382 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16384 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
16385 (vceq_s16): Likewise.
16386 (vceq_s32): Likewise.
16387 (vceq_u8): Likewise.
16388 (vceq_u16): Likewise.
16389 (vceq_u32): Likewise.
16390 (vceq_p8): Likewise.
16391 (vceqq_s8): Likewise.
16392 (vceqq_s16): Likewise.
16393 (vceqq_s32): Likewise.
16394 (vceqq_u8): Likewise.
16395 (vceqq_u16): Likewise.
16396 (vceqq_u32): Likewise.
16397 (vceqq_p8): Likewise.
16398 (vceq_f32): Gate __a == __b on __FAST_MATH__.
16399 (vceqq_f32): Likewise.
16400 (vceq_f16): Likewise.
16401 (vceqq_f16): Likewise.
16403 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16406 * config/arm/iterators.md (NEON_VACMP): Remove.
16407 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
16409 (neon_vca<cmp_op><mode>_insn): Likewise.
16410 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
16413 2021-06-21 Richard Biener <rguenther@suse.de>
16415 PR tree-optimization/101121
16416 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
16417 when we just lack a stmt with the desired op when doing permutation.
16418 (vect_build_slp_tree): When caching a failed SLP build attempt
16419 assert that at least one lane is marked as not matching.
16421 2021-06-21 liuhongt <hongtao.liu@intel.com>
16424 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
16425 register alternative.
16426 (*and<mode>_1): Ditto.
16428 (*andn<mode>_1): Ditto.
16429 (*<code><mode>_1): Ditto.
16430 (*<code>qi_1): Ditto.
16431 (*one_cmpl<mode>2_1): Ditto.
16432 (*one_cmplsi2_1_zext): Ditto.
16433 (*one_cmplqi2_1): Ditto.
16434 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
16435 the order of mask registers to be before general registers.
16437 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
16440 * config/i386/i386.md: New define_peephole2s to shrink writing
16441 1, 2 or 4 consecutive zeros to memory when optimizing for size.
16443 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
16445 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
16446 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
16447 more efficient code when the source can be trivially simplified.
16449 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
16451 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
16452 a range if global is not available.
16453 (ranger_cache::entry_range): Fallback to range_of_def.
16454 * gimple-range-cache.h (range_of_def): Adjust prototype.
16456 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
16458 PR tree-optimization/101014
16459 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
16461 (ranger_cache::~ranger_cache): Ditto.
16462 (ranger_cache::enable_new_values): Delete.
16463 (ranger_cache::push_poor_value): Delete.
16464 (ranger_cache::range_of_def): Remove poor value processing.
16465 (ranger_cache::entry_range): Ditto.
16466 (ranger_cache::fill_block_cache): Ditto.
16467 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
16468 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
16469 * gimple-range.h (class gimple_ranger): Adjust.
16471 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16474 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
16475 derived from arm_canon_arch.
16476 (arm_canon_arch_option): Call it.
16477 (arm_canon_arch_multilib_option): New function.
16478 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
16479 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
16480 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
16481 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
16482 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
16483 * config/arm/arm.opt (mlibarch): New option.
16484 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
16485 of march on RHS with mlibarch.
16487 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
16489 * config.in: Regenerate.
16490 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
16492 * configure: Regenerate.
16493 * configure.ac: Fix for global_load assembler functions.
16495 2021-06-18 Richard Biener <rguenther@suse.de>
16497 PR tree-optimization/101112
16498 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
16499 to lookup a pattern stmt def.
16501 2021-06-18 Jakub Jelinek <jakub@redhat.com>
16503 PR middle-end/101062
16504 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
16505 representatives in QUAL_UNION_TYPE.
16507 2021-06-18 Andrew Pinski <apinski@marvell.com>
16509 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
16510 Add counting of how many times it is done.
16511 (factor_out_conditional_conversion): Likewise.
16512 (match_simplify_replacement): Likewise.
16513 (value_replacement): Likewise.
16514 (spaceship_replacement): Likewise.
16515 (cond_store_replacement): Likewise.
16516 (cond_if_else_store_replacement_1): Likewise.
16517 (hoist_adjacent_loads): Likewise.
16519 2021-06-18 Andrew Pinski <apinski@marvell.com>
16521 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
16522 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
16523 (verify_gimple_assign_binary): Reject point and offset types on
16524 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
16525 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
16526 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
16528 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
16530 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
16531 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
16532 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
16535 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
16537 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
16538 earlyclobber to alts 0/1.
16539 (gen_addadd): Add earlyclobber to alts 0/1.
16540 * config/rs6000/fusion.md: Regenerate file.
16542 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16544 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
16546 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
16548 * gimple-range-cache.cc: Comment cleanups.
16549 * gimple-range-gori.cc: Comment cleanups.
16550 * gimple-range.cc: Comment/spacing cleanups
16551 * value-range.h: Comment cleanups.
16553 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
16556 * calls.c (expand_call): Replace PUSH_ARGS with
16557 targetm.calls.push_argument (0).
16558 (emit_library_call_value_1): Likewise.
16559 * defaults.h (PUSH_ARGS): Removed.
16560 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
16561 targetm.calls.push_argument (0).
16562 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
16563 (emit_push_insn): Pass the number bytes to push to
16564 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
16565 * hooks.c (hook_bool_uint_true): New.
16566 * hooks.h (hook_bool_uint_true): Likewise.
16567 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
16568 targetm.calls.push_argument (0).
16569 * target.def (push_argument): Add a targetm.calls hook.
16570 * targhooks.c (default_push_argument): New.
16571 * targhooks.h (default_push_argument): Likewise.
16572 * config/bpf/bpf.h (PUSH_ARGS): Removed.
16573 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
16574 * config/cr16/cr16.h (PUSH_ARGS): Removed.
16575 * config/i386/i386.c (ix86_push_argument): New.
16576 (TARGET_PUSH_ARGUMENT): Likewise.
16577 * config/i386/i386.h (PUSH_ARGS): Removed.
16578 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
16579 * config/m32c/m32c.h (PUSH_ARGS): Removed.
16580 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
16581 * config/pru/pru.h (PUSH_ARGS): Likewise.
16582 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
16583 TARGET_PUSH_ARGUMENT hook.
16584 * doc/tm.texi: Regenerated.
16586 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
16589 * config/i386/i386-expand.c (expand_vector_set_var):
16590 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
16591 ix86_expand_vector_init_duplicate.
16592 (ix86_expand_vector_init_duplicate): Emit insv_1 for
16593 QImode for !TARGET_PARTIAL_REG_STALL.
16594 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
16595 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
16596 as operand 2 predicate. Call ix86_expand_vector_set_var
16597 for non-constant index operand.
16598 (vec_setv2si): Ditto.
16599 (vec_setv4hi): Ditto.
16600 (vec_setv8qi): ditto.
16602 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
16604 PR tree-optimization/100790
16605 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
16608 2021-06-17 Martin Liska <mliska@suse.cz>
16610 * doc/invoke.texi: Use consistently -O1 instead of -O.
16612 2021-06-17 Martin Liska <mliska@suse.cz>
16614 * gcov-io.h: Update documentation entry about string format.
16616 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
16619 * config/s390/vecintrin.h (vec_doublee): Fix to use
16620 __builtin_s390_vflls.
16621 (vec_floate): Fix to use __builtin_s390_vflrd.
16623 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16625 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
16626 * dominance.h (get_dominated_to_depth): Likewise.
16627 (get_all_dominated_blocks): Likewise.
16628 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
16629 * gcse.c (hoist_code): Likewise.
16630 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
16631 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
16632 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
16633 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
16635 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16637 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
16638 * dominance.h (get_dominated_by_region): Likewise.
16639 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
16640 (gimple_duplicate_sese_tail): Likewise.
16641 (move_sese_region_to_fn): Likewise.
16643 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16645 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
16646 * dominance.h (get_dominated_by): Likewise.
16647 * auto-profile.c (afdo_find_equiv_class): Adjust.
16648 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
16649 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
16650 * tree-cfg.c (test_linear_chain): Likewise.
16651 (test_diamond): Likewise.
16653 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16655 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
16656 * cfgloopanal.c (get_loop_hot_path): Likewise.
16657 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
16659 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16661 * cgraph.c (cgraph_node::collect_callers): Return
16662 auto_vec<cgraph_edge *>.
16663 * cgraph.h (cgraph_node::collect_callers): Likewise.
16664 * ipa-cp.c (create_specialized_node): Adjust.
16665 (decide_about_value): Likewise.
16666 (decide_whether_version_node): Likewise.
16667 * ipa-sra.c (process_isra_node_results): Likewise.
16669 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
16671 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
16672 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
16674 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
16677 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
16679 * gimple-range.cc (debug_seed_ranger): New.
16680 (dump_ranger): New.
16681 (debug_ranger): New.
16683 2021-06-17 Richard Biener <rguenther@suse.de>
16685 PR tree-optimization/54400
16686 * tree-vectorizer.h (enum slp_instance_kind): Add
16687 slp_inst_kind_bb_reduc.
16688 (reduction_fn_for_scalar_code): Declare.
16689 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
16690 Check SLP_INSTANCE_KIND instead of looking at the
16692 (vect_slp_analyze_instance_alignment): Likewise.
16693 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
16694 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
16695 chain linearization from vect_build_slp_tree_2 and generalize
16696 for the use of BB reduction vectorization.
16697 (vect_build_slp_tree_2): Adjust accordingly.
16698 (vect_optimize_slp): Elide permutes at the root of BB reduction
16700 (vectorizable_bb_reduc_epilogue): New function.
16701 (vect_slp_prune_covered_roots): Likewise.
16702 (vect_slp_analyze_operations): Use them.
16703 (vect_slp_check_for_constructors): Recognize associatable
16704 chains for BB reduction vectorization.
16705 (vectorize_slp_instance_root_stmt): Generate code for the
16706 BB reduction epilogue.
16708 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
16710 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
16712 (gori_compute::may_recompute_p): New.
16713 (gori_compute::outgoing_edge_range_p): Perform recomputations.
16714 * gimple-range-gori.h (class gori_compute): Add prototype.
16716 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
16718 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
16719 true when a range can be calculated.
16720 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
16722 2021-06-16 Martin Sebor <msebor@redhat.com>
16724 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
16725 Correct documented defaults.
16727 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
16729 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
16730 m_new_value_p directly.
16732 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
16735 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
16736 Handle 64bit modes for TARGET_SSE4_1.
16737 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
16738 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
16739 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
16740 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
16741 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
16742 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
16744 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
16746 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
16747 Change to an expander that emits the correct instruction
16748 depending on endianness.
16749 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
16750 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
16752 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
16754 * config/aarch64/aarch64-simd-builtins.def: Split generator
16755 for aarch64_<su>qmovn builtins into scalar and vector
16757 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
16759 (aarch64_<su>qmovn<mode>_insn_be): Define.
16760 (aarch64_<su>qmovn<mode>): Split into scalar and vector
16761 variants. Change vector variant to an expander that emits the
16762 correct instruction depending on endianness.
16764 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
16766 * config/aarch64/aarch64-simd-builtins.def: Split generator
16767 for aarch64_sqmovun builtins into scalar and vector variants.
16768 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
16769 Split into scalar and vector variants. Change vector variant
16770 to an expander that emits the correct instruction depending
16772 (aarch64_sqmovun<mode>_insn_le): Define.
16773 (aarch64_sqmovun<mode>_insn_be): Define.
16775 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
16777 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
16778 Define - modeling zero-high-half semantics.
16779 (aarch64_xtn<mode>): Change to an expander that emits the
16780 appropriate instruction depending on endianness.
16781 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
16783 (aarch64_xtn2<mode>_le): Rename to...
16784 (aarch64_xtn2<mode>_insn_le): This.
16785 (aarch64_xtn2<mode>_be): Rename to...
16786 (aarch64_xtn2<mode>_insn_be): This.
16787 (vec_pack_trunc_<mode>): Emit truncation instruction instead
16789 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
16790 attribute iterator.
16792 2021-06-16 Martin Jambor <mjambor@suse.cz>
16794 PR tree-optimization/100453
16795 * tree-sra.c (create_access): Disqualify any const candidates
16796 which are written to.
16797 (sra_modify_expr): Do not store sub-replacements back to a const base.
16798 (handle_unscalarized_data_in_subtree): Likewise.
16799 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
16800 instead of constant_decl_p.
16802 2021-06-16 Jakub Jelinek <jakub@redhat.com>
16804 PR middle-end/101062
16805 * stor-layout.c (finish_bitfield_representative): For fields in unions
16806 assume nextf is always NULL.
16807 (finish_bitfield_layout): Compute bit field representatives also in
16808 unions, but handle it as if each bitfield was the only field in the
16811 2021-06-16 Richard Biener <rguenther@suse.de>
16813 PR tree-optimization/101088
16814 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
16815 supported refs on edges. Do not assert same ref but
16816 different kind stores are unsuported but mark them so.
16817 (hoist_memory_references): Only look for supported refs
16820 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
16822 PR rtl-optimization/46235
16823 * config/i386/i386.md: New define_split for bt followed by cmov.
16824 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
16825 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
16826 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
16827 by setnc with zero extension.
16829 2021-06-16 Richard Biener <rguenther@suse.de>
16831 PR tree-optimization/101083
16832 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
16833 vectype as argument.
16834 (vect_build_slp_tree_2): Adjust.
16836 2021-06-15 Martin Sebor <msebor@redhat.com>
16838 PR middle-end/100876
16839 * builtins.c: (gimple_call_return_array): Account for size_t
16840 mangling as either unsigned int or unsigned long
16842 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
16844 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
16845 up before eliminating comparisons.
16847 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
16849 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
16850 nonzero|X is nonzero.
16851 (range_op_bitwise_and_tests): Add tests for above.
16853 2021-06-15 Carl Love <cel@us.ibm.com>
16856 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
16858 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
16861 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
16864 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
16865 (struct gimplify_omp_ctx): Extend defaultmap array by one.
16866 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
16867 (omp_notice_variable): Update type classification for Fortran.
16868 (gimplify_scan_omp_clauses): Update calls for new argument; handle
16869 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
16870 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
16871 * langhooks.c (lhd_omp_scalar_p): Likewise.
16872 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
16873 (LANG_HOOKS_DECLS): Add them.
16874 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
16875 omp_scalar_p pointer type to include the new bool argument.
16877 2021-06-15 David Malcolm <dmalcolm@redhat.com>
16879 * doc/analyzer.texi
16880 (Special Functions for Debugging the Analyzer): Add
16881 __analyzer_dump_capacity.
16883 2021-06-15 Jakub Jelinek <jakub@redhat.com>
16886 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
16887 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
16889 2021-06-15 Richard Biener <rguenther@suse.de>
16891 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
16892 check to identify loop latches.
16893 * cfgloop.c (verify_loop_structure): Likewise.
16894 * loop-init.c (apply_loop_flags): Allow marked irreducible
16895 regions even with multiple latches.
16896 * predict.c (rebuild_frequencies): Simplify.
16898 2021-06-15 Richard Biener <rguenther@suse.de>
16900 * tree-ssa-threadupdate.c
16901 (jump_thread_path_registry::mark_threaded_blocks): Assert we
16902 have marked irreducible regions.
16904 2021-06-14 Martin Sebor <msebor@redhat.com>
16907 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
16908 Handle calls to placement new.
16909 (ndecl_dealloc_argno): Avoid placement delete.
16911 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
16914 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
16915 create_tmp_reg_or_ssa_name().
16917 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
16919 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
16920 (ranger_cache::enable_new_values): Set to specified value and
16921 return the old value.
16922 (ranger_cache::disable_new_values): Delete.
16923 (ranger_cache::fill_block_cache): Disable non 1st order derived
16925 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
16926 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
16928 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
16931 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
16932 Return true early when testing with V2HImode.
16933 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
16935 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
16937 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
16938 (mve_vec_unpack<US>_hi_<mode>): New pattern.
16939 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
16940 (mve_vmovntq_<supf><mode>): Prefix with '@'.
16941 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
16943 (vec_unpack<US>_lo_<mode>): Likewise.
16944 (vec_pack_trunc_<mode>): Rename to
16945 neon_quad_vec_pack_trunc_<mode>.
16946 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
16948 (vec_unpack<US>_lo_<mode>): New.
16949 (vec_pack_trunc_<mode>): New.
16951 2021-06-14 Richard Biener <rguenther@suse.de>
16953 PR tree-optimization/100934
16954 * tree-ssa-dom.c (pass_dominator::execute): Properly
16955 mark irreducible regions.
16957 2021-06-14 Martin Liska <mliska@suse.cz>
16959 * doc/invoke.texi: Put r{...} on the same line as @item.
16961 2021-06-14 Martin Liska <mliska@suse.cz>
16963 * doc/invoke.texi: Add missing newline.
16965 2021-06-14 Martin Liska <mliska@suse.cz>
16967 * doc/invoke.texi: Remove '+' charasters.
16969 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
16971 * config.gcc (arc): Add support for with_cpu option.
16972 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
16974 2021-06-14 Richard Biener <rguenther@suse.de>
16976 PR tree-optimization/101031
16977 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
16978 instead of size when accounting for a possibly string
16981 2021-06-14 Martin Liska <mliska@suse.cz>
16983 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
16985 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
16987 * value-query.cc (gimple_range_global): Call get_range_global
16988 if called after inlining.
16990 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
16993 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
16994 Emit constant permutation insn directly from here.
16996 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
16998 * attribs.c (find_attribute_namespace): Iterate over vec<> with
17000 * auto-profile.c (afdo_find_equiv_class): Likewise.
17001 * gcc.c (do_specs_vec): Likewise.
17002 (do_spec_1): Likewise.
17003 (driver::set_up_specs): Likewise.
17004 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
17005 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
17006 (imm_store_chain_info::try_coalesce_bswap): Likewise.
17007 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
17008 (get_location_for_stmts): Likewise.
17009 * graphite-poly.c (print_iteration_domains): Likewise.
17010 (free_poly_bb): Likewise.
17011 (remove_gbbs_in_scop): Likewise.
17012 (free_scop): Likewise.
17013 (dump_gbb_cases): Likewise.
17014 (dump_gbb_conditions): Likewise.
17015 (print_pdrs): Likewise.
17016 (print_scop): Likewise.
17017 * ifcvt.c (cond_move_process_if_block): Likewise.
17018 * lower-subreg.c (decompose_multiword_subregs): Likewise.
17019 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
17020 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
17021 * sel-sched-dump.c (dump_insn_vector): Likewise.
17022 * store-motion.c (store_ops_ok): Likewise.
17023 (store_killed_in_insn): Likewise.
17024 * timevar.c (timer::named_items::print): Likewise.
17025 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
17026 (cleanup_tree_cfg_noloop): Likewise.
17027 * tree-data-ref.c (dump_data_references): Likewise.
17028 (print_dir_vectors): Likewise.
17029 (print_dist_vectors): Likewise.
17030 (dump_data_dependence_relations): Likewise.
17031 (dump_dist_dir_vectors): Likewise.
17032 (dump_ddrs): Likewise.
17033 (create_runtime_alias_checks): Likewise.
17034 (free_subscripts): Likewise.
17035 (save_dist_v): Likewise.
17036 (save_dir_v): Likewise.
17037 (invariant_access_functions): Likewise.
17038 (same_access_functions): Likewise.
17039 (access_functions_are_affine_or_constant_p): Likewise.
17040 (find_data_references_in_stmt): Likewise.
17041 (graphite_find_data_references_in_stmt): Likewise.
17042 (free_dependence_relations): Likewise.
17043 (free_data_refs): Likewise.
17044 * tree-inline.c (copy_debug_stmts): Likewise.
17045 * tree-into-ssa.c (dump_currdefs): Likewise.
17046 (rewrite_update_phi_arguments): Likewise.
17047 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
17048 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
17050 (vect_slp_analyze_node_dependences): Likewise.
17051 (vect_slp_analyze_instance_dependence): Likewise.
17052 (vect_record_base_alignments): Likewise.
17053 (vect_get_peeling_costs_all_drs): Likewise.
17054 (vect_peeling_supportable): Likewise.
17055 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
17056 (vec_info::free_stmt_vec_infos): Likewise.
17058 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
17060 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
17061 (andqi3_1<cczn>): Removed.
17062 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
17063 (H8/SX bit logicals): Split out from other patterns.
17064 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
17065 mulqihi3_const_clobber_flags.
17066 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
17068 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
17071 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
17072 to true if red zone is used.
17073 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
17074 ix86_red_zone_used.
17075 * config/i386/i386.h (machine_function): Add red_zone_used.
17076 (ix86_red_zone_size): Removed.
17077 (ix86_red_zone_used): New.
17078 * config/i386/i386.md (peephole2 patterns): Replace
17079 ix86_red_zone_size with ix86_red_zone_used.
17081 2021-06-12 Jason Merrill <jason@redhat.com>
17083 * doc/extend.texi (unused variable attribute): Applies to
17084 structure fields as well.
17086 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
17088 * auto-profile.c (read_profile): fix a typo in an error string
17090 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
17092 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
17094 * tree-pretty-print.c (dump_omp_clauses): Update.
17095 (dump_generic_node) <OMP_CLAUSE>: Use it.
17097 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
17100 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
17101 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
17102 the polymorphic variants matching code.
17103 (__arm_vld1q_z): Likewise.
17104 (__arm_vld2q): Likewise.
17105 (__arm_vld4q): Likewise.
17106 (__arm_vldrbq_gather_offset): Likewise.
17107 (__arm_vldrbq_gather_offset_z): Likewise.
17109 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
17111 PR tree-optimization/96392
17112 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
17114 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
17116 PR tree-optimization/96392
17117 * fold-const.c (fold_real_zero_addition_p): Take both arguments
17118 of the addition or subtraction, not just the zero. Use this
17119 other argument in tests for signaling NaNs and signed zeros.
17120 (tree_expr_maybe_real_minus_zero_p): New predicate.
17121 * fold-const.h (fold_real_zero_addition_p): Update prototype.
17122 (tree_expr_maybe_real_minus_zero_p): New function prototype.
17123 * match.pd: Update calls to fold_real_zero_addition_p.
17124 Replace HONOR_NANS with tree_expr_maybe_nan_p.
17125 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
17126 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
17127 * tree-ssa-reassoc.c (eliminate_using_constants): Update
17128 call to fold_real_zero_addition_p.
17130 2021-06-11 Richard Biener <rguenther@suse.de>
17132 PR tree-optimization/101025
17133 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
17134 all refs that require dependence checking.
17136 2021-06-11 Richard Biener <rguenther@suse.de>
17138 PR tree-optimization/101028
17139 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
17140 reassoc discovery fails fatally, mark appropriate lanes
17143 2021-06-11 Richard Biener <rguenther@suse.de>
17145 PR tree-optimization/101026
17146 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
17147 have a representative for the associated chain nodes.
17149 2021-06-11 Jakub Jelinek <jakub@redhat.com>
17151 PR rtl-optimization/101008
17152 * simplify-rtx.c (relational_result): New function.
17153 (simplify_logical_relational_operation,
17154 simplify_relational_operation): Use it.
17156 2021-06-11 Jakub Jelinek <jakub@redhat.com>
17159 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
17161 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
17164 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
17165 false if the permutation can be implemented with constant
17166 permutation instruction in wider mode.
17167 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
17168 Handle V8QImode and V4HImode.
17170 2021-06-11 Martin Liska <mliska@suse.cz>
17172 PR gcov-profile/100788
17173 * common.opt: Add new option.
17174 * coverage.c (coverage_begin_function): Emit warning instead on
17175 the internal compiler error.
17176 * doc/invoke.texi: Document the option.
17177 * toplev.c (process_options): Enable it by default.
17179 2021-06-11 Richard Biener <rguenther@suse.de>
17181 PR middle-end/101009
17182 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
17183 to set *init_b to true when we encounter a constant equal
17185 (compute_affine_dependence): Also dump the actual DR_REF.
17187 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
17189 PR tree-optimization/100984
17190 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
17191 replacements table.
17192 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
17194 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
17196 * config/rs6000/rs6000.md
17197 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
17198 define_insn_and_split.
17200 2021-06-11 Richard Biener <rguenther@suse.de>
17202 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
17203 to sort operands of the associative chain.
17205 2021-06-11 Richard Biener <rguenther@suse.de>
17207 * system.h (gcc_stablesort_r): Declare.
17208 * sort.cc (gcc_sort_r): Support stable sort.
17209 (gcc_stablesort_r): Define.
17210 * vec.h (vec<>::stablesort): Add.
17212 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
17215 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
17216 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
17217 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
17218 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
17219 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
17220 (expand_vec_perm_interleave2): Handle 64bit modes.
17221 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
17222 (expand_vec_perm_even_odd_1): Ditto.
17223 (ix86_vectorize_vec_perm_const): Ditto.
17224 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
17225 * config/i386/sse.md: ... here.
17226 * config/i386/mmx.md (*vec_interleave_lowv2sf):
17227 New insn_and_split pattern.
17228 (*vec_interleave_highv2sf): Ditto.
17229 (mmx_pshufbv8qi3): New insn pattern.
17230 (*mmx_pblendw): Ditto.
17232 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
17234 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
17235 (build_acc): Likewise.
17236 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
17237 source operands in little-endian mode.
17238 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
17239 (mma_init_builtins): Likewise.
17240 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
17241 ordering for the MMA assemble and build source operands.
17242 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
17244 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
17247 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
17249 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
17251 * config/h8300/extensions.md: Replace _clobber_flags patterns
17254 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
17256 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
17257 (vcond_mask_<mode><tointvec>): this.
17259 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
17260 Thomas Schwinge <thomas@codesourcery.com>
17262 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
17263 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
17264 * gimple.h (enum gf_mask): Split
17265 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
17266 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
17267 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
17268 (is_gimple_omp_oacc): Update.
17269 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
17270 * gimplify.c (gimplify_omp_target_update): Likewise.
17271 * omp-expand.c (expand_omp_target, build_omp_regions_1)
17272 (omp_make_gimple_edges): Likewise.
17273 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
17276 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
17278 * value-query.cc (value_query::value_on_edge): Rename name to
17280 (range_query::range_on_edge): Same.
17281 (range_query::value_of_expr): Same.
17282 (range_query::value_on_edge): Same.
17283 * value-query.h (class value_query): Same.
17284 (class range_query): Same.
17286 2021-06-10 Richard Biener <rguenther@suse.de>
17288 PR tree-optimization/101003
17289 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
17290 use the pattern stmt defs when linearizing a chain.
17292 2021-06-10 Jakub Jelinek <jakub@redhat.com>
17295 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
17296 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
17298 2021-06-10 Clement Chigot <clement.chigot@atos.net>
17300 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
17301 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
17303 2021-06-09 Andrew Pinski <apinski@marvell.com>
17305 PR tree-optimization/100925
17306 * match.pd (a ? CST1 : CST2): Limit transformations
17307 that would produce a negative to integeral types only.
17308 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
17310 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17313 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17315 * doc/tm.texi: Correctly update.
17317 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17319 * doc/tm.texi: Correctly update.
17321 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
17324 * doc/tm.texi.in (Trampolines): Add a missing blank line.
17326 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
17329 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
17330 and -ftrampolines work only with Ada.
17331 * doc/tm.texi.in (Trampolines): Likewise.
17332 * doc/tm.texi: Regenerated.
17334 2021-06-09 Carl Love <cel@us.ibm.com>
17336 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
17337 Add define for new builtins.
17338 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
17339 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
17340 overloaded builtin definitions.
17341 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
17342 VSIGNEXTSD2Q): Add builtin expansions.
17343 (SIGNEXT): Add P10 overload definition.
17344 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
17345 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
17346 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
17347 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
17348 vsignextend_si_v2di)[VIlong]: Add define_expand.
17349 Make define_insn vsx_sign_extend_si_v2di visible.
17350 * doc/extend.texi: Add documentation for the vec_signexti,
17351 vec_signextll builtins and vec_signextq.
17353 2021-06-09 Carl Love <cel@us.ibm.com>
17355 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
17356 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
17357 __floattikf_sw, __floatuntikf_sw respectively.
17358 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
17359 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
17360 define_insn for mode IEEE 128.
17362 2021-06-09 Carl Love <cel@us.ibm.com>
17364 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
17365 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
17366 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
17367 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
17368 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
17369 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
17370 uses of VSX_TI to VEC_TI.
17372 2021-06-09 Carl Love <cel@us.ibm.com>
17374 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
17376 2021-06-09 Carl Love <cel@us.ibm.com>
17378 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
17380 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
17381 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
17382 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
17383 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
17384 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
17385 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
17386 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
17388 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
17389 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
17390 altivec_vrlqnm): New define_expands.
17391 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
17392 VCMPGTUT_P): Add macro expansions.
17393 (BU_P10V_AV_P): Add builtin predicate definition.
17394 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
17395 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
17396 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
17397 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
17398 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
17399 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
17400 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
17401 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
17402 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
17403 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
17404 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
17405 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
17406 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
17407 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
17408 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
17409 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
17410 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
17411 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
17412 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
17413 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
17414 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
17415 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
17416 P10V_BUILTIN_MODU_V1TI):
17417 New overloaded definitions.
17418 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
17419 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
17420 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
17421 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
17422 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
17423 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
17425 (altivec_init_builtins): New E_V1TImode case statement.
17426 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
17427 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
17428 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
17429 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
17430 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
17431 E_V1TImode]: New case statements.
17432 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
17433 value RS6000_BTI_bool_V1TI.
17434 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
17435 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
17436 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
17437 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
17438 vlshrv1ti3, vashrv1ti3): New define_expands.
17439 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
17440 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
17441 UNSPEC_VSX_MODUQ): New unspecs.
17442 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
17443 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
17445 (vcmpnet): New define_expand.
17446 * doc/extend.texi: Add documentation for the new builtins vec_rl,
17447 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
17448 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
17449 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
17450 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
17451 vec_any_ge, vec_any_le.
17453 2021-06-09 Carl Love <cel@us.ibm.com>
17455 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
17456 bug in argument generation.
17458 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
17460 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
17462 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
17463 remove <supf> iterator.
17464 (mve_vclzq_u<mode>): New.
17465 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
17466 (neon_vclz<mode): Move to ...
17467 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
17468 * config/arm/vec-common.md: ... here. Add support for MVE.
17470 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
17472 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
17473 (@mve_vrhaddq_<supf><mode): Likewise.
17474 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
17475 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
17476 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
17478 2021-06-09 imba-tjd <109224573@qq.com>
17480 * doc/invoke.texi: Fix typo.
17482 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
17484 PR middle-end/53267
17485 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
17486 Support evaluation of fmod/fmodf/fmodl at compile-time.
17488 2021-06-09 Richard Biener <rguenther@suse.de>
17490 PR tree-optimization/100981
17491 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
17492 gimple_get_lhs to also handle calls.
17493 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
17496 2021-06-09 Richard Biener <rguenther@suse.de>
17498 PR tree-optimization/97832
17499 * tree-vectorizer.h (_slp_tree::failed): New.
17500 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
17502 (_slp_tree::~_slp_tree): Free failed.
17503 (vect_build_slp_tree): Retain failed nodes and record
17504 matches in them, copying that back out when running
17505 into a cached fail. Dump start and end of discovery.
17506 (dt_sort_cmp): New.
17507 (vect_build_slp_tree_2): Handle associatable chains
17508 together doing more aggressive operand swapping.
17510 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
17513 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
17515 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
17516 Linux and GNU targets.
17518 2021-06-09 Richard Biener <rguenther@suse.de>
17520 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
17523 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
17525 * config/arc/arc.md (loop_end): Change it to
17526 define_insn_and_split.
17528 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
17530 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
17531 (machi): New pattern.
17532 (umaddhisi4): Use VMAC2HU instruction.
17533 (umachi): New pattern.
17535 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
17537 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
17538 * config/arc/arc.c (arc_split_move_p): New function.
17539 (arc_split_move): Clean up.
17540 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
17541 (movdf_insn): Likewise.
17542 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
17544 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
17547 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
17548 argument to "raw". Do not emit segment overrides when "raw" is true.
17550 2021-06-09 Martin Liska <mliska@suse.cz>
17552 * doc/gcov.texi: Create a proper JSON files.
17553 * doc/invoke.texi: Remove dots in order to make it a valid
17556 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
17558 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
17559 (insn_is_load_p): Use pattern_is_rotate64.
17560 (insn_is_swap_p): Likewise.
17561 (quad_aligned_load_p): Likewise.
17562 (const_load_sequence_p): Likewise.
17563 (replace_swapped_aligned_load): Likewise.
17564 (recombine_lvx_pattern): Likewise.
17565 (recombine_stvx_pattern): Likewise.
17567 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
17569 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
17570 fur_stmt source record.
17571 * gimple-range.cc (fur_source::get_operand): Generic range query.
17572 (fur_source::get_phi_operand): New.
17573 (fur_source::register_dependency): New.
17574 (fur_source::query): New.
17575 (class fur_edge): New. Edge source for operands.
17576 (fur_edge::fur_edge): New.
17577 (fur_edge::get_operand): New.
17578 (fur_edge::get_phi_operand): New.
17579 (fur_edge::query): New.
17580 (fur_stmt::fur_stmt): New.
17581 (fur_stmt::get_operand): New.
17582 (fur_stmt::get_phi_operand): New.
17583 (fur_stmt::query): New.
17584 (class fur_depend): New. Statement source and process dependencies.
17585 (fur_depend::fur_depend): New.
17586 (fur_depend::register_dependency): New.
17587 (class fur_list): New. List source for operands.
17588 (fur_list::fur_list): New.
17589 (fur_list::get_operand): New.
17590 (fur_list::get_phi_operand): New.
17591 (fold_range): New. Instantiate appropriate fur_source class and fold.
17592 (fold_using_range::range_of_range_op): Use new API.
17593 (fold_using_range::range_of_address): Ditto.
17594 (fold_using_range::range_of_phi): Ditto.
17595 (imple_ranger::fold_range_internal): Use fur_depend class.
17596 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
17597 * gimple-range.h (class fur_source): Now a base class.
17598 (class fur_stmt): New.
17599 (fold_range): New prototypes.
17600 (fur_source::fur_source): Delete.
17602 2021-06-08 Andrew Pinski <apinski@marvell.com>
17604 PR tree-optimization/25290
17605 * tree-ssa-phiopt.c (xor_replacement): Delete.
17606 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
17607 (match_simplify_replacement): Allow one cheap preparation
17608 statement that can be moved to before the if.
17610 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
17612 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
17613 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
17615 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
17617 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
17618 Create length attribute on define_insn_and_split. Only split for cases which we
17620 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
17621 fix length computation.
17622 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
17624 2021-06-08 Richard Biener <rguenther@suse.de>
17626 PR tree-optimization/100923
17627 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
17628 the operand vector to be valueized.
17629 (valueize_refs): Likewise.
17630 (valueize_shared_reference_ops_from_ref): Adjust.
17631 (valueize_shared_reference_ops_from_call): Likewise.
17632 (vn_reference_lookup_3): Likewise.
17633 (vn_reference_lookup_pieces): Likewise. Re-valueize
17634 with honoring availability when we are about to create
17635 the ao_ref and valueized before.
17636 (vn_reference_lookup): Likewise.
17637 (vn_reference_insert_pieces): Adjust.
17639 2021-06-08 Richard Biener <rguenther@suse.de>
17641 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
17642 (_slp_instance::root_stmts): ... a vector.
17643 (SLP_INSTANCE_ROOT_STMT): Rename to ...
17644 (SLP_INSTANCE_ROOT_STMTS): ... this.
17645 (slp_root::root): Change to...
17646 (slp_root::roots): ... a vector.
17647 (slp_root::slp_root): Adjust.
17648 * tree-vect-slp.c (_slp_instance::location): Adjust.
17649 (vect_free_slp_instance): Release the root stmt vector.
17650 (vect_build_slp_instance): Adjust.
17651 (vect_analyze_slp): Likewise.
17652 (_bb_vec_info::~_bb_vec_info): Likewise.
17653 (vect_slp_analyze_operations): Likewise.
17654 (vect_bb_vectorization_profitable_p): Likewise. Adjust
17655 costs for the root stmt.
17656 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
17658 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
17660 (vectorize_slp_instance_root_stmt): Adjust.
17661 (vect_schedule_slp): Likewise.
17663 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
17665 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
17666 (ssa_equiv_stack::ssa_equiv_stack): New.
17667 (ssa_equiv_stack::~ssa_equiv_stack): New.
17668 (ssa_equiv_stack::enter): New.
17669 (ssa_equiv_stack::leave): New.
17670 (ssa_equiv_stack::push_replacement): New.
17671 (ssa_equiv_stack::get_replacement): New.
17672 (is_pointer_ssa): New.
17673 (class pointer_equiv_analyzer): New.
17674 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
17675 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
17676 (pointer_equiv_analyzer::set_global_equiv): New.
17677 (pointer_equiv_analyzer::set_cond_equiv): New.
17678 (pointer_equiv_analyzer::get_equiv): New.
17679 (pointer_equiv_analyzer::enter): New.
17680 (pointer_equiv_analyzer::leave): New.
17681 (pointer_equiv_analyzer::get_equiv_expr): New.
17682 (pta_valueize): New.
17683 (pointer_equiv_analyzer::visit_stmt): New.
17684 (pointer_equiv_analyzer::visit_edge): New.
17685 (hybrid_folder::value_of_expr): Call PTA.
17686 (hybrid_folder::value_on_edge): Same.
17687 (hybrid_folder::pre_fold_bb): New.
17688 (hybrid_folder::post_fold_bb): New.
17689 (hybrid_folder::pre_fold_stmt): New.
17690 (rvrp_folder::pre_fold_bb): New.
17691 (rvrp_folder::post_fold_bb): New.
17692 (rvrp_folder::pre_fold_stmt): New.
17693 (rvrp_folder::value_of_expr): Call PTA.
17694 (rvrp_folder::value_on_edge): Same.
17696 2021-06-08 Jakub Jelinek <jakub@redhat.com>
17699 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
17700 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
17702 2021-06-08 Richard Biener <rguenther@suse.de>
17704 PR middle-end/100951
17705 * tree-vect-generic.c (expand_vector_piecewise): Build a
17706 VECTOR_CST if all elements are constant.
17707 (expand_vector_condition): Likewise.
17708 (lower_vec_perm): Likewise.
17709 (expand_vector_conversion): Likewise.
17711 2021-06-08 Martin Liska <mliska@suse.cz>
17713 * doc/invoke.texi: Document new param evrp-sparse-threshold.
17715 2021-06-08 Martin Liska <mliska@suse.cz>
17717 * genautomata.c (create_automata): Fix typo.
17719 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
17721 PR tree-optimization/100794
17722 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
17723 allow_unroll_p and only allow unrolling when it's true.
17724 (tree_predictive_commoning): Add parameter allow_unroll_p and
17726 (run_tree_predictive_commoning): Likewise.
17727 (pass_predcom::gate): Check flag_tree_loop_vectorize and
17728 global_options_set.x_flag_predictive_commoning.
17729 (pass_predcom::execute): Adjust for allow_unroll_p.
17731 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
17733 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
17734 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
17735 lambda function cleanup, remove scev_reset call, and adjust return
17737 (tree_predictive_commoning): Adjust for different changed values,
17738 only set flag TODO_update_ssa_only_virtuals if changed.
17739 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
17740 from todo_flags_finish.
17742 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
17744 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
17745 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
17746 (sbr_sparse_bitmap::bitmap_set_quad): New.
17747 (sbr_sparse_bitmap::bitmap_get_quad): New.
17748 (sbr_sparse_bitmap::set_bb_range): New.
17749 (sbr_sparse_bitmap::get_bb_range): New.
17750 (sbr_sparse_bitmap::bb_range_p): New.
17751 (block_range_cache::block_range_cache): initialize bitmap obstack.
17752 (block_range_cache::~block_range_cache): Destruct obstack.
17753 (block_range_cache::set_bb_range): Decide when to utilze the
17754 sparse on entry cache.
17755 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
17756 * params.opt (-param=evrp-sparse-threshold): New.
17758 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
17760 * bitmap.c (bitmap_set_aligned_chunk): New.
17761 (bitmap_get_aligned_chunk): New.
17762 (test_aligned_chunk): New.
17763 (bitmap_c_tests): Call test_aligned_chunk.
17764 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
17766 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
17769 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
17771 (ix86_expand_vector_init_one_nonzero): Ditto.
17772 (ix86_expand_vector_init_one_var): Ditto.
17773 (ix86_expand_vector_init_general): Ditto.
17774 * config/i386/mmx.md (vec_initv4qiqi): New expander.
17776 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
17778 * config/h8300/movepush.md: Change most _clobber_flags
17779 patterns to instead use <cczn> subst.
17780 (movsi_cczn): New pattern with usable CC cases split out.
17781 (movsi_h8sx_cczn): Likewise.
17783 2021-06-07 Martin Liska <mliska@suse.cz>
17785 * common/common-target.def: Split long lines and replace them
17787 * target.def: Likewise.
17788 * doc/tm.texi: Re-generated.
17790 2021-06-07 Jakub Jelinek <jakub@redhat.com>
17793 * fold-const.c (fold_read_from_vector): Return NULL if trying to
17794 read from a CONSTRUCTOR with vector type elements.
17796 2021-06-07 Jakub Jelinek <jakub@redhat.com>
17798 PR middle-end/100898
17799 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
17800 should copy any arguments. Don't call gimple_call_num_args
17801 on id->call_stmt or call_stmt more than once.
17803 2021-06-07 liuhongt <hongtao.liu@intel.com>
17806 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
17808 (<insn>v4siv4di2): Delete constraints for define_expand.
17810 2021-06-07 liuhongt <hongtao.liu@intel.com>
17813 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
17814 assignment of cfun->machine->has_explicit_vzeroupper.
17815 * config/i386/i386-features.c
17816 (ix86_add_reg_usage_to_vzerouppers): Delete.
17817 (ix86_add_reg_usage_to_vzeroupper): Ditto.
17818 (rest_of_handle_insert_vzeroupper): Remove
17819 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
17821 (gate): Remove cfun->machine->has_explicit_vzeroupper.
17822 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
17824 * config/i386/i386.c (ix86_insn_callee_abi): New function.
17825 (ix86_initialize_callee_abi): Ditto.
17826 (ix86_expand_avx_vzeroupper): Ditto.
17827 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
17829 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
17830 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
17832 * config/i386/i386.h (struct GTY(()) machine_function): Delete
17833 has_explicit_vzeroupper.
17834 * config/i386/i386.md (enum unspec): New member
17836 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
17837 define_constants for insn callee abi index.
17838 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
17839 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
17840 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
17841 (*avx_vzeroupper): Rename to ..
17842 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
17843 call_insn which has a special vzeroupper ABI.
17844 (*avx_vzeroupper_1): Deleted.
17846 2021-06-07 liuhongt <hongtao.liu@intel.com>
17849 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
17850 it won't use stack pointer reg.
17851 * final.c (leaf_function_p): When call_insn is a fake call, it
17852 won't affect caller as a leaf function.
17853 * reg-stack.c (callee_clobbers_any_stack_reg): New.
17854 (subst_stack_regs): When call_insn doesn't clobber any stack
17855 reg, don't clear the arguments.
17856 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
17858 * shrink-wrap.c (requires_stack_frame_p): No need for stack
17859 frame for a fake call.
17860 * rtl.h (FAKE_CALL_P): New macro.
17862 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
17864 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
17866 (sparc_order_regs_for_local_alloc): ...this.
17867 (sparc_leaf_reg_remap): Declare.
17868 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
17869 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
17870 * config/sparc/sparc.c (leaf_reg_remap): Delete.
17871 (order_regs_for_local_alloc): Rename to...
17872 (sparc_order_regs_for_local_alloc): ...this.
17873 (sparc_leaf_reg_remap): New function.
17874 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
17876 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
17878 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
17879 Use assemble_name to output BSS section name.
17881 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
17883 * config/i386/constraints.md (Bs):
17884 Remove boolean operators from match_test RTX.
17887 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
17890 2021-06-06 Martin Liska <mliska@suse.cz>
17892 * doc/extend.texi: Add missing @headitem.
17893 * doc/invoke.texi: Likewise.
17894 * doc/objc.texi: Likewise.
17896 2021-06-06 Martin Liska <mliska@suse.cz>
17898 * genhooks.c (emit_findices): Remove unused function.
17899 (emit_documentation): Do not call emit_findices
17900 and do not search for @Fcode directives.
17902 2021-06-06 Martin Liska <mliska@suse.cz>
17904 * doc/invoke.texi: Remove extra character.
17906 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
17908 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
17910 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
17912 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
17913 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
17914 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
17915 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
17917 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
17919 * config/or1k/or1k.md (*movdi): Fix empty split condition.
17921 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
17923 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
17926 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
17928 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
17929 *zero_extendsidi2): Fix empty split condition.
17931 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
17933 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
17935 * config/h8300/bitfield.md: Likewise.
17936 * config/h8300/combiner.md: Likewise.
17937 * config/h8300/divmod.md: Likewise.
17938 * config/h8300/extensions.md: Likewise.
17939 * config/h8300/jumpcall.md: Likewise.
17940 * config/h8300/movepush.md: Likewise.
17941 * config/h8300/multiply.md: Likewise.
17942 * config/h8300/other.md: Likewise.
17943 * config/h8300/shiftrotate.md: Likewise.
17944 * config/h8300/logical.md: Likewise. Fix split pattern to use
17945 code iterator that somehow slipped through.
17947 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
17949 PR middle-end/100905
17950 * tree-nested.c (convert_nonlocal_omp_clauses,
17951 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
17953 2021-06-04 Martin Sebor <msebor@redhat.com>
17955 PR middle-end/100732
17956 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
17957 with either source or destination argument of invalid type.
17958 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
17959 calls with arguments of invalid type.
17961 2021-06-04 Martin Sebor <msebor@redhat.com>
17963 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
17965 (attr_access::vla_bounds): Also handle VLA bounds.
17967 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
17969 * config/i386/predicates.md (GOT_memory_operand):
17970 Implement using match_code RTXes.
17971 (GOT32_symbol_operand): Ditto.
17973 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
17976 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
17978 (ix86_expand_vector_init_general): Ditto.
17979 Use SImode instead of word_mode for logic operations
17980 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
17981 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
17982 implemented by expand_vec_perm_1.
17983 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
17984 should be implemented using standard shuffle patterns.
17985 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
17986 V2HI modes to modes, implementable with shuffle for one operand.
17987 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
17988 (*pshufw_1): New insn pattern.
17989 (*vec_dupv2hi): Ditto.
17990 (vec_initv2hihi): New expander.
17992 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
17994 * config/arm/vfp.md (no_literal_pool_df_immediate,
17995 no_literal_pool_sf_immediate): Fix empty split condition.
17997 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
17999 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
18000 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
18001 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
18002 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
18003 *<sse4_1_avx2>_pblendvb_lt): Likewise.
18005 2021-06-04 Jakub Jelinek <jakub@redhat.com>
18008 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
18009 concatenation from half-sized modes with TImode elements.
18011 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
18013 * config/arc/arc.c (arc_override_options): Disable millicode
18014 thunks when RF16 is on.
18016 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
18018 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
18020 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
18022 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
18023 Replace PROMOTE_MODE marco with its content.
18025 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
18027 * config/cris/cris.md (*addi_reload): Fix empty split condition.
18029 2021-06-03 Jim Wilson <jimw@sifive.com>
18031 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
18032 turn it on for all riscv targets.
18034 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
18037 * config/i386/i386-expand.c (ix86_expand_vector_set):
18038 Handle V2HI and V4QI modes.
18039 (ix86_expand_vector_extract): Ditto.
18040 * config/i386/mmx.md (*pinsrw): New insn pattern.
18043 (*pextrw_zext): Ditto.
18045 (*pextrb_zext): Ditto.
18046 (vec_setv2hi): New expander.
18047 (vec_extractv2hihi): Ditto.
18048 (vec_setv4qi): Ditto.
18049 (vec_extractv4qiqi): Ditto.
18050 (vec_setv8qi): Enable only for TARGET_SSE4_1.
18051 (vec_extractv8qiqi): Ditto.
18053 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
18055 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
18056 order to subf instruction.
18057 * config/rs6000/fusion.md: Regenerate.
18059 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18061 * calls.c (get_size_range): Use range_of_expr instead of
18062 determine_value_range.
18063 * tree-affine.c (expr_to_aff_combination): Same.
18064 * tree-data-ref.c (split_constant_offset): Same.
18065 * tree-vrp.c (determine_value_range_1): Remove.
18066 (determine_value_range): Remove.
18067 * tree-vrp.h (determine_value_range): Remove.
18069 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18071 * function-tests.c (test_ranges): Call gimple_range_tests.
18072 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
18074 * gimple-range.cc (fur_source::get_operand): Do not call
18075 get_tree_range or gimple_range_global.
18077 (get_tree_range): Move to value-query.cc.
18078 Call get_arith_expr_range.
18079 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
18080 Include gimple-range-tests.cc.
18081 * gimple-range.h (fold_range): Add argument.
18082 (get_tree_range): Remove.
18083 * selftest.h (gimple_range_tests): New.
18084 * value-query.cc (global_range_query::range_of_expr): Add
18086 (range_query::get_tree_range): Move from gimple-range.cc.
18087 * value-query.h (class range_query): Add get_tree_range and
18088 get_arith_expr_range. Make fur_source a friend.
18089 * vr-values.c (vr_values::range_of_expr): Pass stmt to
18091 * gimple-range-tests.cc: New file.
18093 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18095 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
18096 update_global_range.
18097 * value-query.cc (update_global_range): New.
18098 * value-query.h (update_global_range): New.
18100 2021-06-03 David Malcolm <dmalcolm@redhat.com>
18102 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
18103 printing the same location twice if there are fix-it hints,
18104 multiple locations, or a label.
18106 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
18108 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
18109 thresholds to narrow the upper bound on epilogue iterations.
18111 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
18113 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
18114 (mve_vabsq_s<mode>): Likewise.
18115 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
18116 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
18117 * config/arm/vec-common.md (neg<mode>2): Rename to
18118 <absneg_str><mode>2.
18120 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
18122 * common/config/arc/arc-common.c (arc_option_optimization_table):
18123 Remove malign-call.
18124 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
18125 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
18126 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
18127 * config/arc/arc.md (abssi2_mixed): Remove pattern.
18128 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
18129 (malign-call): Likewise.
18130 (mmixed-code): Likewise.
18131 * doc/invoke.texi (ARC): Update doc.
18133 2021-06-03 Martin Liska <mliska@suse.cz>
18135 * common.opt: Use proper Enum values.
18136 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
18137 (parse_sanitizer_options): Handle only sanitizer_opts.
18138 (common_handle_option): Just assign value.
18140 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18143 * tree-inline.c (inline_forbidden_p): Remove test on return type.
18145 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18147 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
18148 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
18149 (type_byte_size): Inline into...
18150 (add_byte_size_attribute): ...this and call add_scalar_info.
18152 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18154 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
18155 (typed_binop_from_tree): New function.
18156 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
18157 turn a divide by a power of 2 into a shift.
18158 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
18159 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
18160 typed divide by calling typed_binop_from_tree.
18162 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18164 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
18165 (is_handled_procedure_type): Likewise.
18166 (struct loc_descr_context): Add strict_signedness field.
18167 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
18168 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
18169 (resolve_args_picking): Minor tweak.
18170 (function_to_dwarf_procedure): Initialize strict_signedness field.
18171 (type_byte_size): Likewise.
18172 (field_byte_offset): Likewise.
18173 (gen_descr_array_type_die): Likewise.
18174 (gen_variant_part): Likewise.
18175 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
18176 to true when a context is present before evaluating the arguments.
18177 <COND_EXPR>: Do not generate a useless comparison with zero.
18178 When dereferencing an address, if strict_signedness is true and the
18179 type is small and signed, use DW_OP_deref_type to do the dereference
18180 and then DW_OP_convert to convert back to the generic type.
18182 2021-06-03 Jakub Jelinek <jakub@redhat.com>
18185 * tree-inline.c (copy_tree_body_r): Handle iterators on
18186 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
18188 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
18190 * config/arc/arc.md (*bbit_di): Remove.
18192 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
18194 PR rtl-optimization/100264
18195 * ree.c (get_sub_rtx): Ignore SET expressions without register
18196 destinations and remove assertion, as it is not valid anymore
18197 with this new behaviour.
18198 (merge_def_and_ext): Eliminate destination check for register
18199 as such SET expressions can't occur anymore.
18200 (combine_reaching_defs): Likewise.
18202 2021-06-02 Jakub Jelinek <jakub@redhat.com>
18205 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
18206 -Wtype-limits warnings.
18207 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
18208 in operands to avoid -Wsign-compare warnings.
18210 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
18212 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
18215 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
18217 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
18219 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
18221 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
18223 * config/s390/subst.md(cconly_subst): Use a single constraint
18224 in (match_scratch).
18226 2021-06-02 Martin Liska <mliska@suse.cz>
18228 * ipa-icf.h: Use auto_vec for memory_access_types.
18230 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
18232 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
18233 argument from prototype.
18234 (output_logical_op): Add rtx_code argument.
18235 (compute_logical_op_length): Likewise.
18236 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
18237 to compute_a_shift_length.
18238 (output_logical_op); New argument with the rtx code rather than
18239 extracting it from an operand. Handle QImode too.
18240 (compute_logical_op_length): Similary.
18241 (compute_a_shift_length): Drop unused argument.
18242 * config/h8300/h8300.md (logicals): New code iterator.
18243 * config/h8300/logical.md (<code><mode>3 expander): Combine
18244 the "and" expander with the "ior"/"xor" expander.
18245 (bclr<mode>msx): Combine the QI/HI mode patterns.
18246 (<logical><mode>3 insns): Use code iterator rather than match_operator.
18247 Handle QImode as well. Update call to output_logical_op and
18248 compute_logical_op_length to pass in rtx_code
18249 Fix split condition on all define_insn_and_split patterns.
18250 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
18251 the flags and setting ZN via existing define_subst.
18252 * config/h8300/shiftrotate.md: Drop unused argument from
18253 calls to compute_a_shift_length.
18254 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
18256 2021-06-01 Andrew Pinski <apinski@marvell.com>
18258 PR tree-optimization/25290
18259 * tree-ssa-phiopt.c (match_simplify_replacement):
18261 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
18262 (two_value_replacement): Change the comment about
18263 conditional_replacement.
18264 (conditional_replacement): Delete.
18266 2021-06-01 Andrew Pinski <apinski@marvell.com>
18268 PR tree-optimization/95481
18269 * tree-tailcall.c (find_tail_calls): Handle empty typed
18272 2021-06-01 Andrew Pinski <apinski@marvell.com>
18274 * gimplify.c (zero_sized_field_decl): Delete
18275 (zero_sized_type): Delete
18276 (gimplify_init_ctor_eval): Use is_empty_type instead
18277 of zero_sized_field_decl.
18278 (gimplify_modify_expr): Use is_empty_type instead of
18281 2021-06-01 Jason Merrill <jason@redhat.com>
18284 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
18286 2021-06-01 Jason Merrill <jason@redhat.com>
18289 * diagnostic.h (warning_enabled_at): Declare.
18290 * diagnostic.c (diagnostic_enabled): Factor out from...
18291 (diagnostic_report_diagnostic): ...here.
18292 (warning_enabled_at): New.
18294 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
18296 * gimple-ssa-evrp.c: Enable exporting of global ranges.
18298 2021-06-01 Martin Liska <mliska@suse.cz>
18301 * doc/invoke.texi: Mention that -fgcse-after-reload
18302 is enabled with -O3.
18304 2021-06-01 liuhongt <hongtao.liu@intel.com>
18306 PR tree-optimization/98365
18307 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
18308 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
18309 (convert_scalar_cond_reduction): Ditto.
18310 (predicate_scalar_phi): Ditto.
18312 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18314 PR tree-optimization/100781
18315 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
18316 value calculation by default.
18317 (ranger_cache::enable_new_values): New.
18318 (ranger_cache::disable_new_values): New.
18319 (ranger_cache::push_poor_value): Check if new values are allowed.
18320 * gimple-range-cache.h (class ranger_cache): New member/methods.
18321 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
18322 statement, and disable/renable new value calculation.
18324 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18326 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
18327 (ranger_cache::range_of_def): New.
18328 (ranger_cache::entry_range): New.
18329 (ranger_cache::exit_range): New.
18330 (ranger_cache::range_of_expr): Adjust.
18331 (ranger_cache::range_on_edge): Adjust.
18332 (ranger_cache::propagate_cache): Call exit_range directly.
18333 * gimple-range-cache.h (class ranger_cache): Adjust.
18335 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18337 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
18338 gori_compute being a member rather than base class.
18339 dervied call to member call.
18340 (ranger_cache::dump): No longer dump gori_map.
18341 (ranger_cache::dump_bb): New.
18342 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
18343 being a member rather than base class.
18344 (ranger_cache::set_global_range): Ditto.
18345 (ranger_cache::ssa_range_in_bb): Ditto.
18346 (ranger_cache::range_of_expr): New.
18347 (ranger_cache::range_on_edge): New.
18348 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
18349 (ranger_cache::propagate_cache): Adjust debugging output.
18350 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
18352 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
18353 member, and inherit from range_query instead.
18354 (ranger_cache::dump_bb): New. split from dump.
18355 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
18356 (gori_compute::expr_range_at_stmt): Delete.
18357 (gori_compute::compute_name_range_op): Delete.
18358 (gori_compute::compute_operand_range_switch): Add fur_source.
18359 (gori_compute::compute_operand_range): Add fur_source param, inline
18360 old compute_name_range_op and optimize_logical_operands.
18361 (struct tf_range): Delete.
18362 (gori_compute::logical_combine): Adjust
18363 (gori_compute::optimize_logical_operands): Delete.
18364 (gori_compute::compute_logical_operands_in_chain): Delete.
18365 (gori_compute::compute_logical_operands): Adjust.
18366 (gori_compute::compute_operand1_range): Adjust to fur_source.
18367 (gori_compute::compute_operand2_range): Ditto.
18368 (gori_compute::compute_operand1_and_operand2_range): Ditto.
18369 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
18370 and adjust to fur_source.
18371 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
18372 range_query and fur_source.
18373 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
18374 from the ranger_cache..
18375 (gimple_ranger::fold_range_internal): Adjust to base class change of
18377 (gimple_ranger::dump_bb): Adjust dump.
18378 * gimple-range.h (gimple_ranger):export gori computes object.
18380 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18382 PR tree-optimization/100774
18383 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
18384 Constant values are also not stale.
18385 (ranger_cache::set_global_range): Range invariant values should also
18386 have the correct timestamp.
18388 2021-05-31 Martin Liska <mliska@suse.cz>
18390 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
18391 Unpack FUNCTION_DECL_DECL_TYPE.
18392 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
18393 Stream FUNCTION_DECL_DECL_TYPE instead of
18394 DECL_IS_OPERATOR_NEW_P.
18395 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
18397 (DECL_IS_OPERATOR_NEW_P): Likewise.
18398 (DECL_IS_OPERATOR_DELETE_P): Likewise.
18399 (DECL_LAMBDA_FUNCTION_P): Likewise.
18401 2021-05-31 Richard Biener <rguenther@suse.de>
18404 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
18405 * internal-fn.def (SHUFFLEVECTOR): New.
18406 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
18407 * doc/extend.texi: Document __builtin_shufflevector.
18409 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
18412 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
18413 indexed form addresses.
18415 2021-05-29 Jeff Law <jlaw@tachyum.com>
18417 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
18418 parameter. Call callers fixed.
18420 (output_plussi): Add FALLTHRU markers.
18421 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
18423 2021-05-29 Jakub Jelinek <jakub@redhat.com>
18425 PR middle-end/99928
18426 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
18427 combined with parallel, make sure to add shared clause to
18428 parallel for explicit linear clause.
18430 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
18432 PR tree-optimization/100787
18433 * gimple-ssa-evrp.c: Disable exporting of global ranges.
18435 2021-05-28 Jason Merrill <jason@redhat.com>
18437 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
18438 operator--, operator*, operator==, and operator!=.
18439 (class tsi_range): New.
18441 2021-05-28 Richard Biener <rguenther@suse.de>
18443 PR tree-optimization/100778
18444 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
18445 trapping ops in different BBs.
18447 2021-05-28 Richard Biener <rguenther@suse.de>
18450 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
18451 copy fntype from original call.
18453 2021-05-28 Martin Liska <mliska@suse.cz>
18455 PR gcov-profile/100751
18456 * doc/gcov.texi: Revert partially a hunk that was wrong.
18458 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
18460 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
18462 (HAVE_sync_compare_and_swaphi): Likewise.
18463 (HAVE_sync_compare_and_swapsi): Likewise.
18465 2021-05-28 Jakub Jelinek <jakub@redhat.com>
18467 PR middle-end/99928
18468 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
18470 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
18472 * gimplify.c (gimplify_omp_affinity): New.
18473 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
18474 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
18475 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
18476 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
18477 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
18479 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
18480 Richard Biener <rguenther@suse.de>
18482 * match.pd <popcount & / + pattern matching>:
18483 When generating popcount directly fails, try doing it in two halves.
18485 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
18487 * Makefile.in (generated_files): Add gimple-match.c and
18490 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
18492 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
18494 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
18496 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
18498 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
18500 PR tree-optimization/99398
18501 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
18502 where the fed operands are CTOR/CST and propagated through
18503 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
18504 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
18506 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
18509 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
18511 * config/i386/mmx.md (addv2sf3): Do not call
18512 ix86_fixup_binary_operands_no_copy.
18515 (<smaxmin:code>v2sf3): Ditto.
18516 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
18517 (<plusminus:insn><VI_32:mode>3): Remove expander.
18518 (<plusminus:insn><VI_32:mode>3): Rename from
18519 "*<plusminus:insn><VI_32:mode>3".
18520 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
18521 (mulv2hi3): Remove expander.
18522 (mulv2hi3): Rename from *mulv2hi3.
18523 (<s>mulv2hi3_highpart): Remove expander.
18524 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
18525 (<smaxmin:code><MMXMODE14:mode>3): Rename from
18526 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
18527 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
18528 (SMAXMIN_MMXMODEI): Remove mode iterator.
18529 (<smaxmin:code>v4hi3): New expander.
18530 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
18531 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
18532 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
18533 (SMAXMIN_VI_32): Remove mode iterator.
18534 (<umaxmin:code><MMXMODE24:mode>3): Rename from
18535 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
18536 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
18537 (UMAXMIN_MMXMODEI): Remove mode iterator.
18538 (<umaxmin:code>v8qi3): New expander.
18539 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
18540 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
18541 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
18542 (UMAXMIN_VI_32): Remove mode iterator.
18543 (<any_shift:insn>v2hi3): Remove expander.
18544 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
18545 (<any_logic:code><MMXMODEI:mode>3): Do not call
18546 ix86_fixup_binary_operands_no_copy.
18547 (<any_logic:code><VI_32:mode>3): Remove expander.
18548 (<any_logic:code><VI_32:mode>3): Rename from
18549 "*<any_logic:code><VI_32:mode>3".
18550 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
18551 * config/i386/sse.md (div<VF2:mode>3): Do not call
18552 ix86_fixup_binary_operands_no_copy.
18553 (div<VF1:mode>3): Ditto.
18554 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
18555 (smulhrsv4hi3): Ditto.
18556 (smulhrsv2hi3): Ditto.
18558 2021-05-27 Martin Sebor <msebor@redhat.com>
18560 * ggc.h (gt_ggc_mx): Add overloads for all integers.
18562 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
18564 (hash_map::operator==): New function.
18566 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
18569 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
18570 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
18571 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
18572 (*xop_maskcmp<VI_32:mode>3): Ditto.
18573 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
18574 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
18576 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
18579 * config/arm/arm.c (arm_configure_build_target): Remove parameter
18580 opts_set, directly check opts parameters for being non-null.
18581 (arm_option_restore): Update call to arm_configure_build_target.
18582 (arm_option_override): Likewise.
18583 (arm_can_inline_p): Likewise.
18584 (arm_valid_target_attribute_tree): Likewise.
18585 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
18586 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
18589 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
18591 * vr-values.c (simplify_conversion_using_ranges): Use
18592 get_range_query instead of get_global_range_query.
18594 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
18596 * gimple-range.cc (get_range_global): Move to value-query.cc.
18597 (gimple_range_global): Same.
18598 (get_global_range_query): Same.
18599 (global_range_query::range_of_expr): Same.
18600 * gimple-range.h (class global_range_query): Move to
18602 (gimple_range_global): Same.
18603 * tree-ssanames.c (get_range_info): Move to value-query.cc.
18604 (get_ptr_nonnull): Same.
18605 * tree-ssanames.h (get_range_info): Remove.
18606 (get_ptr_nonnull): Remove.
18607 * value-query.cc (get_ssa_name_range_info): Move from
18609 (get_ssa_name_ptr_info_nonnull): Same.
18610 (get_range_global): Move from gimple-range.cc.
18611 (gimple_range_global): Same.
18612 (get_global_range_query): Same.
18613 (global_range_query::range_of_expr): Same.
18614 * value-query.h (class global_range_query): Move from
18616 (gimple_range_global): Same.
18618 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
18621 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
18622 (uavgv2hi3_ceil): Ditto.
18624 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
18627 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
18629 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18631 * tree-ssanames.c (get_range_info): Merge both copies of
18632 get_range_info into one that works with irange.
18633 * tree-ssanames.h (get_range_info): Remove version that works on
18636 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18638 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
18639 (expand_builtin_strnlen): Same.
18640 (determine_block_size): Same.
18641 * fold-const.c (expr_not_equal_to): Same.
18642 * gimple-fold.c (size_must_be_zero_p): Same.
18643 * gimple-match-head.c: Include gimple-range.h.
18644 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
18645 * gimple-ssa-warn-restrict.c
18646 (builtin_memref::extend_offset_range): Same.
18647 * graphite-sese-to-poly.c (add_param_constraints): Same.
18648 * internal-fn.c (get_min_precision): Same.
18649 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
18650 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
18652 * tree-data-ref.c (split_constant_offset): Same.
18653 (dr_step_indicator): Same.
18654 * tree-dfa.c (get_ref_base_and_extent): Same.
18655 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
18656 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
18657 (determine_value_range): Same.
18658 (record_nonwrapping_iv): Same.
18659 (infer_loop_bounds_from_signedness): Same.
18660 (scev_var_range_cant_overflow): Same.
18661 * tree-ssa-phiopt.c (two_value_replacement): Same.
18662 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
18663 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
18664 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
18666 (dump_strlen_info): Same.
18667 (set_strlen_range): Same.
18668 (maybe_diag_stxncpy_trunc): Same.
18669 (get_len_or_size): Same.
18670 (handle_integral_assign): Same.
18671 * tree-ssa-structalias.c (find_what_p_points_to): Same.
18672 * tree-ssa-uninit.c (find_var_cmp_const): Same.
18673 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
18674 * tree-vect-patterns.c (vect_get_range_info): Same.
18675 (vect_recog_divmod_pattern): Same.
18676 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
18677 (register_edge_assert_for_2): Same.
18678 (determine_value_range_1): Same.
18679 * tree.c (get_range_pos_neg): Same.
18680 * vr-values.c (vr_values::get_lattice_entry): Same.
18681 (vr_values::update_value_range): Same.
18682 (simplify_conversion_using_ranges): Same.
18684 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18686 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
18687 get_range_query instead of query argument.
18688 (pass_walloca::execute): Enable and disable global ranger.
18690 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18692 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
18694 (rvrp_folder::~rvrp_folder): Call disable_ranger.
18695 (hybrid_folder::hybrid_folder): Call enable_ranger.
18696 (hybrid_folder::~hybrid_folder): Call disable_ranger.
18698 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
18700 * function.c (allocate_struct_function): Set cfun->x_range_query.
18701 * function.h (struct function): Declare x_range_query.
18702 (get_range_query): New.
18703 (get_global_range_query): New.
18704 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
18705 Remove call to safe_grow_cleared.
18706 * gimple-range.cc (get_range_global): New.
18707 (gimple_range_global): Move from gimple-range.h.
18708 (get_global_range_query): New.
18709 (global_range_query::range_of_expr): New.
18710 (enable_ranger): New.
18711 (disable_ranger): New.
18712 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
18713 (class global_range_query): New.
18714 (enable_ranger): New.
18715 (disable_ranger): New.
18716 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
18717 dump_all_value_ranges to dump.
18718 * tree-vrp.c (vrp_prop::finalize): Same.
18719 * value-query.cc (range_query::dump): New.
18720 * value-query.h (range_query::dump): New.
18721 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
18722 (vr_values::dump): ...this.
18723 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
18724 dump and make virtual.
18726 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
18728 * config/i386/i386.c (ix86_autovectorize_vector_modes):
18729 Add V4QImode and V16QImode for TARGET_SSE2.
18730 * doc/sourcebuild.texi (Vector-specific attributes):
18731 Add vect64 and vect32 description.
18733 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
18735 * gimple-range-gori.cc (range_def_chain::register_dependency):
18736 Resize m_def_chain when needed.
18738 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
18740 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
18741 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
18742 * config/arm/vec-common.md: .. here. Add support for MVE.
18744 2021-05-26 Jakub Jelinek <jakub@redhat.com>
18746 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
18748 * config/microblaze/microblaze.c (microblaze_legitimize_address,
18750 microblaze_option_override, print_operand): Likewise.
18751 * config/microblaze/microblaze.md (call_internal_plt,
18752 call_value_intern_plt, call_value_intern): Likewise.
18753 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
18754 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
18755 call_value_multiple_internal1): Likewise.
18756 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
18758 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
18760 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
18761 arc_ccfsm_advance, symbolic_reference_mentioned_p,
18762 arc_raw_symbolic_reference_mentioned_p): Remove register
18765 2021-05-26 Jakub Jelinek <jakub@redhat.com>
18768 * omp-low.c: Include omp-offload.h.
18769 (create_omp_child_function): If current_function_decl has
18770 "omp declare target" attribute and is_gimple_omp_offloaded,
18771 remove that attribute from the copy of attribute list and
18772 add "omp target entrypoint" attribute instead.
18773 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
18774 variables for offloading if in omp_maybe_offloaded_ctx.
18775 * omp-offload.c (pass_omp_target_link::execute): Nullify second
18776 argument to GOMP_target_data_ext in offloaded code.
18778 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18780 * config/csky/csky.c (csky_can_change_mode_class): Delete.
18781 For csky, HF/SF mode use the low bits of VREGS.
18783 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
18785 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
18786 DECL which is a reference for OMP.
18788 2021-05-26 Martin Liska <mliska@suse.cz>
18790 PR gcov-profile/100751
18791 * doc/gcov.texi: Document that __gcov_dump can be called just
18792 once and that __gcov_reset resets run-time counters.
18794 2021-05-26 Martin Liska <mliska@suse.cz>
18796 * doc/install.texi: Port relevant part from install-old.texi
18797 and re-generate list of CPUs and systems.
18799 2021-05-26 Martin Liska <mliska@suse.cz>
18801 * Makefile.in: Remove it.
18802 * doc/include/fdl.texi: Update next/previous chapters.
18803 * doc/install.texi: Likewise.
18804 * doc/install-old.texi: Removed.
18806 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18808 * config/csky/csky.c (ck810_legitimate_index_p): Support
18809 "base + index" with DF mode.
18810 * config/csky/constraints.md ("Y"): New constraint for memory operands
18811 without index register.
18812 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
18813 when mov between memory and general registers, and lower their priority.
18814 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
18816 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18818 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
18820 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18822 * config/csky/csky.md (untyped_call): Emit clobber for return
18823 registers to mark them used.
18825 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
18827 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
18829 2021-05-26 Andrew Pinski <apinski@marvell.com>
18831 * match.pd (x < 0 ? ~y : y): New patterns.
18833 2021-05-26 Andrew Pinski <apinski@marvell.com>
18835 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
18836 A?POW2:0 and A?0:POW2.
18838 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
18840 * gimple-range-gori.cc (class logical_stmt_cache): Delete
18841 (logical_stmt_cache::logical_stmt_cache ): Delete.
18842 (logical_stmt_cache::~logical_stmt_cache): Delete.
18843 (logical_stmt_cache::cache_entry::dump): Delete.
18844 (logical_stmt_cache::get_range): Delete.
18845 (logical_stmt_cache::cached_name ): Delete.
18846 (logical_stmt_cache::same_cached_name): Delete.
18847 (logical_stmt_cache::cacheable_p): Delete.
18848 (logical_stmt_cache::slot_diagnostics ): Delete.
18849 (logical_stmt_cache::dump): Delete.
18850 (gori_compute_cache::gori_compute_cache): Delete.
18851 (gori_compute_cache::~gori_compute_cache): Delete.
18852 (gori_compute_cache::compute_operand_range): Delete.
18853 (gori_compute_cache::cache_stmt): Delete.
18854 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
18856 (class gori_compute_cache): Delete.
18858 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
18860 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
18862 (fold_using_range::range_of_address): Adjust.
18863 (fold_using_range::range_of_phi): Adjust.
18864 * gimple-range.h (class fur_source): Adjust.
18865 (fur_source::fur_source): Adjust.
18867 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
18869 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
18870 from expr_range_in_bb and adjust.
18871 (gori_compute::compute_name_range_op): Adjust.
18872 (gori_compute::optimize_logical_operands): Adjust.
18873 (gori_compute::compute_logical_operands_in_chain): Adjust.
18874 (gori_compute::compute_operand1_range): Adjust.
18875 (gori_compute::compute_operand2_range): Adjust.
18876 (ori_compute_cache::cache_stmt): Adjust.
18877 * gimple-range-gori.h (gori_compute): Rename prototype.
18879 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
18881 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
18882 checked only after range_of_stmt, not range_on_entry.
18883 (gimple_ranger::range_on_entry): Check for non-null in any
18884 predecessor block, if it is not already non-null.
18885 (gimple_ranger::range_on_exit): DOnt check for non-null after
18886 range on entry call.
18887 (gimple_ranger::dump_bb): New. Split from dump.
18888 (gimple_ranger::dump): Adjust.
18889 * gimple-range.h (class gimple_ranger): Adjust.
18891 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
18893 * gimple-range-cache.cc (struct range_timestamp): Delete.
18894 (class temporal_cache): Adjust.
18895 (temporal_cache::get_timestamp): Delete.
18896 (temporal_cache::set_dependency): Delete.
18897 (temporal_cache::temporal_value): Adjust.
18898 (temporal_cache::current_p): Take dependencies as params.
18899 (temporal_cache::set_timestamp): Adjust.
18900 (temporal_cache::set_always_current): Adjust.
18901 (ranger_cache::get_non_stale_global_range): Adjust.
18902 (ranger_cache::register_dependency): Delete.
18903 * gimple-range-cache.h (class range_cache): Adjust.
18905 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
18907 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
18909 (range_def_chain::~range_def_chain): Dispose of obstack rather than
18910 each individual bitmap.
18911 (range_def_chain::set_import): New.
18912 (range_def_chain::get_imports): New.
18913 (range_def_chain::chain_import_p): New.
18914 (range_def_chain::register_dependency): Rename from build_def_chain
18916 (range_def_chain::def_chain_in_bitmap_p): New.
18917 (range_def_chain::add_def_chain_to_bitmap): New.
18918 (range_def_chain::has_def_chain): Just check first depenedence.
18919 (range_def_chain::get_def_chain): Process imports, use generic
18920 register_dependency routine.
18921 (range_def_chain::dump): New.
18922 (gori_map::gori_map): Allocate import list.
18923 (gori_map::~gori_map): Release imports.
18924 (gori_map::exports): Check for past allocated block size.
18925 (gori_map::imports): New.
18926 (gori_map::def_chain_in_export_p): Delete.
18927 (gori_map::is_import_p): New.
18928 (gori_map::maybe_add_gori): Handle imports.
18929 (gori_map::dump): Adjust output, add imports.
18930 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
18931 (gori_export_iterator::gori_export_iterator): New.
18932 (gori_export_iterator::next): New.
18933 (gori_export_iterator::get_name): New.
18934 * gimple-range-gori.h (range_def_chain): Add imports and direct
18935 dependecies via struct rdc.
18936 (range_def_chain::depend1): New.
18937 (range_def_chain::depend2): New.
18938 (class gori_map): Adjust.
18939 (FOR_EACH_GORI_IMPORT_NAME): New.
18940 (FOR_EACH_GORI_EXPORT_NAME): New.
18941 (class gori_export_iterator): New.
18943 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
18945 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
18946 export cache filling to here.
18947 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
18949 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
18951 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
18952 (gori_map): Move to gimple-range-gori.h.
18953 (gori_compute::gori_compute): Adjust.
18954 (gori_compute::~gori_compute): Delete.
18955 (gori_compute::compute_operand_range_switch): Adjust.
18956 (gori_compute::compute_operand_range): Adjust.
18957 (gori_compute::compute_logical_operands): Adjust.
18958 (gori_compute::has_edge_range_p ): Adjust.
18959 (gori_compute::set_range_invariant): Delete.
18960 (gori_compute::dump): Adjust.
18961 (gori_compute::outgoing_edge_range_p): Adjust.
18962 * gimple-range-gori.h (class range_def_chain): Relocate here.
18963 (class gori_map): Relocate here.
18964 (class gori_compute): Inherit from gori_map, and adjust.
18966 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
18968 * value-range.cc (range_tests_legacy): Use
18969 build_nonstandard_integer_type instead of int and short.
18971 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
18973 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
18974 when really creating an initialization statement for it.
18976 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
18978 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
18980 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
18982 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
18984 2021-05-25 Martin Liska <mliska@suse.cz>
18986 PR tree-optimization/92860
18988 * optc-save-gen.awk: Remove exceptions.
18990 2021-05-25 Martin Liska <mliska@suse.cz>
18992 * asan.h (sanitize_coverage_p): New function.
18993 * doc/extend.texi: Document it.
18994 * fold-const.c (fold_range_test): Use sanitize_flags_p
18995 instead of flag_sanitize_coverage.
18996 (fold_truth_andor): Likewise.
18997 * sancov.c: Likewise.
18998 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
18999 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
19000 -fsanitize-coverage when inlining.
19002 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
19004 * config/csky/csky-modes.def : Fix copyright.
19006 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
19008 * config/csky/csky-modes.def : Amend copyright.
19009 * config/csky/csky_insn_fpuv2.md : Likewise.
19010 * config/csky/csky_insn_fpuv3.md : Likewise.
19012 2021-05-25 Richard Biener <rguenther@suse.de>
19014 PR middle-end/100727
19015 * calls.c (initialize_argument_information): Explicitely test
19016 for WITH_SIZE_EXPR.
19017 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
19019 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19021 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
19022 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
19023 the signle definition. The signle definition may not work well
19024 at simplify_subreg_regno().
19025 (HARD_FRAME_POINTER_REGNUM): New.
19026 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
19027 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
19028 csky_initial_elimination_offset, csky_expand_prologue,
19029 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
19031 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19033 * config/csky/csky.c (csky_option_override):
19034 Init csky_arch_isa_features[] in advance, so TARGET_DSP
19035 and TARGET_DIV can be set well.
19037 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19039 * config/csky/constraints.md ("l", "h"): Delete.
19040 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
19041 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
19042 * config/csky/csky.c (regno_reg_classm,
19043 csky_secondary_reload, csky_register_move_cost):
19044 Use HILO_REGS instead of LO_REGS and HI_REGS.
19046 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19048 * config/csky/constraints.md ("W"): New constriant for mem operand
19049 with base reg, index register.
19050 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
19051 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
19053 ("Dv"): New constraint for const double value that can be used at
19055 * config/csky/csky-modes.def (HFmode): New mode.
19056 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
19057 to "csky_valid_mem_constraint_operand" and support new constraint
19059 (csky_get_movedouble_length): New.
19060 (fpuv3_output_move): New.
19061 (fpuv3_const_double): New.
19062 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
19063 (decompose_csky_address): Refine.
19064 (csky_print_operand): New "CONST_DOUBLE" operand.
19065 (csky_output_move): Support fpv3 instructions.
19066 (csky_get_movedouble_length): New.
19067 (fpuv3_output_move): New.
19068 (fpuv3_const_double): New.
19069 (csky_emit_compare): Cover float comparsion.
19070 (csky_emit_compare_float): Refine.
19071 (csky_vaild_fpuv2_mem_operand): Rename to
19072 "csky_valid_mem_constraint_operand" and support new constraint "W".
19073 (ck860_rtx_costs): New.
19074 (csky_rtx_costs): Add the cost calculation of CK860.
19075 (regno_reg_class): New vregs for fpuv3.
19076 (csky_dbx_regno): Likewise.
19077 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
19078 (csky_conditional_register_usage): Suporrot fpuv3.
19079 (csky_dwarf_register_span): Suporrot fpuv3.
19080 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
19081 (ck810_legitimate_index_p): Support fp16.
19082 * config/csky/csky.h (TARGET_TLS): ADD CK860.
19083 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
19084 (TARGET_SINGLE_FPU): Support fpuv3.
19085 (TARGET_SUPPORT_FPV3): New.
19086 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
19087 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
19088 REG_CLASS_CONTENTS): Support fpuv3.
19089 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
19090 (csky_movsf_fpv2): Likewise.
19091 (ck801_movsf): Likewise.
19092 (csky_movsf): Likewise.
19094 (csky_movdf_fpv2): Likewise.
19095 (ck801_movdf): Likewise.
19096 (csky_movdf): Likewise.
19097 (movsicc): Refine. Use "comparison_operatior" instead of
19098 "ordered_comparison_operatior".
19099 (addsicc): Likewise.
19100 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
19101 (call_value_internal_vh): New.
19102 * config/csky/csky_cores.def (CK860): New arch and cpu.
19107 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
19108 into emit-patterns and match-patterns, remain the emit-patterns here,
19109 and move the match-patterns to csky_insn_fpuv2.md or
19110 csky_insn_fpuv3.md.
19111 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
19112 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
19114 * config/csky/csky_isa.def (fcr): New.
19119 (CK860): New definition for ck860.
19120 * config/csky/csky_tables.opt (ck860): New processors ck860,
19121 ck860f. And new arch ck860.
19126 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
19127 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
19128 * config/csky/t-csky-elf: Support 860.
19129 * config/csky/t-csky-linux: Likewise.
19130 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
19132 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
19134 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
19135 add generation of logical-add and add-logical fusion pairs.
19136 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
19138 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
19139 logical-add and add-logical fusion by default.
19140 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
19141 -mpower10-fusion-add-logical options.
19142 * config/rs6000/fusion.md: Regenerate file.
19144 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
19146 * value-range.cc (irange::legacy_equal_p): Check type when
19147 comparing VR_VARYING types.
19148 (range_tests_legacy): Test comparing VARYING ranges of different
19151 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
19153 * config/aarch64/aarch64.c (neoversen1_tunings):
19154 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
19156 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
19158 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
19159 extern weak symbols. Limit symbol offsets for non-GOT symbols with
19162 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
19164 * config/arm/neon.md (vec_load_lanesxi<mode>)
19165 (vec_store_lanexoi<mode>): Move ...
19166 * config/arm/vec-common.md: here.
19168 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
19170 * config/arm/neon.md (vec_load_lanesoi<mode>)
19171 (vec_store_lanesoi<mode>): Move ...
19172 * config/arm/vec-common.md: here.
19174 2021-05-24 liuhongt <hongtao.liu@intel.com>
19177 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
19178 stmt with GIMPLE_NOP when lhs doesn't exist.
19180 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
19183 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
19184 New instruction pattern.
19185 (*push<VI_32:mode>2): Ditto.
19186 (push splitter for SSE registers): New splitter.
19188 2021-05-23 Andrew Pinski <apinski@marvell.com>
19190 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
19192 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
19194 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
19195 * config/rs6000/fusion.md: Regenerate file.
19197 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
19199 * config/rs6000/genfusion.pl (gen_addadd): New function.
19200 * config/rs6000/fusion.md: Regenerate file.
19201 * config/rs6000/rs6000-cpus.def: Add
19202 OPTION_MASK_P10_FUSION_2ADD to masks.
19203 * config/rs6000/rs6000.c (rs6000_option_override_internal):
19204 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
19205 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
19207 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19209 PR middle-end/99928
19210 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
19211 * gimplify.c (enum gimplify_omp_var_data): Fix up
19212 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
19213 (omp_lastprivate_for_combined_outer_constructs): If combined target
19214 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
19215 GOVD_MAP | GOVD_SEEN.
19216 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
19217 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
19218 (gimplify_adjust_omp_clauses): For firstprivate clauses with
19219 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
19220 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
19221 let it be replaced by implicit map clause.
19223 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19225 PR middle-end/99928
19226 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
19228 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
19229 (gimplify_omp_for): Likewise.
19231 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19233 PR middle-end/90115
19234 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
19235 'external' in blocks.
19237 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19239 PR middle-end/90115
19240 * flag-types.h (enum openacc_privatization): New.
19241 * params.opt (-param=openacc-privatization): New.
19242 * doc/invoke.texi (openacc-privatization): Document it.
19243 * omp-general.h (get_openacc_privatization_dump_flags): New
19245 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
19246 * omp-offload.c (execute_oacc_device_lower)
19247 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
19248 * target.def (goacc.adjust_private_decl): Add 'location_t'
19250 * doc/tm.texi: Regenerate.
19251 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
19252 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
19253 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
19254 Likewise. Preserve it for...
19255 (nvptx_goacc_expand_var_decl): ... use here.
19257 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19259 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
19262 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19264 PR middle-end/90115
19265 * omp-low.c (oacc_privatization_candidate_p): New function.
19266 (oacc_privatization_scan_clause_chain)
19267 (oacc_privatization_scan_decl_chain): Use it. Also
19268 'gcc_checking_assert' that we're not seeing duplicates.
19270 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19272 PR middle-end/90115
19273 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
19276 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19278 PR middle-end/90115
19279 * omp-offload.c (execute_oacc_device_lower): Explain.
19281 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19283 PR middle-end/90115
19284 * omp-offload.c (execute_oacc_device_lower)
19285 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
19287 * internal-fn.c (expand_UNIQUE): Don't expect
19288 'IFN_UNIQUE_OACC_PRIVATE'.
19290 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19292 PR middle-end/90115
19293 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
19295 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19297 PR middle-end/90115
19298 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
19299 (nvptx_goacc_expand_var_decl): Tighten.
19301 2021-05-21 Julian Brown <julian@codesourcery.com>
19302 Chung-Lin Tang <cltang@codesourcery.com>
19303 Thomas Schwinge <thomas@codesourcery.com>
19305 PR middle-end/90115
19306 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
19307 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
19308 * doc/tm.texi: Regenerate.
19309 * expr.c (expand_expr_real_1): Expand decls using the
19310 expand_var_decl OpenACC hook if defined.
19311 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
19312 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
19313 * omp-low.c (omp_context): Add oacc_privatization_candidates
19315 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
19317 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
19318 private marker's gimple call arguments, and pass it to
19319 lower_oacc_reductions.
19320 (oacc_privatization_scan_clause_chain)
19321 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
19323 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
19324 * omp-offload.c (convert.h): Include.
19325 (oacc_loop_xform_head_tail): Treat private-variable markers like
19326 fork/join when transforming head/tail sequences.
19327 (struct var_decl_rewrite_info): Add struct.
19328 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
19329 (execute_oacc_device_lower): Support rewriting gang-private
19330 variables using target hook, and fix up addr_expr and var_decl
19332 * target.def (adjust_private_decl, expand_var_decl): New hooks.
19333 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
19335 (gcn_goacc_adjust_private_decl): ...this.
19336 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
19338 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
19339 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
19340 definition using gcn_goacc_adjust_gangprivate_decl...
19341 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
19342 gcn_goacc_adjust_private_decl.
19343 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
19344 (gang_private_shared_size): New global variable.
19345 (gang_private_shared_align): Likewise.
19346 (gang_private_shared_sym): Likewise.
19347 (gang_private_shared_hmap): Likewise.
19348 (nvptx_option_override): Initialize these.
19349 (nvptx_file_end): Output gang_private_shared_sym.
19350 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
19352 (nvptx_set_current_function): Clear gang_private_shared_hmap.
19353 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
19354 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
19356 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
19358 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
19360 2021-05-21 Richard Biener <rguenther@suse.de>
19361 H.J. Lu <hjl.tools@gmail.com>
19363 PR middle-end/90773
19364 * expr.c (expand_constructor): Elide expand_constructor if
19365 move by pieces is preferred.
19367 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19369 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
19370 Take a flag and mode value as arguments.
19371 (aarch64_modifies_global_state_p): Likewise.
19372 (aarch64_reads_global_state_p): Likewise.
19373 (aarch64_could_trap_p): Likewise.
19374 (aarch64_get_attributes): Likewise.
19375 (aarch64_init_simd_builtins): Adjust callsite of above.
19376 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
19377 function attributes to apply to builtins.
19378 (aarch64_init_crc32_builtins): Likewise.
19379 (aarch64_init_builtin_rsqrt): Likewise.
19381 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
19383 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
19384 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
19385 (gen_2logical): Use new fusion types.
19386 * config/rs6000/fusion.md: Regenerate.
19388 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
19391 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
19392 Handle V4QI and V2HI modes.
19393 (ix86_expand_sse_movcc): Ditto.
19394 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
19395 New instruction pattern.
19396 (*eq<VI_32:mode>3): Ditto.
19397 (*gt<VI_32:mode>3): Ditto.
19398 (*xop_pcmov_<VI_32:mode>): Ditto.
19399 (mmx_pblendvb32): Ditto.
19400 (mmx_pblendvb64): Rename from mmx_pblendvb.
19401 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
19402 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
19403 (vcond<VI_32:mode><VI_32:mode>): Ditto.
19404 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
19405 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
19407 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19409 PR tree-optimization/94589
19410 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
19411 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
19412 not the same as cmp3) treat <= the same as < and >= the same as >.
19413 Don't require that cond2_phi_edge is true edge, instead take
19414 false/true edges into account based on cmp1/cmp2 comparison kinds.
19416 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
19419 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
19420 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
19421 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
19422 using SMAXMIN_MMXMODEI mode iterator.
19423 (*<smaxmin:code>v4qi3): New insn pattern.
19424 (*<smaxmin:code>v2hi3): Ditto.
19425 (SMAXMIN_VI_32): New mode iterator.
19426 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
19427 (UMAXMIN_MMXMODEI): New mode iterator.
19428 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
19429 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
19430 using UMAXMIN_MMXMODEI mode iterator.
19431 (*<umaxmin:code>v4qi3): New insn pattern.
19432 (*<umaxmin:code>v2hi3): Ditto.
19433 (UMAXMIN_VI_32): New mode iterator.
19434 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
19435 (abs<VI_32:mode>2): New insn pattern.
19436 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
19437 * config/i386/sse.md: ... here.
19439 2021-05-20 Clement Chigot <clement.chigot@atos.net>
19440 David Edelsohn <dje.gcc@gmail.com>
19442 * collect2.c (scan_prog_file): Issue non-fatal warning for
19445 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
19447 * doc/invoke.texi (-Wno-c++11-extensions)
19448 (-Wno-c++14-extensions, -Wno-c++17-extensions)
19449 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
19452 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
19454 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
19455 * config/darwin.c (darwin_override_options): Likewise.
19456 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
19457 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
19458 (DWARF2_FRAME_REG_OUT): Likewise.
19459 * config/mips/mips.c (mips_output_filename): Likewise.
19460 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
19462 (rs6000_dbx_register_number): Likewise.
19463 * dbxout.c: Include flags.h.
19464 * dwarf2cfi.c (cfi_label_required_p): Likewise.
19465 (dwarf2out_do_frame): Likewise.
19466 * except.c: Include flags.h.
19467 * final.c (dwarf2_debug_info_emitted_p): Likewise.
19468 (final_scan_insn_1): Likewise.
19469 * flags.h (dwarf_debuginfo_p): New function declaration.
19470 * opts.c (dwarf_debuginfo_p): New function definition.
19471 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
19472 * toplev.c (process_options): Likewise.
19474 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
19476 * common.opt: Change type to support bitmasks.
19477 * flag-types.h (enum debug_info_type): Rename enumerator constants.
19478 (NO_DEBUG): New bitmask.
19479 (DBX_DEBUG): Likewise.
19480 (DWARF2_DEBUG): Likewise.
19481 (XCOFF_DEBUG): Likewise.
19482 (VMS_DEBUG): Likewise.
19483 (VMS_AND_DWARF2_DEBUG): Likewise.
19484 * flags.h (debug_set_to_format): New function declaration.
19485 (debug_set_count): Likewise.
19486 (debug_set_names): Likewise.
19487 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
19488 (debug_set_to_format): New function definition.
19489 (debug_set_count): Likewise.
19490 (debug_set_names): Likewise.
19491 (set_debug_level): Update access to debug_type_names.
19492 * toplev.c: Likewise.
19494 2021-05-20 Martin Sebor <msebor@redhat.com>
19496 PR middle-end/100684
19497 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
19499 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
19502 * config/i386/i386.md (isa): Remove x64_bmi.
19503 (enabled): Remove x64_bmi.
19504 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
19505 Remove general register alternative.
19506 (*andnot<VI_32:mode>3): Ditto.
19507 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
19508 (*<any_logic:code><VI_32:mode>3): Ditto.
19510 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
19512 * config/arm/arm.c: Include head files tree-vectorizer.h and
19515 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
19518 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
19519 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
19520 (<s>mulv4hi3_highpart): New expander.
19521 (*<s>mulv2hi3_highpart): New insn pattern.
19522 (<s>mulv2hi3_higpart): New expander.
19523 (*<any_shift:insn>v2hi3): New insn pattern.
19524 (<any_shift:insn>v2hi3): New expander.
19525 * config/i386/sse.md (smulhrsv2hi3): New expander.
19526 (*smulhrsv2hi3): New insn pattern.
19528 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
19530 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
19532 * params.opt (vect-inner-loop-cost-factor): New.
19533 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
19534 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
19535 tree-vectorizer.h and its required ones.
19536 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
19537 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
19538 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
19539 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
19540 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
19541 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
19543 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
19544 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
19545 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
19547 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
19548 Torbjörn Svensson <torbjorn.svensson@st.com>
19551 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
19553 2021-05-20 Jakub Jelinek <jakub@redhat.com>
19555 PR middle-end/99928
19556 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
19557 explicit linear clause when combined with target, make it map(tofrom:)
19558 instead of no clause or firstprivate.
19560 2021-05-20 Jakub Jelinek <jakub@redhat.com>
19562 PR tree-optimization/94589
19563 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
19564 of integral conversions.
19566 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
19568 * gimple-range.cc (fur_source::get_operand): New.
19569 (gimple_range_fold): Delete.
19570 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
19571 (fold_using_range::range_of_range_op): Move from gimple_ranger.
19572 (fold_using_range::range_of_address): Ditto.
19573 (fold_using_range::range_of_phi): Ditto.
19574 (fold_using_range::range_of_call): Ditto.
19575 (fold_using_range::range_of_builtin_ubsan_call): Move from
19576 range_of_builtin_ubsan_call.
19577 (fold_using_range::range_of_builtin_call): Move from
19578 range_of_builtin_call.
19579 (gimple_ranger::range_of_builtin_call): Delete.
19580 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
19581 (gimple_ranger::fold_range_internal): New.
19582 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
19583 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
19584 gimple_ranger. Improve ranges of SSA_NAMES when possible.
19585 * gimple-range.h (gimple_ranger): Remove various range_of routines.
19586 (class fur_source): New.
19587 (class fold_using_range): New.
19588 (fur_source::fur_source): New.
19590 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
19591 instead of range_of_builtin_call.
19593 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
19595 * doc/cpp.texi (Common Predefined Macros): Update documentation
19596 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
19598 2021-05-19 Alex Coplan <alex.coplan@arm.com>
19601 * config/arm/arm.md (nonsecure_call_internal): Always ensure
19602 callee's address is in a register.
19604 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
19606 * common/config/riscv/riscv-common.c
19607 (riscv_subset_list::parsing_subset_version): Properly parse the letter
19609 (riscv_subset_list::parse_std_ext,
19610 riscv_subset_list::parse_multiletter_ext): To handle errors generated
19611 in riscv_subset_list::parsing_subset_version.
19613 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19615 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
19616 type attribute in patterns generating XTN(2).
19618 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19620 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
19621 Remove as duplicate of...
19622 (aarch64_xtn<mode>): This.
19623 (aarch64_xtn2<mode>_le): Move position in file.
19624 (aarch64_xtn2<mode>_be): Move position in file.
19625 (aarch64_xtn2<mode>): Move position in file.
19626 (vec_pack_trunc_<mode>): Define as an expander.
19628 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19630 * config/aarch64/aarch64-simd-builtins.def: Split builtin
19631 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
19632 separate scalar and vector generators.
19633 * config/aarch64/aarch64-simd.md
19634 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
19636 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
19637 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
19638 * config/aarch64/iterators.md: Define SD_HSDI iterator.
19640 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
19642 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
19644 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
19646 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19648 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
19649 Implement as an expand emitting a big/little endian
19650 instruction pattern.
19651 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
19652 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
19654 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
19656 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
19657 Implement as an expand emitting a big/little endian
19658 instruction pattern.
19659 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
19660 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
19661 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
19662 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
19664 2021-05-19 Richard Biener <rguenther@suse.de>
19666 PR middle-end/100672
19667 * fold-const.c (fold_negate_expr_1): Use element_precision.
19668 (negate_expr_p): Likewise.
19670 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
19672 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
19673 (pred_load): New int attribute.
19674 * config/aarch64/aarch64-sve.md
19675 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
19676 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
19677 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
19678 code_for_aarch64_load.
19680 2021-05-19 Richard Biener <rguenther@suse.de>
19682 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
19683 sure TARGET_MEM_REF bases are expanded as memory.
19684 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
19685 Do not mark TARGET_MEM_REF bases addressable.
19686 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
19687 TARGET_MEM_REF bases as never rewritable.
19688 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
19689 walk TARGET_MEM_REF bases as address-takens.
19690 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
19692 2021-05-19 Richard Biener <rguenther@suse.de>
19694 * builtins.c (get_object_alignment_1): Strip outer
19696 * tree-dfa.c (get_ref_base_and_extent): Handle outer
19697 WITH_SIZE_EXPR for size processing and process the
19699 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
19700 outer WITH_SIZE_EXPR.
19701 (ao_ref_base_alias_ptr_type): Likewise.
19702 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
19703 and handle that accordingly, stripping it for the
19704 core alias workers.
19705 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
19706 looking through it instead of returning NULL.
19708 2021-05-19 Jakub Jelinek <jakub@redhat.com>
19710 PR middle-end/100576
19711 * builtins.c (check_read_access): Convert bound to size_type_node if
19714 2021-05-19 Richard Biener <rguenther@suse.de>
19716 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
19717 (verify_types_in_gimple_reference): ... here. Sanitize.
19718 (verify_gimple_call): Verify references in LHS and arguments.
19719 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
19721 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
19723 * config/i386/i386.h (VALID_INT_MODE_P):
19724 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
19725 * config/i386/i386.md (isa): Add x64_bmi.
19726 (enabled): Handle x64_bmi.
19727 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
19728 Add alternative using 64bit general registers.
19729 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
19731 2021-05-19 Jakub Jelinek <jakub@redhat.com>
19733 PR middle-end/99928
19734 * tree.h (OMP_MASTER_COMBINED): Define.
19735 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
19736 handling for outer combined/composite constructs to a loop.
19737 Handle lastprivate on combined target.
19738 (gimplify_expr): Formatting fix.
19740 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
19742 * passes.def: Add sink_code pass before store_merging.
19743 * tree-ssa-sink.c (pass_sink_code:clone): New.
19745 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
19747 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
19748 rs6000_special_adjust_field_align_p.
19749 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
19750 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
19751 Remove ABI warning.
19752 (rs6000_function_arg): Likewise.
19753 * config/rs6000/rs6000-protos.h
19754 (rs6000_special_adjust_field_align_p): Remove prototype.
19755 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
19757 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
19758 rs6000_special_adjust_field_align_p.
19760 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
19763 * config/i386/i386.h (VALID_SSE2_REG_MODE):
19764 Add V4QI and V2HI modes.
19765 (VALID_INT_MODE_P): Ditto.
19766 * config/i386/mmx.md (VI_32): New mode iterator.
19767 (mmxvecsize): Handle V4QI and V2HI.
19769 (mov<VI_32:mode>): New expander.
19770 (*mov<mode>_internal): New insn pattern.
19771 (movmisalign<VI_32:mode>): New expander.
19772 (neg<VI_32:mode>): New expander.
19773 (<plusminus:insn><VI_32:mode>3): New expander.
19774 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
19775 (mulv2hi3): New expander.
19776 (*mulv2hi3): New insn pattern.
19777 (one_cmpl<VI_32:mode>2): New expander.
19778 (*andnot<VI_32:mode>3): New insn pattern.
19779 (<any_logic:code><VI_32:mode>3): New expander.
19780 (*<any_logic:code><VI_32:mode>3): New insn pattern.
19782 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
19784 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
19785 Fix a mode mismatch with operand 1.
19787 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
19790 * config/i386/i386-expand.c (split_double_mode): Return
19791 temporary register when simplify_gen_subreg fails with
19792 the high half od the paradoxical subreg.
19794 2021-05-18 Richard Biener <rguenther@suse.de>
19796 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
19797 and honor it when expanding.
19798 (expand_used_vars_for_block): Pass through forced_stack_var.
19799 (expand_used_vars): Likewise.
19800 (discover_nonconstant_array_refs_r): Set bits in
19801 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
19802 (avoid_type_punning_on_regs): Likewise.
19803 (discover_nonconstant_array_refs): Likewise.
19804 (pass_expand::execute): Create and pass down forced_stack_var
19805 bitmap. For parameters and returns temporarily set
19806 TREE_ADDRESSABLE when expand_function_start.
19808 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
19810 * doc/sourcebuild.texi: Document 'dg-note'.
19812 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
19815 * configure: Regenerate.
19816 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
19818 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
19820 * gimple.h (is_gimple_omp_oacc): Tighten.
19821 * omp-low.c (check_omp_nesting_restrictions): Adjust.
19823 2021-05-18 Richard Biener <rguenther@suse.de>
19825 * tree-ssa-operands.c (mark_address_taken): Simplify.
19827 2021-05-18 Martin Liska <mliska@suse.cz>
19829 * config/gcn/mkoffload.c (STR): Redefine.
19830 * config/i386/intelmic-mkoffload.c (STR): Likewise.
19831 * config/nvptx/mkoffload.c (STR): Likewise.
19833 2021-05-18 Martin Liska <mliska@suse.cz>
19835 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
19836 Use startswith function instead of strncmp.
19837 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
19838 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
19839 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
19840 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
19841 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
19842 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
19843 * config/arm/arm.c (arm_file_start): Likewise.
19844 (arm_valid_target_attribute_rec): Likewise.
19845 (thumb1_md_asm_adjust): Likewise.
19846 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
19847 * config/avr/avr.c (STR_PREFIX_P): Likewise.
19848 (avr_set_current_function): Likewise.
19849 (avr_handle_addr_attribute): Likewise.
19850 (avr_asm_output_aligned_decl_common): Likewise.
19851 (avr_asm_named_section): Likewise.
19852 (avr_section_type_flags): Likewise.
19853 (avr_asm_select_section): Likewise.
19854 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
19855 (c6x_section_type_flags): Likewise.
19856 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
19857 (darwin_objc_declare_unresolved_class_reference): Likewise.
19858 (darwin_objc_declare_class_definition): Likewise.
19859 * config/darwin.c (indirect_data): Likewise.
19860 (darwin_encode_section_info): Likewise.
19861 (darwin_objc2_section): Likewise.
19862 (darwin_objc1_section): Likewise.
19863 (machopic_select_section): Likewise.
19864 (darwin_globalize_label): Likewise.
19865 (darwin_label_is_anonymous_local_objc_name): Likewise.
19866 (darwin_asm_named_section): Likewise.
19867 (darwin_asm_output_dwarf_offset): Likewise.
19868 * config/frv/frv.c (frv_string_begins_with): Likewise.
19869 (frv_in_small_data_p): Likewise.
19870 * config/gcn/mkoffload.c (STR): Likewise.
19872 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
19873 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
19874 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
19875 (ix86_md_asm_adjust): Likewise.
19876 * config/i386/intelmic-mkoffload.c (STR): Likewise.
19877 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
19878 (i386_pe_file_end): Likewise.
19879 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
19880 (ia64_section_type_flags): Likewise.
19881 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
19882 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
19883 (mips16_stub_function_p): Likewise.
19884 (mips_function_rodata_section): Likewise.
19885 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
19886 (msp430_function_section): Likewise.
19887 (msp430_section_type_flags): Likewise.
19888 (msp430_expand_helper): Likewise.
19889 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
19890 (nios2_valid_target_attribute_rec): Likewise.
19891 * config/nvptx/mkoffload.c (process): Likewise.
19893 * config/pa/som.h: Likewise.
19894 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
19895 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
19896 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
19897 (rs6000_inner_target_options): Likewise.
19898 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
19899 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
19900 * config/vax/vax.c (vax_output_int_move): Likewise.
19901 * config/vms/vms-ld.c (startswith): Likewise.
19902 (process_args): Likewise.
19904 * config/vms/vms.c: Likewise.
19906 2021-05-18 Jakub Jelinek <jakub@redhat.com>
19908 PR rtl-optimization/100590
19909 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
19910 they are NONJUMP_INSN_P.
19912 2021-05-18 Jakub Jelinek <jakub@redhat.com>
19915 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
19916 DECL_ASSEMBLER_NAME on the fn_decl.
19918 2021-05-18 Jakub Jelinek <jakub@redhat.com>
19920 PR tree-optimization/94589
19921 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
19922 phi result used in (res & ~1) == 0 comparison as res >= 0 as
19923 res == 2 would be UB with -ffinite-math-only.
19925 2021-05-18 Martin Liska <mliska@suse.cz>
19927 * Makefile.in: genversion.o should depend on DATESTAMP.
19929 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
19931 * config/arc/simdext.md (negv2si2): Remove round bracket.
19933 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
19935 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
19936 _Bool as macro expanding to _Bool.
19938 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
19941 * tree.c (build_reference_type_for_mode)
19942 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
19944 (build_reference_type, build_pointer_type): Invoke
19945 build_*_type_for_mode with VOIDmode.
19947 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
19949 PR tree-optimization/100512
19950 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
19951 and non-zero pointer ranges as invariant.
19952 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
19953 processing from here.
19955 2021-05-17 Tom de Vries <tdevries@suse.de>
19958 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
19959 * config/nvptx/nvptx.c (nvptx_output_barrier)
19960 (nvptx_output_atomic_insn): New function.
19961 (nvptx_print_operand): Add support for 'B'.
19962 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
19965 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
19967 PR tree-optimization/100349
19968 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
19971 2021-05-17 Tamar Christina <tamar.christina@arm.com>
19973 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
19974 (host_detect_local_cpu): Use it.
19976 2021-05-17 Martin Liska <mliska@suse.cz>
19978 * doc/invoke.texi: Add 2 missing dots.
19980 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
19982 PR bootstrap/100552
19983 * configure.ac: Replace pattern substitution with call to sed.
19984 * configure: Regenerate.
19986 2021-05-17 Richard Biener <rguenther@suse.de>
19988 PR middle-end/100582
19989 * tree.c (array_at_struct_end_p): Get to the base of the
19990 reference before looking for the underlying decl.
19992 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
19994 * genoutput.c (validate_insn_alternatives) Make "wrong number of
19995 alternatives" message more specific, and remove assumption on where
19998 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20000 * config/arm/iterators.md (V16): New iterator.
20001 (VH_cvtto): New iterator.
20002 (v_cmp_result): Added V4HF and V8HF support.
20003 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
20004 (vcond<mode><mode>): Likewise.
20005 (vcond_mask_<mode><v_cmp_result>): Likewise.
20006 (vcond<VH_cvtto><mode>): New expander.
20008 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20010 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
20012 * config/arm/arm.c (arm_expand_vector_compare): Add support for
20014 (arm_expand_vcond): Likewise.
20015 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
20016 VCMPEQQ_N_S, VCMPNEQ_N_S.
20017 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
20018 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
20019 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
20020 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
20021 (@mve_vpselq_<supf><mode>): Likewise.
20022 (@mve_vpselq_f<mode>"): Likewise.
20023 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
20024 and move to vec-common.md.
20025 (vec_cmpu<mode><mode>): Likewise.
20026 (vcond<mode><mode>): Likewise.
20027 (vcond<V_cvtto><mode>): Likewise.
20028 (vcondu<mode><v_cmp_result>): Likewise.
20029 (vcond_mask_<mode><v_cmp_result>): Likewise.
20030 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
20031 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
20032 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
20033 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
20034 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
20035 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
20037 (vec_cmpu<mode><mode>): Likewise.
20038 (vcond<mode><mode>): Likewise.
20039 (vcond<V_cvtto><mode>): Likewise.
20040 (vcondu<mode><v_cmp_result>): Likewise.
20041 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
20044 2021-05-17 liuhongt <hongtao.liu@intel.com>
20047 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
20048 gsi_insert_seq_before instead.
20050 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20052 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
20053 (arm_sat_ok): ...this.
20055 2021-05-17 Martin Liska <mliska@suse.cz>
20057 * lto-wrapper.c (merge_flto_options): Factor out a new function.
20058 (merge_and_complain): Use it.
20059 (run_gcc): Merge also linker command line -flto=foo argument
20062 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
20064 * config/arm/arm.h (CPP_SPEC): Remove error message about
20065 -mlittle-endian/-mbig-endian conflict.
20067 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20069 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
20070 __ROP_PROTECT__ if -mrop-protect is selected.
20072 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20074 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
20075 rop_hash_save_offset and rop_hash_size.
20076 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
20077 rop_hash_size and rop_hash_save_offset.
20078 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
20079 (rs6000_emit_prologue): Emit hashst[p] in prologue.
20080 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
20081 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
20083 (hashst): New define_insn.
20084 (hashchk): Likewise.
20086 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20088 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20089 Disable shrink wrap when inserting ROP-protect instructions.
20090 * config/rs6000/rs6000.opt (mrop-protect): New option.
20091 (mprivileged): Likewise.
20092 * doc/invoke.texi: Document mrop-protect and mprivileged.
20094 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
20096 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
20099 2021-05-15 Martin Jambor <mjambor@suse.cz>
20102 2021-05-13 Martin Jambor <mjambor@suse.cz>
20104 PR tree-optimization/100453
20105 * tree-sra.c (sra_modify_assign): All const base accesses do not
20106 need refreshing, not just those from decl_pool.
20107 (sra_modify_assign): Do not refresh into a const base decl.
20109 2021-05-15 Jakub Jelinek <jakub@redhat.com>
20111 PR rtl-optimization/100342
20112 * regcprop.c (copy_value): When copying a source reg in a wider
20113 mode than it has recorded for the value, adjust recorded destination
20114 mode too or punt if !REG_CAN_CHANGE_MODE_P.
20116 2021-05-14 Jason Merrill <jason@redhat.com>
20118 * intl.h: Add comments.
20120 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20122 * config/aarch64/aarch64-simd.md
20123 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
20124 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
20125 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
20126 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
20127 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
20128 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
20129 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
20130 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
20131 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
20133 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20136 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
20137 boolean logic equivalent.
20138 (vtst_s16): Likewise.
20139 (vtst_s32): Likewise.
20140 (vtst_u8): Likewise.
20141 (vtst_u16): Likewise.
20142 (vtst_u32): Likewise.
20143 (vtst_p8): Likewise.
20144 (vtst_p16): Likewise.
20145 (vtstq_s8): Likewise.
20146 (vtstq_s16): Likewise.
20147 (vtstq_s32): Likewise.
20148 (vtstq_u8): Likewise.
20149 (vtstq_u16): Likewise.
20150 (vtstq_u32): Likewise.
20151 (vtstq_p8): Likewise.
20152 (vtstq_p16): Likewise.
20153 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
20154 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
20156 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20158 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
20159 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
20160 (aarch64_sqdmlsl2<mode>): Delete.
20161 (aarch64_sqdmlal2_lane<mode>): Merge this...
20162 (aarch64_sqdmlsl2_lane<mode>): ... And this...
20163 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
20164 (aarch64_sqdmlal2_laneq<mode>): Merge this...
20165 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
20166 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
20167 (aarch64_sqdmlal2_n<mode>): Merge this...
20168 (aarch64_sqdmlsl2_n<mode>): ... And this...
20169 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
20171 2021-05-13 Martin Sebor <msebor@redhat.com>
20173 PR middle-end/100574
20174 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
20175 all null arguments.
20177 2021-05-13 Martin Sebor <msebor@redhat.com>
20179 PR tree-optimization/93100
20180 PR middle-end/98583
20181 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
20182 don't modify referenced objects.
20184 2021-05-13 Martin Jambor <mjambor@suse.cz>
20186 PR tree-optimization/100453
20187 * tree-sra.c (sra_modify_assign): All const base accesses do not
20188 need refreshing, not just those from decl_pool.
20189 (sra_modify_assign): Do not refresh into a const base decl.
20191 2021-05-13 Martin Liska <mliska@suse.cz>
20193 * tree-ssa-dom.c: Remove m_simplifier.
20195 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
20198 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
20199 canonicalize DImode inequality comparisons against the
20200 maximum integral value.
20202 2021-05-13 Jakub Jelinek <jakub@redhat.com>
20204 PR tree-optimization/98856
20205 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
20206 Expect V2DI and V4DI arithmetic right shifts to be emulated.
20207 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
20209 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
20210 expand_vec_perm_2perm_pblendv): New functions.
20211 (ix86_expand_vec_perm_const_1): Use them.
20212 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
20213 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
20214 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
20215 (ashrv4di3): New define_expand.
20216 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
20217 and !TARGET_AVX512VL expansion.
20219 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
20222 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
20223 sizes < 16 to a register when constructing vpcmov pattern.
20224 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
20226 2021-05-13 Martin Liska <mliska@suse.cz>
20228 * gcov-io.c (gcov_write_block): Remove.
20229 (gcov_write_words): Likewise.
20230 (gcov_read_words): Re-implement using gcov_read_bytes.
20231 (gcov_allocate): Remove.
20232 (GCOV_BLOCK_SIZE): Likewise.
20233 (struct gcov_var): Remove most of the fields.
20234 (gcov_position): Implement with ftell.
20235 (gcov_rewrite): Remove setting of start and offset fields.
20236 (from_file): Re-format.
20237 (gcov_open): Remove setbuf call. It should not be needed.
20238 (gcov_close): Remove internal buffer handling.
20239 (gcov_magic): Use __builtin_bswap32.
20240 (gcov_write_counter): Use directly gcov_write_unsigned.
20241 (gcov_write_string): Use direct fwrite and do not round
20243 (gcov_seek): Use directly fseek.
20244 (gcov_write_tag): Use gcov_write_unsigned directly.
20245 (gcov_write_length): Likewise.
20246 (gcov_write_tag_length): Likewise.
20247 (gcov_read_bytes): Use directly fread.
20248 (gcov_read_unsigned): Use gcov_read_words.
20249 (gcov_read_counter): Likewise.
20250 (gcov_read_string): Use gcov_read_bytes.
20251 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
20252 that size is not in bytes, but words (4B).
20253 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
20254 (GCOV_TAG_ARCS_LENGTH): Likewise.
20255 (GCOV_TAG_ARCS_NUM): Likewise.
20256 (GCOV_TAG_COUNTER_LENGTH): Likewise.
20257 (GCOV_TAG_COUNTER_NUM): Likewise.
20258 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
20260 2021-05-13 liuhongt <hongtao.liu@intel.com>
20263 * config/i386/sse.md (ssedoublevecmode): Add attribute for
20264 V64QI/V32HI/V16SI/V4DI.
20265 (ssehalfvecmode): Add attribute for V2DI/V2DF.
20266 (*vec_concatv4si_0): Extend to VI124_128.
20267 (*vec_concat<mode>_0): New pre-reload splitter.
20268 * config/i386/predicates.md (movq_parallel): New predicate.
20270 2021-05-13 Alexandre Oliva <oliva@adacore.com>
20272 * targhooks.c (default_zero_call_used_regs): Retry using
20273 successfully-zeroed registers as sources.
20275 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
20277 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
20279 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
20282 * gimple-range.cc (range_of_builtin_call): Skip out on
20283 processing __builtin_clz when varying.
20285 2021-05-12 Tom de Vries <tdevries@suse.de>
20288 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
20289 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
20290 to ptx_version_option.
20291 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
20292 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
20293 (define_insn "nvptx_vote_ballot"): Use sync variant for
20295 * config/nvptx/nvptx.opt (ptx_version): Add enum.
20296 (mptx): Add option.
20297 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
20299 2021-05-12 Richard Biener <rguenther@suse.de>
20301 PR tree-optimization/100566
20302 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
20303 allow_back for all edge queries.
20305 2021-05-12 liuhongt <hongtao.liu@intel.com>
20308 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
20309 splitters for pblendvb of NOT mask register.
20311 2021-05-12 Richard Biener <rguenther@suse.de>
20313 PR tree-optimization/100519
20314 * tree-ssa-reassoc.c (can_associate_p): Split into...
20315 (can_associate_op_p): ... this
20316 (can_associate_type_p): ... and this.
20317 (is_reassociable_op): Call can_associate_op_p.
20318 (break_up_subtract_bb): Call the appropriate predicates.
20319 (reassociate_bb): Likewise.
20321 2021-05-12 Martin Liska <mliska@suse.cz>
20323 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
20324 (run_gcc): Use -flto argument detection for merged
20327 2021-05-12 Martin Liska <mliska@suse.cz>
20329 * lto-wrapper.c (print_lto_docs_link): New function.
20330 (run_gcc): Print warning about missing job server detection
20331 after we know NR of partitions. Do the same for -flto{,=1}.
20332 * opts.c (get_option_html_page): Support -flto option.
20334 2021-05-12 Martin Liska <mliska@suse.cz>
20336 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
20338 (append_option): Remove.
20339 (find_option): Rework to use the vector type.
20340 (remove_option): Remove.
20341 (merge_and_complain): Use vectors for cl_decoded_option data
20343 (append_compiler_options): Likewise.
20344 (append_diag_options): Likewise.
20345 (append_linker_options): Likewise.
20346 (append_offload_options): Likewise.
20347 (compile_offload_image): Likewise.
20348 (compile_images_for_offload_targets): Likewise.
20349 (find_and_merge_options): Likewise.
20350 (run_gcc): Likewise.
20352 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
20355 * dwarf2out.c (dwarf2out_finish): Set
20356 have_multiple_function_sections with multi-range text_section.
20358 2021-05-12 Martin Liska <mliska@suse.cz>
20360 PR bootstrap/100560
20361 * Makefile.in: Remove version.h from linker command line.
20363 2021-05-12 Richard Biener <rguenther@suse.de>
20365 PR middle-end/100547
20366 * rtl.h (rtvec_alloc): Make argument size_t.
20367 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
20369 2021-05-12 Jakub Jelinek <jakub@redhat.com>
20371 PR middle-end/100508
20372 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
20373 type, don't reuse DECL_RTL if it has different mode, instead force
20374 creation of a new DEBUG_EXPR.
20376 2021-05-12 Jakub Jelinek <jakub@redhat.com>
20377 Marc Glisse <marc.glisse@inria.fr>
20379 PR tree-optimization/94589
20380 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
20381 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
20383 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
20386 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
20387 * config/i386/mmx.md (MMXMODE124): New mode iterator.
20389 (mmxintvecmode): New mode attribute.
20390 (mmxintvecmodelower): Ditto.
20391 (*mmx_maskcmpv2sf3_comm): New insn pattern.
20392 (*mmx_maskcmpv2sf3): Ditto.
20393 (vec_cmpv2sfv2si): New expander.
20394 (vcond<V2FI:mode>v2si): Ditto.
20395 (mmx_vlendvps): New insn pattern.
20396 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
20397 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
20398 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
20400 2021-05-11 Martin Sebor <msebor@redhat.com>
20402 PR middle-end/21433
20403 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
20405 2021-05-11 Richard Biener <rguenther@suse.de>
20407 * gimple-fold.c (gimple_fold_call): Do not call
20408 maybe_fold_reference on call arguments or the static chain.
20409 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
20412 2021-05-11 Martin Liska <mliska@suse.cz>
20414 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
20415 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
20416 (DEF_HSAIL_SAT_BUILTIN): Likewise.
20417 (DEF_HSAIL_INTR_BUILTIN): Likewise.
20418 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
20419 * doc/frontends.texi: Remove BRIG.
20420 * doc/install.texi: Likewise.
20421 * doc/invoke.texi: Likewise.
20422 * doc/standards.texi: Likewise.
20423 * brig-builtins.def: Removed.
20424 * brig/ChangeLog: Removed.
20425 * brig/Make-lang.in: Removed.
20426 * brig/brig-builtins.h: Removed.
20427 * brig/brig-c.h: Removed.
20428 * brig/brig-lang.c: Removed.
20429 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
20430 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
20431 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
20432 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
20433 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
20434 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
20435 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
20436 * brig/brigfrontend/brig-comment-handler.cc: Removed.
20437 * brig/brigfrontend/brig-control-handler.cc: Removed.
20438 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
20439 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
20440 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
20441 * brig/brigfrontend/brig-function-handler.cc: Removed.
20442 * brig/brigfrontend/brig-function.cc: Removed.
20443 * brig/brigfrontend/brig-function.h: Removed.
20444 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
20445 * brig/brigfrontend/brig-label-handler.cc: Removed.
20446 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
20447 * brig/brigfrontend/brig-machine.c: Removed.
20448 * brig/brigfrontend/brig-machine.h: Removed.
20449 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
20450 * brig/brigfrontend/brig-module-handler.cc: Removed.
20451 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
20452 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
20453 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
20454 * brig/brigfrontend/brig-to-generic.cc: Removed.
20455 * brig/brigfrontend/brig-to-generic.h: Removed.
20456 * brig/brigfrontend/brig-util.cc: Removed.
20457 * brig/brigfrontend/brig-util.h: Removed.
20458 * brig/brigfrontend/brig-variable-handler.cc: Removed.
20459 * brig/brigfrontend/hsa-brig-format.h: Removed.
20460 * brig/brigfrontend/phsa.h: Removed.
20461 * brig/brigspec.c: Removed.
20462 * brig/config-lang.in: Removed.
20463 * brig/gccbrig.texi: Removed.
20464 * brig/lang-specs.h: Removed.
20465 * brig/lang.opt: Removed.
20467 2021-05-11 Richard Biener <rguenther@suse.de>
20470 * ipa-param-manipulation.c
20471 (ipa_param_body_adjustments::modify_call_stmt): Avoid
20472 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
20473 via gimple_call_lhs_ptr.
20475 2021-05-11 Alex Coplan <alex.coplan@arm.com>
20478 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
20479 Avoid emitting CFA adjusts on the sp if we have the fp.
20481 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
20483 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
20484 (VMULD): New iterator.
20485 (VCOND): Handle V4HF and V8HF.
20486 (VCONQ): Fix entry for V2SF.
20487 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
20488 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
20489 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
20490 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
20491 Use a 128-bit vector mode for the indexed operand. Use stype for
20492 the scheduling type.
20494 2021-05-11 Richard Biener <rguenther@suse.de>
20496 * gimple-fold.c (maybe_fold_reference): Only return
20497 is_gimple_min_invariant values.
20499 2021-05-11 Richard Biener <rguenther@suse.de>
20501 PR middle-end/100509
20502 * gimple-fold.c (fold_gimple_assign): Only call
20503 get_symbol_constant_value on register type symbols.
20505 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
20506 Joe Ramsay <joe.ramsay@arm.com>
20509 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
20510 (__arm_vcmpneq): Remove duplicate definition.
20511 (__arm_vstrwq_scatter_offset_p): Likewise.
20512 (__arm_vmaxq_x): Likewise.
20513 (__arm_vmlsdavaq): Likewise.
20514 (__arm_vmlsdavaxq): Likewise.
20515 (__arm_vmlsdavq_p): Likewise.
20516 (__arm_vmlsdavxq_p): Likewise.
20517 (__arm_vrmlaldavhaq): Likewise.
20518 (__arm_vstrbq_p): Likewise.
20519 (__arm_vstrbq_scatter_offset): Likewise.
20520 (__arm_vstrbq_scatter_offset_p): Likewise.
20521 (__arm_vstrdq_scatter_offset): Likewise.
20522 (__arm_vstrdq_scatter_offset_p): Likewise.
20523 (__arm_vstrdq_scatter_shifted_offset): Likewise.
20524 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
20526 2021-05-11 Jakub Jelinek <jakub@redhat.com>
20528 PR middle-end/100471
20529 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
20530 is 0, bypass the reduction loop including
20531 GOMP_taskgroup_reduction_unregister call.
20533 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
20535 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
20536 costing_for_scalar.
20537 (rs6000_density_test): Early return if costing_for_scalar is true.
20538 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
20540 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
20542 * doc/tm.texi: Regenerated.
20543 * target.def (init_cost): Add new parameter costing_for_scalar.
20544 * targhooks.c (default_init_cost): Adjust for new parameter.
20545 * targhooks.h (default_init_cost): Likewise.
20546 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
20547 (vect_compute_single_scalar_iteration_cost): Likewise.
20548 (vect_analyze_loop_2): Likewise.
20549 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
20550 (vect_bb_vectorization_profitable_p): Likewise.
20551 * tree-vectorizer.h (init_cost): Likewise.
20552 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
20553 * config/i386/i386.c (ix86_init_cost): Likewise.
20554 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
20556 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
20558 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
20559 vect_nonmem and moved into...
20560 (struct rs6000_cost_data): ...here.
20561 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
20562 (rs6000_add_stmt_cost): Likewise.
20563 (rs6000_finish_cost): Likewise.
20565 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
20567 * range-op.cc (get_bool_state): Adjust head comment.
20568 (operator_not_equal::op1_range): Fix comment.
20569 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
20571 2021-05-10 Martin Sebor <msebor@redhat.com>
20573 PR middle-end/100425
20574 PR middle-end/100510
20575 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
20576 (pass_walloca::xlimit_certain_p): ...to this.
20577 (pass_walloca::gate): Execute for any kind of handled warning.
20578 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
20579 warnings when xlimit_certain_p is set.
20581 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
20583 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
20584 Return ALTIVEC_REGS if that is best_class.
20585 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
20587 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20589 * config/arm/arm.h (CPP_SPEC): Remove error message about
20592 2021-05-10 Martin Jambor <mjambor@suse.cz>
20594 * ipa-prop.h (IPA_NODE_REF): Removed.
20595 (IPA_NODE_REF_GET_CREATE): Likewise.
20596 (IPA_EDGE_REF): Likewise.
20597 (IPA_EDGE_REF_GET_CREATE): Likewise.
20598 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
20599 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
20600 use of ipa_node_params_sum.
20601 (ipcp_versionable_function_p): Likewise.
20602 (push_node_to_stack): Likewise.
20603 (pop_node_from_stack): Likewise.
20604 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
20605 direct use of ipa_node_params_sum.
20606 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
20607 ipa_node_params_sum.
20608 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
20610 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
20611 use of ipa_node_params_sum.
20612 (self_recursively_generated_p): Likewise.
20613 (propagate_scalar_across_jump_function): Likewise.
20614 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
20615 direct use of ipa_edge_args_sum, moved the lookup after the early
20616 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
20617 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
20618 direct uses of ipa_node_params_sum.
20619 (propagate_vr_across_jump_function): Likewise.
20620 (propagate_aggregate_lattice): Likewise.
20621 (propagate_aggs_across_jump_function): Likewise.
20622 (propagate_constants_across_call): Likewise, also replaced
20623 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
20624 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
20625 of ipa_node_params_sum.
20626 (estimate_local_effects): Likewise.
20627 (add_all_node_vals_to_toposort): Likewise.
20628 (propagate_constants_topo): Likewise.
20629 (ipcp_propagate_stage): Likewise.
20630 (ipcp_discover_new_direct_edges): Likewise.
20631 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
20632 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
20633 (get_info_about_necessary_edges): Likewise.
20634 (want_remove_some_param_p): Likewise.
20635 (create_specialized_node): Likewise.
20636 (self_recursive_pass_through_p): Likewise.
20637 (self_recursive_agg_pass_through_p): Likewise.
20638 (find_more_scalar_values_for_callers_subset): Likewise and also
20639 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
20640 case replacing two of those with a single query.
20641 (find_more_contexts_for_caller_subset): Likewise for the
20642 ipa_polymorphic_call_context overload.
20643 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
20644 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
20645 ipa_node_params_sum.
20646 (find_aggregate_values_for_callers_subset): Likewise, also reusing
20647 results of ipa_edge_args_sum->get.
20648 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
20649 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
20650 direct use of ipa_edge_args_sum.
20651 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
20652 summary query after the early exit and reused the result later.
20653 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
20654 ipa_node_params_sum.
20655 (decide_whether_version_node): Likewise. Removed re-querying for
20656 summaries after cloning.
20657 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
20658 ipa_node_params_sum.
20659 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
20661 (identify_dead_nodes): Likewise.
20662 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
20663 ipa_node_params_sum.
20664 (ipcp_store_vr_results): Likewise.
20665 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
20666 (ipa_fn_summary_t::duplicate): Likewise.
20667 (analyze_function_body): Likewise.
20668 (estimate_calls_size_and_time): Likewise.
20669 (ipa_cached_call_context::duplicate_from): Likewise.
20670 (ipa_call_context::equal_to): Likewise.
20671 (remap_edge_params): Likewise.
20672 (ipa_merge_fn_summary_after_inlining): Likewise.
20673 (inline_read_section): Likewise.
20674 * ipa-icf.c (sem_function::param_used_p): Likewise.
20675 * ipa-modref.c (compute_parm_map): Likewise.
20676 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
20678 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
20679 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
20681 * ipa-profile.c (check_argument_count): Likewise.
20682 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
20683 with a direct use of ipa_node_params_sum.
20684 (ipa_initialize_node_params): Likewise.
20685 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
20686 direct use of ipa_edge_args_sum and reused the query result.
20687 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
20688 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
20689 direct use of ipa_edge_args_sum.
20690 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
20691 ipa_node_params_sum and reused the result of the query.
20692 (ipa_analyze_node): Likewise.
20693 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
20694 of ipa_node_params_sum.
20695 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
20696 direct uses of ipa_edge_args_sum.
20697 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
20698 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
20699 direct use of ipa_edge_args_sum. Removed superficial re-querying the
20701 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
20702 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
20704 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
20705 direct use of ipa_edge_args_sum.
20706 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
20707 use of ipa_node_params_sum.
20708 (ipa_print_node_params): Likewise.
20709 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
20710 direct uses of ipa_edge_args_sum.
20711 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
20713 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
20714 ipa_node_params_sum.
20715 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
20716 scopes where it is used.
20718 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
20720 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
20721 <case E_V2SImode>: Force op_true to register.
20723 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20725 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
20726 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
20727 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
20729 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
20730 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
20731 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
20732 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
20733 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
20734 (mve_vcmpneq_n_f<mode>): Remove.
20735 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
20736 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
20737 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
20739 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20741 * config/arm/iterators.md (MVE_COMPARISONS): New.
20743 (mve_cmp_type): New.
20744 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
20746 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
20747 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
20748 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
20749 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
20750 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
20751 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
20752 (mve_vcmpneq_n_<mode>): Remove.
20754 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20756 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
20757 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
20758 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
20761 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20763 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
20764 (vcmpneq_n_u): Likewise.
20765 (vcmpeqq_u,): Likewise.
20766 (vcmpeqq_n_u): Likewise.
20767 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
20768 VCMPEQQ_N_U and VCMPNEQ_N_U.
20769 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
20770 (mve_vcmpeqq_n): Likewise.
20771 (mve_vcmpeqq): Likewise.
20772 (mve_vcmpneq_n): Likewise.
20774 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
20776 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
20777 the 's' version of the builtin.
20779 2021-05-10 Richard Biener <rguenther@suse.de>
20781 PR tree-optimization/100492
20782 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
20783 Find nothing when the loop contains an irreducible region.
20785 2021-05-10 Richard Biener <rguenther@suse.de>
20787 PR middle-end/100464
20789 * gimple-fold.c (canonicalize_constructor_val): Do not set
20792 2021-05-10 Richard Biener <rguenther@suse.de>
20794 PR tree-optimization/100434
20795 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
20797 (dse_optimize_stmt): Handle call LHS by dropping the
20798 LHS or the whole call if it doesn't have other
20800 (pass_dse::execute): Adjust.
20802 2021-05-10 Martin Liska <mliska@suse.cz>
20804 * Makefile.in: Add missing genversion rule.
20806 2021-05-10 Alex Coplan <alex.coplan@arm.com>
20809 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
20810 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
20812 2021-05-10 Martin Liska <mliska@suse.cz>
20814 * builtins.c (is_builtin_name): Use startswith
20815 function instead of strncmp.
20816 * collect2.c (main): Likewise.
20817 (has_lto_section): Likewise.
20818 (scan_libraries): Likewise.
20819 * coverage.c (coverage_checksum_string): Likewise.
20820 (coverage_init): Likewise.
20821 * dwarf2out.c (is_cxx): Likewise.
20822 (gen_compile_unit_die): Likewise.
20823 * gcc-ar.c (main): Likewise.
20824 * gcc.c (init_spec): Likewise.
20825 (read_specs): Likewise.
20826 (execute): Likewise.
20827 (check_live_switch): Likewise.
20828 * genattrtab.c (write_attr_case): Likewise.
20829 (IS_ATTR_GROUP): Likewise.
20830 * gencfn-macros.c (main): Likewise.
20831 * gengtype.c (type_for_name): Likewise.
20832 (gen_rtx_next): Likewise.
20833 (get_file_langdir): Likewise.
20834 (write_local): Likewise.
20835 * genmatch.c (get_operator): Likewise.
20836 (get_operand_type): Likewise.
20837 (expr::gen_transform): Likewise.
20838 * genoutput.c (validate_optab_operands): Likewise.
20839 * incpath.c (add_sysroot_to_chain): Likewise.
20840 * langhooks.c (lang_GNU_C): Likewise.
20841 (lang_GNU_CXX): Likewise.
20842 (lang_GNU_Fortran): Likewise.
20843 (lang_GNU_OBJC): Likewise.
20844 * lto-wrapper.c (run_gcc): Likewise.
20845 * omp-general.c (omp_max_simt_vf): Likewise.
20846 * omp-low.c (omp_runtime_api_call): Likewise.
20847 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
20848 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
20849 * real.c (real_from_string): Likewise.
20850 * selftest.c (assert_str_startswith): Likewise.
20851 * timevar.c (timer::validate_phases): Likewise.
20852 * tree.c (get_file_function_name): Likewise.
20853 * ubsan.c (ubsan_use_new_style_p): Likewise.
20854 * varasm.c (default_function_rodata_section): Likewise.
20855 (incorporeal_function_p): Likewise.
20856 (default_section_type_flags): Likewise.
20857 * system.h (startswith): Define startswith.
20859 2021-05-10 Martin Liska <mliska@suse.cz>
20861 * bitmap.h (class auto_bitmap): Remove
20862 __cplusplus >= 201103.
20863 * config/aarch64/aarch64.c: Likewise.
20864 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
20866 * sbitmap.h: Likewise.
20868 2021-05-10 Martin Liska <mliska@suse.cz>
20870 * Makefile.in: Rename gcov-iov to genversion and depend
20871 on version.h (instead of gcov-iov.h).
20872 * gcov-io.h: Include version.h instread of gcov-iov.h.
20873 * gengtype-state.c (read_state_version): Likewise.
20874 * gcov-iov.c: Moved to...
20875 * genversion.c: ...here.
20876 * lto-streamer.h (LTO_major_version): Define it with
20878 * version.c: Removed.
20879 * version.h: Removed.
20881 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
20883 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
20884 * config/arc/simdext.md (VCT): Add predicates for iterator
20887 (voptab): Likewise.
20888 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
20889 (<voptab>v2si3): New patterns.
20891 (reduc_plus_scal_v4hi): Likewise.
20892 (reduc_plus_scal_v2si): Likewise.
20893 (vec_duplicatev2si): Likewise.
20894 (vec_duplicatev4hi): Likewise.
20896 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
20898 * config/arc/simdext.md: Format and cleanup file.
20900 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
20902 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
20903 only when munaligned-access option is on.
20904 (movmisalign<mode>): Likewise.
20906 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
20908 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
20910 * config/arc/arc.c (arc_reorg): Remove underscore from string.
20912 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
20914 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
20915 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
20916 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
20917 (norm_f): Likewise.
20920 (clzsi2): Use fls instruction when available.
20921 (arc_clzsi2): Likewise.
20923 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
20925 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
20927 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
20929 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
20931 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
20933 PR middle-end/100467
20934 * toplev.c (compile_file): Call insn_locations_init before
20935 targetm.asm_out.code_end.
20937 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
20940 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
20942 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
20944 2021-05-07 Jakub Jelinek <jakub@redhat.com>
20945 Andrew Stubbs <amd@codesourcery.com>
20948 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
20949 emit_move_insn operands.
20951 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
20953 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
20954 location for the outgoing edges of an empty block.
20955 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
20956 type by means of the get_array_descr_info langhook, if it is set and
20957 returns true. Remove obsolete code dealing with unnamed subtypes.
20959 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
20961 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
20962 (sbr_vector): Renamed from ssa_block_cache.
20963 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
20964 (ssa_block_ranges::~ssa_block_ranges): Remove.
20965 (sbr_vector::set_bb_range): Use varying and undefined cached values.
20966 (ssa_block_ranges::set_bb_varying): Remove.
20967 (sbr_vector::get_bb_range): Adjust assert.
20968 (sbr_vector::bb_range_p): Adjust assert.
20969 (~block_range_cache): No freeing loop required.
20970 (block_range_cache::get_block_ranges): Remove.
20971 (block_range_cache::set_bb_range): Inline get_block_ranges.
20972 (block_range_cache::set_bb_varying): Remove.
20973 * gimple-range-cache.h (set_bb_varying): Remove prototype.
20974 * value-range.h (irange_allocator::get_memory): New.
20976 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
20978 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
20979 dominator tree is available and requested.
20980 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
20981 (ranger_cache::fill_block_cache): Don't search dom tree here either.
20982 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
20984 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
20986 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
20987 only PHI nodes better.
20989 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
20991 * gimple-range-edge.h (gimple_outgoing_range): Rename from
20993 (gcond_edge_range): Export prototype.
20994 * gimple-range-edge.cc (gcond_edge_range): New.
20995 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
20996 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
20998 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21000 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
21001 default range into a temp and allocate only what is needed.
21003 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21005 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
21007 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21009 * gimple-range.h (gimple_range_global): Pick up parameter initial
21010 values, and use-before defined locals are UNDEFINED.
21012 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
21014 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
21016 * tree.h (reverse_storage_order_for_component_p): Return false if
21017 the type is a pointer.
21019 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21021 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
21023 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
21026 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
21027 Handle V8QI, V4HI and V2SI modes.
21028 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
21029 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
21030 * config/i386/i386.md (unspec): ... here.
21032 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
21033 Tom de Vries <tdevries@suse.de>
21035 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
21036 a truth_value_p reduction variable is nonintegral.
21038 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
21041 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
21042 Return false for mode sizes < 16.
21044 2021-05-07 Jakub Jelinek <jakub@redhat.com>
21047 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
21049 2021-05-06 Martin Jambor <mjambor@suse.cz>
21051 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
21052 when there is no function summary.
21053 (ipa_sra_summarize_function): produce edge summaries even when
21056 2021-05-06 Tom Tromey <tom@tromey.com>
21058 * godump.c (string_hash_eq): Remove.
21059 (go_finish): Use htab_eq_string.
21061 2021-05-06 Tom Tromey <tom@tromey.com>
21063 * gengtype-state.c (read_state): Use htab_eq_string.
21064 (string_eq): Remove.
21066 2021-05-06 Tom Tromey <tom@tromey.com>
21068 * gensupport.c (htab_eq_string): Remove.
21070 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
21073 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
21074 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
21075 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
21076 set_ignored_loc callbacks.
21077 * debug.c (do_nothing_debug_hooks): Likewise.
21078 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21079 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
21080 (in_text_section_p, last_text_label, last_cold_label,
21081 switch_text_ranges, switch_cold_ranges): New data items.
21082 (dwarf2out_note_section_used): Remove.
21083 (dwarf2out_begin_prologue): Set fde->ignored_debug and
21085 (mark_ignored_debug_section): New helper function.
21086 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
21087 mark_ignored_debug_section.
21088 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
21089 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
21090 (size_of_aranges): Adjust formula for multi-part text ranges size.
21091 (output_aranges): Output multi-part text ranges.
21092 (dwarf2out_set_ignored_loc): New callback function.
21093 (dwarf2out_finish): Output multi-part text ranges.
21094 (dwarf2out_c_finalize): Clear new data items.
21095 * final.c (final_start_function_1): Call set_ignored_loc callback.
21096 (final_scan_insn_1): Likewise.
21097 * ggc-page.c (gt_ggc_mx): New helper function.
21098 * stringpool.c (gt_pch_nx): Likewise.
21100 2021-05-06 Richard Biener <rguenther@suse.de>
21102 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
21103 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
21104 (TV_TREE_INTO_SSA): New.
21105 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
21106 (rewrite_blocks): Likewise.
21107 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
21109 2021-05-06 Jakub Jelinek <jakub@redhat.com>
21111 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
21112 abs_replacement, xor_replacement,
21113 cond_removal_in_popcount_clz_ctz_pattern,
21114 replace_phi_edge_with_variable): Change type of phi argument from
21115 gimple * to gphi *.
21117 2021-05-06 Richard Biener <rguenther@suse.de>
21119 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
21120 Output an opt-info message.
21121 (do_split_loop_on_cond): Likewise.
21122 (tree_ssa_split_loops): Update SSA form here.
21124 2021-05-06 Richard Biener <rguenther@suse.de>
21126 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
21127 return variable removal.
21129 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
21131 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
21132 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
21133 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
21135 * config/s390/s390.c (s390_const_operand_ok): Remove unused
21138 2021-05-06 Jakub Jelinek <jakub@redhat.com>
21140 PR tree-optimization/94589
21141 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
21142 spaceship_replacement.
21143 (cond_only_block_p, spaceship_replacement): New functions.
21145 2021-05-06 Richard Biener <rguenther@suse.de>
21148 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
21149 dealing with a debug use and only query existing addresses
21151 (lower_emutls_1): Avoid splitting out addresses for debug
21152 stmts, reset the debug stmt when we fail to find existing
21154 (lower_emutls_phi_arg): Set wi.stmt.
21156 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
21159 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
21160 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
21161 (stack_protect_test): Use cbranch helper.
21163 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21166 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21167 always return the establisher frame for __builtin_frame_address (0).
21169 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
21172 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
21173 (ix86_cpu_model_var): Likewise.
21174 (ix86_cpu_features2_type_node): Likewise.
21175 (ix86_cpu_features2_var): Likewise.
21176 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
21179 2021-05-05 Martin Sebor <msebor@redhat.com>
21181 * passes.def (pass_warn_printf): Run after SSA.
21183 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21185 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
21186 * config/arm/predicates.md (minus_one_operand): New predicate.
21188 2021-05-05 Jeff Law <jlaw@tachyum.com>
21190 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
21192 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
21194 PR rtl-optimization/100263
21195 * postreload.c (move2add_valid_value_p): Ensure register can
21198 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21200 PR rtl-optimization/100411
21201 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
21202 and beginning of function markers.
21204 2021-05-05 Jeff Law <jlaw@tachyum.com>
21206 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
21207 * config/cr16/cr16.c (notice_update_cc): Remove.
21208 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
21210 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
21213 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
21214 Handle V8QI, V4HI and V2SI modes.
21215 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
21216 (ix86_build_signbit_mask): Ditto.
21217 * config/i386/mmx.md (MMXMODE14): New mode iterator.
21218 (<smaxmin:code><MMXMODE14:mode>3): New expander.
21219 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
21220 (<umaxmin:code><MMXMODE24:mode>3): New expander.
21221 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
21222 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
21223 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21224 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21225 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21226 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21228 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21230 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
21231 not expand the VALUE_EXPR of variables put in the non-local frame.
21232 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
21233 to be ignored for debug info, ensure its variable offsets are not.
21235 2021-05-05 Richard Biener <rguenther@suse.de>
21237 PR tree-optimization/79333
21238 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
21239 Fold stmt following SSA edges.
21241 2021-05-05 Richard Biener <rguenther@suse.de>
21243 PR middle-end/100394
21244 * calls.c (expand_call): Preserve possibly throwing calls.
21245 * cfgexpand.c (expand_call_stmt): When a call can throw signal
21246 RTL expansion there are side-effects.
21247 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
21248 mark all possibly throwing stmts necessary unless we can elide
21250 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
21251 -fdelete-dead-exceptions.
21252 * tree.h (DECL_PURE_P): Add note about exceptions.
21254 2021-05-05 Alexandre Oliva <oliva@adacore.com>
21256 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
21259 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
21261 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
21262 get_fnname_from_decl for name of thunk.
21263 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
21264 and ASM_OUTPUT_LABEL.
21265 (rs6000_xcoff_declare_function_name): Use assemble_name and
21267 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
21268 (rs6000_xcoff_encode_section_info): Don't add mapping class
21269 for aliases. Always add [DS] mapping class to primary
21271 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
21273 2021-05-04 Martin Sebor <msebor@redhat.com>
21275 PR middle-end/100307
21276 * builtins.c (compute_objsize_r): Clear base0 for pointers.
21278 2021-05-04 Jeff Law <jlaw@tachyum.com>
21280 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
21282 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
21284 * caller-save.c: Remove CC0.
21285 * cfgcleanup.c: Remove CC0.
21286 * cfgrtl.c: Remove CC0.
21287 * combine.c: Remove CC0.
21288 * compare-elim.c: Remove CC0.
21289 * conditions.h: Remove CC0.
21290 * config/h8300/h8300.h: Remove CC0.
21291 * config/h8300/h8300-protos.h: Remove CC0.
21292 * config/h8300/peepholes.md: Remove CC0.
21293 * config/i386/x86-tune-sched.c: Remove CC0.
21294 * config/m68k/m68k.c: Remove CC0.
21295 * config/rl78/rl78.c: Remove CC0.
21296 * config/sparc/sparc.c: Remove CC0.
21297 * config/xtensa/xtensa.c: Remove CC0.
21298 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
21299 RTL where that is used as a placeholder only.
21300 * cprop.c: Remove CC0.
21301 * cse.c: Remove CC0.
21302 * cselib.c: Remove CC0.
21303 * df-problems.c: Remove CC0.
21304 * df-scan.c: Remove CC0.
21305 * doc/md.texi: Remove CC0. Adjust an example.
21306 * doc/rtl.texi: Remove CC0. Adjust an example.
21307 * doc/tm.texi: Regenerate.
21308 * doc/tm.texi.in: Remove CC0.
21309 * emit-rtl.c: Remove CC0.
21310 * final.c: Remove CC0.
21311 * fwprop.c: Remove CC0.
21312 * gcse-common.c: Remove CC0.
21313 * gcse.c: Remove CC0.
21314 * genattrtab.c: Remove CC0.
21315 * genconfig.c: Remove CC0.
21316 * genemit.c: Remove CC0.
21317 * genextract.c: Remove CC0.
21318 * gengenrtl.c: Remove CC0.
21319 * genrecog.c: Remove CC0.
21320 * haifa-sched.c: Remove CC0.
21321 * ifcvt.c: Remove CC0.
21322 * ira-costs.c: Remove CC0.
21323 * ira.c: Remove CC0.
21324 * jump.c: Remove CC0.
21325 * loop-invariant.c: Remove CC0.
21326 * lra-constraints.c: Remove CC0.
21327 * lra-eliminations.c: Remove CC0.
21328 * optabs.c: Remove CC0.
21329 * postreload-gcse.c: Remove CC0.
21330 * postreload.c: Remove CC0.
21331 * print-rtl.c: Remove CC0.
21332 * read-rtl-function.c: Remove CC0.
21333 * reg-notes.def: Remove CC0.
21334 * reg-stack.c: Remove CC0.
21335 * reginfo.c: Remove CC0.
21336 * regrename.c: Remove CC0.
21337 * reload.c: Remove CC0.
21338 * reload1.c: Remove CC0.
21339 * reorg.c: Remove CC0.
21340 * resource.c: Remove CC0.
21341 * rtl.c: Remove CC0.
21342 * rtl.def: Remove CC0.
21343 * rtl.h: Remove CC0.
21344 * rtlanal.c: Remove CC0.
21345 * sched-deps.c: Remove CC0.
21346 * sched-rgn.c: Remove CC0.
21347 * shrink-wrap.c: Remove CC0.
21348 * simplify-rtx.c: Remove CC0.
21349 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
21350 CC_STATUS_MDEP, and CC_STATUS.
21351 * target.def: Remove CC0.
21352 * valtrack.c: Remove CC0.
21353 * var-tracking.c: Remove CC0.
21355 2021-05-04 Richard Biener <rguenther@suse.de>
21357 PR tree-optimization/100414
21358 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
21360 (tree_ssa_phiopt_worker): But unconditionally here.
21362 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
21364 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
21365 && and || with floating-point and complex arguments.
21367 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
21369 * tree-inline.c (insert_debug_decl_map): Delete.
21370 (copy_debug_stmt): Minor tweak.
21371 (setup_one_parameter): Do not use a variable if the value is either
21372 a read-only DECL or a non-addressable local variable in the caller.
21373 In this case, insert the debug-only variable in the map manually.
21374 (expand_call_inline): Do not generate a CLOBBER for these values.
21375 * tree-inline.h (debug_map): Minor tweak.
21377 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
21379 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
21380 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
21382 2021-05-04 Richard Biener <rguenther@suse.de>
21384 PR tree-optimization/100329
21385 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
21387 (insert_stmt_after): Assert we're not running into asm goto.
21389 2021-05-04 Richard Biener <rguenther@suse.de>
21391 PR tree-optimization/100398
21392 * tree-ssa-dse.c (pass_dse::execute): Preserve control
21395 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21397 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
21399 2021-05-04 Alexandre Oliva <oliva@adacore.com>
21401 * builtins.c (try_store_by_multiple_pieces): New.
21402 (expand_builtin_memset_args): Use it. If target_char_cast
21403 fails, proceed as for non-constant val. Pass len's ctz to...
21404 * expr.c (clear_storage_hints): ... this. Try store by
21405 multiple pieces after setmem.
21406 (clear_storage): Adjust.
21407 * expr.h (clear_storage_hints): Likewise.
21408 (try_store_by_multiple_pieces): Declare.
21409 * passes.def: Replace the last copy_prop with ccp.
21411 2021-05-03 Tom de Vries <tdevries@suse.de>
21414 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
21417 2021-05-03 Richard Biener <rguenther@suse.de>
21419 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
21421 2021-05-03 Richard Biener <rguenther@suse.de>
21423 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
21424 (dse_dom_walker): Remove.
21425 (dse_dom_walker::dse_optimize_stmt): Rename...
21426 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
21427 (dse_dom_walker::before_dom_children): Inline ...
21428 (pass_dse::execute): ... here. Perform a reverse program
21431 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
21434 * configure: Regenerated.
21436 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
21439 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
21440 (s390_md_asm_adjust): Handle hard registers.
21442 2021-05-03 Jakub Jelinek <jakub@redhat.com>
21444 PR tree-optimization/100382
21445 * tree-ssa-dse.c: Include tree-eh.h.
21446 (dse_dom_walker::before_dom_children): Don't remove stmts if
21447 stmt_unremovable_because_of_non_call_eh_p is true.
21449 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
21451 * varasm.c (compute_reloc_for_var): Split out from...
21452 (get_variable_section): Use it.
21453 * output.h (compute_reloc_for_var): Declare.
21454 * config/rs6000/rs6000-protos.h
21455 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
21457 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
21458 Don't append storage mapping class to symbol.
21459 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
21460 Don't convert TLS BSS to common.
21461 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
21462 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
21464 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
21466 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
21467 If align is 0 from TLS class, use the same rules as varasm.c
21468 If not common, switch to BSS section manually.
21469 If common, emit appropriate comm or lcomm directive.
21470 (rs6000_xcoff_encode_section_info): Add logic to append all
21471 storage mapping classes.
21472 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
21473 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
21474 rs6000_xcoff_asm_output_aligned_decl_common.
21475 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
21476 rs6000_xcoff_asm_output_aligned_decl_common.
21477 (ASM_OUTPUT_TLS_COMMON): Use
21478 rs6000_xcoff_asm_output_aligned_decl_common.
21480 2021-05-02 Jakub Jelinek <jakub@redhat.com>
21483 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
21484 as first argument of pseudo_node_t constructors.
21486 2021-05-02 Jakub Jelinek <jakub@redhat.com>
21489 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
21491 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
21493 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
21494 (gt_pch_nx (int_range<1> *&)): New.
21495 (gt_ggc_mx (int_range<1> *&)): New.
21496 * value-range.h (class irange): Add GTY support for
21499 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
21501 * doc/options.texi (Negative): Change either or to both and.
21503 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21505 * config/aarch64/aarch64-simd-builtins.def: Add
21506 float_ml[as][q]_laneq builtin generator macros.
21507 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
21508 (aarch64_float_mla_laneq<mode>): Define.
21509 (aarch64_float_mls_laneq<mode>): Define.
21510 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
21511 instead of GCC vector extensions.
21512 (vmlaq_laneq_f32): Likewise.
21513 (vmls_laneq_f32): Likewise.
21514 (vmlsq_laneq_f32): Likewise.
21516 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21518 * config/aarch64/aarch64-simd-builtins.def: Add
21519 float_ml[as]_lane builtin generator macros.
21520 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
21522 (mul_lane<mode>3): This, and re-order arguments.
21523 (aarch64_float_mla_lane<mode>): Define.
21524 (aarch64_float_mls_lane<mode>): Define.
21525 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
21526 instead of GCC vector extensions.
21527 (vmlaq_lane_f32): Likewise.
21528 (vmls_lane_f32): Likewise.
21529 (vmlsq_lane_f32): Likewise.
21531 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21533 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
21534 builtin generator macros.
21535 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
21537 (aarch64_float_mls<mode>): Define.
21538 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
21539 instead of relying on GCC vector extensions.
21540 (vmla_f64): Likewise.
21541 (vmlaq_f32): Likewise.
21542 (vmlaq_f64): Likewise.
21543 (vmls_f32): Likewise.
21544 (vmls_f64): Likewise.
21545 (vmlsq_f32): Likewise.
21546 (vmlsq_f64): Likewise.
21547 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
21549 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
21551 * config/aarch64/aarch64-simd-builtins.def: Add
21552 float_ml[as]_n_builtin generator macros.
21553 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
21555 (mul_n<mode>3): This, and re-order arguments.
21556 (aarch64_float_mla_n<mode>): Define.
21557 (aarch64_float_mls_n<mode>): Define.
21558 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
21559 instead of inline asm.
21560 (vmlaq_n_f32): Likewise.
21561 (vmls_n_f32): Likewise.
21562 (vmlsq_n_f32): Likewise.
21564 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
21566 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
21567 builtin generator macros.
21568 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
21569 (aarch64_pmull_hiv16qi_insn): Define.
21570 (aarch64_pmull_hiv16qi): Define.
21571 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
21572 instead of inline asm.
21573 (vmull_p8): Likewise.
21575 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
21577 * config/avr/avr.md: Adjust peepholes to match and
21578 generate parallels with clobber of REG_CC.
21579 (mov<mode>_insn): Rename to mov<mode>_insn_split.
21580 (*mov<mode>_insn): Rename to mov<mode>_insn.
21582 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
21584 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
21585 for VAR_DECLs if -fdata-sections enabled.
21587 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
21589 PR bootstrap/100327
21590 * config/rs6000/rs6000.c
21591 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
21592 (rs6000_libgcc_floating_mode_supported_p): New target hook.
21594 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
21596 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
21598 (class back_threader_registry): ...to here...
21599 (class back_threader_profitability): ...and here...
21600 (thread_jumps::thread_through_all_blocks): Remove argument.
21601 (back_threader_registry::back_threader_registry): New.
21602 (back_threader_registry::~back_threader_registry): New.
21603 (back_threader_registry::thread_through_all_blocks): New.
21604 (thread_jumps::profitable_jump_thread_path): Move from here...
21605 (back_threader_profitability::profitable_path_p): ...to here.
21606 (thread_jumps::find_taken_edge): New.
21607 (thread_jumps::convert_and_register_current_path): Move...
21608 (back_threader_registry::register_path): ...to here.
21609 (thread_jumps::register_jump_thread_path_if_profitable): Move...
21610 (thread_jumps::maybe_register_path): ...to here.
21611 (thread_jumps::handle_phi): Call find_taken_edge and
21612 maybe_register_path.
21613 (thread_jumps::handle_assignment): Same.
21614 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
21615 tree argument to handle_phi and handle_assignment.
21616 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
21617 set of m_speed_p and m_max_threaded_paths.
21618 (pass_thread_jumps::execute): Remove second argument from
21619 find_jump_threads_backwards.
21620 (pass_early_thread_jumps::execute): Same.
21622 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
21624 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
21625 (class dom_opt_dom_walker): Initialize some class variables.
21626 (pass_dominator::execute): Pass evrp_range_analyzer and
21627 dom_jump_threader_simplifier to dom_opt_dom_walker.
21628 Adjust for some functions moving into classes.
21629 (simplify_stmt_for_jump_threading): Adjust and move to...
21630 (jump_threader_simplifier::simplify): ...here.
21631 (dom_opt_dom_walker::before_dom_children): Adjust for
21632 m_evrp_range_analyzer.
21633 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
21634 (test_for_singularity): Place in dom_opt_dom_walker class.
21635 (dom_opt_dom_walker::optimize_stmt): The argument
21636 evrp_range_analyzer is now a class field.
21637 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
21638 (thread_jumps::thread_through_all_blocks): New.
21639 (thread_jumps::convert_and_register_current_path): Use m_registry.
21640 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
21641 being in the threader class.
21642 (pass_early_thread_jumps::execute): Same.
21643 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
21644 (jump_threader::jump_threader): ...here.
21645 (threadedge_finalize_values): Move...
21646 (jump_threader::~jump_threader): ...here.
21647 (jump_threader::remove_jump_threads_including): New.
21648 (jump_threader::thread_through_all_blocks): New.
21649 (record_temporary_equivalences_from_phis): Move...
21650 (jump_threader::record_temporary_equivalences_from_phis): ...here.
21651 (record_temporary_equivalences_from_stmts_at_dest): Move...
21652 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
21654 (simplify_control_stmt_condition_1): Move to jump_threader class.
21655 (simplify_control_stmt_condition): Move...
21656 (jump_threader::simplify_control_stmt_condition): ...here.
21657 (thread_around_empty_blocks): Move...
21658 (jump_threader::thread_around_empty_blocks): ...here.
21659 (thread_through_normal_block): Move...
21660 (jump_threader::thread_through_normal_block): ...here.
21661 (thread_across_edge): Move...
21662 (jump_threader::thread_across_edge): ...here.
21663 (thread_outgoing_edges): Move...
21664 (jump_threader::thread_outgoing_edges): ...here.
21665 * tree-ssa-threadedge.h: Move externally facing functings...
21666 (class jump_threader): ...here...
21667 (class jump_threader_simplifier): ...and here.
21668 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
21669 (jump_thread_path_allocator::jump_thread_path_allocator): New.
21670 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
21671 (jump_thread_path_allocator::allocate_thread_edge): New.
21672 (jump_thread_path_allocator::allocate_thread_path): New.
21673 (jump_thread_path_registry::jump_thread_path_registry): New.
21674 (jump_thread_path_registry::~jump_thread_path_registry): New.
21675 (jump_thread_path_registry::allocate_thread_edge): New.
21676 (jump_thread_path_registry::allocate_thread_path): New.
21677 (dump_jump_thread_path): Make extern.
21678 (debug (const vec<jump_thread_edge *> &path)): New.
21679 (struct removed_edges): Move to tree-ssa-threadupdate.h.
21680 (struct thread_stats_d): Remove.
21681 (remove_ctrl_stmt_and_useless_edges): Make static.
21682 (lookup_redirection_data): Move...
21683 (jump_thread_path_registry::lookup_redirection_data): ...here.
21684 (ssa_redirect_edges): Make static.
21685 (thread_block_1): Move...
21686 (jump_thread_path_registry::thread_block_1): ...here.
21687 (thread_block): Move...
21688 (jump_thread_path_registry::thread_block): ...here.
21689 (thread_through_loop_header): Move...
21690 (jump_thread_path_registry::thread_through_loop_header): ...here.
21691 (mark_threaded_blocks): Move...
21692 (jump_thread_path_registry::mark_threaded_blocks): ...here.
21693 (debug_path): Move...
21694 (jump_thread_path_registry::debug_path): ...here.
21695 (debug_all_paths): Move...
21696 (jump_thread_path_registry::dump): ..here.
21697 (rewire_first_differing_edge): Move...
21698 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
21699 (adjust_paths_after_duplication): Move...
21700 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
21701 (duplicate_thread_path): Move...
21702 (jump_thread_path_registry::duplicate_thread_path): ..here.
21703 (remove_jump_threads_including): Move...
21704 (jump_thread_path_registry::remove_jump_threads_including): ...here.
21705 (thread_through_all_blocks): Move to...
21706 (jump_thread_path_registry::thread_through_all_blocks): ...here.
21707 (delete_jump_thread_path): Remove.
21708 (register_jump_thread): Move...
21709 (jump_thread_path_registry::register_jump_thread): ...here.
21710 * tree-ssa-threadupdate.h: Move externally facing functions...
21711 (class jump_thread_path_allocator): ...here...
21712 (class jump_thread_path_registry): ...and here.
21713 (thread_through_all_blocks): Remove.
21714 (struct removed_edges): New.
21715 (register_jump_thread): Remove.
21716 (remove_jump_threads_including): Remove.
21717 (delete_jump_thread_path): Remove.
21718 (remove_ctrl_stmt_and_useless_edges): Remove.
21719 (free_dom_edge_info): New prototype.
21720 * tree-vrp.c: Remove x_vr_values hack.
21721 (class vrp_jump_threader_simplifier): New.
21722 (vrp_jump_threader_simplifier::simplify): New.
21723 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
21724 Remove m_dummy_cond.
21725 Instantiate m_simplifier and m_threader.
21726 (vrp_jump_threader::thread_through_all_blocks): New.
21727 (vrp_jump_threader::simplify_stmt): Remove.
21728 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
21729 Remove x_vr_values hack.
21730 (execute_vrp): Adjust for thread_through_all_blocks being in a
21733 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
21735 * genflags.c (gen_insn): Print failed expansion string.
21737 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
21739 * expr.c (alignment_for_piecewise_move): Call mode_for_size
21740 without limit to MAX_FIXED_MODE_SIZE.
21742 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
21744 PR middle-end/90773
21745 * builtins.c (builtin_memset_gen_str): Don't use return from
21746 simplify_gen_subreg.
21748 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
21751 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
21752 (*addsi3_carry_zext_0r): Ditto.
21753 (*sub<mode>3_carry_0): Ditto.
21754 (*subsi3_carry_zext_0r): Ditto.
21755 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
21757 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
21758 Also consider ix86_carry_flag_unset_operator to calculate
21759 the cost of adc/sbb insn.
21761 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
21763 PR rtl-optimization/100225
21764 PR rtl-optimization/84878
21765 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
21766 where we have an instruction which touches (writes) any hard
21767 register from df->regular_block_artificial_uses set.
21768 Allow not-single-set instruction only right before basic block
21771 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
21773 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
21775 2021-04-30 LevyHsu <admin@levyhsu.com>
21777 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
21778 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
21779 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
21780 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
21782 2021-04-29 Alexandre Oliva <oliva@adacore.com>
21784 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
21786 2021-04-29 Alexandre Oliva <oliva@adacore.com>
21788 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
21789 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
21790 * config/i386/i386.c: Adjust.
21791 * config/i386/i386.md: Adjust.
21792 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
21793 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21794 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21795 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21796 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21797 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21798 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21799 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21800 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21801 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
21802 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
21804 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
21806 * config/i386/i386-expand.c (ix86_expand_int_compare):
21807 Swap operands of GTU and LEU comparison to emit carry flag comparison.
21808 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
21809 predicate to allow more combine opportunities with memory operands.
21810 (*sub<mode>3_carry_0): Ditto.
21812 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
21814 PR rtl-optimization/100303
21815 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
21816 boolean that indicates whether the use will only be used in
21817 debug instructions. Treat it in the same way that existing
21818 cross-EBB debug references would be handled if so.
21819 (function_info::make_uses_available): Likewise.
21820 * rtl-ssa/functions.h (function_info::make_uses_available): Update
21821 prototype accordingly.
21822 (function_info::make_uses_available): Likewise.
21823 * fwprop.c (try_fwprop_subst): Update call accordingly.
21825 2021-04-29 Jeff Law <jlaw@tachyum.com>
21827 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
21830 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
21831 Richard Biener <rguenther@suse.de>
21834 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
21835 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
21836 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
21837 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
21838 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
21839 to PURE_ARGS category.
21840 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
21841 Handle PURE_ARGS category.
21842 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
21844 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
21846 * configure.ac: Check for the presence of sys/locking.h header and
21847 for whether _LK_LOCK is supported by _locking.
21848 * configure: Regenerate.
21849 * config.in: Likewise.
21850 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
21851 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
21852 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
21854 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
21856 * config/i386/predicates.md (fcmov_comparison_operator):
21857 Do not check for trivial FP comparison operator.
21858 <case GEU, case LTU>: Allow CCGZmode.
21859 <case GTU, case LEU>: Do not allow CCCmode.
21860 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
21861 (ix86_carry_flag_operator): Match only LTU and UNLT code.
21862 Do not check for trivial FP comparison operator. Allow CCGZmode.
21864 2021-04-29 Tom de Vries <tdevries@suse.de>
21866 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
21867 fd->loop.step by either step or orig_step.
21869 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
21871 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
21872 (load_got_register): Do the PIC dance here.
21873 (sparc_legitimize_tls_address): Simplify.
21874 (sparc_emit_probe_stack_range): Likewise.
21875 (sparc32_initialize_trampoline): Likewise.
21876 (sparc64_initialize_trampoline): Likewise.
21877 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
21878 (probe_stack_range<P:mode>): Likewise.
21879 (flush<P:mode>): Likewise.
21880 (tgd_hi22<P:mode>): Likewise.
21881 (tgd_lo10<P:mode>): Likewise.
21882 (tgd_add<P:mode>): Likewise.
21883 (tgd_call<P:mode>): Likewise.
21884 (tldm_hi22<P:mode>): Likewise.
21885 (tldm_lo10<P:mode>): Likewise.
21886 (tldm_add<P:mode>): Likewise.
21887 (tldm_call<P:mode>): Likewise.
21888 (tldo_hix22<P:mode>): Likewise.
21889 (tldo_lox10<P:mode>): Likewise.
21890 (tldo_add<P:mode>): Likewise.
21891 (tie_hi22<P:mode>): Likewise.
21892 (tie_lo10<P:mode>): Likewise.
21893 (tie_add<P:mode>): Likewise.
21894 (tle_hix22<P:mode>): Likewise.
21895 (tle_lox10<P:mode>): Likewise.
21896 (stack_protect_setsi): Rename to...
21897 (stack_protect_set32): ...this.
21898 (stack_protect_setdi): Rename to...
21899 (stack_protect_set64): ...this.
21900 (stack_protect_set): Adjust calls to above.
21901 (stack_protect_testsi): Rename to...
21902 (stack_protect_test32): ...this.
21903 (stack_protect_testdi): Rename to...
21904 (stack_protect_test64): ...this.
21905 (stack_protect_test): Adjust calls to above.
21907 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
21909 PR middle-end/90773
21910 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
21911 (builtin_strncpy_read_str): Likewise.
21912 (builtin_memset_read_str): Add an argument for the previous RTL
21913 information and generate the new RTL from the previous RTL info.
21914 (builtin_memset_gen_str): Likewise.
21915 * builtins.h (builtin_strncpy_read_str): Update the prototype.
21916 (builtin_memset_read_str): Likewise.
21917 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
21918 returns true, round up size and alignment to the widest integer
21919 mode for maximum size.
21920 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
21921 and pass it to m_constfn.
21922 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
21923 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
21924 initialize m_push. Initialize m_overlap_op_by_pieces with
21925 targetm.overlap_op_by_pieces_p ().
21926 (op_by_pieces_d::run): Pass the previous RTL information to
21927 pieces_addr::adjust and generate overlapping operations if
21928 m_overlap_op_by_pieces is true.
21930 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
21932 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
21934 (can_store_by_pieces): Use by_pieces_constfn on constfun.
21935 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
21936 for op_by_pieces_d change.
21937 (clear_by_pieces_1): Add a dummy argument.
21938 (clear_by_pieces): Updated for op_by_pieces_d change.
21939 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
21940 (string_cst_read_str): Add a dummy argument.
21941 * expr.h (by_pieces_constfn): Add a dummy argument.
21942 (by_pieces_prev): New.
21943 * target.def (overlap_op_by_pieces_p): New target hook.
21944 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
21945 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
21946 * doc/tm.texi: Regenerated.
21948 2021-04-29 Richard Biener <rguenther@suse.de>
21950 PR tree-optimization/100253
21951 * tree-vect-stmts.c (vectorizable_load): Do not assume
21952 element alignment when DR_MISALIGNMENT is -1.
21953 (vectorizable_store): Likewise.
21955 2021-04-29 Jakub Jelinek <jakub@redhat.com>
21958 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
21959 absu_hwi instead of abs_hwi.
21961 2021-04-29 Richard Biener <rguenther@suse.de>
21963 PR middle-end/38474
21964 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
21965 forwarding when indirect forwarding through ESCAPED
21968 2021-04-29 Tom de Vries <tdevries@suse.de>
21971 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
21972 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
21973 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
21974 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
21976 2021-04-29 Richard Biener <rguenther@suse.de>
21978 PR tree-optimization/99912
21979 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
21980 (dse_dom_walker::todo): Likewise.
21981 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
21983 (dse_dom_walker::before_dom_children): Remove trivially
21984 dead SSA defs and schedule CFG cleanup if we removed all
21986 (pass_dse::execute): Get TODO as computed by the DOM walker
21987 and return it. Wipe dominator info earlier.
21989 2021-04-29 Richard Biener <rguenther@suse.de>
21992 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
21993 Track blocks to cleanup EH in new m_need_eh_cleanup.
21994 (ipcp_modif_dom_walker::cleanup_eh): New.
21995 (ipcp_transform_function): Release dominator info before
21998 2021-04-29 Martin Sebor <msebor@redhat.com>
22000 PR middle-end/100250
22001 * attribs.c (attr_access::array_as_string): Avoid dereferencing
22002 a pointer when it's null.
22004 2021-04-29 Martin Sebor <msebor@redhat.com>
22006 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
22007 * ipa-free-lang-data.cc: New file.
22008 * tree.c: Move pass free_lang_data to file above.
22009 (build_array_type_1): Declare extern.
22010 * tree.h (build_array_type_1): Declare.
22012 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22014 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
22015 make consistent with updated RTL pattern.
22016 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
22017 Implement using ss_truncate and us_truncate rather than
22019 * config/aarch64/iterators.md: Remove redundant unspecs and
22020 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
22022 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22024 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
22025 attributes consistent with those defined in arm_neon.h.
22027 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22029 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
22030 attributes consistent with those defined in arm_neon.h.
22032 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22034 * config/aarch64/aarch64-simd-builtins.def: Add
22035 float_trunc_rodd builtin generator macros.
22036 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
22038 (aarch64_float_trunc_rodd_lo_v2sf): Define.
22039 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
22040 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
22041 (aarch64_float_trunc_rodd_hi_v4sf): Define.
22042 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
22043 instead of inline asm.
22044 (vcvtx_high_f32_f64): Likewise.
22045 (vcvtxd_f32_f64): Likewise.
22046 * config/aarch64/iterators.md: Add FCVTXN unspec.
22048 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22050 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
22052 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
22054 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
22055 instead of inline asm.
22056 (vqtbx1_u8): Likewise.
22057 (vqtbx1_p8): Likewise.
22058 (vqtbx1q_s8): Likewise.
22059 (vqtbx1q_u8): Likewise.
22060 (vqtbx1q_p8): Likewise.
22061 (vtbx2_s8): Likewise.
22062 (vtbx2_u8): Likewise.
22063 (vtbx2_p8): Likewise.
22065 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22067 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
22069 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
22070 instead of inline asm.
22071 (vqtbl1_s8): Likewise.
22072 (vqtbl1_u8): Likewise.
22073 (vqtbl1q_p8): Likewise.
22074 (vqtbl1q_s8): Likewise.
22075 (vqtbl1q_u8): Likewise.
22076 (vtbl1_s8): Likewise.
22077 (vtbl1_u8): Likewise.
22078 (vtbl1_p8): Likewise.
22079 (vtbl2_s8): Likewise.
22080 (vtbl2_u8): Likewise.
22081 (vtbl2_p8): Likewise.
22083 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22085 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
22086 ssri_n buitin generator macro.
22087 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
22088 instead of inline asm.
22089 (vsri_n_p16): Likewise.
22090 (vsri_n_p64): Likewise.
22091 (vsriq_n_p8): Likewise.
22092 (vsriq_n_p16): Likewise.
22093 (vsriq_n_p64): Likewise.
22095 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22097 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
22098 iterator for polynomial ssli_n builtin generator macro.
22099 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
22100 instead of inline asm.
22101 (vsli_n_p16): Likewise.
22102 (vsliq_n_p8): Likewise.
22103 (vsliq_n_p16): Likewise.
22104 * config/aarch64/iterators.md: Define VALLP mode iterator.
22106 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22108 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
22109 iterator to generate [su]adalp RTL builtins.
22110 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
22111 [su]adalp RTL pattern.
22112 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
22113 instead of inline asm.
22114 (vpadal_u32): Likewise.
22116 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22118 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
22119 builtin generator macros.
22120 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
22122 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
22123 instead of inline asm.
22124 (vpaddl_s16): Likewise.
22125 (vpaddl_s32): Likewise.
22126 (vpaddl_u8): Likewise.
22127 (vpaddl_u16): Likewise.
22128 (vpaddl_u32): Likewise.
22129 (vpaddlq_s8): Likewise.
22130 (vpaddlq_s16): Likewise.
22131 (vpaddlq_s32): Likewise.
22132 (vpaddlq_u8): Likewise.
22133 (vpaddlq_u16): Likewise.
22134 (vpaddlq_u32): Liwewise.
22135 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
22136 appropriate attributes.
22138 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22140 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
22141 for aarch64_addp<mode> builtin macro generator.
22142 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
22143 aarch64_addp<mode> RTL pattern.
22144 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
22145 instead of inline asm.
22146 (vpaddq_s16): Likewise.
22147 (vpaddq_s32): Likewise.
22148 (vpaddq_s64): Likewise.
22149 (vpaddq_u8): Likewise.
22150 (vpaddq_u16): Likewise.
22151 (vpaddq_u32): Likewise.
22152 (vpaddq_u64): Likewise.
22154 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22156 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
22157 builtin generator macros.
22158 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
22160 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
22161 instead of inline asm.
22162 (vqdmulh_n_s32): Likewise.
22163 (vqdmulhq_n_s16): Likewise.
22164 (vqdmulhq_n_s32): Likewise.
22165 (vqrdmulh_n_s16): Likewise.
22166 (vqrdmulh_n_s32): Likewise.
22167 (vqrdmulhq_n_s16): Likewise.
22168 (vqrdmulhq_n_s32): Likewise.
22170 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
22172 * doc/install.texi (--enable-offload-defaulted): Document.
22174 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
22176 * config/avr/avr-dimode.md: Turn existing patterns into
22177 define_insn_and_split style patterns where the splitter
22178 adds a clobber of the condition code register. Drop "cc"
22179 attribute. Add new patterns to match output of
22181 * config/avr/avr-fixed.md: Likewise.
22182 * config/avr/avr.c (cc_reg_rtx): New.
22183 (avr_parallel_insn_from_insns): Adjust insn count
22184 for removal of set of cc0.
22185 (avr_is_casesi_sequence): Likewise.
22186 (avr_casei_sequence_check_operands): Likewise.
22187 (avr_optimize_casesi): Likewise. Also insert
22188 new insns after jump_insn.
22189 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
22190 for removal of set of cc0.
22191 (avr_init_expanders): Initialize cc_reg_rtx.
22192 (avr_regno_reg_class): Handle REG_CC.
22193 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
22194 (avr_notice_update_cc): Remove function.
22195 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
22196 (compare_condition): Adjust for PARALLEL with
22198 (out_shift_with_cnt): Likewise.
22199 (ashlhi3_out): Likewise.
22200 (ashrhi3_out): Likewise.
22201 (lshrhi3_out): Likewise.
22202 (avr_class_max_nregs): Return single reg for REG_CC.
22203 (avr_compare_pattern): Check for REG_CC instead
22205 (avr_reorg_remove_redundant_compare): Likewise.
22206 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
22207 (avr_hard_regno_nregs): Return single reg for REG_CC.
22208 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
22209 (avr_md_asm_adjust): Clobber REG_CC.
22210 (TARGET_HARD_REGNO_NREGS): Define.
22211 (TARGET_CLASS_MAX_NREGS): Define.
22212 (TARGET_MD_ASM_ADJUST): Define.
22213 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
22215 (enum reg_class): Add CC_REG class.
22216 (NOTICE_UPDATE_CC): Remove.
22217 (CC_OVERFLOW_UNUSABLE): Remove.
22218 (CC_NO_CARRY): Remove.
22219 * config/avr/avr.md: Turn existing patterns into
22220 define_insn_and_split style patterns where the splitter
22221 adds a clobber of the condition code register. Drop "cc"
22222 attribute. Add new patterns to match output of
22224 (sez): Remove unused pattern.
22226 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
22229 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
22232 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
22235 * config/aarch64/constraints.md (Utq): Require the address to
22236 be valid for both the element mode and for V2DImode.
22238 2021-04-28 Jakub Jelinek <jakub@redhat.com>
22239 Tobias Burnus <tobias@codesourcery.com>
22241 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
22242 * gcc.c (process_command): New variable.
22243 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
22244 set it if -foffload is defaulted.
22245 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
22246 (compile_offload_image): If OFFLOAD_DEFAULTED and
22247 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
22248 if corresponding mkoffload can't be found.
22249 (compile_images_for_offload_targets): Likewise. Free and clear
22250 offload_names if no valid offload is found.
22251 * config.in: Regenerate.
22252 * configure: Regenerate.
22254 2021-04-28 Richard Biener <rguenther@suse.de>
22256 PR tree-optimization/100292
22257 * tree-vect-generic.c (expand_vector_condition): Do not fold
22260 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22262 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
22263 * config/rs6000/aix64.opt (m64): New.
22266 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22268 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
22269 (index_term_p): Handle ASHIFT too.
22271 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22273 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
22274 (sync_lock_test_and_set<mode>): Adjust accordingly.
22275 (sync_lock_release<mode>): Likewise.
22277 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22279 * config/vax/vax-protos.h (adjacent_operands_p): Remove
22281 * config/vax/vax.c (adjacent_operands_p): Remove.
22283 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
22285 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
22286 through to the non-conditional execution case if getting the
22287 condition for conditional execution has failed.
22289 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
22291 PR middle-end/100284
22292 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
22293 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
22294 than asserting on it.
22296 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22298 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
22299 with TARGET_AIX_OS.
22301 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22304 * calls.c (precompute_register_parameters): Additionally test
22305 targetm.precompute_tls_p to pre-compute argument.
22306 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
22307 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
22308 * target.def (precompute_tls_p): New.
22309 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
22310 * doc/tm.texi: Regenerated.
22312 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22315 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
22316 back to HOST_WIDE_INT.
22318 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
22321 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
22322 memory alignment for the outer mode.
22324 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
22326 PR middle-end/90773
22327 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
22328 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
22330 2021-04-27 Alex Coplan <alex.coplan@arm.com>
22333 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
22334 with negative immediates: ensure we expand cbranchsi4_scratch
22335 correctly and ensure we satisfy its constraints.
22336 * config/arm/sync.md
22337 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
22338 attempt to tie two output operands together with constraints;
22339 collapse two alternatives.
22340 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
22341 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
22343 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22346 * config/aarch64/predicates.md (aarch64_sub_immediate,
22347 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
22348 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
22349 * config/aarch64/aarch64.c (aarch64_print_operand,
22350 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
22352 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22354 PR tree-optimization/100239
22355 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
22356 permutations with all indices from the first zero element as vec_shl.
22358 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22360 PR rtl-optimization/100254
22361 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
22362 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
22364 2021-04-27 Richard Biener <rguenther@suse.de>
22366 PR tree-optimization/99912
22367 * passes.def: Add comment about new TODO_remove_unused_locals.
22368 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
22371 2021-04-27 Richard Biener <rguenther@suse.de>
22373 PR tree-optimization/99912
22374 * passes.def (pass_all_optimizations): Add pass_dse before
22375 the first pass_dce, move the first pass_dse before the
22376 pass_dce following pass_pre.
22378 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22380 PR tree-optimization/95527
22381 * generic-match-head.c: Include tm.h.
22382 * gimple-match-head.c: Include tm.h.
22383 * match.pd (CLZ == INTEGER_CST): Don't use
22384 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
22385 if clz == CFN_CLZ. Add missing val declaration.
22386 (CTZ cmp CST): New simplifications.
22388 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22390 PR tree-optimization/96696
22391 * expr.c (expand_expr_divmod): New function.
22392 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
22393 divisions. Formatting fixes.
22394 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
22397 2021-04-27 Martin Jambor <mjambor@suse.cz>
22400 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
22401 If removing a call statement LHS SSA name, release it.
22403 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
22406 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
22407 is valid before including it in the mask.
22409 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
22412 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
22415 2021-04-27 Richard Biener <rguenther@suse.de>
22417 PR tree-optimization/100051
22418 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
22419 disambiguator based on access size vs. decl size.
22421 2021-04-27 Richard Biener <rguenther@suse.de>
22423 PR tree-optimization/100278
22424 * tree-ssa-pre.c (compute_avail): Give up when we cannot
22425 adjust TBAA beacuse of mismatching bases.
22427 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22430 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
22431 For any_rotate define_insn_split and following splitters, use
22432 SWI iterator instead of SWI48.
22434 2021-04-27 Richard Biener <rguenther@suse.de>
22436 PR tree-optimization/99776
22437 * match.pd (bit_field_ref (ctor)): Relax element extract
22438 type compatibility checks.
22440 2021-04-27 Cui,Lili <lili.cui@intel.com>
22442 * common/config/i386/i386-common.c (processor_names):
22443 Sync processor_names with processor_type.
22444 * config/i386/i386-options.c (processor_cost_table):
22445 Sync processor_cost_table with processor_type.
22447 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22449 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
22450 (irange::set): Call irange_set_1bit_anti_range for handling all
22451 1-bit ranges. Fall through on ~[MIN,MAX].
22453 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22455 * value-range.cc (irange::legacy_num_pairs): Remove.
22456 (irange::invert): Change gcc_assert to gcc_checking_assert.
22457 * value-range.h (irange::num_pairs): Adjust for a cached
22458 num_pairs(). Also, rename all gcc_assert's to
22459 gcc_checking_assert's.
22461 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22463 * value-range.cc (irange::operator=): Set m_kind.
22464 (irange::copy_to_legacy): Handle varying and undefined sources
22465 as a legacy copy since they can be easily copied.
22466 (irange::irange_set): Set m_kind.
22467 (irange::irange_set_anti_range): Same.
22468 (irange::set): Rename normalize_min_max to normalize_kind.
22469 (irange::verify_range): Adjust for multi-ranges having the
22471 (irange::irange_union): Set m_kind.
22472 (irange::irange_intersect): Same.
22473 (irange::invert): Same.
22474 * value-range.h (irange::kind): Always return m_kind.
22475 (irange::varying_p): Rename to...
22476 (irange::varying_comptaible_p): ...this.
22477 (irange::undefined_p): Only look at m_kind.
22478 (irange::irange): Always set VR_UNDEFINED if applicable.
22479 (irange::set_undefined): Always set VR_UNDEFINED.
22480 (irange::set_varying): Always set m_kind to VR_VARYING.
22481 (irange::normalize_min_max): Rename to...
22482 (irange::normalize_kind): ...this.
22484 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22486 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
22487 Adjust for constant_p including varying_p.
22488 * tree-vrp.c (vrp_prop::finalize): Same.
22489 (determine_value_range): Same.
22490 * vr-values.c (vr_values::range_of_expr): Same.
22491 * value-range.cc (irange::symbolic_p): Do not check varying_p.
22492 (irange::constant_p): Same.
22494 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
22496 * value-range.cc (irange::legacy_lower_bound): Replace
22497 !undefined_p check with num_ranges > 0.
22498 (irange::legacy_upper_bound): Same.
22499 * value-range.h (irange::type): Same.
22500 (irange::lower_bound): Same.
22501 (irange::upper_bound): Same.
22503 2021-04-26 Richard Biener <rguenther@suse.de>
22505 PR tree-optimization/99956
22506 * gimple-loop-interchange.cc (compute_access_stride):
22507 Try instantiating the access in a shallower loop nest
22508 if instantiating failed.
22509 (compute_access_strides): Pass adjustable loop_nest
22510 to compute_access_stride.
22512 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
22514 * doc/sourcebuild.texi (arm_cmse_hw): Document.
22516 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22518 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
22520 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
22521 Nathan Sidwell <nathan@codesourcery.com>
22522 Tom de Vries <vries@codesourcery.com>
22523 Julian Brown <julian@codesourcery.com>
22524 Kwok Cheung Yeung <kcy@codesourcery.com>
22526 * omp-offload.c (oacc_validate_dims): Implement
22527 '-Wopenacc-parallelism'.
22528 * doc/invoke.texi (-Wopenacc-parallelism): Document.
22530 2021-04-26 Richard Biener <rguenther@suse.de>
22532 * tree-cfg.h (gimplify_build1): Remove.
22533 (gimplify_build2): Likewise.
22534 (gimplify_build3): Likewise.
22535 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
22536 (gimplify_build2): Likewise.
22537 (gimplify_build3): Likewise.
22538 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
22540 (gimplify_build2): Likewise.
22541 (gimplify_build3): Likewise.
22542 (tree_vec_extract): Use resimplify with following SSA edges.
22543 (expand_vector_parallel): Avoid passing NULL size/bitpos
22544 to tree_vec_extract.
22545 * expr.c (store_constructor): Deal with zero-element CTORs.
22546 * match.pd (bit_field_ref <vector CTOR>): Make sure to
22547 produce vector constants when possible.
22549 2021-04-26 Richard Biener <rguenther@suse.de>
22551 * tree-complex.c: Include gimple-fold.h.
22552 (expand_complex_addition): Use gimple_build.
22553 (expand_complex_multiplication_components): Likewise.
22554 (expand_complex_multiplication): Likewise.
22555 (expand_complex_div_straight): Likewise.
22556 (expand_complex_div_wide): Likewise.
22557 (expand_complex_division): Likewise.
22558 (expand_complex_conjugate): Likewise.
22559 (expand_complex_comparison): Likewise.
22561 2021-04-26 Richard Biener <rguenther@suse.de>
22563 * tree-ssa-phiopt.c (two_value_replacement): Remove use
22564 of legacy gimplify_buildN API.
22566 2021-04-26 Richard Biener <rguenther@suse.de>
22568 PR tree-optimization/99473
22569 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
22572 2021-04-26 Richard Biener <rguenther@suse.de>
22574 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
22575 Use replace_call_with_value.
22577 2021-04-26 Richard Biener <rguenther@suse.de>
22579 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
22580 (update_gimple_call): Likewise.
22581 (update_call_from_tree): Likewise.
22582 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
22583 (valid_gimple_call_p): Likewise.
22584 (move_ssa_defining_stmt_for_defs): Likewise.
22585 (finish_update_gimple_call): Likewise.
22586 (update_gimple_call): Likewise.
22587 (update_call_from_tree): Likewise.
22588 (propagate_tree_value_into_stmt): Use replace_call_with_value.
22589 * gimple-fold.h (update_gimple_call): Declare.
22590 * gimple-fold.c (valid_gimple_rhs_p): Move here from
22591 tree-ssa-propagate.c.
22592 (update_gimple_call): Likewise.
22593 (valid_gimple_call_p): Likewise.
22594 (finish_update_gimple_call): Likewise, and simplify.
22595 (gimplify_and_update_call_from_tree): Implement
22596 update_call_from_tree functionality, avoid excessive
22597 push/pop_gimplify_context.
22598 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
22599 (gimple_fold_call): Likewise.
22600 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
22601 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
22602 (pass_fold_builtins::execute): Likewise.
22603 (optimize_stack_restore): Use replace_call_with_value.
22604 * tree-cfg.c (fold_loop_internal_call): Likewise.
22605 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
22606 only gimplify_and_update_call_from_tree.
22607 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
22608 (handle_builtin_strchr): Likewise.
22609 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
22611 2021-04-26 Jakub Jelinek <jakub@redhat.com>
22614 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
22615 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
22618 2021-04-25 liuhongt <hongtao.liu@intel.com>
22621 * config/i386/i386-builtin.def (BDESC): Change the icode of
22622 the following builtins to CODE_FOR_nothing.
22623 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
22624 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
22625 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
22626 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
22627 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
22628 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
22629 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
22630 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
22631 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
22632 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
22633 (sse2_eq<mode>3): Ditto.
22634 (sse4_1_eqv2di3): Ditto.
22635 (sse2_gt<mode>3): Rename to ..
22636 (*sse2_gt<mode>3): .. this.
22638 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
22641 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
22644 * config/darwin.c (darwin_binds_local_p): Assume that any
22645 public symbol might be interposed for PIC code. Update function
22646 header comment to reflect current Darwin capability.
22648 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
22651 * config/darwin.c (darwin_binds_local_p): Assume that any
22652 public symbol might be interposed for PIC code. Update function
22653 header comment to reflect current Darwin capability.
22655 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
22657 * doc/sourcebuild.texi: Document no-opts and any-opts target
22660 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
22662 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
22664 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
22666 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
22668 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
22671 * config/i386/i386-options.c (ix86_option_override_internal):
22672 Error out when -m96bit-long-double is used with 64bit targets.
22673 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
22675 2021-04-23 Martin Liska <mliska@suse.cz>
22677 * lto-wrapper.c: Remove FIXME about usage of
22678 hardware_concurrency. The function is not on par with
22681 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
22684 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
22685 Copy operand 3 to operand 4. Use sse_reg_operand
22686 as operand 3 predicate.
22687 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
22688 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
22689 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
22690 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
22691 Copy operand 1 to operand 0.
22692 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
22693 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
22694 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
22696 2021-04-23 Alex Coplan <alex.coplan@arm.com>
22698 PR rtl-optimization/100230
22699 * early-remat.c (early_remat::sort_candidates): Use delete[]
22700 instead of delete for array allocated with new[].
22702 2021-04-23 Richard Biener <rguenther@suse.de>
22704 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
22705 (capture_info::capture_info): Likewise.
22706 (capture_info::walk_match): Likewise.
22707 (expr::gen_transform): Likewise.
22708 (dt_simplify::gen_1): Likewise.
22709 * gimple-match-head.c (maybe_resimplify_conditional_op):
22710 Remove VEC_COND_EXPR special-casing.
22711 (gimple_simplify): Likewise.
22712 * gimple.c (gimple_could_trap_p_1): Adjust.
22713 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
22714 to participate in PRE.
22716 2021-04-23 Richard Biener <rguenther@suse.de>
22718 * cfganal.c (connect_infinite_loops_to_exit): First call
22719 add_noreturn_fake_exit_edges.
22720 * ipa-sra.c (process_scan_results): Do not call the now redundant
22721 add_noreturn_fake_exit_edges.
22722 * predict.c (tree_estimate_probability): Likewise.
22723 (rebuild_frequencies): Likewise.
22724 * store-motion.c (one_store_motion_pass): Likewise.
22726 2021-04-23 Richard Biener <rguenther@suse.de>
22728 PR tree-optimization/100222
22729 * predict.c (pass_profile::execute): Remove redundant call to
22730 mark_irreducible_loops.
22731 (report_predictor_hitrates): Likewise.
22733 2021-04-23 Richard Biener <rguenther@suse.de>
22735 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
22736 valid_gimple_rhs_p by instead gimplifying to one.
22738 2021-04-23 Richard Biener <rguenther@suse.de>
22740 PR tree-optimization/99971
22741 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
22742 Always use TBAA for loads.
22744 2021-04-23 liuhongt <hongtao.liu@intel.com>
22747 * config/i386/i386-options.c (ix86_option_override_internal):
22748 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
22749 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
22750 by target attribute.
22752 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
22754 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
22756 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
22758 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
22760 * config.gcc (powerpc-ibm-aix6.*): Remove.
22761 * config/rs6000/aix61.h: Delete.
22763 2021-04-22 Martin Liska <mliska@suse.cz>
22765 PR testsuite/100159
22766 PR testsuite/100192
22767 * builtins.c (expand_builtin): Fix typos and missing comments.
22768 * dwarf2out.c (gen_subprogram_die): Likewise.
22769 (gen_struct_or_union_type_die): Likewise.
22771 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
22774 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
22775 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
22777 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
22779 * config/i386/darwin.h (TARGET_64BIT): Remove definition
22780 based on TARGET_ISA_64BIT.
22781 (TARGET_64BIT_P): Remove definition based on
22782 TARGET_ISA_64BIT_P().
22784 2021-04-21 Martin Liska <mliska@suse.cz>
22787 2021-04-21 Martin Liska <mliska@suse.cz>
22789 * lto-wrapper.c (cpuset_popcount): Remove.
22790 (init_num_threads): Remove and use hardware_concurrency.
22792 2021-04-21 Martin Liska <mliska@suse.cz>
22795 * main.c (main): Call toplev::finalize in CHECKING_P mode.
22796 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
22797 when incremental LTO linking happens.
22799 2021-04-21 Martin Liska <mliska@suse.cz>
22801 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
22802 makeserver cannot be detected, then use -flto=N fallback.
22804 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
22806 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
22807 default to yes for aarch64-linux-gnu.
22808 * configure: Regenerate.
22810 2021-04-21 Martin Liska <mliska@suse.cz>
22812 * lto-wrapper.c (cpuset_popcount): Remove.
22813 (init_num_threads): Remove and use hardware_concurrency.
22815 2021-04-21 Martin Liska <mliska@suse.cz>
22817 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
22818 which remains to be '(... || 0)' and clang complains about it.
22819 * dwarf2out.c (AT_vms_delta): Declare conditionally.
22820 (add_AT_vms_delta): Likewise.
22821 * tree.c (fld_simplified_type): Use rather more common pattern
22822 for disabling of something (#if 0).
22823 (get_tree_code_name): Likewise.
22824 (verify_type_variant): Likewise.
22826 2021-04-21 Martin Liska <mliska@suse.cz>
22828 * config/i386/i386-expand.c (decide_alignment): Use newly named
22829 macro TARGET_CPU_P.
22830 * config/i386/i386.c (ix86_decompose_address): Likewise.
22831 (ix86_address_cost): Likewise.
22832 (ix86_lea_outperforms): Likewise.
22833 (ix86_avoid_lea_for_addr): Likewise.
22834 (ix86_add_stmt_cost): Likewise.
22835 * config/i386/i386.h (TARGET_*): Remove.
22836 (TARGET_CPU_P): New macro.
22837 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
22838 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
22839 (swap_top_of_ready_list): Likewise.
22840 (ix86_atom_sched_reorder): Likewise.
22841 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
22842 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
22844 2021-04-21 Martin Liska <mliska@suse.cz>
22846 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
22848 (SET_TARGET_NO_SAHF): Likewise.
22849 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
22850 (SET_TARGET_PREFETCH_SSE): Likewise.
22851 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
22852 (SET_TARGET_NO_TUNE): Likewise.
22853 (TARGET_EXPLICIT_NO_80387_P): Likewise.
22854 (SET_TARGET_NO_80387): Likewise.
22856 * config/i386/i386.h (TARGET_*): Remove.
22857 * opth-gen.awk: Generate new used macros.
22859 2021-04-21 Martin Liska <mliska@suse.cz>
22861 * config/i386/i386.h (PTA_*): Remove.
22862 (enum pta_flag): New.
22863 (DEF_PTA): Generate PTA_* values from i386-isa.def.
22864 * config/i386/i386-isa.def: New file.
22866 2021-04-21 Alex Coplan <alex.coplan@arm.com>
22869 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
22870 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
22871 jump table targets.
22873 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
22875 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
22876 x86_64-*-* targets.
22877 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
22879 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
22880 (ix86_handle_option): Handle -mmwait.
22881 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
22882 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
22883 __builtin_ia32_monitor and __builtin_ia32_mwait.
22884 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
22885 (ix86_valid_target_attribute_inner_p): Likewise.
22886 (ix86_option_override_internal): Enable mwait/monitor
22887 instructions for -msse3.
22888 * config/i386/i386.h (TARGET_MWAIT): New.
22889 (TARGET_MWAIT_P): Likewise.
22890 * config/i386/i386.opt: Add -mmwait.
22891 * config/i386/mwaitintrin.h: New file.
22892 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
22893 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
22895 (@sse3_monitor_<mode>): Likewise.
22896 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
22897 * doc/extend.texi: Document mwait target attribute.
22898 * doc/invoke.texi: Document -mmwait.
22900 2021-04-21 Martin Liska <mliska@suse.cz>
22902 * config/i386/i386-options.c (DEF_ENUM): Remove it.
22903 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
22904 * config/i386/stringop.def (DEF_ENUM): Likewise.
22906 2021-04-21 Martin Liska <mliska@suse.cz>
22908 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
22909 of print_generic_expr.
22911 2021-04-21 Jakub Jelinek <jakub@redhat.com>
22913 PR rtl-optimization/100148
22914 * cprop.c (constprop_register): Use next_nondebug_insn instead of
22917 2021-04-21 Martin Liska <mliska@suse.cz>
22920 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
22921 free_dominance_info calls.
22923 2021-04-21 Richard Biener <rguenther@suse.de>
22925 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
22926 parameter (and assume it to be false).
22927 (fold_gimple_assign): Adjust, remove all callers of
22928 maybe_fold_reference calling it with is_lhs true.
22929 (gimple_fold_call): Likewise.
22930 (fold_stmt_1): Likewise.
22932 2021-04-21 Richard Biener <rguenther@suse.de>
22934 * fold-const.c (pedantic_non_lvalue_loc): Remove.
22935 (fold_binary_loc): Adjust.
22936 (fold_ternary_loc): Likewise.
22938 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
22940 PR middle-end/100130
22941 * varasm.c (get_block_for_decl): Make sure that any use of the
22942 retain attribute matches the section's retain flag.
22943 (switch_to_section): Check for retain mismatches even when
22944 changing sections, but do not warn if the given decl is the
22945 section's named.decl.
22946 (output_object_block): Pass the first decl in the block (if any)
22947 to switch_to_section.
22949 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
22951 * config/i386/i386-c.c (ix86_target_macros_internal): Define
22952 __CRC32__ for -mcrc32.
22953 * config/i386/i386-options.c (ix86_option_override_internal):
22954 Enable crc32 instruction for -msse4.2.
22955 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
22957 (sse4_2_crc32di): Likewise.
22958 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
22961 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
22964 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
22967 2021-04-20 Martin Liska <mliska@suse.cz>
22969 * doc/invoke.texi: Fix typo.
22970 * params.opt: Likewise.
22972 2021-04-20 Martin Liska <mliska@suse.cz>
22974 * doc/invoke.texi: Document new param.
22976 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
22978 PR tree-optimization/100081
22979 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
22980 rather than gori_compute_cache.
22981 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
22982 (range_def_chain::m_logical_depth): New member.
22983 (range_def_chain::range_def_chain): Initialize m_logical_depth.
22984 (range_def_chain::get_def_chain): Don't build defchains through more
22985 than LOGICAL_LIMIT logical expressions.
22986 * params.opt (param_ranger_logical_depth): New.
22988 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
22991 * config/arm/arm.c (arm_configure_build_target): Do not strip
22992 extended FPU/SIMD feature bits from the target ISA when -mfpu
22993 is specified (partial revert of r11-8168).
22995 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
22997 * params.opt (-param=openacc-kernels=): Add.
22998 * omp-oacc-kernels-decompose.cc
22999 (pass_omp_oacc_kernels_decompose::gate): Use it.
23000 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
23001 (--param): ... here, 'openacc-kernels'.
23003 2021-04-19 Martin Liska <mliska@suse.cz>
23006 * gengtype.c (finish_root_table): Align function arguments
23007 in between declaration and definition.
23009 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
23011 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
23012 frames larger than the SEH maximum frame size.
23014 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
23016 PR rtl-optimization/99927
23017 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
23018 is dead, just drop it.
23020 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
23023 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
23024 * doc/tm.texi: Regenerate.
23025 * doc/tm.texi.in (D language and ABI): Add @hook for
23026 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
23028 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
23030 * config/darwin-d.c (darwin_d_handle_target_object_format): New
23032 (darwin_d_register_target_info): New function.
23033 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23034 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
23036 (dragonfly_d_register_target_info): New function.
23037 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23038 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
23040 (freebsd_d_register_target_info): New function.
23041 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23042 * config/glibc-d.c (glibc_d_handle_target_object_format): New
23044 (glibc_d_register_target_info): New function.
23045 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23046 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
23048 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
23049 as handler for objectFormat key.
23050 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
23052 (winnt_d_register_target_info): New function.
23053 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23054 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
23056 (netbsd_d_register_target_info): New function.
23057 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23058 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
23060 (openbsd_d_register_target_info): New function.
23061 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23062 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
23063 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
23064 handler for objectFormat key.
23065 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
23067 (rs6000_d_register_target_info): Add
23068 rs6000_d_handle_target_object_format as handler for objectFormat key.
23069 * config/sol2-d.c (solaris_d_handle_target_object_format): New
23071 (solaris_d_register_target_info): New function.
23072 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23074 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23077 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
23078 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
23080 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
23081 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
23082 the old and new alignment after applying MIN/MAX to it is different.
23084 2021-04-16 Tamar Christina <tamar.christina@arm.com>
23087 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
23088 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
23090 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
23092 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
23094 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
23095 this section and its subsections.
23097 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23100 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
23101 define_insn patterns.
23103 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
23105 PR rtl-optimization/98689
23106 * reg-notes.def (UNTYPED_CALL): New note.
23107 * combine.c (distribute_notes): Handle it.
23108 * emit-rtl.c (try_split): Likewise.
23109 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
23110 that calls with the note implicitly set all return value registers.
23111 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
23114 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
23116 PR rtl-optimization/99596
23117 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
23118 register accesses for const calls. Assume that pure functions
23119 can only read from global registers. Ignore cases in which
23120 the stack pointer has been marked global.
23122 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23125 * tree-vect-loop.c (vect_transform_loop): Don't remove just
23126 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
23127 them by their last argument.
23129 2021-04-15 Martin Liska <mliska@suse.cz>
23131 * doc/invoke.texi: Other params don't use it, remove it.
23133 2021-04-15 Richard Biener <rguenther@suse.de>
23135 * gimple-builder.h: Add deprecation note.
23137 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
23140 * attribs.h (restrict_type_identity_attributes_to): Declare.
23141 * attribs.c (restrict_type_identity_attributes_to): New function.
23143 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
23146 * attribs.h (affects_type_identity_attributes): Declare.
23147 * attribs.c (remove_attributes_matching): New function.
23148 (affects_type_identity_attributes): Likewise.
23150 2021-04-15 Jakub Jelinek <jakub@redhat.com>
23153 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
23154 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
23155 ZERO_EXTEND, SIGN_EXTEND or AND.
23157 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23159 PR rtl-optimization/99929
23160 * rtl.h (same_vector_encodings_p): New function.
23161 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
23162 * cselib.c (rtx_equal_for_cselib_1): Likewise.
23163 * jump.c (rtx_renumbered_equal_p): Likewise.
23164 * lra-constraints.c (operands_match_p): Likewise.
23165 * reload.c (operands_match_p): Likewise.
23166 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
23168 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23170 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
23171 more information about variable-length CONST_VECTORs.
23173 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
23175 PR rtl-optimization/100066
23176 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
23177 ordered modes when choosing splitting mode for hard reg.
23179 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23182 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
23184 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
23186 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
23188 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
23189 for mask operand types.
23190 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
23191 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
23193 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
23195 * config/s390/s390.c (s390_const_operand_ok): Check the new
23196 operand types and generate a list of valid values.
23198 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23200 * doc/tm.texi: Regenerate.
23201 * doc/tm.texi.in (D language and ABI): Add @hook for
23202 TARGET_D_REGISTER_OS_TARGET_INFO.
23204 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23206 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
23208 (aarch64_d_register_target_info): New function.
23209 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
23211 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
23213 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
23214 (arm_d_register_target_info): New function.
23215 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
23216 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23217 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
23218 (ix86_d_register_target_info): New function.
23219 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
23220 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23221 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
23222 (mips_d_register_target_info): New function.
23223 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
23224 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23225 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
23226 (pa_d_register_target_info): New function.
23227 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
23228 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23229 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
23231 (riscv_d_register_target_info): New function.
23232 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
23233 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23234 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
23236 (rs6000_d_register_target_info): New function.
23237 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
23239 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23240 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
23241 (s390_d_register_target_info): New function.
23242 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
23243 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23244 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
23246 (sparc_d_register_target_info): New function.
23247 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
23248 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23249 * doc/tm.texi: Regenerate.
23250 * doc/tm.texi.in (D language and ABI): Add @hook for
23251 TARGET_D_REGISTER_CPU_TARGET_INFO.
23253 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23255 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
23256 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
23257 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
23258 * doc/tm.texi: Regenerate.
23259 * doc/tm.texi.in (D language and ABI): Add @hook for
23260 TARGET_D_HAS_STDCALL_CONVENTION.
23262 2021-04-14 Richard Biener <rguenther@suse.de>
23264 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
23265 VEC_COND_EXPRs have a gimple_val condition.
23266 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
23267 can no longer have a GENERIC condition.
23269 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
23272 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
23273 from the isa_delta when -mfpu has been used.
23274 (arm_options_perform_arch_sanity_checks): It's the architecture that
23275 lacks an FPU not the processor.
23277 2021-04-13 Richard Biener <rguenther@suse.de>
23279 PR tree-optimization/100053
23280 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
23281 not use optimistic dominance queries for backedges to validate
23283 (dominated_by_p_w_unex): Add parameter to ignore executable
23284 state on backedges.
23285 (rpo_elim::eliminate_avail): Adjust.
23287 2021-04-13 Jakub Jelinek <jakub@redhat.com>
23290 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
23291 *aarch64_bfxilsi_extrdi): New define_insn patterns.
23293 2021-04-13 Jakub Jelinek <jakub@redhat.com>
23296 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
23297 outermode, return NULL if the result doesn't encode back to the
23298 original byte sequence.
23299 (simplify_gen_subreg): Don't create SUBREGs from constants to
23300 MODE_COMPOSITE_P outermode.
23302 2021-04-12 Jakub Jelinek <jakub@redhat.com>
23304 PR rtl-optimization/99905
23305 * combine.c (expand_compound_operation): If pos + len > modewidth,
23306 perform the right shift by pos in inner_mode and then convert to mode,
23307 instead of trying to simplify a shift of rtx with inner_mode by pos
23308 as if it was a shift in mode.
23310 2021-04-12 Jakub Jelinek <jakub@redhat.com>
23313 * combine.c (simplify_and_const_int_1): Don't optimize varop
23314 away if it has side-effects.
23316 2021-04-12 Martin Liska <mliska@suse.cz>
23318 * doc/extend.texi: Escape @smallexample content.
23320 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
23322 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
23323 alternative in order to load a DFP zero.
23325 2021-04-12 Martin Liska <mliska@suse.cz>
23327 * doc/extend.texi: Be more precise in documentation
23328 of symver attribute.
23330 2021-04-12 Martin Liska <mliska@suse.cz>
23333 * gimplify.c (gimplify_expr): Right now, we unpoison all
23334 variables before a goto <dest>. We should not do it if we are
23337 2021-04-12 Cui,Lili <lili.cui@intel.com>
23339 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
23341 * common/config/i386/i386-common.c (processor_names): Add
23343 (processor_alias_table): Add rocketlake.
23344 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
23345 INTEL_COREI7_ROCKETLAKE.
23346 * config.gcc: Add -march=rocketlake.
23347 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23349 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
23350 (processor_cost_table): Add rocketlake cost.
23351 * config/i386/i386.h (ix86_size_cost) : Define
23353 (processor_type) : Add PROCESSOR_ROCKETLAKE.
23354 (PTA_ROCKETLAKE): Ditto.
23355 * doc/extend.texi: Add rocketlake.
23356 * doc/invoke.texi: Add rocketlake.
23358 2021-04-12 Cui,Lili <lili.cui@intel.com>
23360 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
23361 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
23362 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
23363 * doc/invoke.texi: Change alderlake ISA list.
23365 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
23367 PR middle-end/98088
23368 * omp-expand.c (expand_oacc_collapse_init): Update condition in
23371 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
23374 * config/i386/serializeintrin.h (_serialize): Defined as macro.
23376 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23379 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
23380 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
23382 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23384 PR middle-end/99989
23385 * gimple-ssa-warn-alloca.c
23386 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
23387 0 with integer precision unconditionally.
23389 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23391 PR rtl-optimization/98601
23392 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
23393 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
23394 unaligned_mems handle VOIDmode like BLKmode.
23396 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
23399 * tree.c (free_lang_data_in_decl): Do not release body of
23400 declare_variant_alt.
23402 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
23404 * config/aarch64/aarch64.c (aarch64_option_restore): If the
23405 architecture was specified explicitly and the tuning wasn't,
23406 tune for the architecture rather than the configured default CPU.
23408 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
23410 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
23411 as the temporary register.
23413 2021-04-09 Martin Liska <mliska@suse.cz>
23415 * doc/extend.texi: Move non-target attributes on the top level.
23417 2021-04-09 Martin Liska <mliska@suse.cz>
23419 * doc/invoke.texi: Document minimum and maximum value of the
23420 argument for both supported compression algorithms.
23422 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
23424 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
23425 TLS BSS before TLS data.
23426 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
23428 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
23430 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
23432 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
23434 * match.pd: Extend vec_cond folds to handle shifts.
23436 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
23438 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
23441 2021-04-08 Alex Coplan <alex.coplan@arm.com>
23444 * config/arm/iterators.md (MVE_vecs): New.
23445 (V_elem): Also handle V2DF.
23446 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
23447 (*mve_vdup<mode>): ... this. Remove second alternative since
23448 vec_duplicate of const_int is not canonical RTL, and we don't
23449 want to match symbol_refs.
23450 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
23452 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
23454 * fold-const.c (fold_single_bit_test): Fix typo.
23455 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
23458 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
23460 PR tree-optimization/97513
23461 * tree-vect-slp.c (vect_add_slp_permutation): New function,
23463 (vectorizable_slp_permutation): ...here. Detect cases in which
23464 all VEC_PERM_EXPRs are guaranteed to have the same stepped
23465 permute vector and only generate one permute vector for that case.
23466 Extend that case to handle variable-length vectors.
23468 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
23470 PR tree-optimization/99873
23471 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
23472 (vect_build_slp_instance): Don't split store groups that could
23473 use IFN_STORE_LANES.
23475 2021-04-07 Jakub Jelinek <jakub@redhat.com>
23478 * varasm.c (output_constant_pool_contents): Don't strip name encoding
23479 from XSTR (desc->sym, 0) or from label before passing those to
23482 2021-04-07 Richard Biener <rguenther@suse.de>
23484 PR tree-optimization/99954
23485 * tree-loop-distribution.c: Include tree-affine.h.
23486 (generate_memcpy_builtin): Try using tree-affine to prove
23488 (loop_distribution::classify_builtin_ldst): Always classify
23491 2021-04-07 Richard Biener <rguenther@suse.de>
23493 PR tree-optimization/99947
23494 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
23495 steps vector to avoid pushing elements from the reallocated
23498 2021-04-07 Richard Biener <rguenther@suse.de>
23500 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
23501 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
23503 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
23505 (debug_vn_reference_ops): New.
23507 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
23509 PR tree-optimization/98736
23510 * tree-loop-distribution.c
23511 * (loop_distribution::bb_top_order_init):
23512 Compute RPO with programing order preserved by calling function
23513 rev_post_order_and_mark_dfs_back_seme.
23515 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
23518 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
23519 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
23521 (process_bb_lives): Don't update biggest mode of hard reg for
23522 implicit in multi-register group. Use the new functions for
23523 updating dead_set and unused_set by register notes.
23525 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
23527 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
23530 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
23532 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
23533 (skylake_memset): Likewise.
23534 (skylake_cost): Change CLEAR_RATIO to 17.
23535 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
23536 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
23537 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
23539 2021-04-06 Richard Biener <rguenther@suse.de>
23541 PR tree-optimization/99880
23542 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
23543 set vectorized defs of relevant PHIs.
23545 2021-04-06 Richard Biener <rguenther@suse.de>
23547 PR tree-optimization/99924
23548 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
23549 nodes w/o scalar stmts as visited.
23551 2021-04-06 Alex Coplan <alex.coplan@arm.com>
23554 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
23555 PCS for [su]fix_optab.
23557 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
23559 * config/darwin.c (machopic_legitimize_pic_address): Check
23560 that the current pic register is one of the hard reg set
23561 before setting liveness.
23563 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
23565 * config/darwin.c (machopic_legitimize_pic_address): Fix
23566 whitespace, remove unused code.
23568 2021-04-03 Jakub Jelinek <jakub@redhat.com>
23570 PR tree-optimization/99882
23571 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
23574 2021-04-03 Jakub Jelinek <jakub@redhat.com>
23576 PR rtl-optimization/99863
23577 * dse.c (replace_read): Drop regs_live argument. Instead of
23578 regs_live, use store_insn->fixed_regs_live if non-NULL,
23579 otherwise punt if insns sequence clobbers or sets any hard
23582 2021-04-03 Jakub Jelinek <jakub@redhat.com>
23585 * targhooks.h (default_print_patchable_function_entry_1): Declare.
23586 * targhooks.c (default_print_patchable_function_entry_1): New function,
23587 copied from default_print_patchable_function_entry with an added flags
23589 (default_print_patchable_function_entry): Rewritten into a small
23590 wrapper around default_print_patchable_function_entry_1.
23591 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23593 (rs6000_print_patchable_function_entry): New function.
23595 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
23597 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
23599 2021-04-01 Jason Merrill <jason@redhat.com>
23602 * common.opt: Document v15 and v16.
23604 2021-04-01 Richard Biener <rguenther@suse.de>
23606 PR tree-optimization/99863
23607 * gimplify.c (gimplify_init_constructor): Recompute vector
23610 2021-04-01 Jakub Jelinek <jakub@redhat.com>
23612 * doc/extend.texi (symver attribute): Fix up syntax errors
23615 2021-04-01 Jakub Jelinek <jakub@redhat.com>
23617 PR tree-optimization/96573
23618 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
23619 also pointer types.
23621 2021-04-01 Richard Biener <rguenther@suse.de>
23623 PR tree-optimization/99856
23624 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
23625 precision to vector element precision.
23627 2021-04-01 Martin Jambor <mjambor@suse.cz>
23629 PR tree-optimization/97009
23630 * tree-sra.c (access_or_its_child_written): New function.
23631 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
23634 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
23637 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
23639 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
23642 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
23643 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
23644 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
23645 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
23646 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
23647 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
23649 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
23650 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
23651 (define_attr "prefixed"): Update initializer.
23653 2021-03-31 Jakub Jelinek <jakub@redhat.com>
23656 * dwarf2out.c (debug_ranges_dwo_section): New variable.
23657 (DW_RANGES_IDX_SKELETON): Define.
23658 (struct dw_ranges): Add begin_entry and end_entry members.
23659 (DEBUG_DWO_RNGLISTS_SECTION): Define.
23660 (add_ranges_num): Adjust r initializer for addition of *_entry
23662 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
23663 set idx to DW_RANGES_IDX_SKELETON.
23664 (use_distinct_base_address_for_range): New function.
23665 (index_rnglists): Don't set r->idx if it is equal to
23666 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
23667 r->end_entry for -gsplit-dwarf if those will be needed by
23669 (output_rnglists): Add DWO argument. If true, switch to
23670 debug_ranges_dwo_section rather than debug_ranges_section.
23671 Adjust l1/l2 label indexes. Only output the offset table when
23672 dwo is true and don't include in there the skeleton range
23673 entry if present. For -gsplit-dwarf, skip ranges that belong
23674 to the other rnglists section. Change return type from void
23675 to bool and return true if there are any range entries for
23676 the other section. For dwarf_split_debug_info use
23677 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
23678 entries instead of DW_RLE_start_end, DW_RLE_start_length and
23679 DW_RLE_base_address. Use use_distinct_base_address_for_range.
23680 (init_sections_and_labels): Initialize debug_ranges_dwo_section
23681 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
23682 and range_base_label indexes.
23683 (dwarf2out_finish): Call index_rnglists earlier before finalizing
23684 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
23685 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
23686 with different dwo arguments.
23687 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
23689 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23691 PR tree-optimization/98268
23692 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
23693 recompute_tree_invariant_for_addr_expr after successfully
23694 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
23696 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23698 PR tree-optimization/99726
23699 * tree-data-ref.c (create_intersect_range_checks_index): Bail
23700 out if there is more than one access function SCEV for the loop
23703 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23705 PR rtl-optimization/97141
23706 PR rtl-optimization/98726
23707 * emit-rtl.c (valid_for_const_vector_p): Return true for
23709 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
23710 poly_wide_int instead of a wide_int.
23711 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
23713 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
23714 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
23715 false for CONST_VECTORs that cannot be forced to memory.
23716 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
23717 is too complex to force to memory, build it up from individual
23720 2021-03-31 Jan Hubicka <jh@suse.cz>
23723 * cgraph.c (cgraph_node::release_body): Fix overactive check.
23725 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
23728 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
23731 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
23733 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
23734 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
23736 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
23737 "rep movsb/stosb" only for known sizes.
23738 * config/i386/i386-options.c (processor_cost_table): Use Ice
23739 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
23740 Rapids and Alder Lake.
23741 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
23742 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
23743 (icelake_memset): Likewise.
23744 (icelake_cost): Likewise.
23745 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
23748 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
23751 * config/aarch64/aarch64.c
23752 (aarch64_vectorize_preferred_vector_alignment): Query the size
23753 of the provided SVE vector; do not assume that all SVE vectors
23754 have the same size.
23756 2021-03-31 Jan Hubicka <jh@suse.cz>
23759 * cgraph.c (cgraph_node::release_body): Remove all callers and
23761 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
23762 * cgraphunit.c (cgraph_node::expand): And here.
23764 2021-03-31 Martin Liska <mliska@suse.cz>
23766 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
23767 and one negated condition.
23769 2021-03-31 Jakub Jelinek <jakub@redhat.com>
23770 Richard Sandiford <richard.sandiford@arm.com>
23773 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
23774 constraints on operands[2] and similarly 0 and rk constraints
23775 on operands[1] corresponding to that.
23777 2021-03-31 Jakub Jelinek <jakub@redhat.com>
23780 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
23781 linker doesn't support DWARF sections new in DWARF5.
23782 * config/i386/i386-options.c (ix86_option_override_internal): Default
23783 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
23785 * config.in: Regenerated.
23786 * configure: Regenerated.
23788 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23791 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
23792 available issue_info before using it.
23794 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23797 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
23800 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
23803 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
23804 (altivec_lvsl_reg_<mode>): ... this.
23805 (altivec_lvsr_reg): Change to ...
23806 (altivec_lvsr_reg_<mode>): ... this.
23807 * config/rs6000/predicates.md (vec_set_index_operand): New.
23808 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23809 Enable 32bit variable vec_insert for all TARGET_VSX.
23810 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
23811 Enable 32bit variable vec_insert for p9 and above.
23812 (rs6000_expand_vector_set_var_p8): Rename to ...
23813 (rs6000_expand_vector_set_var_p7): ... this.
23814 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
23816 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
23817 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
23818 gen_altivec_lvsr_reg_di.
23820 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
23823 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
23824 (__rdtscp): Likewise.
23826 2021-03-30 Tamar Christina <tamar.christina@arm.com>
23828 PR tree-optimization/99825
23829 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
23830 Reject non-mult 2 lanes.
23832 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
23835 * config/arm/arm.c (arm_file_start): Fix emission of
23836 Tag_ABI_VFP_args attribute.
23838 2021-03-30 Richard Biener <rguenther@suse.de>
23840 PR tree-optimization/99824
23841 * stor-layout.c (set_min_and_max_values_for_integral_type):
23842 Assert the precision is within the bounds of
23843 WIDE_INT_MAX_PRECISION.
23844 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
23845 the outermost component ref only to lower the access size
23846 and initialize that from the access type.
23848 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
23851 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
23852 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
23854 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
23856 * config/aarch64/aarch64.md
23857 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
23858 attribute to disambiguate between SIMD and FP variants of the
23861 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
23863 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
23864 (analyze_ssa_name_flags): Fix typo in comment.
23866 2021-03-29 Alex Coplan <alex.coplan@arm.com>
23869 * config/aarch64/aarch64-sve-builtins.cc
23870 (function_builder::add_function): Add placeholder_p argument, use
23871 placeholder decls if this is set.
23872 (function_builder::add_unique_function): Instead of conditionally adding
23873 direct overloads, unconditionally add either a direct overload or a
23875 (function_builder::add_overloaded_function): Set placeholder_p if we're
23876 using C++ overloads. Use the obstack for string storage instead
23877 of relying on the tree nodes.
23878 (function_builder::add_overloaded_functions): Don't return early for
23879 m_direct_overloads: we need to add placeholders.
23880 * config/aarch64/aarch64-sve-builtins.h
23881 (function_builder::add_function): Add placeholder_p argument.
23883 2021-03-29 Richard Biener <rguenther@suse.de>
23885 PR tree-optimization/99807
23886 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
23887 assert below VEC_PERM handling.
23889 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23892 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
23893 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
23894 matching const_int 0.
23895 (move_lo_quad_internal_be_<mode>): Likewise.
23896 (move_lo_quad_<mode>): Update for the above.
23897 * config/aarch64/iterators.md (VQ_2E): Delete.
23899 2021-03-29 Jakub Jelinek <jakub@redhat.com>
23901 PR tree-optimization/99777
23902 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
23903 types other than scalar integral types.
23905 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
23907 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
23908 XCOFF TLS reloc decorations.
23910 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
23912 * doc/analyzer.texi (Analyzer Internals): Update link to
23913 "A Memory Model for Static Analysis of C Programs".
23915 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
23917 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
23918 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
23920 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
23921 (rs6000_special_round_type_align): Recursively check innermost first
23924 2021-03-26 Jakub Jelinek <jakub@redhat.com>
23927 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
23928 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
23929 assignment with drap_reg active, queue reg save for hfp with offset 0
23930 and flush queued reg saves. When handling a push with rule18,
23931 defer queueing reg save for hfp and just assert the offset is 0.
23932 (scan_trace): Assert that fde->rule18 is false.
23934 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
23937 * ira-costs.c (record_reg_classes): Put case with
23938 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
23939 * ira.c (ira_setup_alts): Ditto.
23940 * lra-constraints.c (process_alt_operands): Ditto.
23941 * recog.c (asm_operand_ok): Ditto.
23942 * reload.c (find_reloads): Ditto.
23944 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
23946 * config/aarch64/aarch64-protos.h
23947 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
23948 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
23949 * config/aarch64/aarch64.c (generic_addrcost_table): Update
23950 accordingly, using the same costs as for post_modify.
23951 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
23952 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
23953 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
23954 (a64fx_addrcost_table): Likewise.
23955 (neoversev1_addrcost_table): New.
23956 (neoversev1_tunings): Use neoversev1_addrcost_table.
23957 (aarch64_address_cost): Use the new post_modify costs for CImode
23960 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
23962 * config/aarch64/aarch64.opt
23963 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
23964 * doc/invoke.texi: Document it.
23965 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
23966 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
23967 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
23968 (aarch64_vec_issue_info): New structures.
23969 (cpu_vector_cost): Write comments above the variables rather
23971 (cpu_vector_cost::issue_info): New member variable.
23972 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
23973 and tree-ssa-loop-niter.h.
23974 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
23975 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
23976 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
23977 (thunderx3t110_vector_cost): Initialize issue_info to null.
23978 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
23979 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
23980 (neoversev1_vector_cost): Use them.
23981 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
23982 (aarch64_vector_costs::saw_sve_only_op): New member variable.
23983 (aarch64_vector_costs::num_vector_iterations): Likewise.
23984 (aarch64_vector_costs::scalar_ops): Likewise.
23985 (aarch64_vector_costs::advsimd_ops): Likewise.
23986 (aarch64_vector_costs::sve_ops): Likewise.
23987 (aarch64_vector_costs::seen_loads): Likewise.
23988 (aarch64_simd_vec_costs_for_flags): New function.
23989 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
23990 Count the number of predicate operations required by SVE WHILE
23992 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
23993 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
23994 (aarch64_count_ops): Likewise.
23995 (aarch64_add_stmt_cost): Record whether see an SVE operation
23996 that cannot currently be implementing using Advanced SIMD.
23997 Record issue information about the scalar, Advanced SIMD
23998 and (where relevant) SVE versions of a loop.
23999 (aarch64_vec_op_count::dump): New function.
24000 (aarch64_sve_op_count::dump): Likewise.
24001 (aarch64_estimate_min_cycles_per_iter): Likewise.
24002 (aarch64_adjust_body_cost): If issue information is available,
24003 try to compare the issue rates of the various loop implementations
24004 and increase or decrease the vector body cost accordingly.
24006 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24008 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
24009 Assume a zero cost for induction phis.
24011 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24013 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
24015 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
24016 vector comparisons.
24018 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24020 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
24022 (aarch64_add_stmt_cost): Call it.
24024 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24026 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
24027 New tuning parameter.
24028 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
24029 (aarch64_estimated_sve_vq): New function.
24030 (aarch64_vector_costs::analyzed_vinfo): New member variable.
24031 (aarch64_vector_costs::is_loop): Likewise.
24032 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
24033 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
24034 (aarch64_record_potential_advsimd_unrolling): New function.
24035 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
24036 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
24037 aarch64_analyze_bb_vinfo on the first use of a costs structure.
24038 Detect whether we're vectorizing a loop for SVE that might be
24039 completely unrolled if it used Advanced SIMD instead.
24040 (aarch64_adjust_body_cost_for_latency): New function.
24041 (aarch64_finish_cost): Call it.
24043 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24045 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
24046 (aarch64_init_cost): New function.
24047 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
24048 the default unsigned[3].
24049 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
24050 (TARGET_VECTORIZE_INIT_COST): Override.
24051 (TARGET_VECTORIZE_FINISH_COST): Likewise.
24052 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
24054 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24056 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
24057 (neoversev1_sve_vector_cost): New cost structures.
24058 (neoversev1_vector_cost): Likewise.
24059 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
24061 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24063 * config/aarch64/aarch64-protos.h
24064 (sve_vec_cost::scatter_store_elt_cost): New member variable.
24065 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
24066 accordingly, taking the cost from the cost of a scalar_store.
24067 (a64fx_sve_vector_cost): Likewise.
24068 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
24070 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24072 * config/aarch64/aarch64-protos.h
24073 (simd_vec_cost::store_elt_extra_cost): New member variable.
24074 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24075 accordingly, using the vec_to_scalar cost for the new field.
24076 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24077 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24078 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24079 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24080 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24081 (thunderx3t110_advsimd_vector_cost): Likewise.
24082 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
24084 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24086 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
24087 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
24088 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
24089 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24090 accordingly, using zero for the new costs.
24091 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24092 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24093 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24094 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24095 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24096 (thunderx3t110_advsimd_vector_cost): Likewise.
24097 (aarch64_ld234_st234_vectors): New function.
24098 (aarch64_adjust_stmt_cost): Likewise.
24099 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
24100 the new vector costs.
24102 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24104 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
24105 derived class of simd_vec_cost. Add information about CLAST[AB]
24106 and FADDA instructions.
24107 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
24108 accordingly, using the vec_to_scalar costs for the new fields.
24109 (a64fx_sve_vector_cost): Likewise.
24110 (aarch64_reduc_type): New function.
24111 (aarch64_sve_in_loop_reduction_latency): Likewise.
24112 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
24113 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
24114 that occur in the loop body.
24115 (aarch64_add_stmt_cost): Update call accordingly.
24117 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24119 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
24121 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
24122 above the fields rather than to the right.
24123 (simd_vec_cost::reduc_i8_cost): New member variable.
24124 (simd_vec_cost::reduc_i16_cost): Likewise.
24125 (simd_vec_cost::reduc_i32_cost): Likewise.
24126 (simd_vec_cost::reduc_i64_cost): Likewise.
24127 (simd_vec_cost::reduc_f16_cost): Likewise.
24128 (simd_vec_cost::reduc_f32_cost): Likewise.
24129 (simd_vec_cost::reduc_f64_cost): Likewise.
24130 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24131 accordingly, using the vec_to_scalar_cost for the new fields.
24132 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24133 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24134 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24135 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24136 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24137 (thunderx3t110_advsimd_vector_cost): Likewise.
24138 (aarch64_use_new_vector_costs_p): New function.
24139 (aarch64_simd_vec_costs): New function, split out from...
24140 (aarch64_builtin_vectorization_cost): ...here.
24141 (aarch64_is_reduction): New function.
24142 (aarch64_detect_vector_stmt_subtype): Likewise.
24143 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
24144 using the new vector costs.
24146 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24149 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
24150 TLS declarations as public.
24152 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24154 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
24155 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
24156 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
24157 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
24158 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
24159 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
24160 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
24161 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
24162 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
24164 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24167 * config.gcc (*-*-cygwin*): Add winnt-d.o
24168 (*-*-mingw*): Likewise.
24169 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
24170 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
24171 * config/i386/t-cygming: Add winnt-d.o.
24172 * config/i386/winnt-d.c: New file.
24174 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24176 * config/freebsd-d.c: Include memmodel.h.
24178 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24181 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
24182 * config/t-openbsd: Add openbsd-d.o.
24183 * config/openbsd-d.c: New file.
24185 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
24187 PR tree-optimization/96974
24188 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
24189 with graceful exit.
24191 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24194 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24198 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
24199 always_inline in system headers.
24201 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
24203 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
24205 2021-03-25 Jakub Jelinek <jakub@redhat.com>
24208 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
24209 * fold-const.c (operand_compare::operand_equal_p): Don't compare
24210 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
24212 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24216 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
24217 always_inline in system headers.
24219 2021-03-25 Richard Biener <rguenther@suse.de>
24221 PR tree-optimization/99746
24222 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
24223 the scalar stmt as patterned. Instead set up required things
24226 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
24228 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
24231 2021-03-24 Martin Liska <mliska@suse.cz>
24234 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
24236 * config/i386/i386-options.c (ix86_option_override_internal):
24237 Add run-time assert.
24239 2021-03-24 Martin Jambor <mjambor@suse.cz>
24242 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
24243 parameters with unknown type.
24244 (ipacp_value_safe_for_type): New function.
24245 (propagate_vals_across_arith_jfunc): Verify that the constant type
24246 can be used for a type of the formal parameter.
24247 (propagate_vals_across_ancestor): Likewise.
24248 (propagate_scalar_across_jump_function): Likewise. Pass the type
24249 also to propagate_vals_across_ancestor.
24251 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
24254 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
24256 (movmisalign<mode>_mve_load): Likewise.
24258 2021-03-24 Jakub Jelinek <jakub@redhat.com>
24261 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
24262 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
24264 2021-03-24 Alexandre Oliva <oliva@adacore.com>
24266 * doc/sourcebuild.texi (sysconf): New effective target.
24268 2021-03-24 Alexandre Oliva <oliva@adacore.com>
24270 * config/i386/predicates.md (reg_or_const_vec_operand): New.
24271 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
24272 the now *-prefixed insn_and_split, turn the splitter const vec
24273 into an input for the insn, making it an ignored immediate for
24274 non-split cases, and loaded into the scratch register
24277 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
24280 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
24281 Use define_relaxed_memory_constraint for them.
24283 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
24286 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
24287 colon to the diagnostic message.
24289 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
24291 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
24293 (try_fwprop_subst_note): Use set_info instead of insn_info.
24294 (try_fwprop_subst_pattern): Likewise.
24295 (try_fwprop_subst_notes): Likewise.
24296 (try_fwprop_subst): Likewise.
24297 (forward_propagate_subreg): Likewise.
24298 (forward_propagate_and_simplify): Likewise.
24299 (forward_propagate_into): Likewise.
24300 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
24302 (set_info::single_nondebug_insn_use): Likewise.
24303 (set_info::single_phi_use): Likewise.
24304 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
24306 (set_info::single_nondebug_insn_use): Likewise.
24307 (set_info::single_phi_use): Likewise.
24309 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
24311 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
24313 2021-03-23 Jakub Jelinek <jakub@redhat.com>
24316 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
24317 expand_mult to perform an unsigned rather than a signed
24320 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
24323 * config/i386/cpuid.h (__cpuid): Add __volatile__.
24324 (__cpuid_count): Likewise.
24326 2021-03-23 Richard Biener <rguenther@suse.de>
24328 PR tree-optimization/99721
24329 * tree-vect-slp.c (vect_slp_analyze_node_operations):
24330 Make sure we can schedule the node.
24332 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24334 * config/riscv/riscv.c (riscv_subword): Take endianness into
24335 account when calculating the byte offset.
24337 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24339 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
24340 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
24341 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
24342 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
24343 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
24344 new predicate "subreg_lowpart_operator"
24346 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24348 * config/riscv/riscv.c (riscv_swap_instruction): New function
24349 to byteswap an SImode rtx containing an instruction.
24350 (riscv_trampoline_init): Byteswap the generated instructions
24353 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24355 * common/config/riscv/riscv-common.c
24356 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
24357 * config.gcc (riscv32be-*, riscv64be-*): Set
24358 TARGET_BIG_ENDIAN_DEFAULT to 1.
24359 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
24360 depending on default endianness.
24361 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
24362 * config/riscv/linux.h (LINK_SPEC): Likewise.
24363 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
24364 default endianness.
24365 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
24367 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24369 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
24370 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
24371 * config/riscv/linux.h (LINK_SPEC): Likewise.
24372 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
24374 (BYTES_BIG_ENDIAN): Handle big endian.
24375 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
24376 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
24378 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
24380 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
24382 * regcprop.c (find_oldest_value_reg): Ask target whether
24383 different mode is fine for replacement register.
24385 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
24387 PR tree-optimization/99296
24388 * value-range.cc (irange::irange_set_1bit_anti_range): New.
24389 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
24390 * value-range.h (irange::irange_set_1bit_anti_range): New.
24392 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
24395 * config/aarch64/constraints.md (UtQ): Use
24396 define_relaxed_memory_constraint for it.
24397 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
24398 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24399 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
24400 (have_relaxed_memory_constraints): New static var.
24401 (relaxed_memory_start, relaxed_memory_end): Ditto.
24402 (add_constraint): Add arg is_relaxed_memory. Check name for
24403 relaxed memory. Set up is_relaxed_memory in constraint_data and
24404 have_relaxed_memory_constraints. Adjust calls.
24405 (choose_enum_order): Process relaxed memory.
24406 (write_tm_preds_h): Ditto.
24407 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24408 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24409 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
24410 * ira-lives.c (single_reg_class): Use
24411 insn_extra_relaxed_memory_constraint.
24412 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
24413 * lra-constraints.c (valid_address_p): Use
24414 insn_extra_relaxed_memory_constraint instead of other memory
24416 (process_alt_operands): Process CT_RELAXED_MEMORY.
24417 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
24418 * recog.c (asm_operand_ok, preprocess_constraints): Process
24420 * reload.c (find_reloads): Ditto.
24421 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
24422 * stmt.c (parse_input_constraint): Use
24423 insn_extra_relaxed_memory_constraint.
24425 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
24428 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
24431 2021-03-22 Alex Coplan <alex.coplan@arm.com>
24434 * config/arm/arm-protos.h (neon_make_constant): Add generate
24435 argument to guard emitting insns, default to true.
24436 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
24437 CONST_VECTORs which neon_make_constant can't handle.
24438 (neon_vdup_constant): Add generate argument, avoid emitting
24439 insns if it's not set.
24440 (neon_make_constant): Plumb new generate argument through.
24441 * config/arm/constraints.md (Ui): New. Use it...
24442 * config/arm/mve.md (*mve_mov<mode>): ... here.
24443 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
24444 synthesize constants.
24446 2021-03-22 Richard Biener <rguenther@suse.de>
24448 * debug.h: Add deprecation warning.
24450 2021-03-22 Richard Biener <rguenther@suse.de>
24452 PR tree-optimization/99694
24453 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
24456 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
24459 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
24460 after type checking.
24462 2021-03-22 Jakub Jelinek <jakub@redhat.com>
24466 * dwarf2out.c (get_full_len): Use get_precision rather than
24468 (add_const_value_attribute): Make sure add_AT_wide argument has
24469 precision prec rather than some very wide one.
24471 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
24473 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
24474 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
24475 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
24476 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
24477 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
24478 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
24479 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
24480 *fix_trunc<mode>si2_internal): Fix empty split condition.
24481 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
24482 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
24483 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
24484 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
24486 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
24489 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
24490 Convert idx to DImode.
24491 (rs6000_expand_vector_set_var_p8): Likewise.
24493 2021-03-21 Jakub Jelinek <jakub@redhat.com>
24496 * dwarf2out.c (insert_float): Change return type from void to
24497 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
24498 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
24501 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
24504 * config/i386/i386.c (construct_container): Check cfun != NULL
24505 before accessing silent_p.
24507 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
24509 * asan.c: Fix typos in comments.
24511 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
24513 PR rtl-optimization/99680
24514 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
24515 (process_address_1): Check empty constraint before using
24518 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
24520 * config/rs6000/rs6000.c (power10_cost): New.
24521 (rs6000_option_override_internal): Set Power10 costs.
24522 (rs6000_issue_rate): Set Power10 issue rate.
24523 * config/rs6000/power10.md: Rewrite for Power10.
24525 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
24528 * lra-constraints.c (process_address_1): Don't use unknown
24529 constraint for address constraint.
24531 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
24534 * config.gcc (powerpc-*-darwin8): Delete the reference to
24535 the now removed darwin8.h.
24537 2021-03-19 Olivier Hainque <hainque@adacore.com>
24540 * config/vxworksae.h (VX_CPU_PREFIX): Define.
24542 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
24544 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
24546 2021-03-19 Tamar Christina <tamar.christina@arm.com>
24548 PR tree-optimization/99656
24549 * tree-vect-slp-patterns.c (linear_loads_p,
24550 complex_add_pattern::matches, is_eq_or_top,
24551 vect_validate_multiplication, complex_mul_pattern::matches,
24552 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
24553 * tree-vectorizer.h: (complex_load_perm_t): Removed.
24554 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
24555 complex_load_perm_t.
24557 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
24560 * config/i386/i386-options.c (ix86_init_machine_status): Set
24562 * config/i386/i386.c (init_cumulative_args): Set silent_p to
24564 (construct_container): Return early for return and argument
24565 errors if silent_p is true.
24566 * config/i386/i386.h (machine_function): Add silent_p.
24568 2021-03-19 Jakub Jelinek <jakub@redhat.com>
24571 * config/arm/constraints.md (Ds): New constraint.
24572 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
24573 constraint instead of w,Dm.
24575 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
24577 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
24580 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
24582 PR middle-end/99641
24583 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
24584 array type, do the computation of the current position in sizetype.
24586 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
24589 * lra-constraints.c (process_address_1): Use lookup_constraint
24590 only for a single constraint.
24592 2021-03-18 Martin Sebor <msebor@redhat.com>
24594 PR middle-end/99502
24595 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
24596 (inbounds_memaccess_p): ...to this. Check the ending offset of
24597 the accessed member.
24599 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
24601 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
24602 %> quote markers to error messages.
24603 (gcn_goacc_validate_dims): Likewise.
24604 (gcn_conditional_register_usage): Remove exclaimation mark from error
24606 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
24608 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
24610 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
24613 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
24614 Kito Cheng <kito.cheng@sifive.com>
24616 * config/riscv/riscv.c (riscv_block_move_straight): Change type
24617 to unsigned HOST_WIDE_INT for parameter and local variable with
24618 HOST_WIDE_INT type.
24619 (riscv_adjust_block_mem): Ditto.
24620 (riscv_block_move_loop): Ditto.
24621 (riscv_expand_block_move): Ditto.
24623 2021-03-18 Nick Clifton <nickc@redhat.com>
24625 * config/v850/v850.c (construct_restore_jr): Increase static
24627 (construct_save_jarl): Likewise.
24628 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
24630 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24632 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
24633 (aarch64_override_options_internal): Use it.
24634 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
24637 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
24639 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
24640 error message format issues.
24641 (nios2_option_override): Likewise.
24642 (nios2_expand_fpu_builtin): Likewise.
24643 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
24644 truncation warning.
24645 (nios2_expand_custom_builtin): More error message format fixes.
24646 (nios2_expand_rdwrctl_builtin): Likewise.
24647 (nios2_expand_rdprs_builtin): Likewise.
24648 (nios2_expand_eni_builtin): Likewise.
24649 (nios2_expand_builtin): Likewise.
24650 (nios2_register_custom_code): Likewise.
24651 (nios2_valid_target_attribute_rec): Likewise.
24652 (nios2_add_insn_asm): Fix uninitialized variable warning.
24654 2021-03-17 Jan Hubicka <jh@suse.cz>
24656 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
24657 of gather to match reality.
24658 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
24660 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24662 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
24663 to compare against CC_REG rather than NE.
24665 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
24668 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
24669 inline assembly statements.
24670 (ix86_print_operand): Update 'P' handling for -fno-plt.
24672 2021-03-17 Tamar Christina <tamar.christina@arm.com>
24675 * config/aarch64/aarch64.c
24676 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
24678 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
24681 * config/rs6000/predicates.md (branch_comparison_operator): Allow
24682 ordered and unordered for CCFPmode, if flag_finite_math_only.
24684 2021-03-16 Jakub Jelinek <jakub@redhat.com>
24687 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
24688 rather than ASHIFT.
24689 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
24691 2021-03-16 Martin Liska <mliska@suse.cz>
24694 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
24695 cl_optimization_compare function.
24697 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
24699 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
24702 2021-03-16 Jakub Jelinek <jakub@redhat.com>
24705 * config/i386/i386.h (struct machine_function): Add
24706 has_explicit_vzeroupper bitfield.
24707 * config/i386/i386-expand.c (ix86_expand_builtin): Set
24708 cfun->machine->has_explicit_vzeroupper when expanding
24709 IX86_BUILTIN_VZEROUPPER.
24710 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
24711 Do the mode switching only when TARGET_VZEROUPPER, expensive
24712 optimizations turned on and not optimizing for size.
24713 (pass_insert_vzeroupper::gate): Enable even when
24714 cfun->machine->has_explicit_vzeroupper is set.
24716 2021-03-16 Jakub Jelinek <jakub@redhat.com>
24719 * config/aarch64/aarch64.c
24720 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
24721 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
24722 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
24724 2021-03-15 Richard Biener <rguenther@suse.de>
24726 PR tree-optimization/98834
24727 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
24728 subsetting by truncating the access size.
24730 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
24732 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
24733 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
24736 2021-03-15 Martin Liska <mliska@suse.cz>
24738 * spellcheck.c: Add missing comma in initialization.
24740 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
24742 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
24743 alternative 2 and alternative 1 with alternative 3 using
24744 YW register constraint.
24745 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
24746 using YW register constraint.
24747 (*vec_extractv16qi_zext): Ditto.
24748 (*vec_extractv4si): Merge alternatives 4 and 5
24749 using Yw register constraint.
24750 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
24752 2021-03-13 Martin Sebor <msebor@redhat.com>
24754 PR tree-optimization/99489
24755 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
24756 is not a call statement.
24758 2021-03-13 Jakub Jelinek <jakub@redhat.com>
24760 PR tree-optimization/99544
24761 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
24762 if for vector types multiplication can't be done in type's mode.
24764 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
24767 * config/sparc/constraints.md (w): Rename to...
24768 (W): ... this and ditch previous implementation.
24769 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
24770 (*movdf_insn_sp64): Likewise.
24771 (*mov<VM64:mode>_insn_sp64): Likewise.
24772 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
24774 (atomic_compare_and_swap_leon3_1): Likewise.
24775 (*atomic_compare_and_swapdi_v8plus): Likewise.
24776 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
24777 architecture and add missing address validity check during LRA.
24779 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
24782 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
24783 occuring for assumed-size arrays in use_device_{ptr,addr}.
24785 2021-03-12 Jakub Jelinek <jakub@redhat.com>
24788 * config/i386/constraints.md (YW): New internal constraint.
24789 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
24790 (*<sse2_avx2>_<insn><mode>3<mask_name>,
24791 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
24792 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
24794 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
24795 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
24796 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
24797 into one, use Yw instead of former x,v.
24798 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
24799 the last alternative.
24800 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
24801 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
24802 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
24803 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
24804 into one, use <v_Yw> instead of former x,v.
24805 (avx2_interleave_highv32qi<mask_name>,
24806 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
24807 constraints. Add && <mask_avx512bw_condition> to condition.
24808 (avx2_interleave_lowv32qi<mask_name>,
24809 vec_interleave_lowv16qi<mask_name>,
24810 avx2_interleave_highv16hi<mask_name>,
24811 vec_interleave_highv8hi<mask_name>,
24812 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
24813 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
24814 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
24815 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
24816 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
24817 Yw instead of v in constraints.
24818 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
24819 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
24820 instead of Yv in constraints.
24821 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
24822 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
24823 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
24824 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
24825 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
24827 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
24828 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
24829 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
24830 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
24831 two, one with just x, another isa avx512vl with v.
24833 2021-03-12 Martin Liska <mliska@suse.cz>
24835 * doc/invoke.texi: Add missing param documentation.
24837 2021-03-11 David Malcolm <dmalcolm@redhat.com>
24840 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
24841 analyzer/trimmed-graph.o.
24842 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
24843 feasibility checking to reflect new implementation.
24844 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
24846 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
24848 2021-03-11 David Malcolm <dmalcolm@redhat.com>
24850 * digraph.cc (selftest::test_shortest_paths): Update
24851 shortest_paths init for new param. Add test of
24852 SPS_TO_GIVEN_TARGET.
24853 * shortest-paths.h (enum shortest_path_sense): New.
24854 (shortest_paths::shortest_paths): Add "sense" param.
24855 Update for renamings. Generalize to use "sense" param.
24856 (shortest_paths::get_shortest_path): Rename param.
24857 (shortest_paths::m_sense): New field.
24858 (shortest_paths::m_prev): Rename...
24859 (shortest_paths::m_best_edge): ...to this.
24860 (shortest_paths::get_shortest_path): Update for renamings.
24861 Conditionalize flipping of path on sense of traversal.
24863 2021-03-11 David Malcolm <dmalcolm@redhat.com>
24865 * digraph.cc (selftest::test_shortest_paths): Add test coverage
24866 for paths from B and C.
24867 * shortest-paths.h (shortest_paths::shortest_paths): Handle
24868 unreachable nodes, rather than asserting.
24870 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
24873 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
24874 xcoff_tbss_section_name.
24875 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
24876 * xcoffout.c (xcoff_tbss_section_name): Delete.
24877 * xcoffout.h (xcoff_tbss_section_name): Delete.
24879 2021-03-11 Richard Biener <rguenther@suse.de>
24881 PR tree-optimization/99523
24882 * tree-cfg.c (dump_function_to_file): Dump SSA names
24883 w/o identifier to the decls section as well, not only those
24884 without a VAR_DECL.
24886 2021-03-11 Jakub Jelinek <jakub@redhat.com>
24889 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
24890 function calls with lhs fail if the lhs don't have compatible types.
24892 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
24894 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
24895 Change FRAME_POINTER_REGNUM to correspond to a new faked
24896 register faked_fp, part of GENNONACR_REGS like faked_ap.
24897 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
24898 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
24899 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
24900 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
24901 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
24903 (CRIS_REAL_FP_REGNUM): New constant.
24904 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
24905 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
24906 (cris_initial_elimination_offset): Handle elimination changes
24907 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
24908 and add one from FRAME_POINTER_REGNUM to
24909 HARD_FRAME_POINTER_REGNUM.
24910 (cris_expand_prologue, cris_expand_epilogue): Emit code for
24911 hard_frame_pointer_rtx instead of frame_pointer_rtx.
24913 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
24916 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
24917 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
24919 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
24922 * lra-constraints.c (process_address_1): Don't check unknown
24923 constraint, use X for empty constraint.
24925 2021-03-10 Alex Coplan <alex.coplan@arm.com>
24927 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
24928 Fix typo in comment describing "is_ha" argument.
24930 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
24932 * doc/sourcebuild.texi: Document LRA target selector.
24934 2021-03-10 David Malcolm <dmalcolm@redhat.com>
24936 * doc/ux.texi: Add subsection contrasting interactive versus
24937 batch usage of GCC.
24939 2021-03-10 Joel Hutton <joel.hutton@arm.com>
24942 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
24944 (vectorizable_load): Fix gather load mask check condition.
24946 2021-03-10 Richard Biener <rguenther@suse.de>
24948 PR tree-optimization/99510
24949 * tree.c (check_aligned_type): Check that the candidate
24950 has TYPE_USER_ALIGN set instead of matching with the
24953 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
24955 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
24956 float and vector integer modes only if the mode is not larger.
24958 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
24960 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
24962 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
24964 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
24966 * ira-lives.c (single_reg_class): Ditto.
24968 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
24970 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
24971 the architecture-specific rtems.h.
24972 (aarch64-*-rtems*): Likewise.
24973 (arm*-*-rtems*): Likewise.
24974 (epiphany-*-rtems*): Likewise.
24975 (riscv*-*-rtems*): Likewise.
24977 2021-03-09 Jakub Jelinek <jakub@redhat.com>
24979 PR tree-optimization/99305
24980 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
24981 before integer_all_onesp instead of vice versa.
24983 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
24985 * common/config/arm/arm-common.c (arm_config_default): Change type
24986 of 'i' to unsigned.
24988 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
24991 * lra-constraints.c (process_address_1): Process constraint 'g'
24992 separately and digital constraints containing more one digit.
24994 2021-03-09 Nick Clifton <nickc@redhat.com>
24996 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
24997 (DWARF"_DEBUGGING_INFO): Define.
24999 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
25002 * calls.c (initialize_argument_information): When the argument
25003 is passed by reference, do not make a copy in a thunk only if
25004 the argument is already in memory. Remove redundant test for
25005 the case of callee copy.
25007 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25010 * lra-constraints.c (process_address_1): Process 0..9 constraints
25011 in process_address_1.
25013 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
25015 * config/s390/s390.c (struct s390_processor processor_table):
25016 Binutils name string must not be empty.
25018 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
25020 * config/arc/arc.c (arc_attr_type): Remove function.
25022 2021-03-09 Martin Liska <mliska@suse.cz>
25025 * config/i386/i386-options.c (ix86_option_override_internal):
25026 Set isa_flags for OPTS argument and not for the global
25029 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
25031 * config/rs6000/predicates.md (ds_form_mem_operand): Check
25034 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
25037 * config/rs6000/predicates.md (ds_form_mem_operand) New
25039 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
25040 ds_form_mem_operand in ld/lwa patterns.
25041 * config/rs6000/fusion.md: Regenerate file.
25043 2021-03-08 Martin Sebor <msebor@redhat.com>
25045 PR middle-end/98266
25046 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
25047 (array_bounds_checker::check_array_bounds): Call it.
25049 2021-03-08 Martin Sebor <msebor@redhat.com>
25051 PR middle-end/97631
25052 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
25053 (handle_builtin_stxncpy_strncat): Rename locals. Determine
25054 destination size from allocation calls. Issue a more appropriate
25056 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
25057 (handle_builtin_memset): Same.
25059 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
25062 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
25063 to ensure we do not have an Altivec style address.
25064 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
25065 an Altivec style address.
25066 (*vsx_le_perm_store_<mode>): Likewise.
25067 (splitters after *vsx_le_perm_store_<mode>): Likewise.
25068 (vsx_load_<mode>): Disable special expander if passed an Altivec
25070 (vsx_store_<mode>): Likewise.
25072 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25075 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
25076 (aarch64_simd_shift_imm_vec_hi): Likewise.
25077 (aarch64_simd_shift_imm_vec_si): Likewise.
25078 (aarch64_simd_shift_imm_vec_di): Likewise.
25079 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
25080 predicate from above.
25081 (aarch64_shrn<mode>_insn_be): Likewise.
25082 (aarch64_rshrn<mode>_insn_le): Likewise.
25083 (aarch64_rshrn<mode>_insn_be): Likewise.
25084 (aarch64_shrn2<mode>_insn_le): Likewise.
25085 (aarch64_shrn2<mode>_insn_be): Likewise.
25086 (aarch64_rshrn2<mode>_insn_le): Likewise.
25087 (aarch64_rshrn2<mode>_insn_be): Likewise.
25089 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
25092 * lra-constraints.c (skip_contraint_modifiers): New function.
25093 (process_address_1): Use it before lookup_constraint call.
25095 2021-03-08 Martin Liska <mliska@suse.cz>
25098 * config/i386/i386-options.c (ix86_option_override_internal):
25099 Enable UINTR and HRESET for -march that supports it.
25101 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
25103 * config/s390/s390.c (f_constraint_p): New function.
25104 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
25105 (TARGET_MD_ASM_ADJUST): Likewise.
25107 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
25110 * tree-nested.c (convert_local_reference_stmt): Avoid calling
25111 lookup_field_for_decl for Fortran module (= namespace context).
25113 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
25115 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
25116 comparison with arithmetic right shift.
25117 (s390_expand_vcond): No need for a force_reg anymore.
25118 s390_vec_compare will do it.
25119 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
25120 immediate operands.
25122 2021-03-07 Jakub Jelinek <jakub@redhat.com>
25125 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
25126 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
25128 * config/i386/sse.md (v_Yw): New define_mode_attr.
25129 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
25130 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
25132 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
25133 xYw in constraints.
25135 2021-03-06 Julian Brown <julian@codesourcery.com>
25137 * tree-pretty-print.c (dump_generic_node): Emit non-generic
25138 address space info for aggregates.
25140 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
25142 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
25144 2021-03-05 Jakub Jelinek <jakub@redhat.com>
25146 PR middle-end/99322
25147 * tree-cfg.c (bb_to_omp_idx): New variable.
25148 (execute_build_cfg): Release the bb_to_omp_idx vector after
25149 cleanup_tree_cfg returns.
25150 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
25151 for bb_to_omp_idx being a vec<int> instead of pointer to array
25153 (make_edges): Remove bb_to_omp_idx local variable, don't pass
25154 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
25155 vec<int> instead of pointer to array of ints and don't free/release
25157 (remove_bb): When removing a bb and placing forced label somewhere
25158 else, ensure it is put into the same OpenMP region during cfg
25159 pass if possible or to entry successor as fallback. Unregister
25160 bb from bb_to_omp_idx.
25162 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
25165 * lra-constraints.c (process_address_1): Skip decomposing address
25166 for asm insn operand with unknown constraint.
25168 2021-03-05 Martin Jambor <mjambor@suse.cz>
25171 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
25172 corresponding speculative edges if we are about to resolve
25173 sepculation. Make edge direct (and so resolve speculations) before
25174 removing it from call_site_hash.
25175 (cgraph_edge::make_direct): Relax the initial assert to allow calling
25176 the function on speculative direct edges.
25178 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
25180 PR rtl-optimization/99376
25181 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
25182 of low-order zero bits is too large, set the result to 0 directly.
25184 2021-03-04 Jakub Jelinek <jakub@redhat.com>
25186 PR middle-end/93235
25187 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
25188 SUBREG and a SUBREG to op_mode can't be created.
25190 2021-03-04 Alex Coplan <alex.coplan@arm.com>
25193 * config/aarch64/aarch64-sve-builtins.cc
25194 (function_resolver::require_vector_type): Handle error_mark_node.
25196 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
25198 * cfgexpand.c (expand_asm_loc): Pass new parameter.
25199 (expand_asm_stmt): Likewise.
25200 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
25202 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
25203 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
25204 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
25205 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
25206 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
25207 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
25208 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
25209 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
25210 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
25211 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
25212 * doc/tm.texi (md_asm_adjust): Likewise.
25213 * target.def (md_asm_adjust): Likewise.
25215 2021-03-04 Richard Biener <rguenther@suse.de>
25217 PR middle-end/97855
25218 * tree-pretty-print.c: Poison pp_printf.
25219 (dump_decl_name): Avoid use of pp_printf.
25220 (dump_block_node): Likewise.
25221 (dump_generic_node): Likewise.
25223 2021-03-04 Martin Sebor <msebor@redhat.com>
25225 PR middle-end/96963
25226 PR middle-end/94655
25227 * builtins.c (handle_array_ref): New helper.
25228 (handle_mem_ref): New helper.
25229 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
25230 into new helper functions. Correct a workaround for vectorized
25233 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
25235 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
25236 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
25237 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
25238 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
25239 attribute for Power10.
25240 * config/rs6000/mma.md (*movoo): Likewise.
25241 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
25242 (define_mode_attr bits): Add DD/TD modes.
25243 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
25244 store_conditionalpti): Update size attribute for Power10.
25246 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25249 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
25250 -Wuninitialized, -Wmaybe-uninitialized.
25251 (wide-int.o-warn): Likewise.
25253 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
25255 * common/config/arm/arm-common.c: Include configargs.h.
25256 (arm_config_default): New function.
25257 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
25258 processors that do not support Thumb. Take into account the
25259 --with-mode configuration setting for selecting the default.
25260 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
25261 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
25263 2021-03-03 Martin Liska <mliska@suse.cz>
25265 PR gcov-profile/97461
25266 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
25268 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
25271 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
25272 point back the hard frame pointer to its default location when the
25273 frame is larger than SEH_MAX_FRAME_SIZE.
25275 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25278 * config/i386/predicates.md (logic_operator): New define_predicate.
25279 * config/i386/i386.md (mov + mem using comm arith peephole2):
25280 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
25281 and the inner mode is [QH]Imode.
25283 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25286 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
25287 (new_loc_list): Clear end_entry.
25288 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
25289 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
25291 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
25292 initialize also end_entry.
25294 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25297 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
25298 partitions, if in non-layout mode after reorder_blocks also move
25299 affected blocks to ensure a single partition transition.
25301 2021-03-03 Jason Merrill <jason@redhat.com>
25304 * cgraphunit.c (process_function_and_variable_attributes): Don't
25305 warn about flatten on an alias if the target also has it.
25306 * cgraph.h (symtab_node::get_alias_target_tree): New.
25308 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
25310 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
25311 period to symbol name.
25312 (tls_get_addr_internal<mode>): Same.
25314 2021-03-02 David Malcolm <dmalcolm@redhat.com>
25317 * diagnostic-show-locus.c
25318 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
25321 2021-03-02 Martin Sebor <msebor@redhat.com>
25323 PR middle-end/99276
25324 * builtins.c (warn_for_access): Remove stray warning text.
25326 2021-03-02 Martin Sebor <msebor@redhat.com>
25328 PR middle-end/99295
25329 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
25332 2021-03-02 Jakub Jelinek <jakub@redhat.com>
25335 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
25336 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
25337 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
25338 DW_MACRO_define_strx and DW_MACRO_undef_strx.
25339 (save_macinfo_strings): Use DW_MACRO_*_str* even with
25340 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
25341 DW_MACRO_undef_strx.
25343 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
25345 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
25347 (BT_FN_V8HI_V8HI_UINT): Likewise.
25348 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
25349 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
25350 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
25351 New builtin definitions.
25352 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
25353 vector extension version.
25354 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
25355 available with current -march level.
25356 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
25357 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
25358 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
25359 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
25360 (vec_extend_to_fp32_lo): Likewise.
25361 (vec_round_from_fp32): Likewise.
25362 (vec_convert_to_fp16): Likewise.
25363 (vec_convert_from_fp16): Likewise.
25364 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
25365 (vclfnls_v8hi): Likewise.
25366 (vcrnfs_v8hi): Likewise.
25367 (vcfn_v8hi): Likewise.
25368 (vcnf_v8hi): Likewise.
25370 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
25372 * common/config/s390/s390-common.c (processor_flags_table): New entry.
25373 * config.gcc: Enable arch14 for --with-arch and --with-tune.
25374 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
25375 arch14 for unknown CPU models.
25376 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
25377 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
25378 (s390_get_sched_attrmask): Likewise.
25379 (s390_get_unit_mask): Likewise.
25380 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
25381 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
25382 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
25383 (TARGET_NNPA_P): New macro definitions.
25384 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
25385 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
25387 2021-03-02 Jakub Jelinek <jakub@redhat.com>
25389 PR middle-end/95757
25390 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
25391 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
25392 != 1 comparisons if name is lhs of a comparison.
25394 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
25398 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
25399 * config/darwin.c (darwin_should_restore_cfa_state): New.
25400 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
25401 * doc/tm.texi: Regenerated.
25402 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
25403 * dwarf2cfi.c (connect_traces): If the target requests, restore
25404 the CFA expression after a DW_CFA_restore.
25405 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
25407 2021-03-01 Martin Liska <mliska@suse.cz>
25410 * optc-save-gen.awk: Add 4 more exceptions.
25412 2021-03-01 Nathan Sidwell <nathan@acm.org>
25415 * tree.h (TYPE_ALIGN_RAW): New accessor.
25416 (TYPE_ALIGN): Use it.
25418 2021-03-01 Jan Hubicka <jh@suse.cz>
25421 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
25423 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
25426 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
25427 point the hard frame pointer to the SSE register save area instead
25428 of the general register save area. Perform only minimal adjustment
25429 for small frames if it is initially not correctly aligned.
25430 (ix86_expand_prologue): Remove early saves for a SEH target.
25431 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
25433 2021-02-28 Jakub Jelinek <jakub@redhat.com>
25436 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
25437 typo - referneced -> referenced.
25438 * tree.c (component_ref_size): Fix comment typo -
25439 refernce -> reference.
25440 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
25441 traling -> trailing.
25442 (aliasing_component_refs_p): Fix comment typos -
25443 refernce -> reference and refernece -> reference and
25444 traling -> trailing.
25445 (nonoverlapping_refs_since_match_p): Fix comment typo -
25446 referneces -> references.
25447 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
25448 referneces -> references.
25450 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
25452 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
25453 diagnostic message to avoid use of a contraction and format
25456 2021-02-27 Jakub Jelinek <jakub@redhat.com>
25459 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
25460 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
25461 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
25462 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
25465 2021-02-26 Richard Biener <rguenther@suse.de>
25467 PR middle-end/99281
25468 * expr.c (store_field): For calls with return-slot optimization
25469 and addressable return type expand the store directly.
25471 2021-02-26 Richard Biener <rguenther@suse.de>
25474 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
25476 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
25479 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
25482 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
25484 * config.gcc: Add rs6000-pcrel-opt.o.
25485 * config/rs6000/rs6000-pcrel-opt.c: New file.
25486 * config/rs6000/pcrel-opt.md: New file.
25487 * config/rs6000/predicates.md: Add d_form_memory predicate.
25488 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
25489 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
25490 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
25491 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
25492 and make_pass_pcrel_opt().
25493 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
25494 (rs6000_option_override_internal): Add pcrel-opt.
25495 (rs6000_delegitimize_address): Support pcrel-opt.
25496 (rs6000_opt_masks): Add pcrel-opt.
25497 (pcrel_opt_valid_mem_p): New function.
25498 (reg_to_non_prefixed): Make global.
25499 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
25500 (output_pcrel_opt_reloc): New function.
25501 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
25502 (pcrel_extern_addr): Set loads_extern_addr.
25503 Add include for pcrel-opt.md.
25504 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
25505 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
25508 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
25511 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
25512 If TARGET_64BIT and dest is SUBREG, we check the width, if it
25513 equal to SImode, we use SImode operation, just like what we are
25516 2021-02-26 Marek Polacek <polacek@redhat.com>
25518 * builtins.c (warn_for_access): Fix typos.
25520 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
25522 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
25523 mark in front of the immediate quantity.
25524 (<optab>_rolsi3_uxtw): Likewise.
25526 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
25529 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
25530 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
25531 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
25532 address and disable when the FPCXT is not available.
25533 (nonsecure_call_value_reg_thumb2): Likewise.
25535 2021-02-25 Nathan Sidwell <nathan@acm.org>
25538 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
25540 2021-02-25 Tamar Christina <tamar.christina@arm.com>
25542 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
25544 2021-02-25 Richard Biener <rguenther@suse.de>
25546 PR tree-optimization/99253
25547 * tree-vect-loop.c (check_reduction_path): First compute
25548 code, then verify out-of-loop uses.
25550 2021-02-25 Jakub Jelinek <jakub@redhat.com>
25553 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
25555 2021-02-25 Jakub Jelinek <jakub@redhat.com>
25557 PR tree-optimization/80635
25558 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
25559 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
25560 has mode precision.
25562 2021-02-25 Richard Biener <rguenther@suse.de>
25564 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
25565 load_map population.
25566 (vect_match_slp_patterns_2): Revert part of last change.
25567 (vect_analyze_slp): Do not interleave optimize_load_redistribution
25568 with pattern detection but do it afterwards. Dump the
25569 whole SLP graph after pattern recognition and load
25570 redistribution optimization finished.
25572 2021-02-24 Jakub Jelinek <jakub@redhat.com>
25575 * omp-low.c (struct omp_context): Add teams_nested_p and
25576 nonteams_nested_p members.
25577 (scan_omp_target): Diagnose teams nested inside of target with other
25578 directives strictly nested inside of the same target.
25579 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
25580 ctx->nonteams_nested_p as needed.
25582 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
25584 PR inline-asm/99123
25585 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
25587 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
25589 * config/cris/cris.c (cris_expand_prologue): Set
25590 current_function_static_stack_size, if flag_stack_usage_info.
25592 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
25594 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
25595 (rs6000_final_prescan_insn): Adjust.
25596 (rs6000_asm_output_opcode): Likewise.
25598 2021-02-24 Martin Sebor <msebor@redhat.com>
25600 PR middle-end/97172
25601 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
25602 from function arguments.
25604 2021-02-24 Tamar Christina <tamar.christina@arm.com>
25606 PR tree-optimization/99220
25607 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
25608 node from cache when it's about to be deleted.
25610 2021-02-24 Jakub Jelinek <jakub@redhat.com>
25612 PR tree-optimization/99225
25613 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
25614 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
25615 build_int_cst (..., 1). Formatting fixes.
25617 2021-02-24 Tamar Christina <tamar.christina@arm.com>
25619 PR tree-optimization/99149
25620 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
25622 (vect_slp_reset_pattern): Remove.
25623 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
25624 (complex_mul_pattern::build, complex_fma_pattern::build,
25625 complex_fms_pattern::build): Fix ref counts.
25626 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
25627 when node is being deleted.
25628 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
25629 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
25631 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
25633 2021-02-24 Matthias Klose <doko@ubuntu.com>
25636 2020-12-07 Matthias Klose <doko@ubuntu.com>
25638 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
25639 and ENABLE_RTL_FLAG_CHECKING.
25641 2021-02-24 Richard Biener <rguenther@suse.de>
25644 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
25646 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
25648 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
25649 (vsx_assemble_pair): ...to this.
25650 (*mma_assemble_pair): Rename from this...
25651 (*vsx_assemble_pair): ...to this.
25652 (mma_disassemble_pair): Rename from this...
25653 (vsx_disassemble_pair): ...to this.
25654 (*mma_disassemble_pair): Rename from this...
25655 (*vsx_disassemble_pair): ...to this.
25656 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
25657 BU_COMPAT): New macros.
25658 (mma_assemble_pair): Rename from this...
25659 (vsx_assemble_pair): ...to this.
25660 (mma_disassemble_pair): Rename from this...
25661 (vsx_disassemble_pair): ...to this.
25662 (mma_assemble_pair): New compatibility built-in.
25663 (mma_disassemble_pair): Likewise.
25664 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
25665 (RS6000_BUILTIN_COMPAT): Define.
25666 (bdesc_compat): New.
25667 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
25668 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
25669 and VSX_BUILTIN_ASSEMBLE_PAIR.
25670 (rs6000_init_builtins): Register compatibility built-ins.
25671 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
25672 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
25673 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
25674 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
25675 (__builtin_vsx_assemble_pair): ...to this.
25676 (__builtin_mma_disassemble_pair): Rename from this...
25677 (__builtin_vsx_disassemble_pair): ...to this.
25679 2021-02-23 Martin Liska <mliska@suse.cz>
25682 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
25683 with different alignment. That leads to an invalid red zone
25684 size allocated in runtime.
25686 2021-02-23 Jakub Jelinek <jakub@redhat.com>
25688 PR tree-optimization/99204
25689 * fold-const.c (fold_read_from_constant_string): Check that
25690 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
25692 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
25693 Kewen Lin <linkw@gcc.gnu.org>
25695 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
25696 (rotl<mode>3_insert_3): ...this.
25697 (plus_ior_xor): New code_iterator.
25698 (define_split for GPR rl*imi): New splitter.
25699 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
25700 for integer merging.
25702 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25704 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
25706 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
25707 into a register when the above is enabled.
25708 * config/aarch64/aarch64.c (neoversev1_tunings):
25709 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
25710 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
25712 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
25714 * config/cris/cris.c (cris_print_operand) <'T'>: Change
25715 valid operand from is now an addi mult-value to shift-value.
25716 * config/cris/cris.md (*addi): Change expression of scaled
25717 operand from mult to ashift.
25718 * config/cris/cris.md (*addi_reload): New insn_and_split.
25720 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
25723 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
25724 hook_bool_const_tree_hwi_hwi_const_tree_true.
25725 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
25727 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
25729 PR rtl-optimization/98791
25730 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
25731 for unordered modes.
25733 2021-02-22 Martin Liska <mliska@suse.cz>
25735 * tree-inline.c (inline_forbidden_p): Set
25736 inline_forbidden_reason.
25738 2021-02-22 Richard Biener <rguenther@suse.de>
25740 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
25743 2021-02-22 Richard Biener <rguenther@suse.de>
25745 PR tree-optimization/99165
25746 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
25747 Accumulate changed to ret.
25749 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
25752 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
25754 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
25756 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
25759 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
25761 (trunctf<DFP_ALL:mode>2): Likewise.
25762 (trunctdtf2_vr): Likewise.
25763 (trunctdtf2): Likewise.
25764 (extend<DFP_ALL:mode>tf2_vr): Likewise.
25765 (extend<DFP_ALL:mode>tf2): Likewise.
25766 (extendtftd2_vr): Likewise.
25767 (extendtftd2): Likewise.
25769 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
25771 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
25772 add memory alternative.
25773 (tf_to_fprx2): New pattern.
25775 2021-02-19 Martin Sebor <msebor@redhat.com>
25778 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
25779 (attr_access::free_lang_data): Remove a spurious test.
25781 2021-02-19 Nathan Sidwell <nathan@acm.org>
25783 * doc/invoke.texi (flang-info-module-read): Document.
25785 2021-02-19 Martin Liska <mliska@suse.cz>
25787 PR translation/99167
25788 * params.opt: Fix typo.
25790 2021-02-19 Richard Biener <rguenther@suse.de>
25792 PR middle-end/99122
25793 * tree-inline.c (inline_forbidden_p): Do not inline functions
25794 with VLA arguments or return value.
25796 2021-02-19 Jakub Jelinek <jakub@redhat.com>
25799 * config/arm/arm.md (*stack_protect_combined_set_insn,
25800 *stack_protect_combined_test_insn): If force_const_mem result
25801 is not valid general operand, force its address into the destination
25804 2021-02-19 Jakub Jelinek <jakub@redhat.com>
25807 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
25808 pad or non-local label, put FORCED_LABELs from bb b after that label
25809 rather than before it.
25811 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
25814 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
25815 expand_vector_broadcast' to emit the vec_duplicate operand.
25817 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
25819 PR rtl-optimization/96264
25820 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
25823 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
25826 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
25827 looking up the retain attribute.
25828 (resolve_unique_section): Likewise.
25829 (get_variable_section): Likewise.
25830 (switch_to_section): Likewise. Warn when a symbol without the
25831 retain attribute and a symbol with the retain attribute are
25832 placed in the section with the same name, instead of the used
25834 * doc/extend.texi: Document the "retain" attribute.
25836 2021-02-18 Nathan Sidwell <nathan@acm.org>
25839 * doc/invoke.texi (flang-info-include-translate): Document header
25842 2021-02-18 Richard Biener <rguenther@suse.de>
25844 PR middle-end/99122
25845 * ipa-fnsummary.c (analyze_function_body): Set
25846 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
25847 * tree-inline.c (insert_init_debug_bind): Pass NULL for
25848 error_mark_node values.
25849 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
25851 (setup_one_parameter): Delay force_value_to_type until when
25854 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
25856 PR tree-optimization/99142
25857 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
25859 2021-02-18 Jakub Jelinek <jakub@redhat.com>
25861 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
25862 wide_int_bitmask::wide_int_bitmask (uint64_t),
25863 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
25864 wide_int_bitmask::operator ~ () const,
25865 wide_int_bitmask::operator | (wide_int_bitmask) const,
25866 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
25868 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
25869 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
25870 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
25871 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
25872 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
25873 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
25874 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
25875 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
25876 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
25877 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
25878 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
25879 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
25880 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
25881 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
25882 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
25883 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
25884 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
25885 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
25886 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
25887 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
25888 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
25889 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
25890 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
25891 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
25893 2021-02-18 Jakub Jelinek <jakub@redhat.com>
25895 PR middle-end/99109
25896 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
25897 (build_printable_array_type): ... this. Add nelts argument. For
25898 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
25899 nelts, call build_array_type_nelts.
25900 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
25901 instead of build_zero_elt_array_type and build_array_type_nelts.
25903 2021-02-18 Jakub Jelinek <jakub@redhat.com>
25906 * config/i386/i386.c (distance_non_agu_define): Don't call
25907 extract_insn_cached here.
25908 (ix86_lea_outperforms): Save and restore recog_data around call
25909 to distance_non_agu_define and distance_agu_use.
25910 (ix86_ok_to_clobber_flags): Remove.
25911 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
25912 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
25913 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
25914 into define_insn. Move the splitting to define_peephole2 and
25915 check there using peep2_regno_dead_p if FLAGS_REG is dead.
25917 2021-02-17 Julian Brown <julian@codesourcery.com>
25919 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
25922 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
25925 * config/mips/mips.c (mips_symbol_insns): Do not use
25926 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
25928 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
25930 PR inline-asm/98096
25931 * stmt.c (resolve_operand_name_1): Take inout operands into account
25932 for access to labels by names.
25933 * doc/extend.texi: Describe counting operands for accessing labels.
25935 2021-02-16 Richard Biener <rguenther@suse.de>
25937 PR tree-optimization/38474
25938 * tree-ssa-structalias.c (variable_info::address_taken): New.
25939 (new_var_info): Initialize address_taken.
25940 (process_constraint): Set address_taken.
25941 (solve_constraints): Use the new address_taken flag rather
25942 than is_reg_var for sorting variables.
25943 (dump_constraint): Dump the variable number if the name
25946 2021-02-16 Jakub Jelinek <jakub@redhat.com>
25949 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
25950 multiply by 4096 and for inbranch by 8192.
25951 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
25952 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
25954 2021-02-15 Maya Rashish <coypu@sdf.org>
25956 * config/aarch64/aarch64.c (aarch64_init_builtins):
25957 Call SUBTARGET_INIT_BUILTINS.
25959 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
25961 PR rtl-optimization/98872
25962 * init-regs.c (initialize_uninitialized_regs): Skip initialization
25963 if CONST0_RTX is NULL.
25965 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
25967 PR rtl-optimization/98863
25968 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
25969 (function_info::build_info): Turn into a declaration, moving the
25970 definition to internals.h.
25971 (function_info::bb_walker): Declare.
25972 (function_info::create_reg_use): Likewise.
25973 (function_info::calculate_potential_phi_regs): Take a build_info
25975 (function_info::place_phis, function_info::create_ebbs): Declare.
25976 (function_info::calculate_ebb_live_in_for_debug): Likewise.
25977 (function_info::populate_backedge_phis): Delete.
25978 (function_info::start_block, function_info::end_block): Declare.
25979 (function_info::populate_phi_inputs): Delete.
25980 (function_info::m_potential_phi_regs): Move information to build_info.
25981 * rtl-ssa/internals.h: New file.
25982 (function_info::bb_phi_info): New class.
25983 (function_info::build_info): Moved from functions.h.
25984 Add a constructor and destructor.
25985 (function_info::build_info::ebb_use): Delete.
25986 (function_info::build_info::ebb_def): Likewise.
25987 (function_info::build_info::bb_live_out): Likewise.
25988 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
25989 (function_info::build_info::potential_phi_regs): Likewise.
25990 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
25991 (function_info::build_info::ebb_def_regs): Likewise.
25992 (function_info::build_info::bb_phis): Likewise.
25993 (function_info::build_info::bb_mem_live_out): Likewise.
25994 (function_info::build_info::bb_to_rpo): Likewise.
25995 (function_info::build_info::def_stack): Likewise.
25996 (function_info::build_info::old_def_stack_limit): Likewise.
25997 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
25998 Remove the regno argument. Push the previous definition onto the
25999 definition stack where necessary.
26000 * rtl-ssa/accesses.cc: Include internals.h.
26001 * rtl-ssa/changes.cc: Likewise.
26002 * rtl-ssa/blocks.cc: Likewise.
26003 (function_info::build_info::build_info): Define.
26004 (function_info::build_info::~build_info): Likewise.
26005 (function_info::bb_walker): New class.
26006 (function_info::bb_walker::bb_walker): Define.
26007 (function_info::add_live_out_use): Convert a logarithmic-complexity
26008 test into a linear one. Allow the same definition to be passed
26010 (function_info::calculate_potential_phi_regs): Moved from
26011 functions.cc. Take a build_info parameter and store the
26012 information there instead.
26013 (function_info::place_phis): New function.
26014 (function_info::add_entry_block_defs): Update call to record_reg_def.
26015 (function_info::calculate_ebb_live_in_for_debug): New function.
26016 (function_info::add_phi_nodes): Use bb_phis to decide which
26017 registers need phi nodes and initialize ebb_def_regs accordingly.
26018 Do not add degenerate phis here.
26019 (function_info::add_artificial_accesses): Use create_reg_use.
26020 Assert that all definitions are listed in the DF LR sets.
26021 Update call to record_reg_def.
26022 (function_info::record_block_live_out): Record live-out register
26023 values in the phis of successor blocks. Use the live-out set
26024 when processing the last block in an EBB, instead of always
26025 using the live-in sets of successor blocks. AND the live sets
26026 with the set of registers that have been defined in the EBB,
26027 rather than with all potential phi registers. Cope correctly
26028 with branches back to the start of the current EBB.
26029 (function_info::start_block): New function.
26030 (function_info::end_block): Likewise.
26031 (function_info::populate_phi_inputs): Likewise.
26032 (function_info::create_ebbs): Likewise.
26033 (function_info::process_all_blocks): Rewrite into a multi-phase
26035 * rtl-ssa/functions.cc: Include internals.h.
26036 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
26037 (function_info::init_function_data): Remove caller.
26038 * rtl-ssa/insns.cc: Include internals.h
26039 (function_info::create_reg_use): New function. Lazily any
26040 degenerate phis needed by the linear RPO view.
26041 (function_info::record_use): Use create_reg_use. When processing
26042 debug uses, use potential_phi_regs and test it before checking
26043 whether the register is live on entry to the current EBB. Lazily
26044 calculate ebb_live_in_for_debug.
26045 (function_info::record_call_clobbers): Update call to record_reg_def.
26046 (function_info::record_def): Likewise.
26048 2021-02-15 Martin Liska <mliska@suse.cz>
26050 * toplev.c (init_asm_output): Free output of
26051 gen_command_line_string function.
26052 (process_options): Likewise.
26054 2021-02-15 Martin Liska <mliska@suse.cz>
26056 * params.opt: Add 2 missing Param keywords.
26058 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
26060 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
26062 2021-02-15 Jakub Jelinek <jakub@redhat.com>
26064 PR tree-optimization/99079
26065 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
26066 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
26067 require both type and TREE_TYPE (@1) to be integral types and either
26068 type having smaller or equal precision, or TREE_TYPE (@1) being
26069 unsigned type, or type being signed type. If TREE_TYPE (@1)
26070 doesn't have wrapping overflow, perform the subtraction of one in
26073 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
26074 Richard Biener <rguether@suse.de>
26077 * ipa-reference.c (ipa_init): Only conditinally initialize
26078 reference_vars_to_consider.
26079 (propagate): Conditionally deninitialize reference_vars_to_consider.
26080 (ipa_reference_write_optimization_summary): Sanity check that
26081 reference_vars_to_consider is not allocated.
26083 2021-02-13 Levy Hsu <admin@levyhsu.com>
26086 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
26087 extend parameter to get_si_mem_base_reg declaration.
26088 (get_si_mem_base_reg): Add extend parameter. Set it.
26089 (analyze): Pass extend arg to get_si_mem_base_reg.
26090 (transform): Likewise. Use it when rewriting mems.
26091 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
26092 loads and emit sign/zero extending load followed by subreg move.
26094 2021-02-13 Jim Wilson <jimw@sifive.com>
26097 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
26098 exit when !reload_completed. Only perform check for compressed reg
26099 if reload_completed.
26100 (riscv_rtx_costs): In MEM case, when optimizing for size and
26101 shorten memrefs, if not compressible, then increase cost.
26103 2021-02-13 Jakub Jelinek <jakub@redhat.com>
26105 PR rtl-optimization/98439
26106 * recog.c (pass_split_before_regstack::gate): Enable even when
26107 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
26110 2021-02-13 Jakub Jelinek <jakub@redhat.com>
26113 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
26114 swap of V2SImode elements in memory into DImode memory rotate by 32.
26116 2021-02-12 Martin Sebor <msebor@redhat.com>
26118 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
26120 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
26122 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
26123 m_temp_obstack rather than m_obstack to allocate the temporary use.
26125 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
26127 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
26128 as read-modify operations.
26130 2021-02-12 Richard Biener <rguenther@suse.de>
26132 PR middle-end/38474
26133 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
26134 fbi->aa_walk_budget is bigger than zero. Update
26135 fbi->aa_walk_budget.
26136 (param_change_prob): Likewise.
26137 * ipa-prop.c (detect_type_change_from_memory_writes):
26138 Properly account walk_aliased_vdefs.
26139 (parm_preserved_before_stmt_p): Canonicalize updates.
26140 (parm_ref_data_preserved_p): Likewise.
26141 (parm_ref_data_pass_through_p): Likewise.
26142 (determine_known_aggregate_parts): Account own alias queries.
26144 2021-02-12 Martin Liska <mliska@suse.cz>
26146 * opts-common.c (decode_cmdline_option): Release werror_arg.
26147 * opts.c (gen_producer_string): Release output of
26148 gen_command_line_string.
26150 2021-02-12 Richard Biener <rguenther@suse.de>
26152 PR tree-optimization/38474
26153 * params.opt (-param=max-store-chains-to-track=): New param.
26154 (-param=max-stores-to-track=): Likewise.
26155 * doc/invoke.texi (max-store-chains-to-track): Document.
26156 (max-stores-to-track): Likewise.
26157 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
26159 (pass_store_merging::m_n_stores): Likewise.
26160 (pass_store_merging::terminate_and_process_chain): Update
26161 m_n_stores and m_n_chains.
26162 (pass_store_merging::process_store): Likewise. Terminate
26163 oldest chains if the number of stores or chains get too large.
26164 (imm_store_chain_info::terminate_and_process_chain): Dump
26167 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
26169 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
26170 the cold section, emit a nop before the directive if the previous
26171 active instruction can throw.
26173 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
26176 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
26177 memory addresses that are legal for quad word accesses.
26179 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
26182 * config/arm/thumb2.md (*doloop_end_internal): Generate
26183 alternative sequence to handle long range branches.
26185 2021-02-11 Joel Hutton <joel.hutton@arm.com>
26187 PR tree-optimization/98772
26188 * optabs-tree.c (supportable_half_widening_operation): New function
26189 to check for supportable V8QI->V8HI widening patterns.
26190 * optabs-tree.h (supportable_half_widening_operation): New function.
26191 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
26192 to create promotion stmts for V8QI->V8HI widening patterns.
26193 (vectorizable_conversion): Add case for V8QI->V8HI.
26195 2021-02-11 Richard Biener <rguenther@suse.de>
26197 * sparseset.h (SPARSESET_ELT_BITS): Remove.
26198 (SPARSESET_ELT_TYPE): Use unsigned int.
26199 * fwprop.c: Do not include sparseset.h.
26201 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26204 * varasm.c (declare_weak): For -fsyntax-only, allow even
26205 TREE_ASM_WRITTEN function decls.
26207 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26210 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
26211 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
26212 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
26213 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
26214 calling simplify_gen_subreg on it.
26216 2021-02-10 Martin Liska <mliska@suse.cz>
26218 * config/nvptx/nvptx.c (nvptx_option_override): Use
26219 flag_patchable_function_entry instead of the removed
26220 function_entry_patch_area_size.
26222 2021-02-10 Martin Liska <mliska@suse.cz>
26224 PR tree-optimization/99002
26225 PR tree-optimization/99026
26226 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
26227 leak when adjacent cases are merged.
26228 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
26230 (make_pass_lower_switch): Remove trailing whitespace.
26231 * tree-switch-conversion.h (release_clusters): New.
26233 2021-02-10 Richard Biener <rguenther@suse.de>
26235 PR rtl-optimization/99054
26236 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
26237 (fixup_partitions): Adjust.
26238 (rtl_verify_edges): Likewise.
26240 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26242 PR middle-end/99007
26243 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
26244 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
26247 2021-02-10 Richard Biener <rguenther@suse.de>
26250 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
26253 2021-02-10 Richard Biener <rguenther@suse.de>
26255 PR tree-optimization/99024
26256 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
26257 clear loop->aux if it is associated with the destroyed loop_vinfo.
26259 2021-02-10 Martin Liska <mliska@suse.cz>
26261 PR tree-optimization/99002
26262 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
26265 2021-02-10 Martin Liska <mliska@suse.cz>
26268 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
26269 a reference exists.
26271 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26274 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
26275 at class scope for DWARF5+.
26277 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
26279 PR rtl-optimization/96015
26280 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
26281 (relax_delay_slots): When deleting a jump to the next active
26282 instruction over a barrier, first delete the barrier if the
26283 jump is the only way to reach the target label.
26285 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
26287 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
26288 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
26289 vector multiplies and vect.alu for SSRA.
26290 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
26291 vect.mul cost field.
26292 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
26293 * config/arm/arm.c: Likewise.
26295 2021-02-09 Richard Biener <rguenther@suse.de>
26297 PR tree-optimization/98863
26298 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
26299 * tree-ssa-sccvn.c (last_pushed_avail): New global.
26300 (rpo_elim::eliminate_push_avail): Chain pushed avails.
26301 (unwind_state::avail_top): Add.
26302 (do_unwind): Rewrite unwinding of avail entries.
26303 (do_rpo_vn): Initialize last_pushed_avail and
26304 avail_top of the undo state.
26306 2021-02-09 Jakub Jelinek <jakub@redhat.com>
26308 PR middle-end/99004
26309 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
26310 const char * to char * and free those pointers after use.
26312 2021-02-09 Richard Biener <rguenther@suse.de>
26314 PR tree-optimization/99017
26315 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
26316 zero vector cost entries.
26318 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
26320 PR middle-end/98974
26321 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
26322 parameter in vectorizable_condition.
26324 2021-02-08 Richard Biener <rguenther@suse.de>
26327 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
26329 2021-02-08 Martin Liska <mliska@suse.cz>
26332 * cfgexpand.c (pass_expand::execute): Parse per-function option
26333 flag_patchable_function_entry and use it.
26334 * common.opt: Remove function_entry_patch_area_size and
26335 function_entry_patch_area_start global variables.
26336 * opts.c (parse_and_check_patch_area): New function.
26337 (common_handle_option): Use it.
26338 * opts.h (parse_and_check_patch_area): New function.
26339 * toplev.c (process_options): Parse and use
26340 function_entry_patch_area_size.
26342 2021-02-08 Martin Sebor <msebor@redhat.com>
26344 * doc/extend.texi (attribute malloc): Correct typos.
26346 2021-02-05 Nathan Sidwell <nathan@acm.org>
26349 * gcc.c (driver::maybe_run_linker): Check for input file
26350 accessibility if not linking.
26352 2021-02-05 Richard Biener <rguenther@suse.de>
26354 PR tree-optimization/98855
26355 * tree-vectorizer.h (add_stmt_cost): New overload.
26356 * tree-vect-slp.c (li_cost_vec_cmp): New.
26357 (vect_bb_slp_scalar_cost): Cost individual loop regions
26358 separately. Account for the scalar instance root stmt.
26360 2021-02-05 Tom de Vries <tdevries@suse.de>
26363 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
26365 (bit_test_cluster::emit): Reuse location_t for newly created
26367 (switch_decision_tree::try_switch_expansion): Preserve
26369 * tree-switch-conversion.h: Change function signatures.
26371 2021-02-05 Jakub Jelinek <jakub@redhat.com>
26374 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
26375 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
26376 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
26377 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
26379 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26381 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
26382 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
26383 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
26384 (vget_high_f16): Reimplement using new builtin.
26385 (vget_high_f32): Likewise.
26386 (vget_high_f64): Likewise.
26387 (vget_high_p8): Likewise.
26388 (vget_high_p16): Likewise.
26389 (vget_high_p64): Likewise.
26390 (vget_high_s8): Likewise.
26391 (vget_high_s16): Likewise.
26392 (vget_high_s32): Likewise.
26393 (vget_high_s64): Likewise.
26394 (vget_high_u8): Likewise.
26395 (vget_high_u16): Likewise.
26396 (vget_high_u32): Likewise.
26397 (vget_high_u64): Likewise.
26399 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26401 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
26402 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
26403 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
26404 (vget_low_f16): Reimplement using new builtin.
26405 (vget_low_f32): Likewise.
26406 (vget_low_f64): Likewise.
26407 (vget_low_p8): Likewise.
26408 (vget_low_p16): Likewise.
26409 (vget_low_p64): Likewise.
26410 (vget_low_s8): Likewise.
26411 (vget_low_s16): Likewise.
26412 (vget_low_s32): Likewise.
26413 (vget_low_s64): Likewise.
26414 (vget_low_u8): Likewise.
26415 (vget_low_u16): Likewise.
26416 (vget_low_u32): Likewise.
26417 (vget_low_u64): Likewise.
26419 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
26421 * gcc.c (print_multilib_info): Check all required argument is provided
26424 2021-02-05 liuhongt <hongtao.liu@intel.com>
26427 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
26428 generate integer mask comparison for 128/256-bits vector when
26429 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
26430 delete redundant !maskcmp condition.
26431 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
26433 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
26434 condition directly to if (maskcmp), add extra check for
26435 cmpmode, it should be MODE_INT.
26436 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
26437 parameters op_true/op_false.
26438 (ix86_use_mask_cmp_p): New.
26440 2021-02-05 liuhongt <hongtao.liu@intel.com>
26443 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
26444 Remove m_GENERIC from ~list.
26445 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
26447 2021-02-04 David Malcolm <dmalcolm@redhat.com>
26450 * diagnostic-show-locus.c (compatible_locations_p): Require
26451 locations in the same macro map to be either both from the
26452 macro definition, or both from the macro arguments.
26454 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
26456 * config/aarch64/aarch64-simd-builtins.def: Add
26457 [su]mull_hi_lane[q] builtin generator macros.
26458 * config/aarch64/aarch64-simd.md
26459 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
26460 (aarch64_<su>mull_hi_lane<mode>): Define.
26461 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
26462 (aarch64_<su>mull_hi_laneq<mode>): Define.
26463 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
26464 builtin instead of inline asm.
26465 (vmull_high_lane_s32): Likewise.
26466 (vmull_high_lane_u16): Likewise.
26467 (vmull_high_lane_u32): Likewise.
26468 (vmull_high_laneq_s16): Likewise.
26469 (vmull_high_laneq_s32): Likewise.
26470 (vmull_high_laneq_u16): Likewise.
26471 (vmull_high_laneq_u32): Liekwise.
26473 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
26475 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
26476 builtin generator macros.
26477 * config/aarch64/aarch64-simd.md
26478 (aarch64_<su>mull_hi_n<mode>_insn): Define.
26479 (aarch64_<su>mull_hi_n<mode>): Define.
26480 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
26481 instead of inline asm.
26482 (vmull_high_n_s32): Likewise.
26483 (vmull_high_n_u16): Likewise.
26484 (vmull_high_n_u32): Likewise.
26486 2021-02-04 Richard Biener <rguenther@suse.de>
26488 PR tree-optimization/98855
26489 * tree-vect-loop.c (vectorizable_phi): Do not cost
26490 single-argument PHIs.
26491 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
26492 * tree-vect-stmts.c (vectorizable_bswap): Also perform
26493 costing for SLP operation.
26495 2021-02-04 Martin Liska <mliska@suse.cz>
26497 * doc/extend.texi: Mention -mprefer-vector-width in target
26500 2021-02-03 Martin Sebor <msebor@redhat.com>
26502 PR tree-optimization/98937
26503 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
26504 Flush pointer_query cache.
26506 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
26508 * config/rs6000/genfusion.pl (gen_2logical): Add missing
26509 fixes based on patch review.
26510 * config/rs6000/fusion.md: Regenerate file.
26512 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
26514 * config/rs6000/t-rs6000: Comment out auto generation of
26517 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
26519 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
26520 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
26521 (output_file_start): Add gfx908.
26522 * config/gcn/gcn.opt (gpu_type): Add gfx908.
26523 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
26524 (MULTILIB_DIRNAMES): Add gfx908.
26525 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
26526 (main): Recognize gfx908.
26527 * config/gcn/t-omp-device: Add gfx908.
26529 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26531 * config/aarch64/aarch64-simd-builtins.def: Add
26532 [su]mlsl_hi_lane[q] builtin macro generators.
26533 * config/aarch64/aarch64-simd.md
26534 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
26535 (aarch64_<su>mlsl_hi_lane<mode>): Define.
26536 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
26537 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
26538 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
26539 builtin instead of inline asm.
26540 (vmlsl_high_lane_s32): Likewise.
26541 (vmlsl_high_lane_u16): Likewise.
26542 (vmlsl_high_lane_u32): Likewise.
26543 (vmlsl_high_laneq_s16): Likewise.
26544 (vmlsl_high_laneq_s32): Likewise.
26545 (vmlsl_high_laneq_u16): Likewise.
26546 (vmlsl_high_laneq_u32): Likewise.
26547 (vmlal_high_laneq_u32): Likewise.
26549 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26551 * config/aarch64/aarch64-simd-builtins.def: Add
26552 [su]mlal_hi_lane[q] builtin generator macros.
26553 * config/aarch64/aarch64-simd.md
26554 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
26555 (aarch64_<su>mlal_hi_lane<mode>): Define.
26556 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
26557 (aarch64_<su>mlal_hi_laneq<mode>): Define.
26558 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
26559 builtin instead of inline asm.
26560 (vmlal_high_lane_s32): Likewise.
26561 (vmlal_high_lane_u16): Likewise.
26562 (vmlal_high_lane_u32): Likewise.
26563 (vmlal_high_laneq_s16): Likewise.
26564 (vmlal_high_laneq_s32): Likewise.
26565 (vmlal_high_laneq_u16): Likewise.
26566 (vmlal_high_laneq_u32): Likewise.
26568 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26570 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
26571 builtin generator macros.
26572 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
26574 (aarch64_<su>mlsl_hi_n<mode>): Define.
26575 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
26576 instead of inline asm.
26577 (vmlsl_high_n_s32): Likewise.
26578 (vmlsl_high_n_u16): Likewise.
26579 (vmlsl_high_n_u32): Likewise.
26581 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26583 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
26584 builtin generator macros.
26585 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
26587 (aarch64_<su>mlal_hi_n<mode>): Define.
26588 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
26589 instead of inline asm.
26590 (vmlal_high_n_s32): Likewise.
26591 (vmlal_high_n_u16): Likewise.
26592 (vmlal_high_n_u32): Likewise.
26594 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
26596 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
26598 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
26600 (aarch64_<su>mlal_hi<mode>_insn): This.
26601 (aarch64_<su>mlal_hi<mode>): Define.
26602 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
26603 instead of inline asm.
26604 (vmlal_high_s16): Likewise.
26605 (vmlal_high_s32): Likewise.
26606 (vmlal_high_u8): Likewise.
26607 (vmlal_high_u16): Likewise.
26608 (vmlal_high_u32): Likewise.
26610 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
26612 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
26613 after calling alter_subreg() on a (mem).
26615 2021-02-03 Martin Liska <mliska@suse.cz>
26618 * lto-streamer-out.c (produce_lto_section): Fill up missing
26620 * lto-streamer.h (struct lto_section): Add _padding field.
26622 2021-02-03 Richard Biener <rguenther@suse.de>
26624 * lto-streamer.c (lto_get_section_name): Free temporary
26626 * tree-loop-distribution.c
26627 (loop_distribution::merge_dep_scc_partitions): Free edge data.
26629 2021-02-03 Jakub Jelinek <jakub@redhat.com>
26631 PR middle-end/97487
26632 * ifcvt.c (noce_can_force_operand): New function.
26633 (noce_emit_move_insn): Use it.
26634 (noce_try_sign_mask): Likewise. Formatting fix.
26636 2021-02-03 Jakub Jelinek <jakub@redhat.com>
26638 PR middle-end/97971
26639 * lra-constraints.c (process_alt_operands): For inline asm, don't call
26640 fatal_insn, but instead return false.
26642 2021-02-03 Jakub Jelinek <jakub@redhat.com>
26644 PR tree-optimization/98287
26645 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
26648 2021-02-03 Tamar Christina <tamar.christina@arm.com>
26650 PR tree-optimization/98928
26651 * tree-vect-loop.c (vect_analyze_loop_2): Change
26652 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
26653 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
26654 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
26655 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
26657 2021-02-02 Richard Biener <rguenther@suse.de>
26659 * gimple-loop-interchange.cc (prepare_data_references):
26661 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
26662 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
26663 * tree-vect-stmts.c (vectorizable_condition): Do not
26665 (vectorizable_comparison): Likewise.
26667 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26669 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
26670 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
26671 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
26672 (vrsqrteq_u32): Likewise.
26674 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26676 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
26677 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
26678 (aarch64_sqxtun2<mode>_be): Likewise.
26679 (aarch64_sqxtun2<mode>): Likewise.
26680 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
26681 (vqmovun_high_s32): Likewise.
26682 (vqmovun_high_s64): Likewise.
26683 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
26685 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26687 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
26689 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
26691 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26693 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
26694 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
26695 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
26696 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
26697 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
26698 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
26699 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
26700 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
26702 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26704 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
26705 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
26706 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
26708 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26710 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
26711 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
26713 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26715 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
26716 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
26717 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
26718 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
26719 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
26720 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
26721 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
26722 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
26723 NONE builtin flags.
26725 2021-02-02 Jakub Jelinek <jakub@redhat.com>
26727 PR tree-optimization/98848
26728 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
26729 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
26731 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
26734 * expr.c: Check mode before calling store_expr.
26736 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
26738 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
26740 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
26741 instruction using expression ior.
26742 (mve_vornq_u<mode>): New expander.
26743 (mve_vornq_f<mode>): Use ior code instead of unspec.
26744 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
26746 2021-02-02 Alexandre Oliva <oliva@adacore.com>
26748 * tree-nested.c (convert_nonlocal_reference_op): Move
26749 current_function_decl restore after re-gimplification.
26750 (convert_local_reference_op): Likewise.
26752 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26754 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
26756 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
26758 (aarch64_rshrn<mode>_insn_be): Likewise.
26759 (aarch64_rshrn<mode>): Likewise.
26760 (aarch64_rshrn2<mode>_insn_le): Likewise.
26761 (aarch64_rshrn2<mode>_insn_be): Likewise.
26762 (aarch64_rshrn2<mode>): Likewise.
26763 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
26764 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
26766 (vrshrn_high_n_s32): Likewise.
26767 (vrshrn_high_n_s64): Likewise.
26768 (vrshrn_high_n_u16): Likewise.
26769 (vrshrn_high_n_u32): Likewise.
26770 (vrshrn_high_n_u64): Likewise.
26771 (vrshrn_n_s16): Likewise.
26772 (vrshrn_n_s32): Likewise.
26773 (vrshrn_n_s64): Likewise.
26774 (vrshrn_n_u16): Likewise.
26775 (vrshrn_n_u32): Likewise.
26776 (vrshrn_n_u64): Likewise.
26778 2021-02-01 Sergei Trofimovich <siarheit@google.com>
26780 PR tree-optimization/98499
26781 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
26782 conservatively and assume all possible side-effects.
26784 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26786 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
26787 vec_unpacku_hi_): Define builtins.
26788 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
26790 (vmovl_high_s16): Likewise.
26791 (vmovl_high_s32): Likewise.
26792 (vmovl_high_u8): Likewise.
26793 (vmovl_high_u16): Likewise.
26794 (vmovl_high_u32): Likewise.
26796 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26798 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
26800 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
26802 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
26804 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
26806 (vabdl_s16): Likewise.
26807 (vabdl_s32): Likewise.
26808 (vabdl_u8): Likewise.
26809 (vabdl_u16): Likewise.
26810 (vabdl_u32): Likewise.
26811 * config/aarch64/iterators.md (ABDL): New int iterator.
26812 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
26814 2021-02-01 Martin Sebor <msebor@redhat.com>
26816 * tree.h (BLOCK_VARS): Add comment.
26817 (BLOCK_SUBBLOCKS): Same.
26818 (BLOCK_SUPERCONTEXT): Same.
26819 (BLOCK_ABSTRACT_ORIGIN): Same.
26820 (inlined_function_outer_scope_p): Same.
26822 2021-02-01 Martin Sebor <msebor@redhat.com>
26824 PR middle-end/97172
26825 * attribs.c (attr_access::free_lang_data): Define new function.
26826 * attribs.h (attr_access::free_lang_data): Declare new function.
26828 2021-02-01 Richard Biener <rguenther@suse.de>
26830 * vec.h (auto_vec::auto_vec): Add memory stat parameters
26832 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
26834 2021-02-01 Tamar Christina <tamar.christina@arm.com>
26836 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
26837 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
26839 2021-02-01 Richard Biener <rguenther@suse.de>
26841 PR rtl-optimization/98863
26842 * config/i386/i386-features.c (convert_scalars_to_vector):
26843 Set DF_RD_PRUNE_DEAD_DEFS.
26845 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
26847 * system.h (SIZE_MAX): Define if not already defined.
26849 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
26851 * config/rs6000/genfusion.pl (gen_2logical): New function to
26852 generate patterns for logical-logical fusion.
26853 * config/rs6000/fusion.md: Regenerated patterns.
26854 * config/rs6000/rs6000-cpus.def: Add
26855 OPTION_MASK_P10_FUSION_2LOGICAL.
26856 * config/rs6000/rs6000.c (rs6000_option_override_internal):
26857 Enable logical-logical fusion for p10.
26858 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
26860 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
26862 * config/rs6000/rs6000.opt: Add periods to new AIX options.
26864 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
26866 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
26867 (mabi=vec-default): New.
26868 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
26869 __EXTABI__ for AIX Vector extended ABI.
26870 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
26872 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
26874 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
26876 2021-01-30 Jakub Jelinek <jakub@redhat.com>
26878 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
26879 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
26881 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
26884 * lra-constraints.c (in_class_p): Don't narrow class only for REG
26887 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
26889 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
26890 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
26891 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
26893 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
26895 PR tree-optimization/98866
26896 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
26897 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
26898 (gori_map::m_maybe_invariant): Rename from all_outgoing.
26899 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
26900 (gori_map::is_export_p): Ditto.
26901 (gori_map::calculate_gori): Ditto.
26902 (gori_compute::set_range_invariant): New.
26903 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
26904 invariant for pointers evaluating to [1, +INF].
26906 2021-01-29 Richard Biener <rguenther@suse.de>
26908 PR rtl-optimization/98863
26909 * config/i386/i386-features.c (remove_partial_avx_dependency):
26910 Do not perform DF analysis.
26911 (pass_data_remove_partial_avx_dependency): Remove
26914 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
26916 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
26917 builtin generator macros.
26918 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
26920 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
26921 instead of inline asm.
26922 (vmull_n_s32): Likewise.
26923 (vmull_n_u16): Likewise.
26924 (vmull_n_u32): Likewise.
26926 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26928 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
26930 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
26932 (aarch64_<sur>abdl2<mode>): ... This.
26933 (<sur>sadv16qi): Adjust use of above.
26934 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
26936 (vabdl_high_s16): Likewise.
26937 (vabdl_high_s32): Likewise.
26938 (vabdl_high_u8): Likewise.
26939 (vabdl_high_u16): Likewise.
26940 (vabdl_high_u32): Likewise.
26942 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26944 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
26946 (uabal2): Likewise.
26947 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
26949 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
26951 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
26953 (vabal_high_s16): Likewise.
26954 (vabal_high_s32): Likewise.
26955 (vabal_high_u8): Likewise.
26956 (vabal_high_u16): Likewise.
26957 (vabal_high_u32): Likewise.
26958 * config/aarch64/iterators.md (ABAL2): New mode iterator.
26959 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
26961 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26963 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
26966 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
26968 (aarch64_<sur>abal<mode>): ... This
26969 (<sur>sadv16qi): Adust use of the above.
26970 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
26972 (vabal_s16): Likewise.
26973 (vabal_s32): Likewise.
26974 (vabal_u8): Likewise.
26975 (vabal_u16): Likewise.
26976 (vabal_u32): Likewise.
26978 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26980 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
26982 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
26984 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
26986 (vaddlv_s16): Likewise.
26987 (vaddlv_u8): Likewise.
26988 (vaddlv_u16): Likewise.
26989 (vaddlvq_s8): Likewise.
26990 (vaddlvq_s16): Likewise.
26991 (vaddlvq_s32): Likewise.
26992 (vaddlvq_u8): Likewise.
26993 (vaddlvq_u16): Likewise.
26994 (vaddlvq_u32): Likewise.
26995 (vaddlv_s32): Likewise.
26996 (vaddlv_u32): Likewise.
26997 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
26998 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
26999 (Vwstype): New mode attribute.
27001 (VWIDE_S): Likewise.
27002 (USADDLV): New int iterator.
27003 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
27005 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
27007 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
27008 builtin generator macros.
27009 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
27011 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
27012 instead of inline asm.
27013 (vmlsl_lane_s32): Likewise.
27014 (vmlsl_lane_u16): Likewise.
27015 (vmlsl_lane_u32): Likewise.
27016 (vmlsl_laneq_s16): Likewise.
27017 (vmlsl_laneq_s32): Likewise.
27018 (vmlsl_laneq_u16): Likewise.
27019 (vmlsl_laneq_u32): Likewise.
27021 2021-01-29 Richard Biener <rguenther@suse.de>
27023 * doc/invoke.texi (--param max-gcse-memory): Document unit
27025 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
27026 * params.opt (--param max-gcse-memory): Adjust default and
27027 document unit of size.
27029 2021-01-29 Richard Biener <rguenther@suse.de>
27031 PR rtl-optimization/98863
27032 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
27033 HOST_WIDE_INT for the memory estimate.
27035 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
27036 Richard Biener <rguenther@suse.de>
27038 PR tree-optimization/97627
27039 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
27040 Do not analyze fake edges.
27042 2021-01-29 Richard Biener <rguenther@suse.de>
27044 PR rtl-optimization/98144
27045 * df.h (df_mir_bb_info): Add con_visited member.
27046 * df-problems.c (df_mir_alloc): Initialize con_visited,
27047 do not fully populate IN and OUT.
27048 (df_mir_reset): Likewise.
27049 (df_mir_confluence_0): Set con_visited.
27050 (df_mir_confluence_n): Properly handle implicitely
27051 fully populated IN and OUT as designated by con_visited
27052 and update con_visited accordingly.
27054 2021-01-29 Jakub Jelinek <jakub@redhat.com>
27057 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
27058 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
27059 && !TARGET_REALLY_IWMMXT to conditions.
27061 2021-01-29 Jakub Jelinek <jakub@redhat.com>
27064 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
27067 2021-01-28 Marek Polacek <polacek@redhat.com>
27070 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
27071 the main variant, maybe reset it in its variants too.
27072 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
27073 (check_aligned_type): Check if TYPE_USER_ALIGN match.
27075 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
27078 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
27079 of constant zero for comparisons.
27081 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
27083 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
27084 support for mapping built-in function names for long double
27085 built-in functions if long double is IEEE 128-bit.
27087 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
27089 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
27090 builtin generator macros.
27091 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
27093 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
27094 instead of inline asm.
27095 (vmlsl_n_s32): Likewise.
27096 (vmlsl_n_u16): Likewise.
27097 (vmlsl_n_u32): Likewise.
27099 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
27101 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
27102 builtin generator macros.
27103 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
27105 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
27106 instead of inline asm.
27107 (vmlal_n_s32): Likewise.
27108 (vmlal_n_u16): Likewise.
27109 (vmlal_n_u32): Likewise.
27111 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27113 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
27115 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
27117 (aarch64_shrn2<mode>_insn_be): Likewise.
27118 (aarch64_shrn2<mode>): Likewise.
27119 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
27121 (vshrn_high_n_s32): Likewise.
27122 (vshrn_high_n_s64): Likewise.
27123 (vshrn_high_n_u16): Likewise.
27124 (vshrn_high_n_u32): Likewise.
27125 (vshrn_high_n_u64): Likewise.
27127 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27129 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
27131 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
27133 (aarch64_shrn<mode>_insn_be): Likewise.
27134 (aarch64_shrn<mode>): Likewise.
27135 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
27137 (vshrn_n_s32): Likewise.
27138 (vshrn_n_s64): Likewise.
27139 (vshrn_n_u16): Likewise.
27140 (vshrn_n_u32): Likewise.
27141 (vshrn_n_u64): Likewise.
27142 * config/aarch64/iterators.md (vn_mode): New mode attribute.
27144 2021-01-28 Richard Biener <rguenther@suse.de>
27146 PR rtl-optimization/80960
27147 * dse.c (check_mem_read_rtx): Call get_addr on the
27150 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
27151 David Edelsohn <dje.gcc@gmail.com>
27154 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27155 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
27157 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
27159 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
27160 wrapper call rs6000_expand_vector_set_var for cleanup. Call
27161 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
27163 (rs6000_expand_vector_set_var): Delete.
27164 (rs6000_expand_vector_set_var_p9): Make static.
27165 (rs6000_expand_vector_set_var_p8): Make static.
27167 2021-01-28 Xing GUO <higuoxing@gmail.com>
27169 * common/config/riscv/riscv-common.c
27170 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
27171 when `p` extension exists.
27173 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
27175 PR rtl-optimization/97684
27176 * ira.c (ira): Call ira_set_pseudo_classes before
27177 update_equiv_regs when it is necessary.
27179 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27182 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
27183 %w0, %w1 and %2 instead of %0, %1 and %2.
27185 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
27187 * config/rs6000/genfusion.pl: New script to generate
27188 define_insn_and_split patterns so combine can arrange fused
27189 instructions next to each other.
27190 * config/rs6000/fusion.md: New file, generated fused instruction
27191 patterns for combine.
27192 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
27193 (non_update_memory_operand): New predicate.
27194 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
27195 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
27197 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
27199 * config/rs6000/rs6000.c (rs6000_option_override_internal):
27200 Automatically set OPTION_MASK_P10_FUSION and
27201 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
27202 (rs600_opt_masks): Allow -mpower10-fusion
27203 in function attributes.
27204 (address_is_non_pfx_d_or_x): New function.
27205 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
27206 * config/rs6000/rs6000.md: Include fusion.md.
27207 * config/rs6000/rs6000.opt: Add -mpower10-fusion
27208 and -mpower10-fusion-ld-cmpi.
27209 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
27211 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27213 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
27214 builtin generator macros.
27215 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
27217 (aarch64_<su>mlal<mode>): This.
27218 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
27219 instead of inline asm.
27220 (vmlal_s16): Likewise.
27221 (vmlal_s32): Likewise.
27222 (vmlal_u8): Likewise.
27223 (vmlal_u16): Likewise.
27224 (vmlal_u32): Likewise.
27226 2021-01-27 Richard Biener <rguenther@suse.de>
27228 PR tree-optimization/98854
27229 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
27230 PHIs from scalars when the number of CTORs matches the
27231 number of children.
27233 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27235 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
27237 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
27239 (aarch64_mls_n<mode>): This.
27240 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
27242 (vmls_n_s32): Likewise.
27243 (vmls_n_u16): Likewise.
27244 (vmls_n_u32): Likewise.
27245 (vmlsq_n_s16): Likewise.
27246 (vmlsq_n_s32): Likewise.
27247 (vmlsq_n_u16): Likewise.
27248 (vmlsq_n_u32): Likewise.
27250 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27252 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
27254 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
27256 (vmls_s16): Likewise.
27257 (vmls_s32): Likewise.
27258 (vmls_u8): Likewise.
27259 (vmls_u16): Likewise.
27260 (vmls_u32): Likewise.
27261 (vmlsq_s8): Likewise.
27262 (vmlsq_s16): Likewise.
27263 (vmlsq_s32): Likewise.
27264 (vmlsq_u8): Likewise.
27265 (vmlsq_u16): Likewise.
27266 (vmlsq_u32): Likewise.
27268 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27270 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
27272 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
27274 (aarch64_mla_n<mode>): This.
27275 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
27277 (vmla_n_s32): Likewise.
27278 (vmla_n_u16): Likewise.
27279 (vmla_n_u32): Likewise.
27280 (vmlaq_n_s16): Likewise.
27281 (vmlaq_n_s32): Likewise.
27282 (vmlaq_n_u16): Likewise.
27283 (vmlaq_n_u32): Likewise.
27285 2021-01-27 liuhongt <hongtao.liu@intel.com>
27288 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
27289 (*sse2_eq<mode>3): Ditto.
27291 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27293 * tree-pass.h (PROP_trees): Rename to ...
27294 (PROP_gimple): ... this.
27295 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
27296 * passes.c (execute_function_dump, execute_function_todo,
27297 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
27298 * varpool.c (ctor_for_folding): Likewise.
27300 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27302 PR tree-optimization/97260
27303 * varpool.c: Include tree-pass.h.
27304 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
27305 non-TREE_SIDE_EFFECTS automatic variables.
27307 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
27309 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
27311 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
27312 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
27313 (gen_compile_unit_die): Recognise C++23.
27315 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27318 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
27321 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27324 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
27325 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
27326 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
27328 2021-01-26 Richard Biener <rguenther@suse.de>
27330 * gimple-pretty-print.c (dump_binary_rhs): Handle
27331 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
27333 2021-01-26 Richard Biener <rguenther@suse.de>
27335 PR middle-end/98726
27336 * tree.h (vector_cst_int_elt): Remove.
27337 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
27340 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
27342 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
27343 for V64DFmode min/max reductions.
27345 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27347 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
27348 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
27349 two halves, one with x and the other with const0_rtx, ordered
27350 depending on endianity.
27352 2021-01-26 Alexandre Oliva <oliva@adacore.com>
27354 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
27355 temporaries not seen in binding block, and not about to be
27356 added as gimple variables.
27358 2021-01-25 Martin Sebor <msebor@redhat.com>
27361 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
27363 2021-01-25 Martin Liska <mliska@suse.cz>
27365 * value-prof.c (get_nth_most_common_value): Use %s instead
27368 2021-01-25 Jakub Jelinek <jakub@redhat.com>
27371 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
27372 readelf -wi is able to read the emitted .debug_info back.
27373 * configure: Regenerated.
27375 2021-01-25 Martin Liska <mliska@suse.cz>
27377 PR gcov-profile/98739
27378 * common.opt: Add missing sign symbol.
27379 * value-prof.c (get_nth_most_common_value): Restore handling
27380 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
27381 PROFILE_REPRODUCIBILITY_MULTITHREADED.
27383 2021-01-25 Richard Biener <rguenther@suse.de>
27385 PR middle-end/98807
27386 * tree.c (vector_element_bits): Always use precision of
27387 the element type for boolean vectors.
27389 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
27391 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
27392 (ENDFILE_SPEC): Evaluate qnolinkcmds.
27394 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
27396 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
27397 nostartfiles handling since this is already done by
27398 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
27399 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
27400 is already done by LINK_COMMAND_SPEC.
27401 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
27402 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
27405 2021-01-25 Jakub Jelinek <jakub@redhat.com>
27408 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
27409 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
27410 value rather than host size_t.
27411 (fold_const_call): Change type of s2 from size_t to
27412 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
27413 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
27414 instead of s2 as last argument.
27416 2021-01-25 Tamar Christina <tamar.christina@arm.com>
27418 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
27419 VCMLA_OP, VCMUL_OP): New.
27420 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
27421 * config/arm/neon.md (cmul<conj_op><mode>3): New.
27422 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
27423 UNSPEC_VCMUL_CONJ): New.
27424 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
27425 cml<fcmac1><conj_op><mode>4): New.
27427 2021-01-23 Jakub Jelinek <jakub@redhat.com>
27430 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
27432 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
27434 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
27436 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
27438 (vmla_s16): Likewise.
27439 (vmla_s32): Likewise.
27440 (vmla_u8): Likewise.
27441 (vmla_u16): Likewise.
27442 (vmla_u32): Likewise.
27443 (vmlaq_s8): Likewise.
27444 (vmlaq_s16): Likewise.
27445 (vmlaq_s32): Likewise.
27446 (vmlaq_u8): Likewise.
27447 (vmlaq_u16): Likewise.
27448 (vmlaq_u32): Likewise.
27450 2021-01-22 David Malcolm <dmalcolm@redhat.com>
27452 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
27455 2021-01-22 Jakub Jelinek <jakub@redhat.com>
27458 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
27459 filenames to emit, still emit the required 0 index directory and
27460 filename entries that match DW_AT_comp_dir and DW_AT_name of the
27463 2021-01-22 Marek Polacek <polacek@redhat.com>
27466 * doc/invoke.texi: Update C++ ABI Version 15 description.
27468 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27470 PR tree-optimization/98766
27471 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
27472 comparing against type size with param_avoid_fma_max_bits.
27474 2021-01-22 Richard Biener <rguenther@suse.de>
27476 PR middle-end/98793
27477 * tree.c (vector_element_bits): Key single-bit bool vector on
27478 integer mode rather than not vector mode.
27480 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
27483 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27484 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
27486 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
27487 to call different path for P8 and P9.
27488 (rs6000_expand_vector_set_var_p9): New function.
27489 (rs6000_expand_vector_set_var_p8): New function.
27491 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
27495 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27496 Ajdust variable index vec_insert from address dereference to
27497 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
27498 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
27500 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
27502 2021-01-22 Martin Liska <mliska@suse.cz>
27504 PR gcov-profile/98739
27505 * profile.c (compute_value_histograms): Drop time profile for
27506 -fprofile-reproducible=multithreaded.
27508 2021-01-22 Nathan Sidwell <nathan@acm.org>
27510 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
27513 2021-01-22 Richard Biener <rguenther@suse.de>
27515 PR middle-end/98773
27516 * tree-data-ref.c (initalize_matrix_A): Revert previous
27517 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
27519 2021-01-22 Jakub Jelinek <jakub@redhat.com>
27521 PR tree-optimization/90248
27522 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
27523 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
27525 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
27526 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
27528 2021-01-22 Jakub Jelinek <jakub@redhat.com>
27530 PR tree-optimization/98255
27531 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
27532 extend index - low_bound from sizetype's precision rather than index
27534 (get_addr_base_and_unit_offset_1): Likewise.
27535 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
27536 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
27538 2021-01-22 Richard Biener <rguenther@suse.de>
27540 PR tree-optimization/98786
27541 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
27542 adding new uses of abnormals. Verify we deal with a conditional
27545 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27548 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
27550 2021-01-22 liuhongt <hongtao.liu@intel.com>
27554 * config/i386/sse.md (VI_128_256): New mode iterator.
27555 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
27556 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
27557 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
27558 define_insn_and_split to lower avx512 vector comparison to avx
27559 version when dest is vector.
27560 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
27561 define_insn_and_split for negating the comparison result.
27562 * config/i386/predicates.md (float_vector_all_ones_operand):
27564 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
27565 general NOT operator without UNSPEC_MASKOP.
27567 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
27569 PR rtl-optimization/98777
27570 * lra-int.h (lra_pmode_pseudo): New extern.
27571 * lra.c (lra_pmode_pseudo): New global.
27573 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
27575 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
27577 * fwprop.c (fwprop_propagation::classify_result): Allow
27578 (subreg (mem)) simplifications.
27580 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27582 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
27584 (aarch64_sqdmlal<mode>): ... This...
27585 (aarch64_sqdmlsl<mode>): ... And this.
27586 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
27587 (aarch64_sqdmlal_lane<mode>): ... This...
27588 (aarch64_sqdmlsl_lane<mode>): ... And this.
27589 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
27590 (aarch64_sqdmlsl_laneq<mode>): ... This...
27591 (aarch64_sqdmlal_laneq<mode>): ... And this.
27592 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
27593 (aarch64_sqdmlsl_n<mode>): ... This...
27594 (aarch64_sqdmlal_n<mode>): ... And this.
27595 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
27596 (aarch64_sqdmlal2<mode>_internal): ... This...
27597 (aarch64_sqdmlsl2<mode>_internal): ... And this.
27599 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
27601 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
27603 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
27606 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
27608 2021-01-21 liuhongt <hongtao.liu@intel.com>
27610 PR rtl-optimization/98694
27611 * regcprop.c (copy_value): If SRC had been assigned a mode
27612 narrower than the copy, we can't link DEST into the chain even
27613 they have same hard_regno_nregs(i.e. HImode/SImode in i386
27616 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27618 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
27619 Convert to define_insn_and_split. Split into simple move when moving
27622 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
27624 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
27625 Adjust comment. Simplify code.
27627 2021-01-20 Jakub Jelinek <jakub@redhat.com>
27630 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
27631 with DW_FORM_line_strp form.
27632 (prune_unused_types_update_strings): Don't add into debug_str_hash
27633 indirect strings with DW_FORM_line_strp form.
27634 (adjust_name_comp_dir): New function.
27635 (dwarf2out_finish): Call it on CU DIEs after resetting
27636 debug_line_str_hash.
27638 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
27640 PR rtl-optimization/98722
27641 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
27642 has no 3-op add insn to transform insns containing two pluses.
27644 2021-01-20 Richard Biener <rguenther@suse.de>
27646 * hwint.h (add_hwi): New function.
27647 (mul_hwi): Likewise.
27648 * tree-data-ref.c (initialize_matrix_A): Properly translate
27649 tree constants and avoid HOST_WIDE_INT_MIN.
27650 (lambda_matrix_row_add): Avoid undefined integer overflow
27651 and return true on such overflow.
27652 (lambda_matrix_right_hermite): Handle overflow from
27653 lambda_matrix_row_add gracefully. Simplify previous fix.
27654 (analyze_subscript_affine_affine): Likewise.
27656 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
27658 PR tree-optimization/96674
27659 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
27660 x >= y && y != XXX_MIN --> x > y - 1
27662 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
27664 PR tree-optimization/98535
27665 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
27666 If the high and low permutes are the same, remove the high permutes
27667 from the working set and only continue with the low ones.
27669 2021-01-20 Jakub Jelinek <jakub@redhat.com>
27671 PR tree-optimization/98721
27672 * builtins.c (access_ref::inform_access): Don't assume
27673 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
27674 object whenever allocfn is NULL, rather than only when DECL_P
27675 is true. Use %qE instead of %qD for that. Formatting fixes.
27677 2021-01-20 Richard Biener <rguenther@suse.de>
27679 PR tree-optimization/98758
27680 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
27681 (lambda_matrix_right_hermite): Avoid undefinedness with
27682 signed integer abs and multiplication.
27683 (analyze_subscript_affine_affine): Use lambda_int.
27685 2021-01-20 David Malcolm <dmalcolm@redhat.com>
27688 * dwarf2out.c (output_line_info): Rename static variable
27689 "generation", moving it out of the function to...
27690 (output_line_info_generation): New.
27691 (init_sections_and_labels): Likewise, renaming the variable to...
27692 (init_sections_and_labels_generation): New.
27693 (dwarf2out_c_finalize): Reset the new variables.
27695 2021-01-19 Martin Sebor <msebor@redhat.com>
27697 PR middle-end/98664
27698 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
27699 all functions, even if they're not declared artificial or inline.
27700 * tree.c (tree_inlined_location): Use macro expansion location
27701 only if scope traversal fails to expose one.
27703 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
27705 PR rtl-optimization/92294
27706 * alias.c (compare_base_symbol_refs): Take an extra parameter
27707 and add the distance between two symbols to it. Enshrine in
27708 comments that -1 means "either 0 or 1, but we can't tell
27709 which at compile time".
27710 (memrefs_conflict_p): Update call accordingly.
27711 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
27714 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27716 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
27717 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
27718 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
27719 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
27720 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
27721 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
27722 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
27723 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
27724 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
27725 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
27726 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
27727 sqmovun): Use NONE flags.
27729 2021-01-19 Richard Biener <rguenther@suse.de>
27732 * ipa-modref.c (analyze_stmt): Only record a summary for a
27735 2021-01-19 Richard Biener <rguenther@suse.de>
27737 PR middle-end/98638
27738 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
27740 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
27742 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
27743 built-in define __FIX_LEON3FT_TN0018.
27745 2021-01-19 Richard Biener <rguenther@suse.de>
27748 * tree-inline.c (tree_function_versioning): Set input_location
27749 to UNKNOWN_LOCATION throughout the function.
27751 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
27754 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
27756 2021-01-19 Martin Jambor <mjambor@suse.cz>
27759 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
27760 whether non-call exceptions allow removal of a statement.
27761 (isra_analyze_call): Pass the appropriate function to
27762 ssa_name_only_returned_p.
27764 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
27766 * config/riscv/arch-canonicalize (longext_sort): New function for
27767 sorting 'multi-letter'.
27768 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
27769 'alts'. The 'arch' may not be the first of 'alts'.
27770 (_expand_combination): Add underline for the 'ext' without '*'.
27771 This is because, a single-letter extension can always be treated well
27772 with a '_' prefix, but it cannot be separated out if it is appended
27775 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
27778 * ira.c (ira): Skip abnormal critical edge splitting.
27780 2021-01-18 Jakub Jelinek <jakub@redhat.com>
27782 PR tree-optimization/98727
27783 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
27784 second .MUL_OVERFLOW operand for signed multiplication with overflow
27785 checking if the second operand of multiplication is not constant.
27787 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
27789 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
27790 defaults to version 4.
27792 2021-01-18 David Malcolm <dmalcolm@redhat.com>
27794 * attribs.h (fndecl_dealloc_argno): New decl.
27795 * builtins.c (call_dealloc_argno): Split out second half of
27797 (fndecl_dealloc_argno): New.
27798 * doc/extend.texi (Common Function Attributes): Document the
27799 interaction between the analyzer and the malloc attribute.
27800 * doc/invoke.texi (Static Analyzer Options): Likewise.
27802 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
27804 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
27805 dwarf_version to 4.
27806 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
27808 2021-01-17 Martin Jambor <mjambor@suse.cz>
27811 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
27814 2021-01-17 Mark Wielaard <mark@klomp.org>
27816 * common.opt (gdwarf-): Init(5).
27817 * doc/invoke.texi (-gdwarf): Document default to 5.
27819 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
27821 * builtin-types.def
27822 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
27824 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
27825 ...this. Add extra argument.
27826 * gimplify.c (omp_default_clause): Ensure that event handle is
27827 firstprivate in a task region.
27828 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
27829 (gimplify_adjust_omp_clauses): Likewise.
27830 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
27831 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
27832 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
27833 if detach clause specified. Add detach argument when generating
27835 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
27837 (finish_taskreg_scan): Move field for variable containing the event
27838 handle to the front of the struct.
27839 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
27841 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
27842 OMP_CLAUSE_DETACH clause.
27843 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
27844 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
27845 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
27847 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
27849 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
27851 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
27853 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
27854 multilib directory names. Use MULTILIB_REQUIRED instead of
27855 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
27856 -mcustom-fpu-cfg=fph2 multilib.
27858 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
27860 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
27861 (nios2_init_fpu_configs): Provide register values for new
27862 -mcustom-fpu-cfg=fph2 option variant.
27863 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
27866 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
27868 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
27869 custom instruction warnings.
27871 2021-01-16 Jakub Jelinek <jakub@redhat.com>
27873 PR tree-optimization/96669
27874 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
27876 2021-01-16 Jakub Jelinek <jakub@redhat.com>
27878 PR tree-optimization/96271
27879 * passes.def: Pass false argument to first two pass_cd_dce
27880 instances and true to last instance. Add comment that
27881 last instance rewrites no longer addressed locals.
27882 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
27884 (pass_cd_dce::set_pass_param): New method.
27885 (pass_cd_dce::execute): Return TODO_update_address_taken from
27886 last cd_dce instance.
27888 2021-01-15 Carl Love <cel@us.ibm.com>
27890 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
27892 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
27893 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
27894 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
27895 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
27896 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
27897 Add builtin define.
27898 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
27899 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
27900 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
27901 New overloaded definitions.
27902 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
27903 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
27904 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
27905 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
27906 P10V_BUILTIN_MULHU_V4SI]: Add case
27907 statement for builtins.
27908 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
27909 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
27910 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
27911 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
27912 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
27913 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
27914 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
27915 Add define_insn, mode is VIlong.
27916 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
27917 Add builtin descriptions.
27919 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
27921 * final.c (final_start_function_1): Reset force_source_line.
27923 2021-01-15 Jakub Jelinek <jakub@redhat.com>
27925 PR tree-optimization/96669
27926 * match.pd (((1 << A) & 1) != 0 -> A == 0,
27927 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
27928 possibly different power of two constants and to right shift too.
27930 2021-01-15 Jakub Jelinek <jakub@redhat.com>
27932 PR tree-optimization/96681
27933 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
27934 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
27935 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
27936 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
27938 2021-01-15 Alexandre Oliva <oliva@adacore.com>
27940 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
27942 2021-01-15 Tamar Christina <tamar.christina@arm.com>
27944 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
27945 cmul<conj_op><mode>3): New.
27946 * config/aarch64/iterators.md (UNSPEC_FCMUL,
27947 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
27948 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
27949 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
27950 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
27951 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
27952 (rot_op): Renamed to conj_op.
27953 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
27954 cmul<conj_op><mode>3): New.
27955 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
27956 cmul<conj_op><mode>3): New.
27958 2021-01-15 David Malcolm <dmalcolm@redhat.com>
27962 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
27963 Escape the tempfile name when constructing the expected output.
27965 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27967 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
27969 (aarch64_<su>mlsl_hi<mode>): ... This.
27970 (aarch64_<su>mlsl_hi<mode>): Define.
27971 (*aarch64_<su>mlsl<mode): Rename to...
27972 (aarch64_<su>mlsl<mode): ... This.
27973 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
27974 smlsl_hi, umlsl_hi): Define builtins.
27975 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
27976 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
27977 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
27978 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
27980 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
27982 * config/i386/i386-c.c (ix86_target_macros):
27983 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
27985 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
27988 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
27989 * Makefile.in (RTL_SSA_H): New variable.
27990 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
27991 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
27992 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
27994 * config/aarch64/aarch64-cc-fusion.cc: New file.
27996 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
27998 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
27999 calling cancel_changes for changes that no longer exist.
28001 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28003 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
28004 (function_info::reg_defs): ...this.
28005 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
28006 (function_info::reg_defs): ...this.
28008 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28011 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28013 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28016 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28019 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28021 2021-01-15 Richard Biener <rguenther@suse.de>
28023 PR tree-optimization/96376
28024 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
28025 for VMAT_INVARIANT.
28027 2021-01-15 Martin Liska <mliska@suse.cz>
28029 * doc/install.texi: Document that some tests need pytest module.
28030 * doc/sourcebuild.texi: Likewise.
28032 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28035 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28037 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28039 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
28040 (mve_vshrq_n_u<mode>_imm): Likewise.
28041 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
28042 * config/arm/vec-common.md: ... here.
28044 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28046 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
28048 * config/arm/neon.md (vashl<mode>3): Delete.
28049 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
28050 (vasl<mode>3): New expander.
28052 2021-01-15 Richard Biener <rguenther@suse.de>
28054 PR tree-optimization/98685
28055 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
28056 of vector extern defs.
28058 2021-01-14 David Malcolm <dmalcolm@redhat.com>
28061 * diagnostic.c (diagnostic_kind_text): Break out this array
28063 (diagnostic_build_prefix): ...here.
28064 (fancy_abort): Detect when diagnostic_initialize has not yet been
28065 called and fall back to a minimal implementation of printing the
28066 ICE, rather than segfaulting in internal_error.
28068 2021-01-14 David Malcolm <dmalcolm@redhat.com>
28070 * diagnostic.c (diagnostic_initialize): Eliminate
28071 parseable_fixits_p in favor of initializing extra_output_kind from
28072 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
28073 (convert_column_unit): New function, split out from...
28074 (diagnostic_converted_column): ...this.
28075 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
28076 Use them to call convert_column_unit on the column values.
28077 (diagnostic_report_diagnostic): Eliminate conditional on
28078 parseable_fixits_p in favor of a switch statement on
28079 extra_output_kind, passing the appropriate values to the new
28080 params of print_parseable_fixits.
28081 (selftest::test_print_parseable_fixits_none): Update for new
28082 params of print_parseable_fixits.
28083 (selftest::test_print_parseable_fixits_insert): Likewise.
28084 (selftest::test_print_parseable_fixits_remove): Likewise.
28085 (selftest::test_print_parseable_fixits_replace): Likewise.
28086 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
28088 (selftest::diagnostic_c_tests): Call it.
28089 * diagnostic.h (enum diagnostics_extra_output_kind): New.
28090 (diagnostic_context::parseable_fixits_p): Delete field in favor
28092 (diagnostic_context::extra_output_kind): ...this new field.
28093 * doc/invoke.texi (Environment Variables): Add
28094 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
28095 * opts.c (common_handle_option): Update handling of
28096 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
28099 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28101 * tree-vect-slp-patterns.c (class complex_operations_pattern,
28102 complex_operations_pattern::matches,
28103 complex_operations_pattern::recognize,
28104 complex_operations_pattern::build): New.
28105 (slp_patterns): Use it.
28107 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28109 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
28110 * optabs.def (cmls_optab, cmls_conj_optab): New.
28111 * doc/md.texi: Document them.
28112 * tree-vect-slp-patterns.c (class complex_fms_pattern,
28113 complex_fms_pattern::matches, complex_fms_pattern::recognize,
28114 complex_fms_pattern::build): New.
28116 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28118 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
28119 * optabs.def (cmla_optab, cmla_conj_optab): New.
28120 * doc/md.texi: Document them.
28121 * tree-vect-slp-patterns.c (vect_match_call_p,
28122 class complex_fma_pattern, vect_slp_reset_pattern,
28123 complex_fma_pattern::matches, complex_fma_pattern::recognize,
28124 complex_fma_pattern::build): New.
28126 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28128 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
28129 * optabs.def (cmul_optab, cmul_conj_optab): New.
28130 * doc/md.texi: Document them.
28131 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
28132 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
28133 vect_build_combine_node, class complex_mul_pattern,
28134 complex_mul_pattern::matches, complex_mul_pattern::recognize,
28135 complex_mul_pattern::build): New.
28137 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28139 * tree-vect-slp.c (optimize_load_redistribution_1): New.
28140 (optimize_load_redistribution, vect_is_slp_load_node): New.
28141 (vect_match_slp_patterns): Use it.
28143 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28145 * tree-vect-slp-patterns.c (complex_add_pattern::build):
28148 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
28150 * config/gcn/mkoffload.c (main): Create an offload image only in
28151 64-bit configurations.
28153 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
28156 * config/i386/i386-options.c (ix86_option_override_internal):
28157 Issue an error for -fcf-protection with CF_BRANCH when compiling
28158 for 32-bit non-TARGET_CMOV targets.
28160 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
28163 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
28164 Remove declaration and initialization of shadow variable "ret".
28165 (ix86_option_override_internal): Remove delcaration of
28166 shadow variable "i". Redeclare shadowed variable to unsigned.
28167 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
28168 * config/i386/i386-builtins.c (get_builtin_code_for_version):
28169 Update for redeclaration.
28170 * config/i386/i386.h (pta_size): Ditto.
28172 2021-01-14 Richard Biener <rguenther@suse.de>
28174 PR tree-optimization/98674
28175 * tree-data-ref.c (base_supports_access_fn_components_p): New.
28176 (initialize_data_dependence_relation): For two bases without
28177 possible access fns resort to type size equality when determining
28178 shape compatibility.
28180 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
28183 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
28184 <=, >= operators in vcle and vcge intrinsics respectively.
28185 * config/arm/arm_neon_builtins.def: Remove entry for
28188 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
28191 * config/i386/i386-options.c (ix86_function_specific_save):
28192 Remove redundant assignment to opts->x_ix86_branch_cost.
28193 * config/i386/i386.c (ix86_prefetch_sse):
28194 Rename from x86_prefetch_sse. Update all uses.
28195 * config/i386/i386.h: Update for rename.
28196 * config/i386/i386-options.h: Ditto.
28198 2021-01-14 Jakub Jelinek <jakub@redhat.com>
28201 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
28202 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
28203 Use Bm instead of m for non-avx. Add isa attribute.
28205 2021-01-14 Jakub Jelinek <jakub@redhat.com>
28207 PR tree-optimization/96688
28208 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
28209 ~X can be simplified.
28211 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
28213 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
28214 IFN_LOAD_LANES results.
28216 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28218 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
28220 (aarch64_xtn<mode>): Likewise.
28221 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
28224 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
28226 (vmovl_s16): Likewise.
28227 (vmovl_s32): Likewise.
28228 (vmovl_u8): Likewise.
28229 (vmovl_u16): Likewise.
28230 (vmovl_u32): Likewise.
28231 (vmovn_s16): Likewise.
28232 (vmovn_s32): Likewise.
28233 (vmovn_s64): Likewise.
28234 (vmovn_u16): Likewise.
28235 (vmovn_u32): Likewise.
28236 (vmovn_u64): Likewise.
28238 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28240 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
28242 (aarch64_<su>qxtn2<mode>_be): Likewise.
28243 (aarch64_<su>qxtn2<mode>): Likewise.
28244 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
28246 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
28247 (su): Handle ss_truncate and us_truncate.
28248 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
28250 (vqmovn_high_s32): Likewise.
28251 (vqmovn_high_s64): Likewise.
28252 (vqmovn_high_u16): Likewise.
28253 (vqmovn_high_u32): Likewise.
28254 (vqmovn_high_u64): Likewise.
28256 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28258 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
28260 (aarch64_xtn2<mode>_be): Likewise.
28261 (aarch64_xtn2<mode>): Likewise.
28262 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
28264 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
28266 (vmovn_high_s32): Likewise.
28267 (vmovn_high_s64): Likewise.
28268 (vmovn_high_u16): Likewise.
28269 (vmovn_high_u32): Likewise.
28270 (vmovn_high_u64): Likewise.
28272 2021-01-13 Stafford Horne <shorne@gmail.com>
28274 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
28276 2021-01-13 Stafford Horne <shorne@gmail.com>
28278 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
28280 2021-01-13 Stafford Horne <shorne@gmail.com>
28282 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
28283 define for __or1k_hard_float__.
28285 2021-01-13 Stafford Horne <shorne@gmail.com>
28287 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
28288 (PROFILE_HOOK): Define to call _mcount.
28289 (FUNCTION_PROFILER): Change from abort to no-op.
28291 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28293 PR tree-optimization/96691
28294 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
28295 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
28296 (~D ^ C) or (D ^ C) can be simplified.
28298 2021-01-13 Richard Biener <rguenther@suse.de>
28300 PR tree-optimization/92645
28301 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
28302 until after vector lowering.
28304 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28306 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
28308 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
28309 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
28311 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28313 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
28315 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
28316 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
28318 2021-01-13 Richard Biener <rguenther@suse.de>
28320 PR tree-optimization/92645
28321 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
28322 BIT_FIELD_REF argument.
28323 (vect_build_slp_tree_2): Record the desired vector type
28324 on the external vector def.
28325 (vectorizable_slp_permutation): Handle required punning
28326 of existing vector defs.
28328 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28330 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
28332 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28334 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
28336 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
28338 * config.gcc [$target == *-*-gnu*]: Enable
28339 'default_gnu_indirect_function'.
28341 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28344 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
28345 registers before calling targetm.vectorize.vec_perm_const, only after
28347 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
28348 two argument permutation when one operand is zero vector and only
28349 after that force operands into registers.
28350 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
28351 define_insn_and_split pattern.
28352 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
28353 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
28354 (*avx2_zero_extendv8hiv8si2_1): Likewise.
28355 (*avx512f_zero_extendv8siv8di2_1): Likewise.
28356 (*avx2_zero_extendv4siv4di2_1): Likewise.
28357 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
28359 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
28360 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
28361 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
28362 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
28363 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
28364 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
28366 2021-01-13 Martin Liska <mliska@suse.cz>
28368 PR tree-optimization/98455
28369 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
28370 Record also virtual PHIs.
28371 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
28374 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
28376 * doc/invoke.texi (C++ Modules): Fix typos.
28378 2021-01-13 Richard Biener <rguenther@suse.de>
28380 PR tree-optimization/98640
28381 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
28382 handle plus or minus from a truncated operand to be
28385 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28388 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
28389 define_insn_and_split patterns.
28390 (splitter after *btr<mode>_2): New splitter.
28392 2021-01-13 Martin Liska <mliska@suse.cz>
28395 * cgraphunit.c (analyze_functions): Remove dead code.
28397 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
28399 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
28400 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
28401 (a64fx_regmove_cost, a64fx_vector_cost): New.
28402 (a64fx_tunings): Use the new added cost tables.
28404 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28407 * config/i386/predicates.md (pmovzx_parallel): New predicate.
28408 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
28409 define_insn_and_split pattern.
28410 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
28411 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
28413 2021-01-13 Julian Brown <julian@codesourcery.com>
28415 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
28416 to fix v0 register.
28418 2021-01-13 Julian Brown <julian@codesourcery.com>
28420 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
28423 2021-01-13 Julian Brown <julian@codesourcery.com>
28425 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
28426 for reciprocal-approximation instructions.
28427 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
28428 refinement and division result.
28429 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
28431 2021-01-13 Julian Brown <julian@codesourcery.com>
28433 * config/gcn/gcn-valu.md (subdf): Rename to...
28436 2021-01-12 Martin Liska <mliska@suse.cz>
28438 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
28440 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
28442 * function-abi.h: Fix typo.
28444 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
28448 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
28449 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
28450 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
28451 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
28452 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
28453 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
28454 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
28455 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
28456 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
28457 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
28458 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
28459 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
28460 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
28461 (ARM_HAVE_V2DI_LDST): Likewise.
28462 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
28463 (*movmisalign<mode>_mve_load): New pattern.
28464 * config/arm/neon.md (movmisalign<mode>): Move to ...
28465 * config/arm/vec-common.md: ... here.
28467 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
28470 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
28471 of pattern 'plus (plus (hard reg, const), pseudo)'.
28473 2021-01-12 Richard Biener <rguenther@suse.de>
28475 PR tree-optimization/98550
28476 * tree-vect-slp.c (vect_record_max_nunits): Check whether
28477 the group size is a multiple of the vector element count.
28478 (vect_build_slp_tree_1): When we need to fail because
28479 the vector type choosen causes unrolling do so lazily
28480 without affecting matches only at the end to guide group splitting.
28482 2021-01-12 Martin Liska <mliska@suse.cz>
28485 * optc-save-gen.awk: Compare also n_target_save vars with
28488 2021-01-12 Martin Liska <mliska@suse.cz>
28490 * gcov.c (source_info::debug): New.
28491 (print_usage): Add --debug (-D) option.
28492 (process_args): Likewise.
28493 (generate_results): Call src->debug after
28494 accumulate_line_counts.
28495 (read_graph_file): Properly assign id for EXIT_BLOCK.
28496 * profile.c (branch_prob): Dump function body before it is
28499 2021-01-12 Jakub Jelinek <jakub@redhat.com>
28501 PR tree-optimization/98629
28502 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
28503 unless returning non-zero.
28505 2021-01-12 Jakub Jelinek <jakub@redhat.com>
28507 PR tree-optimization/95731
28508 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
28509 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
28510 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
28511 only after optimize_range_tests_var_bound.
28513 2021-01-12 Jakub Jelinek <jakub@redhat.com>
28515 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
28516 * configure: Regenerated.
28518 2021-01-12 liuhongt <hongtao.liu@intel.com>
28521 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
28523 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
28526 2021-01-12 Alexandre Oliva <oliva@adacore.com>
28528 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
28530 (auto_end_imm_use_stmt_traverse): New struct.
28531 (FOR_EACH_IMM_USE_STMT): Use it.
28532 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
28534 * gimple-ssa-strength-reduction.c: ... here, ...
28535 * graphite-scop-detection.c: ... here, ...
28536 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
28537 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
28538 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
28539 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
28540 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
28541 * tree-vect-slp.c: ... and here, ...
28542 * doc/tree-ssa.texi: ... and the example here.
28544 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28546 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
28547 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
28548 (*sdiv_pow2<mode>3): New pattern.
28549 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
28550 Wrap the ASRD in an UNSPEC_PRED_X.
28551 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
28552 predicate with a constant PTRUE, if it isn't already.
28553 (*cond_<sve_int_op><mode>_z): Replace with...
28554 (*cond_<sve_int_op><mode>_any): ...this new pattern.
28556 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28558 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
28559 SVE_FULL_I to SVE_I.
28560 (*cond_bic<mode>_any): Likewise.
28562 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28564 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
28565 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
28568 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28570 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
28571 SVE_FULL_I to SVE_I.
28572 (*aarch64_cond_<su>abd<mode>_2): Likewise.
28573 (*aarch64_cond_<su>abd<mode>_any): Likewise.
28574 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
28575 for the max and min but not for the minus.
28576 (*aarch64_cond_<su>abd<mode>_3): New pattern.
28578 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28580 * config/aarch64/iterators.md (SVE_24I): New iterator.
28581 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
28582 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
28584 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28586 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
28587 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
28589 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
28590 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
28591 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
28592 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
28594 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28596 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
28597 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
28598 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
28601 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
28603 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
28604 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
28605 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
28607 2021-01-11 Martin Liska <mliska@suse.cz>
28610 * symtab-clones.h (clone_info::release): Release
28611 symtab::m_clones with ggc_delete as it's a GGC memory.
28613 2021-01-11 Matthias Klose <doko@ubuntu.com>
28615 * Makefile.in (LINK_PROGRESS): Show the link target.
28617 2021-01-11 Richard Biener <rguenther@suse.de>
28619 PR tree-optimization/91403
28620 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
28621 single-element interleaving group size at 4096 elements.
28623 2021-01-11 Richard Biener <rguenther@suse.de>
28625 PR tree-optimization/98526
28626 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
28627 of the actual reduction op for the regular case.
28628 (vectorizable_reduction): Cost the stmts
28629 vect_transform_reduction produces here.
28631 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
28633 * tree-ssa-forwprop.c (simplify_vector_constructor): For
28634 big-endian, use UNPACK[_FLOAT]_HI.
28636 2021-01-11 Tamar Christina <tamar.christina@arm.com>
28638 * tree-vect-slp-patterns.c (class complex_pattern,
28639 class complex_add_pattern): Add parameters to matches.
28640 (complex_add_pattern::build): Free memory.
28641 (complex_add_pattern::matches): Move validation end of match.
28642 (complex_add_pattern::recognize): Likewise.
28644 2021-01-11 Tamar Christina <tamar.christina@arm.com>
28646 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
28648 2021-01-11 Tamar Christina <tamar.christina@arm.com>
28650 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
28652 2021-01-11 Jakub Jelinek <jakub@redhat.com>
28654 PR tree-optimization/95867
28655 * tree-ssa-math-opts.h: New header.
28656 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
28657 (powi_as_mults): No longer static. Use build_one_cst instead of
28658 build_real. Formatting fix.
28659 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
28660 (attempt_builtin_powi): Handle multiplication reassociation without
28661 powi_fndecl using powi_as_mults.
28662 (reassociate_bb): For integral types don't require
28663 -funsafe-math-optimizations to call attempt_builtin_powi.
28665 2021-01-11 Jakub Jelinek <jakub@redhat.com>
28667 PR tree-optimization/95852
28668 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
28669 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
28670 allow in the bb any of the stmts in that vector, div_stmt and
28671 up to 3 cast stmts.
28672 (arith_cast_equal_p): New function.
28673 (arith_overflow_check_p): Add cast_stmt argument, handle signed
28674 multiply overflow checks.
28675 (match_arith_overflow): Adjust caller. Handle signed multiply
28678 2021-01-11 Jakub Jelinek <jakub@redhat.com>
28680 PR tree-optimization/95852
28681 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
28682 (uaddsub_overflow_check_p): Renamed to ...
28683 (arith_overflow_check_p): ... this. Handle also multiplication
28684 with overflow check.
28685 (match_uaddsub_overflow): Renamed to ...
28686 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
28687 also multiplication with overflow check. Adjust function comment.
28688 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
28689 match_arith_overflow also for MULT_EXPR.
28691 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28693 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
28694 __builtin_convertvector.
28695 (vmovl_s16): Likewise.
28696 (vmovl_s32): Likewise.
28697 (vmovl_u8): Likewise.
28698 (vmovl_u16): Likewise.
28699 (vmovl_u32): Likewise.
28700 (vmovn_s16): Likewise.
28701 (vmovn_s32): Likewise.
28702 (vmovn_s64): Likewise.
28703 (vmovn_u16): Likewise.
28704 (vmovn_u32): Likewise.
28705 (vmovn_u64): Likewise.
28707 2021-01-11 Martin Liska <mliska@suse.cz>
28709 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
28710 (if_chain::is_beneficial): Delete clusters
28711 (find_conditions): Make second argument of conditions_in_bbs a
28712 pointer so that we control over it's lifetime.
28713 (pass_if_to_switch::execute): Delete them.
28715 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
28717 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
28720 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28722 * config/vax/vax.md (cc): Remove mode attribute.
28723 (subst_<cc>, subst_f<cc>): Rename to...
28724 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
28725 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
28726 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
28727 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
28729 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28731 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
28732 `const_double_zero'.
28734 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28736 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
28737 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
28740 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
28742 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
28744 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
28745 with `const_double_zero'.
28746 * doc/rtl.texi (Constant Expression Types): Document it.
28748 2021-01-09 Jakub Jelinek <jakub@redhat.com>
28751 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
28752 POINTER_DIFF_EXPR to be any integral type.
28754 2021-01-09 Jakub Jelinek <jakub@redhat.com>
28756 PR rtl-optimization/98603
28757 * function.c (instantiate_virtual_regs_in_insn): For asm goto
28758 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
28759 if any, set ASM_OPERANDS mode to VOIDmode and change
28760 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
28762 2021-01-09 Alexandre Oliva <oliva@gnu.org>
28765 * final.c (notice_source_line): Narrow down the condition to
28766 skip a line-0 marker.
28768 2021-01-08 Sergei Trofimovich <siarheit@google.com>
28770 * ipa-modref.c (merge_call_side_effects): Fix
28771 linebreak split by reordering two print calls.
28773 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
28775 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
28776 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
28777 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
28778 consistency, fix constraint.
28780 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
28782 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
28783 callables instead of mask values.
28784 (struct target_flag_set_p): New predicate.
28785 (s390_cpu_cpp_builtins_internal): Define or undefine
28786 __LONG_DOUBLE_VX__ macro.
28788 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
28791 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
28792 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
28795 2021-01-08 Richard Biener <rguenther@suse.de>
28797 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
28799 2021-01-08 Richard Biener <rguenther@suse.de>
28801 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
28802 (vect_build_slp_tree): On cache hit release the matched
28803 scalar stmts vector.
28804 * tree-vect-stmts.c (vectorizable_store): Properly free
28805 vec_oprnds before possibly gathering them again.
28807 2021-01-08 Richard Biener <rguenther@suse.de>
28809 PR tree-optimization/98544
28810 * tree-vect-slp.c (vect_optimize_slp): Always materialize
28811 permutes at a permute node.
28813 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
28816 * config/i386/i386.c (x86_function_profiler): Use R10 to call
28817 mcount in large model. Sorry for large model with PIC.
28819 2021-01-08 Jakub Jelinek <jakub@redhat.com>
28822 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
28823 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
28824 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
28825 TargetSave and initialize for variables with enum types.
28826 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
28827 mstack-protector-guard-symbol=): Add Save.
28828 * config/i386/i386-options.c (ix86_function_specific_save,
28829 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
28830 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
28831 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
28832 x_ix86_veclibabi_type.
28834 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
28836 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
28837 SVE_FULL_I to SVE_I.
28838 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
28840 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
28842 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
28843 SVE_FULL_I to SVE_I.
28844 (*cond_uxt<mode>_any): Likewise.
28846 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28848 * config/aarch64/iterators.md (Vwhalf): New iterator.
28849 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
28851 (aarch64_<sur>adalp<mode>): ... This. Make more
28853 (<sur>sadv16qi): Adjust callsite of the above.
28854 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
28856 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
28858 (vpadal_s16): Likewise.
28859 (vpadal_u8): Likewise.
28860 (vpadal_u16): Likewise.
28861 (vpadalq_s8): Likewise.
28862 (vpadalq_s16): Likewise.
28863 (vpadalq_s32): Likewise.
28864 (vpadalq_u8): Likewise.
28865 (vpadalq_u16): Likewise.
28866 (vpadalq_u32): Likewise.
28868 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28870 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
28872 (aarch64_<su>abd<mode>): ... This.
28873 (<sur>sadv16qi): Adjust callsite of the above.
28874 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
28876 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
28878 (vabd_s16): Likewise.
28879 (vabd_s32): Likewise.
28880 (vabd_u8): Likewise.
28881 (vabd_u16): Likewise.
28882 (vabd_u32): Likewise.
28883 (vabdq_s8): Likewise.
28884 (vabdq_s16): Likewise.
28885 (vabdq_s32): Likewise.
28886 (vabdq_u8): Likewise.
28887 (vabdq_u16): Likewise.
28888 (vabdq_u32): Likewise.
28890 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28892 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
28894 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
28895 (vaba_s16): Likewise.
28896 (vaba_s32): Likewise.
28897 (vaba_u8): Likewise.
28898 (vaba_u16): Likewise.
28899 (vaba_u32): Likewise.
28900 (vabaq_s8): Likewise.
28901 (vabaq_s16): Likewise.
28902 (vabaq_s32): Likewise.
28903 (vabaq_u8): Likewise.
28904 (vabaq_u16): Likewise.
28905 (vabaq_u32): Likewise.
28907 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28909 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
28910 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
28911 Change RTL pattern to match.
28913 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
28915 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
28916 * config/riscv/riscv-c.c (riscv-subset.h): New.
28917 (INCLUDE_STRING): Define.
28918 (riscv_cpu_cpp_builtins): Add new style architecture extension
28920 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
28921 (riscv_subset_list::end): New.
28922 (riscv_current_subset_list): New.
28924 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
28926 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
28927 Move to riscv-subset.h.
28928 (struct riscv_subset_t): Ditto.
28929 (class riscv_subset_list): Ditto.
28930 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
28931 from riscv-common.c.
28932 (struct riscv_subset_t): Ditto.
28933 (class riscv_subset_list): Ditto.
28934 * config/riscv/t-riscv ($(common_out_file)): Add file
28937 2021-01-07 Jakub Jelinek <jakub@redhat.com>
28940 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
28941 New define_insn patterns.
28943 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
28945 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
28946 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
28947 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
28949 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
28951 PR tree-optimization/98560
28952 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
28953 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
28955 (vec_cond_direct): Likewise argument 2.
28956 (vec_condu_direct, vec_condeq_direct): Delete.
28957 (expand_vect_cond_optab_fn): Rename to...
28958 (expand_vec_cond_optab_fn): ...this, replacing old macro.
28959 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
28960 (expand_vect_cond_mask_optab_fn): Rename to...
28961 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
28962 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
28964 (direct_vec_cond_optab_supported_p): Likewise.
28965 (direct_vec_condu_optab_supported_p): Delete.
28966 (direct_vec_condeq_optab_supported_p): Delete.
28967 * gimple-isel.cc: Include internal-fn.h.
28968 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
28971 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
28973 PR tree-optimization/98560
28974 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
28975 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
28977 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
28979 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
28980 rotate_insn and optab code attributes.
28981 Update all uses to merged code attribute.
28982 * config/i386/sse.md: Update all uses to merged code attribute.
28983 * config/i386/mmx.md: Update all uses to merged code attribute.
28985 2021-01-07 Jakub Jelinek <jakub@redhat.com>
28987 PR tree-optimization/98568
28988 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
28989 (bswap_replace): Use it.
28991 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
28993 PR rtl-optimization/97978
28994 * lra-int.h (lra_hard_reg_split_p): New external.
28995 * lra.c (lra_hard_reg_split_p): New global.
28996 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
28997 * lra-assigns.c (lra_assign): Don't check allocation correctness
28998 after hard reg splitting.
29000 2021-01-06 Martin Sebor <msebor@redhat.com>
29003 * builtins.c (new_delete_mismatch_p): New overload.
29004 (new_delete_mismatch_p (tree, tree)): Call it.
29006 2021-01-06 Alexandre Oliva <oliva@adacore.com>
29008 * Makefile.in (T_GLIMITS_H): New.
29009 (stmp-int-hdrs): Depend on it, use it.
29010 * config/t-vxworks (T_GLIMITS_H): Override it.
29011 (vxw-glimits.h): New.
29013 2021-01-06 Richard Biener <rguenther@suse.de>
29015 PR tree-optimization/98513
29016 * value-range.cc (intersect_ranges): Compare the upper bounds
29017 for the expected relation.
29019 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
29022 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
29024 * doc/standards.texi (HSAIL): Remove section.
29026 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
29028 * configure: Re-generate.
29030 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29032 * doc/invoke.texi (-std=c++20): Adjust for the publication of
29033 ISO 14882:2020 standard.
29034 * doc/standards.texi: Likewise.
29036 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29038 PR tree-optimization/94802
29039 * expr.h (maybe_optimize_sub_cmp_0): Declare.
29040 * expr.c: Include tree-pretty-print.h and flags.h.
29041 (maybe_optimize_sub_cmp_0): New function.
29042 (do_store_flag): Use it.
29043 * cfgexpand.c (expand_gimple_cond): Likewise.
29045 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29047 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
29048 * rtlanal.c (simple_regno_set): Tweak description to clarify the
29051 2021-01-05 Richard Biener <rguenther@suse.de>
29053 PR tree-optimization/98516
29054 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
29055 lanes when materializing on a VEC_PERM node.
29056 (vectorizable_slp_permutation): Dump the permute properly.
29058 2021-01-05 Richard Biener <rguenther@suse.de>
29060 * tree-vect-slp.c (vect_slp_region): Move debug counter
29061 to cover individual subgraphs.
29063 2021-01-05 Richard Biener <rguenther@suse.de>
29065 PR tree-optimization/98428
29066 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
29067 vector lane extracts for loop vectorization.
29069 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29071 PR tree-optimization/98514
29072 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
29074 (operand_rank): Change type from hash_map<tree, long> to
29075 hash_map<tree, int64_t>.
29076 (phi_rank): Change return type from long to int64_t.
29077 (loop_carried_phi): Change block_rank variable type from long to
29079 (propagate_rank): Change return type, rank parameter type and
29080 op_rank variable type from long to int64_t.
29081 (find_operand_rank): Change return type from long to int64_t
29082 and change slot variable type from long * to int64_t *.
29083 (insert_operand_rank): Change rank parameter type from long to
29085 (get_rank): Change return type and rank variable type from long to
29086 int64_t. Use PRId64 instead of ld to print the rank.
29087 (init_reassoc): Change rank variable type from long to int64_t
29088 and adjust correspondingly bb_rank and operand_rank initialization.
29090 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29092 PR tree-optimization/96928
29093 * tree-ssa-phiopt.c (xor_replacement): New function.
29094 (tree_ssa_phiopt_worker): Call it.
29096 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29098 PR tree-optimization/96930
29099 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
29100 from narrower value which has the same type as 1 << B, perform
29101 the right shift on the narrower value followed by extension.
29103 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29105 PR tree-optimization/96239
29106 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
29108 (get_status_for_store_merging): Don't return BB_INVALID for blocks
29109 with potential bswap optimizable CONSTRUCTORs.
29110 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
29113 2021-01-05 Richard Biener <rguenther@suse.de>
29115 PR tree-optimization/98381
29116 * tree.c (vector_element_bits): Properly compute bool vector
29118 * tree-vect-loop.c (vectorizable_live_operation): Properly
29119 compute the last lane bit offset.
29121 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
29124 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
29125 Clear the top 64 bytes of the input XMM register.
29126 (sse_cvttps2pi): Ditto.
29128 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
29131 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
29133 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
29136 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
29139 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
29141 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
29142 (umaddsidi4_split): Likewise.
29144 2021-01-05 liuhongt <hongtao.liu@intel.com>
29147 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
29148 define_insn_and_split for zero_extend of subreg HI of pmovskb
29150 (*sse2_pmovskb_zexthisi): Add new combine splitters for
29151 zero_extend of not of subreg HI of pmovskb result.
29153 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29156 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
29158 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
29159 convert_memory_address to convert symbolic immediates to ptr_mode
29160 before forcing them to memory.
29162 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29164 PR rtl-optimization/97144
29165 * recog.c (constrain_operands): Initialize matching_operand
29166 for each alternative, rather than only doing it once.
29168 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29170 PR rtl-optimization/98403
29171 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
29172 why we don't remove call clobbers.
29173 (function_info::apply_changes_to_insn): Don't attempt to add
29174 call clobbers here.
29176 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29178 PR tree-optimization/98371
29179 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
29180 (vect_analyze_loop): If an epilogue loop appears to be cheaper
29181 than the main loop, re-analyze it as a main loop before adopting
29184 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29187 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
29188 * aclocal.m4, configure: Regenerate.
29189 * Makefile.in (NETLIBS): Define.
29190 (BACKEND): Remove $(CODYLIB).
29192 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29194 PR rtl-optimization/98334
29195 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
29196 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
29198 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
29200 * tree-inline.c (expand_call_inline): Restore input_location.
29201 Return result from recursive call.
29203 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29205 PR tree-optimization/95401
29206 * config/aarch64/aarch64-sve-builtins.cc
29207 (gimple_folder::load_store_cookie): Use bits rather than bytes
29208 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
29209 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
29210 * tree-vect-stmts.c (vectorizable_store): Likewise.
29211 (vectorizable_load): Likewise.
29213 2021-01-04 Richard Biener <rguenther@suse.de>
29215 PR tree-optimization/98308
29216 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
29219 2021-01-04 Jakub Jelinek <jakub@redhat.com>
29221 PR tree-optimization/95771
29222 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
29223 with precision smaller than int's precision and types with precision
29224 twice as large as long long. Formatting fixes.
29226 2021-01-04 Richard Biener <rguenther@suse.de>
29228 PR tree-optimization/98464
29229 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
29230 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
29231 (process_bb): Adjust.
29233 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
29236 * doc/invoke.texi (-fsanitize=address): Fix wording describing
29237 clash with -fsanitize=hwaddress.
29239 2021-01-04 Richard Biener <rguenther@suse.de>
29241 PR tree-optimization/98282
29242 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
29243 invariants as VN_NARY.
29245 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29248 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
29249 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
29250 to handle zero operands.
29252 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29254 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
29255 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
29256 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
29259 2021-01-04 Richard Biener <rguenther@suse.de>
29261 PR tree-optimization/98393
29262 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
29263 when hitting the limit.
29265 2021-01-04 Richard Biener <rguenther@suse.de>
29267 PR tree-optimization/98291
29268 * tree-vect-loop.c (vectorizable_reduction): Bypass
29269 associativity check for SLP reductions with VF 1.
29271 2021-01-04 Jakub Jelinek <jakub@redhat.com>
29273 PR tree-optimization/96782
29274 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
29276 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
29278 * collect-utils.c (collect_execute): Check dumppfx.
29279 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
29280 to collect_execute.
29281 (do_link): Add new parameter atsuffix.
29282 (main): Handle -dumpdir option. Skip one argument for
29283 -o, -isystem and -B options.
29284 * gcc.c (make_at_file): New helper function.
29285 (close_at_file): Use it.
29287 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29289 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
29290 Amend handling for LD64_VERSION fallback defaults.
29292 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29294 * config.gcc: Compute default version information
29295 from the configured target. Likewise defaults for
29297 * config/darwin10.h: Removed.
29298 * config/darwin12.h: Removed.
29299 * config/darwin9.h: Removed.
29300 * config/rs6000/darwin8.h: Removed.
29302 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29304 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
29306 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29308 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
29309 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
29311 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29313 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
29315 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
29317 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29319 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
29320 for the Darwin10 unwinder stub from here ...
29321 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
29323 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29325 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
29326 (ASM_DEBUG_SPEC):Only define if the assembler supports
29328 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
29329 (DARWIN_PREFER_DWARF): Define.
29330 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
29331 (DARWIN_PREFER_DWARF): Likewise
29332 (DSYMUTIL_SPEC): Likewise.
29333 (COLLECT_RUN_DSYMUTIL): Likewise.
29334 (ASM_DEBUG_SPEC): Likewise.
29335 (ASM_DEBUG_OPTION_SPEC): Likewise.
29337 2021-01-02 Jan Hubicka <jh@suse.cz>
29339 * cfg.c (free_block): ggc_free bb.
29341 2021-01-01 Jakub Jelinek <jakub@redhat.com>
29343 * gcc.c (process_command): Update copyright notice dates.
29344 * gcov-dump.c (print_version): Ditto.
29345 * gcov.c (print_version): Ditto.
29346 * gcov-tool.c (print_version): Ditto.
29347 * gengtype.c (create_file): Ditto.
29348 * doc/cpp.texi: Bump @copying's copyright year.
29349 * doc/cppinternals.texi: Ditto.
29350 * doc/gcc.texi: Ditto.
29351 * doc/gccint.texi: Ditto.
29352 * doc/gcov.texi: Ditto.
29353 * doc/install.texi: Ditto.
29354 * doc/invoke.texi: Ditto.
29356 2021-01-01 Jakub Jelinek <jakub@redhat.com>
29358 * ChangeLog-2020: Rotate ChangeLog. New file.
29361 Copyright (C) 2021 Free Software Foundation, Inc.
29363 Copying and distribution of this file, with or without modification,
29364 are permitted in any medium without royalty provided the copyright
29365 notice and this notice are preserved.