1 2020-03-17 Andre Vieira <andre.simoesdiasvieira@arm.com>
2 Mihail Ionescu <mihail.ionescu@arm.com>
3 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
5 * config/arm/arm-builtins.c (UNOP_NONE_NONE_QUALIFIERS): Define macro.
6 (UNOP_NONE_SNONE_QUALIFIERS): Likewise.
7 (UNOP_NONE_UNONE_QUALIFIERS): Likewise.
8 * config/arm/arm_mve.h (vrndxq_f16): Define macro.
9 (vrndxq_f32): Likewise.
11 (vrndq_f32): Likewise.
12 (vrndpq_f16): Likewise.
13 (vrndpq_f32): Likewise.
14 (vrndnq_f16): Likewise.
15 (vrndnq_f32): Likewise.
16 (vrndmq_f16): Likewise.
17 (vrndmq_f32): Likewise.
18 (vrndaq_f16): Likewise.
19 (vrndaq_f32): Likewise.
20 (vrev64q_f16): Likewise.
21 (vrev64q_f32): Likewise.
22 (vnegq_f16): Likewise.
23 (vnegq_f32): Likewise.
24 (vdupq_n_f16): Likewise.
25 (vdupq_n_f32): Likewise.
26 (vabsq_f16): Likewise.
27 (vabsq_f32): Likewise.
28 (vrev32q_f16): Likewise.
29 (vcvttq_f32_f16): Likewise.
30 (vcvtbq_f32_f16): Likewise.
31 (vcvtq_f16_s16): Likewise.
32 (vcvtq_f32_s32): Likewise.
33 (vcvtq_f16_u16): Likewise.
34 (vcvtq_f32_u32): Likewise.
35 (__arm_vrndxq_f16): Define intrinsic.
36 (__arm_vrndxq_f32): Likewise.
37 (__arm_vrndq_f16): Likewise.
38 (__arm_vrndq_f32): Likewise.
39 (__arm_vrndpq_f16): Likewise.
40 (__arm_vrndpq_f32): Likewise.
41 (__arm_vrndnq_f16): Likewise.
42 (__arm_vrndnq_f32): Likewise.
43 (__arm_vrndmq_f16): Likewise.
44 (__arm_vrndmq_f32): Likewise.
45 (__arm_vrndaq_f16): Likewise.
46 (__arm_vrndaq_f32): Likewise.
47 (__arm_vrev64q_f16): Likewise.
48 (__arm_vrev64q_f32): Likewise.
49 (__arm_vnegq_f16): Likewise.
50 (__arm_vnegq_f32): Likewise.
51 (__arm_vdupq_n_f16): Likewise.
52 (__arm_vdupq_n_f32): Likewise.
53 (__arm_vabsq_f16): Likewise.
54 (__arm_vabsq_f32): Likewise.
55 (__arm_vrev32q_f16): Likewise.
56 (__arm_vcvttq_f32_f16): Likewise.
57 (__arm_vcvtbq_f32_f16): Likewise.
58 (__arm_vcvtq_f16_s16): Likewise.
59 (__arm_vcvtq_f32_s32): Likewise.
60 (__arm_vcvtq_f16_u16): Likewise.
61 (__arm_vcvtq_f32_u32): Likewise.
62 (vrndxq): Define polymorphic variants.
72 (vcvtbq_f32): Likewise.
73 (vcvttq_f32): Likewise.
75 * config/arm/arm_mve_builtins.def (VAR2): Define.
77 * config/arm/mve.md (mve_vrndxq_f<mode>): Add RTL pattern.
78 (mve_vrndq_f<mode>): Likewise.
79 (mve_vrndpq_f<mode>): Likewise.
80 (mve_vrndnq_f<mode>): Likewise.
81 (mve_vrndmq_f<mode>): Likewise.
82 (mve_vrndaq_f<mode>): Likewise.
83 (mve_vrev64q_f<mode>): Likewise.
84 (mve_vnegq_f<mode>): Likewise.
85 (mve_vdupq_n_f<mode>): Likewise.
86 (mve_vabsq_f<mode>): Likewise.
87 (mve_vrev32q_fv8hf): Likewise.
88 (mve_vcvttq_f32_f16v4sf): Likewise.
89 (mve_vcvtbq_f32_f16v4sf): Likewise.
90 (mve_vcvtq_to_f_<supf><mode>): Likewise.
92 2020-03-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
93 Mihail Ionescu <mihail.ionescu@arm.com>
94 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
96 * config/arm/arm-builtins.c (CF): Define mve_builtin_data.
98 (ARM_BUILTIN_MVE_PATTERN_START): Define.
99 (arm_init_mve_builtins): Define function.
100 (arm_init_builtins): Add TARGET_HAVE_MVE check.
101 (arm_expand_builtin_1): Check the range of fcode.
102 (arm_expand_mve_builtin): Define function to expand MVE builtins.
103 (arm_expand_builtin): Check the range of fcode.
104 * config/arm/arm_mve.h (__ARM_FEATURE_MVE): Define MVE floating point
106 (__ARM_MVE_PRESERVE_USER_NAMESPACE): Define to protect user namespace.
107 (vst4q_s8): Define macro.
108 (vst4q_s16): Likewise.
109 (vst4q_s32): Likewise.
110 (vst4q_u8): Likewise.
111 (vst4q_u16): Likewise.
112 (vst4q_u32): Likewise.
113 (vst4q_f16): Likewise.
114 (vst4q_f32): Likewise.
115 (__arm_vst4q_s8): Define inline builtin.
116 (__arm_vst4q_s16): Likewise.
117 (__arm_vst4q_s32): Likewise.
118 (__arm_vst4q_u8): Likewise.
119 (__arm_vst4q_u16): Likewise.
120 (__arm_vst4q_u32): Likewise.
121 (__arm_vst4q_f16): Likewise.
122 (__arm_vst4q_f32): Likewise.
123 (__ARM_mve_typeid): Define macro with MVE types.
124 (__ARM_mve_coerce): Define macro with _Generic feature.
125 (vst4q): Define polymorphic variant for different vst4q builtins.
126 * config/arm/arm_mve_builtins.def: New file.
127 * config/arm/iterators.md (VSTRUCT): Modify to allow XI and OI
129 * config/arm/mve.md (MVE_VLD_ST): Define iterator.
130 (unspec): Define unspec.
131 (mve_vst4q<mode>): Define RTL pattern.
132 * config/arm/neon.md (mov<mode>): Modify expand to allow XI and OI
134 (neon_mov<mode>): Modify RTL define_insn to allow XI and OI modes
136 (define_split): Allow OI mode split for MVE after reload.
137 (define_split): Allow XI mode split for MVE after reload.
138 * config/arm/t-arm (arm.o): Add entry for arm_mve_builtins.def.
139 (arm-builtins.o): Likewise.
141 2020-03-17 Christophe Lyon <christophe.lyon@linaro.org>
143 * c-typeck.c (process_init_element): Handle constructor_type with
144 type size represented by POLY_INT_CST.
146 2020-03-17 Jakub Jelinek <jakub@redhat.com>
148 PR tree-optimization/94187
149 * tree-ssa-strlen.c (count_nonzero_bytes): Punt if
150 nchars - offset < nbytes.
153 * builtins.c (expand_builtin_strnlen): Do return NULL_RTX if we would
154 emit a warning if it was enabled and don't depend on TREE_NO_WARNING
157 2020-03-16 Vladimir Makarov <vmakarov@redhat.com>
160 * lra-spills.c (remove_pseudos): Do not reuse insn alternative
161 after changing memory subreg.
163 2020-03-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
164 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
166 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Modify function to add
167 emulator calls for dobule precision arithmetic operations for MVE.
169 2020-03-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
170 Mihail Ionescu <mihail.ionescu@arm.com>
171 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
173 * common/config/arm/arm-common.c (arm_asm_auto_mfpu): When vfp_base
174 feature bit is on and -mfpu=auto is passed as compiler option, do not
175 generate error on not finding any matching fpu. Because in this case
177 * config/arm/arm-cpus.in (vfp_base): Define feature bit, this bit is
178 enabled for MVE and also for all VFP extensions.
179 (VFPv2): Modify fgroup to enable vfp_base feature bit when ever VFPv2
181 (MVE): Define fgroup to enable feature bits mve, vfp_base and armv7em.
182 (MVE_FP): Define fgroup to enable feature bits is fgroup MVE and FPv5
183 along with feature bits mve_float.
184 (mve): Modify add options in armv8.1-m.main arch for MVE.
185 (mve.fp): Modify add options in armv8.1-m.main arch for MVE with
187 * config/arm/arm.c (use_return_insn): Replace the
188 check with TARGET_VFP_BASE.
189 (thumb2_legitimate_index_p): Replace TARGET_HARD_FLOAT with
191 (arm_rtx_costs_internal): Replace "TARGET_HARD_FLOAT || TARGET_HAVE_MVE"
192 with TARGET_VFP_BASE, to allow cost calculations for copies in MVE as
194 (arm_get_vfp_saved_size): Replace TARGET_HARD_FLOAT with
195 TARGET_VFP_BASE, to allow space calculation for VFP registers in MVE
197 (arm_compute_frame_layout): Likewise.
198 (arm_save_coproc_regs): Likewise.
199 (arm_fixed_condition_code_regs): Modify to enable using VFPCC_REGNUM
201 (arm_hard_regno_mode_ok): Replace "TARGET_HARD_FLOAT || TARGET_HAVE_MVE"
202 with equivalent macro TARGET_VFP_BASE.
203 (arm_expand_epilogue_apcs_frame): Likewise.
204 (arm_expand_epilogue): Likewise.
205 (arm_conditional_register_usage): Likewise.
206 (arm_declare_function_name): Add check to skip printing .fpu directive
207 in assembly file when TARGET_VFP_BASE is enabled and fpu_to_print is
209 * config/arm/arm.h (TARGET_VFP_BASE): Define.
210 * config/arm/arm.md (arch): Add "mve" to arch.
211 (eq_attr "arch" "mve"): Enable on TARGET_HAVE_MVE is true.
212 (vfp_pop_multiple_with_writeback): Replace "TARGET_HARD_FLOAT
213 || TARGET_HAVE_MVE" with equivalent macro TARGET_VFP_BASE.
214 * config/arm/constraints.md (Uf): Define to allow modification to FPCCR
216 * config/arm/thumb2.md (thumb2_movsfcc_soft_insn): Modify target guard
217 to not allow for MVE.
218 * config/arm/unspecs.md (UNSPEC_GET_FPSCR): Move to volatile unspecs
220 (VUNSPEC_GET_FPSCR): Define.
221 * config/arm/vfp.md (thumb2_movhi_vfp): Add support for VMSR and VMRS
222 instructions which move to general-purpose Register from Floating-point
223 Special register and vice-versa.
224 (thumb2_movhi_fp16): Likewise.
225 (thumb2_movsi_vfp): Add support for VMSR and VMRS instructions along
226 with MCR and MRC instructions which set and get Floating-point Status
227 and Control Register (FPSCR).
228 (movdi_vfp): Modify pattern to enable Single-precision scalar float move
230 (thumb2_movdf_vfp): Modify pattern to enable Double-precision scalar
231 float move patterns in MVE.
232 (thumb2_movsfcc_vfp): Modify pattern to enable single float conditional
233 code move patterns of VFP also in MVE by adding TARGET_VFP_BASE check.
234 (thumb2_movdfcc_vfp): Modify pattern to enable double float conditional
235 code move patterns of VFP also in MVE by adding TARGET_VFP_BASE check.
236 (push_multi_vfp): Add support to use VFP VPUSH pattern for MVE by adding
237 TARGET_VFP_BASE check.
238 (set_fpscr): Add support to set FPSCR register for MVE. Modify pattern
239 using VFPCC_REGNUM as few MVE intrinsics use carry bit of FPSCR
241 (get_fpscr): Add support to get FPSCR register for MVE. Modify pattern
242 using VFPCC_REGNUM as few MVE intrinsics use carry bit of FPSCR
246 2020-03-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
247 Mihail Ionescu <mihail.ionescu@arm.com>
248 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
250 * config.gcc (arm_mve.h): Include mve intrinsics header file.
251 * config/arm/aout.h (p0): Add new register name for MVE predicated
253 * config/arm-builtins.c (ARM_BUILTIN_SIMD_LANE_CHECK): Define macro
254 common to Neon and MVE.
255 (ARM_BUILTIN_NEON_LANE_CHECK): Renamed to ARM_BUILTIN_SIMD_LANE_CHECK.
256 (arm_init_simd_builtin_types): Disable poly types for MVE.
257 (arm_init_neon_builtins): Move a check to arm_init_builtins function.
258 (arm_init_builtins): Use ARM_BUILTIN_SIMD_LANE_CHECK instead of
259 ARM_BUILTIN_NEON_LANE_CHECK.
260 (mve_dereference_pointer): Add function.
261 (arm_expand_builtin_args): Call to mve_dereference_pointer when MVE is
263 (arm_expand_neon_builtin): Moved to arm_expand_builtin function.
264 (arm_expand_builtin): Moved from arm_expand_neon_builtin function.
265 * config/arm/arm-c.c (__ARM_FEATURE_MVE): Define macro for MVE and MVE
266 with floating point enabled.
267 * config/arm/arm-protos.h (neon_immediate_valid_for_move): Renamed to
268 simd_immediate_valid_for_move.
269 (simd_immediate_valid_for_move): Renamed from
270 neon_immediate_valid_for_move function.
271 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Generate
272 error if vfpv2 feature bit is disabled and mve feature bit is also
273 disabled for HARD_FLOAT_ABI.
274 (use_return_insn): Check to not push VFP regs for MVE.
275 (aapcs_vfp_allocate): Add MVE check to have same Procedure Call Standard
277 (aapcs_vfp_allocate_return_reg): Likewise.
278 (thumb2_legitimate_address_p): Check to return 0 on valid Thumb-2
279 address operand for MVE.
280 (arm_rtx_costs_internal): MVE check to determine cost of rtx.
281 (neon_valid_immediate): Rename to simd_valid_immediate.
282 (simd_valid_immediate): Rename from neon_valid_immediate.
283 (simd_valid_immediate): MVE check on size of vector is 128 bits.
284 (neon_immediate_valid_for_move): Rename to
285 simd_immediate_valid_for_move.
286 (simd_immediate_valid_for_move): Rename from
287 neon_immediate_valid_for_move.
288 (neon_immediate_valid_for_logic): Modify call to neon_valid_immediate
290 (neon_make_constant): Modify call to neon_valid_immediate function.
291 (neon_vector_mem_operand): Return VFP register for POST_INC or PRE_DEC
293 (output_move_neon): Add MVE check to generate vldm/vstm instrcutions.
294 (arm_compute_frame_layout): Calculate space for saved VFP registers for
296 (arm_save_coproc_regs): Save coproc registers for MVE.
297 (arm_print_operand): Add case 'E' to print memory operands for MVE.
298 (arm_print_operand_address): Check to print register number for MVE.
299 (arm_hard_regno_mode_ok): Check for arm hard regno mode ok for MVE.
300 (arm_modes_tieable_p): Check to allow structure mode for MVE.
301 (arm_regno_class): Add VPR_REGNUM check.
302 (arm_expand_epilogue_apcs_frame): MVE check to calculate epilogue code
304 (arm_expand_epilogue): MVE check for enabling pop instructions in
306 (arm_print_asm_arch_directives): Modify function to disable print of
307 .arch_extension "mve" and "fp" for cases where MVE is enabled with
309 (arm_vector_mode_supported_p): Check for modes available in MVE interger
310 and MVE floating point.
311 (arm_array_mode_supported_p): Add TARGET_HAVE_MVE check for array mode
313 (arm_conditional_register_usage): Enable usage of conditional regsiter
315 (fixed_regs[VPR_REGNUM]): Enable VPR_REG for MVE.
316 (arm_declare_function_name): Modify function to disable print of
317 .arch_extension "mve" and "fp" for cases where MVE is enabled with
319 * config/arm/arm.h (TARGET_HAVE_MVE): Disable for soft float abi and
320 when target general registers are required.
321 (TARGET_HAVE_MVE_FLOAT): Likewise.
322 (FIXED_REGISTERS): Add bit for VFP_REG class which is enabled in arm.c
324 (CALL_USED_REGISTERS): Set bit for VFP_REG class in CALL_USED_REGISTERS
325 which indicate this is not available for across function calls.
326 (FIRST_PSEUDO_REGISTER): Modify.
327 (VALID_MVE_MODE): Define valid MVE mode.
328 (VALID_MVE_SI_MODE): Define valid MVE SI mode.
329 (VALID_MVE_SF_MODE): Define valid MVE SF mode.
330 (VALID_MVE_STRUCT_MODE): Define valid MVE struct mode.
331 (VPR_REGNUM): Add Vector Predication Register in arm_regs_in_sequence
333 (IS_VPR_REGNUM): Macro to check for VPR_REG register.
334 (REG_ALLOC_ORDER): Add VPR_REGNUM entry.
335 (enum reg_class): Add VPR_REG entry.
336 (REG_CLASS_NAMES): Add VPR_REG entry.
337 * config/arm/arm.md (VPR_REGNUM): Define.
338 (conds): Check is_mve_type attrbiute to differentiate "conditional" and
339 "unconditional" instructions.
340 (arm_movsf_soft_insn): Modify RTL to not allow for MVE.
341 (movdf_soft_insn): Modify RTL to not allow for MVE.
342 (vfp_pop_multiple_with_writeback): Enable for MVE.
343 (include "mve.md"): Include mve.md file.
344 * config/arm/arm_mve.h: Add MVE intrinsics head file.
345 * config/arm/constraints.md (Up): Constraint to enable "p0" register in MVE
346 for vector predicated operands.
347 * config/arm/iterators.md (VNIM1): Define.
349 (VHFBF_split): Define
350 * config/arm/mve.md: New file.
351 (mve_mov<mode>): Define RTL for move, store and load in MVE.
352 (mve_mov<mode>): Define move RTL pattern with vec_duplicate operator for
354 * config/arm/neon.md (neon_immediate_valid_for_move): Rename with
355 simd_immediate_valid_for_move.
356 (neon_mov<mode>): Split pattern and move expand pattern "movv8hf" which
357 is common to MVE and NEON to vec-common.md file.
358 (vec_init<mode><V_elem_l>): Add TARGET_HAVE_MVE check.
359 * config/arm/predicates.md (vpr_register_operand): Define.
360 * config/arm/t-arm: Add mve.md file.
361 * config/arm/types.md (mve_move): Add MVE instructions mve_move to
363 (mve_store): Add MVE instructions mve_store to attribute "type".
364 (mve_load): Add MVE instructions mve_load to attribute "type".
365 (is_mve_type): Define attribute.
366 * config/arm/vec-common.md (mov<mode>): Modify RTL expand to support
367 standard move patterns in MVE along with NEON and IWMMXT with mode
369 (mov<mode>): Modify RTL expand to support standard move patterns in NEON
370 and IWMMXT with mode iterator V8HF.
371 (movv8hf): Define RTL expand to support standard "movv8hf" pattern in
373 * config/arm/vfp.md (neon_immediate_valid_for_move): Rename to
374 simd_immediate_valid_for_move.
377 2020-03-16 H.J. Lu <hongjiu.lu@intel.com>
380 * config/i386/i386.md (*movsi_internal): Call ix86_output_ssemov
381 for TYPE_SSEMOV. Remove ext_sse_reg_operand and TARGET_AVX512VL
383 * config/i386/predicates.md (ext_sse_reg_operand): Removed.
385 2020-03-16 Jakub Jelinek <jakub@redhat.com>
388 * tree-inline.c (insert_init_stmt): Don't gimple_regimplify_operands
391 PR tree-optimization/94166
392 * tree-ssa-reassoc.c (sort_by_mach_mode): Use SSA_NAME_VERSION
393 as secondary comparison key.
395 2020-03-16 Bin Cheng <bin.cheng@linux.alibaba.com>
397 PR tree-optimization/94125
398 * tree-loop-distribution.c
399 (loop_distribution::break_alias_scc_partitions): Update post order
400 number for merged scc.
402 2020-03-15 H.J. Lu <hongjiu.lu@intel.com>
405 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_SI and
407 * config/i386/i386.md (*movsf_internal): Call ix86_output_ssemov
408 for TYPE_SSEMOV. Remove TARGET_PREFER_AVX256, TARGET_AVX512VL
409 and ext_sse_reg_operand check.
411 2020-03-15 Lewis Hyatt <lhyatt@gmail.com>
413 * common.opt: Avoid redundancy in the help text.
414 * config/arc/arc.opt: Likewise.
415 * config/cr16/cr16.opt: Likewise.
417 2020-03-14 Jakub Jelinek <jakub@redhat.com>
420 * tree-nested.c (convert_nonlocal_omp_clauses,
421 convert_local_omp_clauses): Handle {,in_,task_}reduction clauses
422 with C/C++ array sections.
424 2020-03-14 H.J. Lu <hongjiu.lu@intel.com>
427 * config/i386/i386.md (*movdi_internal): Call ix86_output_ssemov
428 for TYPE_SSEMOV. Remove ext_sse_reg_operand and TARGET_AVX512VL
431 2020-03-14 Jakub Jelinek <jakub@redhat.com>
433 * gimple-fold.c (gimple_fold_builtin_strncpy): Change
434 "a an" to "an" in a comment.
435 * hsa-common.h (is_a_helper): Likewise.
436 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Likewise.
437 * config/arc/arc.c (arc600_corereg_hazard): Likewise.
438 * config/s390/s390.c (s390_indirect_branch_via_thunk): Likewise.
440 2020-03-13 Aaron Sawdey <acsawdey@linux.ibm.com>
443 * config/rs6000/rs6000.c (num_insns_constant_multi): Don't shift a
444 64-bit value by 64 bits (UB).
446 2020-03-13 Vladimir Makarov <vmakarov@redhat.com>
448 PR rtl-optimization/92303
449 * lra-spills.c (remove_pseudos): Try to simplify memory subreg.
451 2020-03-13 Segher Boessenkool <segher@kernel.crashing.org>
453 PR rtl-optimization/94148
454 PR rtl-optimization/94042
455 * df-core.c (BB_LAST_CHANGE_AGE): Delete.
456 (df_worklist_propagate_forward): New parameter last_change_age, use
457 that instead of bb->aux.
458 (df_worklist_propagate_backward): Ditto.
459 (df_worklist_dataflow_doublequeue): Use a local array last_change_age.
461 2020-03-13 Richard Biener <rguenther@suse.de>
463 PR tree-optimization/94163
464 * tree-ssa-pre.c (create_expression_by_pieces): Check
465 whether alignment would be zero.
467 2020-03-13 Martin Liska <mliska@suse.cz>
470 * lto-wrapper.c (run_gcc): Use concat for appending
471 to collect_gcc_options.
473 2020-03-13 Jakub Jelinek <jakub@redhat.com>
476 * config/aarch64/aarch64.c (aarch64_add_offset_1): Use gen_int_mode
479 2020-03-13 H.J. Lu <hongjiu.lu@intel.com>
482 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_DF.
483 * config/i386/i386.md (*movdf_internal): Call ix86_output_ssemov
484 for TYPE_SSEMOV. Remove TARGET_AVX512F, TARGET_PREFER_AVX256,
485 TARGET_AVX512VL and ext_sse_reg_operand check.
487 2020-03-13 Bu Le <bule1@huawei.com>
490 * config/aarch64/aarch64.opt (-param=aarch64-float-recp-precision=)
491 (-param=aarch64-double-recp-precision=): New options.
492 * doc/invoke.texi: Document them.
493 * config/aarch64/aarch64.c (aarch64_emit_approx_div): Use them
494 instead of hard-coding the choice of 1 for float and 2 for double.
496 2020-03-13 Eric Botcazou <ebotcazou@adacore.com>
498 PR rtl-optimization/94119
499 * resource.h (clear_hashed_info_until_next_barrier): Declare.
500 * resource.c (clear_hashed_info_until_next_barrier): New function.
501 * reorg.c (add_to_delay_list): Fix formatting.
502 (relax_delay_slots): Call clear_hashed_info_until_next_barrier on
503 the next instruction after removing a BARRIER.
505 2020-03-13 Eric Botcazou <ebotcazou@adacore.com>
508 * expmed.c (store_integral_bit_field): For fields larger than a word,
509 call extract_bit_field on the value if the mode is BLKmode. Remove
510 specific path for big-endian targets and tidy things up a little bit.
512 2020-03-12 Richard Sandiford <richard.sandiford@arm.com>
514 PR rtl-optimization/90275
515 * cse.c (cse_insn): Delete no-op register moves too.
517 2020-03-12 Darius Galis <darius.galis@cyberthorstudios.com>
519 * config/rx/rx.md (CTRLREG_CPEN): Remove.
520 * config/rx/rx.c (rx_print_operand): Remove CTRLREG_CPEN support.
522 2020-03-12 Richard Biener <rguenther@suse.de>
524 PR tree-optimization/94103
525 * tree-ssa-sccvn.c (visit_reference_op_load): Avoid type
526 punning when the mode precision is not sufficient.
528 2020-03-12 H.J. Lu <hongjiu.lu@intel.com>
531 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_DI,
532 MODE_V1DF and MODE_V2SF.
533 * config/i386/mmx.md (MMXMODE:*mov<mode>_internal): Call
534 ix86_output_ssemov for TYPE_SSEMOV. Remove ext_sse_reg_operand
537 2020-03-12 Jakub Jelinek <jakub@redhat.com>
539 * doc/tm.texi.in (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Change
540 ASM_OUTPUT_ALIGNED_DECL in description to ASM_OUTPUT_ALIGNED_LOCAL
541 and ASM_OUTPUT_DECL to ASM_OUTPUT_LOCAL.
542 * doc/tm.texi: Regenerated.
544 PR tree-optimization/94130
545 * tree-ssa-dse.c: Include gimplify.h.
546 (increment_start_addr): If stmt has lhs, drop the lhs from call and
547 set it after the call to the original value of the first argument.
549 (decrement_count): Formatting fix.
551 2020-03-11 Delia Burduv <delia.burduv@arm.com>
553 * config/arm/arm-builtins.c
554 (arm_init_simd_builtin_scalar_types): New.
555 * config/arm/arm_neon.h (vld2_bf16): Used new builtin type.
556 (vld2q_bf16): Used new builtin type.
557 (vld3_bf16): Used new builtin type.
558 (vld3q_bf16): Used new builtin type.
559 (vld4_bf16): Used new builtin type.
560 (vld4q_bf16): Used new builtin type.
561 (vld2_dup_bf16): Used new builtin type.
562 (vld2q_dup_bf16): Used new builtin type.
563 (vld3_dup_bf16): Used new builtin type.
564 (vld3q_dup_bf16): Used new builtin type.
565 (vld4_dup_bf16): Used new builtin type.
566 (vld4q_dup_bf16): Used new builtin type.
568 2020-03-11 Jakub Jelinek <jakub@redhat.com>
571 * config/pdp11/pdp11.c (pdp11_asm_output_var): Call switch_to_section
572 at the start to switch to data section. Don't print extra newline if
573 .globl directive has not been emitted.
575 2020-03-11 Richard Biener <rguenther@suse.de>
577 * match.pd ((T *)(ptr - ptr-cst) -> &MEM[ptr + -ptr-cst]):
580 2020-03-11 Eric Botcazou <ebotcazou@adacore.com>
583 * tree.c (variably_modified_type_p) <RECORD_TYPE>: Recurse into fields
584 whose type is a qualified union.
586 2020-03-11 Jakub Jelinek <jakub@redhat.com>
589 * config/aarch64/aarch64.c (aarch64_add_offset_1): Use absu_hwi
590 instead of abs_hwi, change moffset type to unsigned HOST_WIDE_INT.
593 * value-prof.c (dump_histogram_value): Use abs_hwi instead of
595 (get_nth_most_common_value): Use abs_hwi instead of abs.
598 * dfp.c (decimal_to_binary): Only use decimal128ToString if from->cl
599 is rvc_normal, otherwise use real_to_decimal to print the number to
602 PR tree-optimization/94114
603 * tree-loop-distribution.c (generate_memset_builtin): Call
604 rewrite_to_non_trapping_overflow even on mem.
605 (generate_memcpy_builtin): Call rewrite_to_non_trapping_overflow even
608 2020-03-10 Jeff Law <law@redhat.com>
610 * config/bfin/bfin.md (movsi_insv): Add length attribute.
612 2020-03-10 Jiufu Guo <guojiufu@linux.ibm.com>
615 * gcc/config/rs6000/rs6000.c (rs6000_emit_p9_fp_minmax): Check
616 NAN and SIGNED_ZEROR for smax/smin.
618 2020-03-10 Will Schmidt <will_schmidt@vnet.ibm.com>
621 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
622 clause to handle P9V_BUILTIN_VEC_LXVL with const arguments.
624 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
626 * loop-iv.c (find_simple_exit): Make it static.
627 * cfgloop.h: Remove the corresponding prototype.
629 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
631 * ddg.c (create_ddg): Fix intendation.
632 (set_recurrence_length): Likewise.
633 (create_ddg_all_sccs): Likewise.
635 2020-03-10 Jakub Jelinek <jakub@redhat.com>
638 * config/i386/i386.md (*testqi_ext_3): Call ix86_match_ccmode with
639 CCZmode instead of CCNOmode if operands[2] has DImode and pos + len
642 2020-03-09 Jason Merrill <jason@redhat.com>
644 * gdbinit.in (pgs): Fix typo in documentation.
646 2020-03-09 Vladimir Makarov <vmakarov@redhat.com>
650 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
652 PR rtl-optimization/93564
653 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
654 do not honor reg alloc order.
656 2020-03-09 Andrew Pinski <apinski@marvell.com>
659 * doc/extend.texi (x86 Operand Modifiers): Fix column
662 2020-03-09 Martin Liska <mliska@suse.cz>
665 * config/rs6000/rs6000.c (rs6000_option_override_internal):
666 Remove set of str_align_loops and str_align_jumps as these
667 should be set in previous 2 conditions in the function.
669 2020-03-09 Jakub Jelinek <jakub@redhat.com>
671 PR rtl-optimization/94045
672 * params.opt (-param=max-find-base-term-values=): New option.
673 * alias.c (find_base_term): Add cut-off for number of visited VALUEs
674 in a single toplevel find_base_term call.
676 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
679 * config/aarch64/aarch64-builtins.c (TYPES_TERNOPU_LANE): Add define.
680 * config/aarch64/aarch64-simd.md
681 (aarch64_vec_<su>mult_lane<Qlane>): Add new insn for widening lane mul.
682 (aarch64_vec_<su>mlal_lane<Qlane>): Likewise.
683 * config/aarch64/aarch64-simd-builtins.def: Add intrinsics.
684 * config/aarch64/arm_neon.h:
685 (vmlal_lane_s16): Expand using intrinsics rather than inline asm.
686 (vmlal_lane_u16): Likewise.
687 (vmlal_lane_s32): Likewise.
688 (vmlal_lane_u32): Likewise.
689 (vmlal_laneq_s16): Likewise.
690 (vmlal_laneq_u16): Likewise.
691 (vmlal_laneq_s32): Likewise.
692 (vmlal_laneq_u32): Likewise.
693 (vmull_lane_s16): Likewise.
694 (vmull_lane_u16): Likewise.
695 (vmull_lane_s32): Likewise.
696 (vmull_lane_u32): Likewise.
697 (vmull_laneq_s16): Likewise.
698 (vmull_laneq_u16): Likewise.
699 (vmull_laneq_s32): Likewise.
700 (vmull_laneq_u32): Likewise.
701 * config/aarch64/iterators.md (Vcondtype): New iterator for lane mul.
704 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
706 * aarch64/aarch64-simd.md (aarch64_mla_elt<mode>): Correct lane syntax.
707 (aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
708 (aarch64_mls_elt<mode>): Likewise.
709 (aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
710 (aarch64_fma4_elt<mode>): Likewise.
711 (aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
712 (aarch64_fma4_elt_to_64v2df): Likewise.
713 (aarch64_fnma4_elt<mode>): Likewise.
714 (aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
715 (aarch64_fnma4_elt_to_64v2df): Likewise.
717 2020-03-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
719 * config/aarch64/aarch64-sve2.md (@aarch64_sve_<sve_int_op><mode>:
720 Specify movprfx attribute.
721 (@aarch64_sve_<sve_int_op>_lane_<mode>): Likewise.
723 2020-03-06 David Edelsohn <dje.gcc@gmail.com>
726 * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
728 (TARGET_NO_FP_IN_TOC): Same.
729 * config/rs6000/aix71.h: Same.
730 * config/rs6000/aix72.h: Same.
732 2020-03-06 Andrew Pinski <apinski@marvell.com>
733 Jeff Law <law@redhat.com>
735 PR rtl-optimization/93996
736 * haifa-sched.c (remove_notes): Be more careful when adding
739 2020-03-06 Delia Burduv <delia.burduv@arm.com>
741 * config/arm/arm_neon.h (vld2_bf16): New.
747 (vld2_dup_bf16): New.
748 (vld2q_dup_bf16): New.
749 (vld3_dup_bf16): New.
750 (vld3q_dup_bf16): New.
751 (vld4_dup_bf16): New.
752 (vld4q_dup_bf16): New.
753 * config/arm/arm_neon_builtins.def
754 (vld2): Changed to VAR13 and added v4bf, v8bf
755 (vld2_dup): Changed to VAR8 and added v4bf, v8bf
756 (vld3): Changed to VAR13 and added v4bf, v8bf
757 (vld3_dup): Changed to VAR8 and added v4bf, v8bf
758 (vld4): Changed to VAR13 and added v4bf, v8bf
759 (vld4_dup): Changed to VAR8 and added v4bf, v8bf
760 * config/arm/iterators.md (VDXBF2): New iterator.
761 *config/arm/neon.md (neon_vld2): Use new iterators.
762 (neon_vld2_dup<mode): Use new iterators.
763 (neon_vld3<mode>): Likewise.
764 (neon_vld3qa<mode>): Likewise.
765 (neon_vld3qb<mode>): Likewise.
766 (neon_vld3_dup<mode>): Likewise.
767 (neon_vld4<mode>): Likewise.
768 (neon_vld4qa<mode>): Likewise.
769 (neon_vld4qb<mode>): Likewise.
770 (neon_vld4_dup<mode>): Likewise.
771 (neon_vld2_dupv8bf): New.
772 (neon_vld3_dupv8bf): Likewise.
773 (neon_vld4_dupv8bf): Likewise.
775 2020-03-06 Delia Burduv <delia.burduv@arm.com>
777 * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
778 (bfloat16x8x2_t): New typedef.
779 (bfloat16x4x3_t): New typedef.
780 (bfloat16x8x3_t): New typedef.
781 (bfloat16x4x4_t): New typedef.
782 (bfloat16x8x4_t): New typedef.
789 * config/arm/arm-builtins.c (v2bf_UP): Define.
791 (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
792 * config/arm/arm-modes.def (V2BF): New mode.
793 * config/arm/arm-simd-builtin-types.def
794 (Bfloat16x2_t): New entry.
795 * config/arm/arm_neon_builtins.def
796 (vst2): Changed to VAR13 and added v4bf, v8bf
797 (vst3): Changed to VAR13 and added v4bf, v8bf
798 (vst4): Changed to VAR13 and added v4bf, v8bf
799 * config/arm/iterators.md (VDXBF): New iterator.
800 (VQ2BF): New iterator.
801 *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
802 (neon_vst2<mode>): Used new iterators.
803 (neon_vst3<mode>): Used new iterators.
804 (neon_vst3<mode>): Used new iterators.
805 (neon_vst3qa<mode>): Used new iterators.
806 (neon_vst3qb<mode>): Used new iterators.
807 (neon_vst4<mode>): Used new iterators.
808 (neon_vst4<mode>): Used new iterators.
809 (neon_vst4qa<mode>): Used new iterators.
810 (neon_vst4qb<mode>): Used new iterators.
812 2020-03-06 Delia Burduv <delia.burduv@arm.com>
814 * config/aarch64/aarch64-simd-builtins.def
815 (bfcvtn): New built-in function.
816 (bfcvtn_q): New built-in function.
817 (bfcvtn2): New built-in function.
818 (bfcvt): New built-in function.
819 * config/aarch64/aarch64-simd.md
820 (aarch64_bfcvtn<q><mode>): New pattern.
821 (aarch64_bfcvtn2v8bf): New pattern.
822 (aarch64_bfcvtbf): New pattern.
823 * config/aarch64/arm_bf16.h (float32_t): New typedef.
824 (vcvth_bf16_f32): New intrinsic.
825 * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
826 (vcvtq_low_bf16_f32): New intrinsic.
827 (vcvtq_high_bf16_f32): New intrinsic.
828 * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
829 (UNSPEC_BFCVTN): New UNSPEC.
830 (UNSPEC_BFCVTN2): New UNSPEC.
831 (UNSPEC_BFCVT): New UNSPEC.
832 * config/arm/types.md (bf_cvt): New type.
834 2020-03-06 Andreas Krebbel <krebbel@linux.ibm.com>
836 * config/s390/s390.md ("tabort"): Get rid of two consecutive
837 blanks in format string.
839 2020-03-05 H.J. Lu <hongjiu.lu@intel.com>
843 * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
844 * config/i386/i386.c (ix86_get_ssemov): New function.
845 (ix86_output_ssemov): Likewise.
846 * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
847 ix86_output_ssemov for TYPE_SSEMOV. Remove TARGET_AVX512VL
849 (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
850 (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
851 Remove ext_sse_reg_operand and TARGET_AVX512VL check.
852 (*movti_internal): Likewise.
853 (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
855 2020-03-05 Jeff Law <law@redhat.com>
857 PR tree-optimization/91890
858 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
859 Use gimple_or_expr_nonartificial_location.
860 (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
861 Use gimple_or_expr_nonartificial_location.
862 * gimple.c (gimple_or_expr_nonartificial_location): New function.
863 * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
864 * tree-ssa-strlen.c (maybe_warn_overflow): Use
865 gimple_or_expr_nonartificial_location.
866 (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
867 (maybe_warn_pointless_strcmp): Likewise.
869 2020-03-05 Jakub Jelinek <jakub@redhat.com>
872 * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
873 SRC and MASK arguments to __m128 from __m128d.
874 (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
876 (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
878 * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
879 argument to __m128i from __m128d.
880 (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
882 (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
883 (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
886 2020-03-05 Delia Burduv <delia.burduv@arm.com>
888 * config/arm/arm_neon.h (vbfmmlaq_f32): New.
889 (vbfmlalbq_f32): New.
890 (vbfmlaltq_f32): New.
891 (vbfmlalbq_lane_f32): New.
892 (vbfmlaltq_lane_f32): New.
893 (vbfmlalbq_laneq_f32): New.
894 (vbfmlaltq_laneq_f32): New.
895 * config/arm/arm_neon_builtins.def (vmmla): New.
902 * config/arm/iterators.md (BF_MA): New int iterator.
903 (bt): New int attribute.
904 (VQXBF): Copy of VQX with V8BF.
905 * config/arm/neon.md (neon_vmmlav8bf): New insn.
906 (neon_vfma<bt>v8bf): New insn.
907 (neon_vfma<bt>_lanev8bf): New insn.
908 (neon_vfma<bt>_laneqv8bf): New expand.
909 (neon_vget_high<mode>): Changed iterator to VQXBF.
910 * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
911 (UNSPEC_BFMAB): New UNSPEC.
912 (UNSPEC_BFMAT): New UNSPEC.
914 2020-03-05 Jakub Jelinek <jakub@redhat.com>
917 * tree-pretty-print.h (pretty_print_string): Declare.
918 * tree-pretty-print.c (pretty_print_string): Remove forward
919 declaration, no longer static. Change nbytes parameter type
920 from unsigned to size_t.
921 * print-rtl.c (print_value) <case CONST_STRING>: Use
922 pretty_print_string and for shrink way too long strings.
924 2020-03-05 Richard Biener <rguenther@suse.de>
925 Jakub Jelinek <jakub@redhat.com>
927 PR tree-optimization/93582
928 * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
929 last operand as signed when looking for memset offset. Formatting
932 2020-03-04 Andrew Pinski <apinski@marvell.com>
935 * value-prof.c (dump_histogram_value): Use std::abs.
937 2020-03-04 Martin Sebor <msebor@redhat.com>
939 PR tree-optimization/93986
940 * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
941 operands to the same precision widest_int to avoid ICEs.
943 2020-03-04 Bill Schmidt <wschmidt@linux.ibm.com>
946 * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
947 * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
948 for OPTION_MASK_ALTIVEC.
950 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
952 * config.gcc: Include the glibc-stdint.h header for zTPF.
954 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
956 * config/s390/s390.c (s390_secondary_memory_needed): Disallow
957 direct FPR-GPR copies.
958 (s390_register_info_gprtofpr): Disallow GPR content to be saved in
961 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
963 * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
964 operands to the prologue_tpf expander.
965 (s390_emit_epilogue): Likewise.
966 (s390_option_override_internal): Do error checking and setup for
968 * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
969 (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
970 (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
971 (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
972 * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
973 operands for the check flag and the branch target.
974 * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
975 ("mtpf-trace-hook-prologue-target")
976 ("mtpf-trace-hook-epilogue-check")
977 ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
979 * doc/invoke.texi: Document -mtpf-trace-skip option. The other
980 options are for debugging purposes and will not be documented
983 2020-03-04 Jakub Jelinek <jakub@redhat.com>
986 * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
988 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
989 argument. Change pd argument so that it can be modified. Turn
990 constant non-CONSTRUCTOR store into non-constant if it is too large.
991 Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
993 (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
996 2020-02-04 Richard Biener <rguenther@suse.de>
998 PR tree-optimization/93964
999 * graphite-isl-ast-to-gimple.c
1000 (gcc_expression_from_isl_ast_expr_id): Add intermediate
1001 conversion for pointer to integer converts.
1002 * graphite-scop-detection.c (assign_parameter_index_in_region):
1005 2020-03-04 Martin Liska <mliska@suse.cz>
1009 * doc/invoke.texi: Clarify --help=language and --help=common
1012 2020-03-04 Jakub Jelinek <jakub@redhat.com>
1014 PR tree-optimization/94001
1015 * tree-tailcall.c (process_assignment): Before comparing op1 to
1016 *ass_var, verify *ass_var is non-NULL.
1018 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
1021 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
1024 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
1026 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
1027 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
1028 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
1029 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
1030 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
1031 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
1032 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
1033 (V_bf_low, V_bf_cvt_m): New mode attributes.
1034 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
1035 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
1036 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
1037 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
1038 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
1040 2020-03-03 Jakub Jelinek <jakub@redhat.com>
1042 PR tree-optimization/93582
1043 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
1044 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
1045 members, initialize them in the constructor and if mask is non-NULL,
1046 artificially push_partial_def {} for the portions of the mask that
1048 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
1049 val and return (void *)-1. Formatting fix.
1050 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
1052 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
1053 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
1055 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
1057 (visit_stmt): Formatting fix.
1059 2020-03-03 Richard Biener <rguenther@suse.de>
1061 PR tree-optimization/93946
1062 * alias.h (refs_same_for_tbaa_p): Declare.
1063 * alias.c (refs_same_for_tbaa_p): New function.
1064 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
1066 * tree-ssa-scopedtables.h
1067 (avail_exprs_stack::lookup_avail_expr): Add output argument
1068 giving access to the hashtable entry.
1069 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
1071 * tree-ssa-dom.c: Include alias.h.
1072 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
1073 removing redundant store.
1074 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
1075 (ao_ref_init_from_vn_reference): Adjust prototype.
1076 (vn_reference_lookup_pieces): Likewise.
1077 (vn_reference_insert_pieces): Likewise.
1078 * tree-ssa-sccvn.c: Track base alias set in addition to alias
1080 (eliminate_dom_walker::eliminate_stmt): Also check base alias
1081 set when removing redundant stores.
1082 (visit_reference_op_store): Likewise.
1083 * dse.c (record_store): Adjust valdity check for redundant
1086 2020-03-03 Jakub Jelinek <jakub@redhat.com>
1089 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
1091 PR rtl-optimization/94002
1092 * explow.c (plus_constant): Punt if cst has VOIDmode and
1093 get_pool_mode is different from mode.
1095 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
1097 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
1098 address has an offset which fits the scalling constraint for a
1099 load/store operation.
1100 (legitimate_scaled_address_p): Update use
1101 leigitimate_small_data_address_p.
1102 (arc_print_operand): Likewise.
1103 (arc_legitimate_address_p): Likewise.
1104 (legitimate_small_data_address_p): Likewise.
1106 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
1108 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
1109 (fnmasf4_fpu): Likewise.
1111 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
1113 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
1116 (adddi3_i): Remove pattern.
1117 (subdi3_i): Likewise.
1119 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
1121 * config/arc/arc.md (eh_return): Add length info.
1123 2020-03-02 David Malcolm <dmalcolm@redhat.com>
1125 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
1127 2020-03-02 David Malcolm <dmalcolm@redhat.com>
1129 * doc/invoke.texi (Static Analyzer Options): Add
1130 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
1133 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
1136 * config/i386/i386.md (movstrict<mode>): Allow only
1137 registers with VALID_INT_MODE_P modes.
1139 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
1141 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
1142 (reduc_insn): Use 'U' and 'B' operand codes.
1143 (reduc_<reduc_op>_scal_<mode>): Allow all types.
1144 (reduc_<reduc_op>_scal_v64di): Delete.
1145 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
1146 (*plus_carry_dpp_shr_v64si): Change to ...
1147 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
1148 (mov_from_lane63_v64di): Change to ...
1149 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
1150 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
1151 Support UNSPEC_MOV_DPP_SHR output formats.
1152 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
1153 Add "use_extends" reductions.
1154 (print_operand_address): Add 'I' and 'U' codes.
1155 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
1157 2020-03-02 Martin Liska <mliska@suse.cz>
1159 * lto-wrapper.c: Fix typo in comment about
1160 C++ standard version.
1162 2020-03-01 Martin Sebor <msebor@redhat.com>
1165 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
1167 2020-03-01 Martin Sebor <msebor@redhat.com>
1170 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
1171 of a pointer in the outermost ADDR_EXPRs.
1173 2020-02-28 Jeff Law <law@redhat.com>
1175 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
1176 * config/v850/v850.c (v850_asm_trampoline_template): Update
1179 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
1182 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
1185 2020-02-28 Martin Liska <mliska@suse.cz>
1188 * configure.ac: Improve detection of ld_date by requiring
1189 either two dashes or none.
1190 * configure: Regenerate.
1192 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
1194 PR rtl-optimization/93564
1195 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
1196 do not honor reg alloc order.
1198 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
1201 * config/aarch64/aarch64.c (aarch64_override_options): Fix
1202 misleading warning string.
1204 2020-02-27 Martin Sebor <msebor@redhat.com>
1206 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
1208 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
1211 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
1212 Split the insn into two parts. This insn only does variable
1213 extract from a register.
1214 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
1215 variable extract from memory.
1216 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
1217 only does variable extract from a register.
1218 (vsx_extract_v4sf_var_load): New insn, do variable extract from
1220 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
1221 into two parts. This insn only does variable extract from a
1223 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
1224 do variable extract from memory.
1226 2020-02-27 Martin Jambor <mjambor@suse.cz>
1227 Feng Xue <fxue@os.amperecomputing.com>
1230 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
1231 new function calls_same_node_or_its_all_contexts_clone_p.
1232 (cgraph_edge_brings_value_p): Use it.
1233 (cgraph_edge_brings_value_p): Likewise.
1234 (self_recursive_pass_through_p): Return false if caller is a clone.
1235 (self_recursive_agg_pass_through_p): Likewise.
1237 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
1240 * alias.c (ends_tbaa_access_path_p): Break out from ...
1241 (component_uses_parent_alias_set_from): ... here.
1242 * alias.h (ends_tbaa_access_path_p): Declare.
1243 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
1244 handle trailing arrays past end of tbaa access path.
1245 (aliasing_component_refs_p): ... here; likewise.
1246 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
1247 path; disambiguate also past end of it.
1248 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
1251 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
1253 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
1254 beginning of the file.
1255 (vcreate_bf16, vcombine_bf16): New.
1256 (vdup_n_bf16, vdupq_n_bf16): New.
1257 (vdup_lane_bf16, vdup_laneq_bf16): New.
1258 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
1259 (vduph_lane_bf16, vduph_laneq_bf16): New.
1260 (vset_lane_bf16, vsetq_lane_bf16): New.
1261 (vget_lane_bf16, vgetq_lane_bf16): New.
1262 (vget_high_bf16, vget_low_bf16): New.
1263 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
1264 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
1265 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
1266 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
1267 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
1268 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
1269 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
1270 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
1271 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
1272 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
1273 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
1274 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
1275 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
1276 (vreinterpretq_bf16_p128): New.
1277 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
1278 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
1279 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
1280 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
1281 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
1282 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
1283 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
1284 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
1285 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
1286 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
1287 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
1288 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
1289 (vreinterpretq_p128_bf16): New.
1290 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
1292 (V_elem_l): Likewise.
1293 (VD_LANE): Likewise.
1295 (V_DOUBLE): Likewise.
1296 (VDQX): Add V4BF and V8BF.
1297 (V_two_elem, V_three_elem, V_four_elem): Likewise.
1300 (V_double_vector_mode): Likewise.
1301 (V_cmp_result): Likewise.
1302 (V_uf_sclr): Likewise.
1303 (V_sz_elem): Likewise.
1304 (Is_d_reg): Likewise.
1305 (V_mode_nunits): Likewise.
1306 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
1308 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
1310 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
1311 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
1312 (<expander><mode>3<exec>): Likewise.
1313 (<expander><mode>3): New.
1314 (v<expander><mode>3): New.
1315 (<expander><mode>3): New.
1316 (<expander><mode>3<exec>): Rename to ...
1317 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
1318 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
1320 2020-02-27 Alexandre Oliva <oliva@adacore.com>
1322 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
1325 2020-02-27 Richard Biener <rguenther@suse.de>
1327 PR tree-optimization/93508
1328 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
1329 non-_CHK variants. Valueize their length arguments.
1331 2020-02-27 Richard Biener <rguenther@suse.de>
1333 PR tree-optimization/93953
1334 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
1335 to the hash-map entry.
1337 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
1339 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
1341 2020-02-27 Mark Williams <mwilliams@fb.com>
1343 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
1344 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
1345 -ffile-prefix-map and -fmacro-prefix-map.
1346 * lto-streamer-out.c: Include file-prefix-map.h.
1347 (lto_output_location): Remap the file part of locations.
1349 2020-02-27 Jakub Jelinek <jakub@redhat.com>
1352 * gimplify.c (gimplify_init_constructor): Don't promote readonly
1353 DECL_REGISTER variables to TREE_STATIC.
1355 PR tree-optimization/93582
1356 PR tree-optimization/93945
1357 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
1358 non-zero INTEGER_CST second argument and ref->offset or ref->size
1359 not a multiple of BITS_PER_UNIT.
1361 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
1363 * doc/install.texi (Binaries): Update description of BullFreeware.
1365 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
1369 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
1370 C++ Language Options, Warning Options, and Static Analyzer
1371 Options lists. Document negative form of options enabled by
1372 default. Move some things around to more accurately sort
1373 warnings by category.
1374 (C++ Dialect Options, Warning Options, Static Analyzer
1375 Options): Document negative form of options when enabled by
1376 default. Move some things around to more accurately sort
1377 warnings by category. Add some missing index entries.
1380 2020-02-26 Carl Love <cel@us.ibm.com>
1383 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
1384 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
1385 for the vector unsigned short arguments. It is also listed as the
1386 name of the built-in for arguments vector unsigned short,
1387 vector unsigned int and vector unsigned long long built-ins. The
1388 name of the builtins for these arguments should be:
1389 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
1390 __builtin_crypto_vpmsumd respectively.
1392 2020-02-26 Richard Biener <rguenther@suse.de>
1394 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
1395 and load permutation.
1397 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
1400 * optabs-tree.c (supportable_convert_operation): Reject types with
1403 2020-02-26 David Malcolm <dmalcolm@redhat.com>
1405 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
1407 2020-02-26 Jakub Jelinek <jakub@redhat.com>
1409 PR tree-optimization/93820
1410 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
1411 argument to ALL_INTEGER_CST_P boolean.
1412 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
1413 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
1414 adjacent INTEGER_CST store into merged_store->only_constants like
1417 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1420 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
1422 * cfghooks.c (verify_flow_info): Likewise.
1423 * predict.c (combine_predictions_for_bb): Likewise.
1424 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
1425 sucessor -> successor.
1426 (find_traces_1_round): Fix comment typo, destinarion -> destination.
1427 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
1429 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
1430 message typo, sucessors -> successors.
1432 2020-02-25 Martin Sebor <msebor@redhat.com>
1434 * doc/extend.texi (attribute access): Correct an example.
1436 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
1438 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
1440 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
1441 (VAR15, VAR16): New.
1442 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
1443 (VD): Enable for V4BF.
1445 (VQ): Enable for V8BF.
1447 (VQ_NO2E): Likewise.
1448 (VDBL, Vdbl): Add V4BF.
1449 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
1450 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
1451 (bfloat16x8x2_t): Likewise.
1452 (bfloat16x4x3_t): Likewise.
1453 (bfloat16x8x3_t): Likewise.
1454 (bfloat16x4x4_t): Likewise.
1455 (bfloat16x8x4_t): Likewise.
1456 (vcombine_bf16): New.
1457 (vld1_bf16, vld1_bf16_x2): New.
1458 (vld1_bf16_x3, vld1_bf16_x4): New.
1459 (vld1q_bf16, vld1q_bf16_x2): New.
1460 (vld1q_bf16_x3, vld1q_bf16_x4): New.
1461 (vld1_lane_bf16): New.
1462 (vld1q_lane_bf16): New.
1463 (vld1_dup_bf16): New.
1464 (vld1q_dup_bf16): New.
1467 (vld2_dup_bf16): New.
1468 (vld2q_dup_bf16): New.
1471 (vld3_dup_bf16): New.
1472 (vld3q_dup_bf16): New.
1475 (vld4_dup_bf16): New.
1476 (vld4q_dup_bf16): New.
1477 (vst1_bf16, vst1_bf16_x2): New.
1478 (vst1_bf16_x3, vst1_bf16_x4): New.
1479 (vst1q_bf16, vst1q_bf16_x2): New.
1480 (vst1q_bf16_x3, vst1q_bf16_x4): New.
1481 (vst1_lane_bf16): New.
1482 (vst1q_lane_bf16): New.
1490 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
1492 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
1493 (VALL_F16): Likewise.
1494 (VALLDI_F16): Likewise.
1497 (vswap_width_name): Likewise.
1498 (VSWAP_WIDTH): Likewise.
1502 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
1503 (vget_lane_bf16, vgetq_lane_bf16): New.
1504 (vcreate_bf16): New.
1505 (vdup_n_bf16, vdupq_n_bf16): New.
1506 (vdup_lane_bf16, vdup_laneq_bf16): New.
1507 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
1508 (vduph_lane_bf16, vduph_laneq_bf16): New.
1509 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
1510 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
1511 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
1512 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
1513 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
1514 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
1515 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
1516 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
1517 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
1518 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
1519 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
1520 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
1521 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
1522 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
1523 (vreinterpretq_bf16_p128): New.
1524 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
1525 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
1526 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
1527 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
1528 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
1529 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
1530 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
1531 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
1532 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
1533 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
1534 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
1535 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
1536 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
1537 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
1538 (vreinterpretq_p128_bf16): New.
1540 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
1542 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
1543 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
1544 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
1545 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
1546 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
1547 * config/arm/iterators.md (VSF2BF): New attribute.
1548 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
1549 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
1550 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
1552 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
1554 * config/arm/arm.md (required_for_purecode): New attribute.
1555 (enabled): Handle required_for_purecode.
1556 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
1557 work with -mpure-code.
1559 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1561 PR rtl-optimization/93908
1562 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
1565 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
1567 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
1569 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
1571 * doc/install.texi (--enable-checking): Adjust wording.
1573 2020-02-25 Richard Biener <rguenther@suse.de>
1575 PR tree-optimization/93868
1576 * tree-vect-slp.c (slp_copy_subtree): New function.
1577 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
1578 re-arranging stmts in it.
1580 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1583 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
1584 dummy function and remove it at the end.
1586 PR translation/93864
1587 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
1588 paramter -> parameter.
1589 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
1590 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
1592 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
1594 * doc/install.texi (--enable-checking): Properly document current
1596 (--enable-stage1-checking): Minor clarification about bootstrap.
1598 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1601 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
1602 -fanalyzer-checker=taint is also required.
1603 (-fanalyzer-checker=): Note that providing this option enables the
1604 given checker, and doing so may be required for checkers that are
1605 disabled by default.
1607 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1609 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
1610 significant control flow events; add a "3" which shows all
1611 control flow events; the old "3" becomes "4".
1613 2020-02-24 Jakub Jelinek <jakub@redhat.com>
1615 PR tree-optimization/93582
1616 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
1617 pd.offset and pd.size to be counted in bits rather than bytes, add
1618 support for maxsizei that is not a multiple of BITS_PER_UNIT and
1619 handle bitfield stores and loads.
1620 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
1621 uncomparable quantities - bytes vs. bits. Allow push_partial_def
1622 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
1623 pd.offset/pd.size to be counted in bits rather than bytes.
1624 Formatting fix. Rename shadowed len variable to buflen.
1626 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1627 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
1630 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
1631 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
1632 * opts-common.c (parse_options_from_collect_gcc_options): New function.
1633 (prepend_xassembler_to_collect_as_options): Likewise.
1634 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
1635 (prepend_xassembler_to_collect_as_options): Likewise.
1636 * lto-opts.c (lto_write_options): Stream assembler options
1637 in COLLECT_AS_OPTIONS.
1638 * lto-wrapper.c (xassembler_options_error): New static variable.
1639 (get_options_from_collect_gcc_options): Move parsing options code to
1640 parse_options_from_collect_gcc_options and call it.
1641 (merge_and_complain): Validate -Xassembler options.
1642 (append_compiler_options): Handle OPT_Xassembler.
1643 (run_gcc): Append command line -Xassembler options to
1644 collect_gcc_options.
1645 * doc/invoke.texi: Add documentation about using Xassembler
1648 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
1650 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
1652 (riscv_rtx_costs): Update cost model for LTGT.
1654 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
1656 PR rtl-optimization/93564
1657 * ira-color.c (struct update_cost_queue_elem): New member start.
1658 (queue_update_cost, get_next_update_cost): Add new arg start.
1659 (allocnos_conflict_p): New function.
1660 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
1661 Add checking conflicts with allocnos_conflict_p.
1662 (update_costs_from_prefs, restore_costs_from_copies): Adjust
1663 update_costs_from_allocno calls.
1664 (update_conflict_hard_regno_costs): Add checking conflicts with
1665 allocnos_conflict_p. Adjust calls of queue_update_cost and
1666 get_next_update_cost.
1667 (assign_hard_reg): Adjust calls of queue_update_cost. Add
1669 (bucket_allocno_compare_func): Restore previous version.
1671 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
1673 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
1674 double-word size when handling aggregate return values.
1675 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
1676 that homogeneous SFmode and DFmode aggregates are passed and returned
1677 in general registers.
1679 2020-02-21 Jakub Jelinek <jakub@redhat.com>
1681 PR translation/93759
1682 * opts.c (print_filtered_help): Translate help before appending
1683 messages to it rather than after that.
1685 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1687 PR rtl-optimization/PR92989
1688 * lra-lives.c (process_bb_lives): Restore the original order
1689 of the bb liveness update. Call make_hard_regno_dead for each
1690 register clobbered at the start of an EH receiver.
1692 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
1695 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
1696 self-recursively generated.
1698 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
1701 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
1704 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
1706 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
1707 Document new target supports option.
1709 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
1711 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
1712 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
1713 * config/arm/iterators.md (MATMUL): New iterator.
1714 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
1715 (mmla_sfx): New attribute.
1716 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
1717 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
1718 (UNSPEC_MATMUL_US): New.
1720 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1722 * config/arm/arm.md: Prevent scalar shifts from being used when big
1725 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
1726 Richard Biener <rguenther@suse.de>
1728 PR tree-optimization/93586
1729 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1730 after mismatched array refs; do not sure type size information to
1731 recover from unmatched referneces with !flag_strict_aliasing_p.
1733 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1735 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1736 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1737 (scatter_store<mode>): Rename to ...
1738 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1739 (scatter<mode>_exec): Delete. Move contents ...
1740 (mask_scatter_store<mode>): ... here, and rename that to ...
1741 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1742 Remove mode conversion.
1743 (mask_gather_load<mode>): Rename to ...
1744 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1745 Remove mode conversion.
1746 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1748 2020-02-21 Martin Jambor <mjambor@suse.cz>
1750 PR tree-optimization/93845
1751 * tree-sra.c (verify_sra_access_forest): Only test access size of
1754 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1756 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1757 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1758 (addv64di3_exec): Likewise.
1759 (subv64di3): Likewise.
1760 (subv64di3_exec): Likewise.
1761 (addv64di3_zext): Likewise.
1762 (addv64di3_zext_exec): Likewise.
1763 (addv64di3_zext_dup): Likewise.
1764 (addv64di3_zext_dup_exec): Likewise.
1765 (addv64di3_zext_dup2): Likewise.
1766 (addv64di3_zext_dup2_exec): Likewise.
1767 (addv64di3_sext_dup2): Likewise.
1768 (addv64di3_sext_dup2_exec): Likewise.
1769 (<expander>v64di3): Likewise.
1770 (<expander>v64di3_exec): Likewise.
1771 (*<reduc_op>_dpp_shr_v64di): Likewise.
1772 (*plus_carry_dpp_shr_v64di): Likewise.
1773 * config/gcn/gcn.md (adddi3): Likewise.
1774 (addptrdi3): Likewise.
1775 (<expander>di3): Likewise.
1777 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1779 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1781 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1783 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1784 support. Use aarch64_emit_mult instead of emitting multiplication
1785 instructions directly.
1786 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1787 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1789 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1791 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1792 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
1793 instead of emitting multiplication instructions directly.
1794 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1795 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1796 (@aarch64_frecps<mode>): New expanders.
1798 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1800 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1801 on and produce uint64_ts rather than ints.
1802 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1803 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1805 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1807 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1808 an unused xmsk register when handling approximate rsqrt.
1810 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1812 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1813 flag_finite_math_only condition.
1815 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
1818 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1819 to destination operand for shufps alternative.
1820 (*vec_extractv2si_1): Ditto.
1822 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
1825 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1828 2020-02-20 Martin Liska <mliska@suse.cz>
1830 PR translation/93831
1831 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1833 2020-02-20 Martin Liska <mliska@suse.cz>
1835 PR translation/93830
1836 * common/config/avr/avr-common.c: Remote trailing "|".
1838 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1840 * collect2.c (maybe_run_lto_and_relink): Fix typo in
1843 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1845 PR tree-optimization/93767
1846 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1847 access-size bias from the offset calculations for negative strides.
1849 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1851 * collect2.c (c_file, o_file): Make const again.
1852 (ldout,lderrout, dump_ld_file): Remove.
1853 (tool_cleanup): Avoid calling not signal-safe functions.
1854 (maybe_run_lto_and_relink): Avoid possible signal handler
1855 access to unintialzed memory (lto_o_files).
1856 (main): Avoid leaking temp files in $TMPDIR.
1857 Initialize c_file/o_file with concat, which avoids exposing
1858 uninitialized memory to signal handler, which calls unlink(!).
1859 Avoid calling maybe_unlink when the main function returns,
1860 since the atexit handler is already doing this.
1861 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1863 2020-02-19 Martin Jambor <mjambor@suse.cz>
1865 PR tree-optimization/93776
1866 * tree-sra.c (create_access): Do not create zero size accesses.
1867 (get_access_for_expr): Do not search for zero sized accesses.
1869 2020-02-19 Martin Jambor <mjambor@suse.cz>
1871 PR tree-optimization/93667
1872 * tree-sra.c (scalarizable_type_p): Return false if record fields
1873 do not follow wach other.
1875 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1877 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1878 rather than fmv.x.s/fmv.s.x.
1880 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
1882 * config/aarch64/aarch64-simd-builtins.def
1883 (intrinsic_vec_smult_lo_): New.
1884 (intrinsic_vec_umult_lo_): Likewise.
1885 (vec_widen_smult_hi_): Likewise.
1886 (vec_widen_umult_hi_): Likewise.
1887 * config/aarch64/aarch64-simd.md
1888 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1889 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1890 (vmull_high_s16): Likewise.
1891 (vmull_high_s32): Likewise.
1892 (vmull_high_u8): Likewise.
1893 (vmull_high_u16): Likewise.
1894 (vmull_high_u32): Likewise.
1895 (vmull_s8): Likewise.
1896 (vmull_s16): Likewise.
1897 (vmull_s32): Likewise.
1898 (vmull_u8): Likewise.
1899 (vmull_u16): Likewise.
1900 (vmull_u32): Likewise.
1902 2020-02-18 Martin Liska <mliska@suse.cz>
1904 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1905 bootstrap by missing removal of invalid sanity check.
1907 2020-02-18 Martin Liska <mliska@suse.cz>
1910 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1911 Always compare LHS of gimple_assign.
1913 2020-02-18 Martin Liska <mliska@suse.cz>
1916 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1917 and return type of functions.
1918 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1919 Drop MALLOC attribute for void functions.
1920 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1921 malloc_state for a new VOID clone.
1923 2020-02-18 Martin Liska <mliska@suse.cz>
1926 * common.opt: Add -fprofile-reproducibility.
1927 * doc/invoke.texi: Document it.
1928 * value-prof.c (dump_histogram_value):
1929 Document and support behavior for counters[0]
1930 being a negative value.
1931 (get_nth_most_common_value): Handle negative
1932 counters[0] in respect to flag_profile_reproducible.
1934 2020-02-18 Jakub Jelinek <jakub@redhat.com>
1937 * cgraph.c (verify_speculative_call): Use speculative_id instead of
1938 speculative_uid in messages. Remove trailing whitespace from error
1939 message. Use num_speculative_call_targets instead of
1940 num_speculative_targets in a message.
1941 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1942 edge messages and stmt instead of cal_stmt in reference message.
1944 PR tree-optimization/93780
1945 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1946 before calling build_vector_type.
1947 (execute_update_addresses_taken): Likewise.
1950 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1951 typo, functoin -> function.
1952 * tree.c (free_lang_data_in_decl): Fix comment typo,
1953 functoin -> function.
1954 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1956 2020-02-17 David Malcolm <dmalcolm@redhat.com>
1958 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1960 (print_option_information): Don't call get_option_url if URLs
1963 2020-02-17 Alexandre Oliva <oliva@adacore.com>
1965 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1966 handling of register_common-less targets.
1968 2020-02-17 Martin Liska <mliska@suse.cz>
1971 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1973 2020-02-17 Martin Liska <mliska@suse.cz>
1975 PR translation/93755
1976 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1979 2020-02-17 Martin Liska <mliska@suse.cz>
1982 * config/rx/elf.opt: Fix typo.
1984 2020-02-17 Richard Biener <rguenther@suse.de>
1987 * opts-global.c (print_ignored_options): Use inform and
1990 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
1993 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1995 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
1998 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1999 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
2001 2020-02-15 Jason Merrill <jason@redhat.com>
2003 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
2005 2020-02-15 Jakub Jelinek <jakub@redhat.com>
2007 PR tree-optimization/93744
2008 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
2009 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
2010 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
2011 sure @2 in the first and @1 in the other patterns has no side-effects.
2013 2020-02-15 David Malcolm <dmalcolm@redhat.com>
2014 Bernd Edlinger <bernd.edlinger@hotmail.de>
2018 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
2019 * configure.ac (--with-diagnostics-urls): New configuration
2020 option, based on --with-diagnostics-color.
2021 (DIAGNOSTICS_URLS_DEFAULT): New define.
2022 * config.h: Regenerate.
2023 * configure: Regenerate.
2024 * diagnostic.c (diagnostic_urls_init): Handle -1 for
2025 DIAGNOSTICS_URLS_DEFAULT from configure-time
2026 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
2027 and TERM_URLS environment variable.
2028 * diagnostic-url.h (diagnostic_url_format): New enum type.
2029 (diagnostic_urls_enabled_p): rename to...
2030 (determine_url_format): ... this, and change return type.
2031 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
2032 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
2033 the linux console, and mingw.
2034 (diagnostic_urls_enabled_p): rename to...
2035 (determine_url_format): ... this, and adjust.
2036 * pretty-print.h (pretty_printer::show_urls): rename to...
2037 (pretty_printer::url_format): ... this, and change to enum.
2038 * pretty-print.c (pretty_printer::pretty_printer,
2039 pp_begin_url, pp_end_url, test_urls): Adjust.
2040 * doc/install.texi (--with-diagnostics-urls): Document the new
2041 configuration option.
2042 (--with-diagnostics-color): Document the existing interaction
2043 with GCC_COLORS better.
2044 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
2045 vindex reference. Update description of defaults based on the above.
2046 (-fdiagnostics-color): Update description of how -fdiagnostics-color
2047 interacts with GCC_COLORS.
2049 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
2052 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
2053 conjunction with TARGET_GNU_TLS in early return.
2055 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
2057 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
2058 the mode is not wider than UNITS_PER_WORD.
2060 2020-02-14 Martin Jambor <mjambor@suse.cz>
2062 PR tree-optimization/93516
2063 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
2064 access of the same type as the parent.
2065 (propagate_subaccesses_from_lhs): Likewise.
2067 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
2070 * config/i386/avx512vbmi2intrin.h
2071 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
2072 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
2073 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
2074 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
2075 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
2076 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
2077 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
2078 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
2079 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
2080 of lacking a closing parenthesis.
2081 * config/i386/avx512vbmi2vlintrin.h
2082 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
2083 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
2084 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
2085 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
2086 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
2087 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
2088 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
2089 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
2090 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
2091 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
2092 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
2093 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
2094 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
2095 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
2096 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
2097 _mm_shldi_epi32, _mm_mask_shldi_epi32,
2098 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
2099 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
2101 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
2104 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
2105 the target function entry.
2107 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
2109 * common/config/arc/arc-common.c (arc_option_optimization_table):
2110 Disable if-conversion step when optimized for size.
2112 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
2114 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
2115 R12-R15 are always in ARCOMPACT16_REGS register class.
2116 * config/arc/arc.opt (mq-class): Deprecate.
2117 * config/arc/constraint.md ("q"): Remove dependency on mq-class
2119 * doc/invoke.texi (mq-class): Update text.
2120 * common/config/arc/arc-common.c (arc_option_optimization_table):
2123 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
2125 * config/arc/arc.c (arc_insn_cost): New function.
2126 (TARGET_INSN_COST): Define.
2127 * config/arc/arc.md (cost): New attribute.
2128 (add_n): Use arc_nonmemory_operand.
2129 (ashlsi3_insn): Likewise, also update constraints.
2130 (ashrsi3_insn): Likewise.
2131 (rotrsi3): Likewise.
2132 (add_shift): Likewise.
2133 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
2135 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
2137 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
2139 (umulsidi_600): Likewise.
2141 2020-02-13 Jakub Jelinek <jakub@redhat.com>
2144 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
2145 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
2146 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
2147 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
2148 pass __A to the builtin followed by __W instead of __A followed by
2150 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
2151 _mm512_mask_popcnt_epi64): Likewise.
2152 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
2153 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
2154 _mm256_mask_popcnt_epi64): Likewise.
2156 PR tree-optimization/93582
2157 * fold-const.h (shift_bytes_in_array_left,
2158 shift_bytes_in_array_right): Declare.
2159 * fold-const.c (shift_bytes_in_array_left,
2160 shift_bytes_in_array_right): New function, moved from
2161 gimple-ssa-store-merging.c, no longer static.
2162 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
2163 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
2164 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
2165 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
2166 shift_bytes_in_array.
2167 (verify_shift_bytes_in_array): Rename to ...
2168 (verify_shift_bytes_in_array_left): ... this. Use
2169 shift_bytes_in_array_left instead of shift_bytes_in_array.
2170 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
2171 instead of verify_shift_bytes_in_array.
2172 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
2173 / native_interpret_expr where the store covers all needed bits,
2174 punt on PDP-endian, otherwise allow all involved offsets and sizes
2175 not to be byte-aligned.
2178 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
2179 use const_0_to_255_operand predicate instead of immediate_operand.
2180 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
2181 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
2182 vgf2p8affineinvqb_<mode><mask_name>,
2183 vgf2p8affineqb_<mode><mask_name>): Drop mode from
2184 const_0_to_255_operand predicated operands.
2186 2020-02-12 Jeff Law <law@redhat.com>
2188 * config/h8300/h8300.md (comparison shortening peepholes): Use
2189 a mode iterator to merge the HImode and SImode peepholes.
2191 2020-02-12 Jakub Jelinek <jakub@redhat.com>
2194 * real.c (is_even): Make static. Function comment fix.
2195 (is_halfway_below): Make static, don't assert R is not inf/nan,
2196 instead return false for those. Small formatting fixes.
2198 2020-02-12 Martin Sebor <msebor@redhat.com>
2201 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
2202 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
2203 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
2204 (strlen_check_and_optimize_call): Adjust callee name.
2206 2020-02-12 Jeff Law <law@redhat.com>
2208 * config/h8300/h8300.md (comparison shortening peepholes): Drop
2209 (and (xor)) variant. Combine other two into single peephole.
2211 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
2213 PR rtl-optimization/93565
2214 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
2216 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
2218 * config/aarch64/aarch64-simd.md
2219 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
2220 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
2221 generating separate ADDV and zero_extend patterns.
2222 * config/aarch64/iterators.md (VDQV_E): New iterator.
2224 2020-02-12 Jeff Law <law@redhat.com>
2226 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
2227 expanders, splits, etc.
2228 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
2229 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
2230 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
2231 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
2232 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
2233 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
2235 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
2237 2020-02-12 Jakub Jelinek <jakub@redhat.com>
2240 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
2241 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
2242 TARGET_AVX512DQ from condition.
2243 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
2244 instead of <mask_mode512bit_condition> in condition. If
2245 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
2247 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
2250 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
2253 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
2255 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
2257 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
2258 where strlen is more legible.
2259 (rs6000_builtin_vectorized_libmass): Ditto.
2260 (rs6000_print_options_internal): Ditto.
2262 2020-02-11 Martin Sebor <msebor@redhat.com>
2264 PR tree-optimization/93683
2265 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
2267 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
2269 * config/rs6000/predicates.md (cint34_operand): Rename the
2270 -mprefixed-addr option to be -mprefixed.
2271 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
2272 the -mprefixed-addr option to be -mprefixed.
2273 (OTHER_FUTURE_MASKS): Likewise.
2274 (POWERPC_MASKS): Likewise.
2275 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
2276 the -mprefixed-addr option to be -mprefixed. Change error
2277 messages to refer to -mprefixed.
2278 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
2280 (rs6000_legitimate_offset_address_p): Likewise.
2281 (rs6000_mode_dependent_address): Likewise.
2282 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
2283 "-mprefixed" for target attributes and pragmas.
2284 (address_to_insn_form): Rename the -mprefixed-addr option to be
2286 (rs6000_adjust_insn_length): Likewise.
2287 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
2288 -mprefixed-addr option to be -mprefixed.
2289 (ASM_OUTPUT_OPCODE): Likewise.
2290 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
2291 -mprefixed-addr option to be -mprefixed.
2292 * config/rs6000/rs6000.opt (-mprefixed): Rename the
2293 -mprefixed-addr option to be prefixed. Change the option from
2294 being undocumented to being documented.
2295 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
2296 -mprefixed option. Update the -mpcrel documentation to mention
2299 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
2301 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
2302 including FIRST_PSEUDO_REGISTER - 1.
2303 * ira-color.c (print_hard_reg_set): Ditto.
2305 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2307 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
2308 (USTERNOP_QUALIFIERS): New define.
2309 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
2310 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
2311 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
2312 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
2313 * config/arm/arm_neon.h (vusdot_s32): New.
2314 (vusdot_lane_s32): New.
2315 (vusdotq_lane_s32): New.
2316 (vsudot_lane_s32): New.
2317 (vsudotq_lane_s32): New.
2318 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
2319 * config/arm/iterators.md (DOTPROD_I8MM): New.
2320 (sup, opsuffix): Add <us/su>.
2321 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
2322 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
2324 2020-02-11 Richard Biener <rguenther@suse.de>
2326 PR tree-optimization/93661
2327 PR tree-optimization/93662
2328 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
2330 * tree-sra.c (get_access_for_expr): Likewise.
2332 2020-02-10 Jakub Jelinek <jakub@redhat.com>
2335 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
2336 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
2337 Change condition from TARGET_AVX2 to TARGET_AVX.
2339 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
2342 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
2343 argument of strncmp.
2345 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
2347 Try to generate zero-based comparisons.
2348 * config/cris/cris.c (cris_reduce_compare): New function.
2349 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
2350 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
2351 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
2353 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
2356 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
2357 in Thumb state and also as a destination in Arm state. Add T16
2360 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
2362 * md.texi (Define Subst): Match closing paren in example.
2364 2020-02-10 Jakub Jelinek <jakub@redhat.com>
2368 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
2369 arguments of strncmp.
2371 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
2374 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
2375 but different source value.
2376 (adjust_callers_for_value_intersection): New function.
2377 (gather_edges_for_value): Adjust order of callers to let a
2378 non-self-recursive caller be the first element.
2379 (self_recursive_pass_through_p): Add a new parameter "simple", and
2380 check generalized self-recursive pass-through jump function.
2381 (self_recursive_agg_pass_through_p): Likewise.
2382 (find_more_scalar_values_for_callers_subset): Compute value from
2383 pass-through jump function for self-recursive.
2384 (intersect_with_plats): Cleanup previous implementation code for value
2385 itersection with self-recursive call edge.
2386 (intersect_with_agg_replacements): Likewise.
2387 (intersect_aggregates_with_edge): Deduce value from pass-through jump
2388 function for self-recursive call edge. Cleanup previous implementation
2389 code for value intersection with self-recursive call edge.
2390 (decide_whether_version_node): Remove dead callers and adjust order
2391 to let a non-self-recursive caller be the first element.
2393 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
2395 * recog.c: Move pass_split_before_sched2 code in front of
2396 pass_split_before_regstack.
2397 (pass_data_split_before_sched2): Rename pass to split3 from split4.
2398 (pass_data_split_before_regstack): Rename pass to split4 from split3.
2399 (rest_of_handle_split_before_sched2): Remove.
2400 (pass_split_before_sched2::execute): Unconditionally call
2402 (enable_split_before_sched2): New function.
2403 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
2404 (pass_split_before_regstack::gate): Ditto.
2405 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
2406 Update name check for renamed split4 pass.
2407 * config/sh/sh.c (register_sh_passes): Update pass insertion
2408 point for renamed split4 pass.
2410 2020-02-09 Jakub Jelinek <jakub@redhat.com>
2412 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
2413 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
2414 copying them around between host and target.
2416 2020-02-08 Andrew Pinski <apinski@marvell.com>
2419 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
2420 STRICT_ALIGNMENT also.
2422 2020-02-08 Jim Wilson <jimw@sifive.com>
2425 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
2427 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
2428 Jakub Jelinek <jakub@redhat.com>
2431 * config/i386/i386.h (CALL_USED_REGISTERS): Make
2432 xmm16-xmm31 call-used even in 64-bit ms-abi.
2434 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
2436 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
2437 (simd_ummla, simd_usmmla): Likewise.
2438 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
2439 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
2440 (vusmmlaq_s32): New.
2442 2020-02-07 Richard Biener <rguenther@suse.de>
2445 * tree-inline.c (fold_marked_statements): Do a PRE walk,
2446 skipping unreachable regions.
2447 (optimize_inline_calls): Skip folding stmts when we didn't
2450 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
2453 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
2454 Don't return aggregates with only SFmode and DFmode in SSE
2456 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
2458 2020-02-07 Jakub Jelinek <jakub@redhat.com>
2461 * config/rs6000/rs6000-logue.c
2462 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
2463 if it fails, move rs into end_addr and retry. Add
2464 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
2465 the insn pattern doesn't describe well what exactly happens to
2469 * config/i386/predicates.md (avx_identity_operand): Remove.
2470 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
2471 (avx_<castmode><avxsizesuffix>_<castmode>,
2472 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
2473 a VEC_CONCAT of the operand and UNSPEC_CAST.
2474 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
2475 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
2479 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
2480 recog_data.insn if distance_non_agu_define changed it.
2482 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
2485 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
2486 we only had X-FORM (reg+reg) addressing for vectors. Also before
2487 ISA 3.0, we only had X-FORM addressing for scalars in the
2488 traditional Altivec registers.
2490 2020-02-06 <zhongyunde@huawei.com>
2491 Vladimir Makarov <vmakarov@redhat.com>
2493 PR rtl-optimization/93561
2494 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
2495 hard register range.
2497 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2499 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
2502 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
2504 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
2505 where the low and the high 32 bits are equal to each other specially,
2506 with an rldimi instruction.
2508 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
2510 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
2512 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
2514 * config/arm/arm-tables.opt: Regenerate.
2516 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2519 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
2520 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
2521 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
2523 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2525 PR rtl-optimization/87763
2526 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
2528 2020-02-06 Delia Burduv <delia.burduv@arm.com>
2530 * config/aarch64/aarch64-simd-builtins.def
2531 (bfmlaq): New built-in function.
2532 (bfmlalb): New built-in function.
2533 (bfmlalt): New built-in function.
2534 (bfmlalb_lane): New built-in function.
2535 (bfmlalt_lane): New built-in function.
2536 * config/aarch64/aarch64-simd.md
2537 (aarch64_bfmmlaqv4sf): New pattern.
2538 (aarch64_bfmlal<bt>v4sf): New pattern.
2539 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
2540 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
2541 (vbfmlalbq_f32): New intrinsic.
2542 (vbfmlaltq_f32): New intrinsic.
2543 (vbfmlalbq_lane_f32): New intrinsic.
2544 (vbfmlaltq_lane_f32): New intrinsic.
2545 (vbfmlalbq_laneq_f32): New intrinsic.
2546 (vbfmlaltq_laneq_f32): New intrinsic.
2547 * config/aarch64/iterators.md (BF_MLA): New int iterator.
2548 (bt): New int attribute.
2550 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
2552 * config/i386/i386.md (*pushtf): Emit "#" instead of
2553 calling gcc_unreachable in insn output.
2556 (*pushsf_rex64): Ditto for alternatives other than 1.
2557 (*pushsf): Ditto for alternatives other than 1.
2559 2020-02-06 Martin Liska <mliska@suse.cz>
2561 PR gcov-profile/91971
2562 PR gcov-profile/93466
2563 * coverage.c (coverage_init): Revert mangling of
2564 path into filename. It can lead to huge filename length.
2565 Creation of subfolders seem more natural.
2567 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2570 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
2571 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
2572 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
2574 2020-02-06 Jakub Jelinek <jakub@redhat.com>
2577 * config/i386/predicates.md (avx_identity_operand): New predicate.
2578 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
2579 define_insn_and_split.
2582 * omp-low.c (use_pointer_for_field): For nested constructs, also
2583 look for map clauses on target construct.
2584 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
2585 taskreg_nesting_level.
2588 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
2589 shared clause, call omp_notice_variable on outer context if any.
2591 2020-02-05 Jason Merrill <jason@redhat.com>
2594 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
2595 non-zero address even if weak and not yet defined.
2597 2020-02-05 Martin Sebor <msebor@redhat.com>
2599 PR tree-optimization/92765
2600 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
2601 * tree-ssa-strlen.c (compute_string_length): Remove.
2602 (determine_min_objsize): Remove.
2603 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
2604 Avoid using type size as the upper bound on string length.
2605 (handle_builtin_string_cmp): Add an argument. Adjust.
2606 (strlen_check_and_optimize_call): Pass additional argument to
2607 handle_builtin_string_cmp.
2609 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
2611 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
2612 (*pushdi2_rex64 peephole2): Unconditionally split after
2614 (*ashl<mode>3_doubleword): Ditto.
2615 (*<shift_insn><mode>3_doubleword): Ditto.
2617 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
2620 * config/rs6000/rs6000.c (get_vector_offset): Fix
2622 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
2624 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
2626 2020-02-05 David Malcolm <dmalcolm@redhat.com>
2629 (Special Functions for Debugging the Analyzer): Update description
2630 of __analyzer_dump_exploded_nodes.
2632 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2635 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
2636 include sets and not clobbers in the vzeroupper pattern.
2637 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
2638 the parallel has 17 (64-bit) or 9 (32-bit) elts.
2639 (*avx_vzeroupper_1): New define_insn_and_split.
2642 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
2643 don't run when !optimize.
2644 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
2647 2020-02-05 Richard Biener <rguenther@suse.de>
2650 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
2651 checks before matching calls.
2653 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2655 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
2656 function comment typo.
2659 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
2660 simd_clone_create failed when i == 0, adjust clone->nargs by
2663 2020-02-05 Martin Liska <mliska@suse.cz>
2666 * doc/invoke.texi: Document that one should
2667 not combine ASLR and -fpch.
2669 2020-02-04 Richard Biener <rguenther@suse.de>
2671 PR tree-optimization/93538
2672 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
2674 2020-02-04 Richard Biener <rguenther@suse.de>
2676 PR tree-optimization/91123
2677 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
2678 (vn_walk_cb_data::last_vuse): New member.
2679 (vn_walk_cb_data::saved_operands): Likewsie.
2680 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
2681 (vn_walk_cb_data::push_partial_def): Use finish.
2682 (vn_reference_lookup_2): Update last_vuse and use finish if
2683 we've saved operands.
2684 (vn_reference_lookup_3): Use finish and update calls to
2685 push_partial_defs everywhere. When translating through
2686 memcpy or aggregate copies save off operands and alias-set.
2687 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
2688 operation for redundant store removal.
2690 2020-02-04 Richard Biener <rguenther@suse.de>
2692 PR tree-optimization/92819
2693 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
2694 generating more stmts than before.
2696 2020-02-04 Martin Liska <mliska@suse.cz>
2698 * config/arm/arm.c (arm_gen_far_branch): Move the function
2699 outside of selftests.
2701 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2703 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
2704 function to adjust PC-relative vector addresses.
2705 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
2706 handle vectors with PC-relative addresses.
2708 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2710 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
2712 (hard_reg_and_mode_to_addr_mask): Delete.
2713 (rs6000_adjust_vec_address): If the original vector address
2714 was REG+REG or REG+OFFSET and the element is not zero, do the add
2715 of the elements in the original address before adding the offset
2716 for the vector element. Use address_to_insn_form to validate the
2717 address using the register being loaded, rather than guessing
2718 whether the address is a DS-FORM or DQ-FORM address.
2720 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2722 * config/rs6000/rs6000.c (get_vector_offset): New helper function
2723 to calculate the offset in memory from the start of a vector of a
2724 particular element. Add code to keep the element number in
2725 bounds if the element number is variable.
2726 (rs6000_adjust_vec_address): Move calculation of offset of the
2727 vector element to get_vector_offset.
2728 (rs6000_split_vec_extract_var): Do not do the initial AND of
2729 element here, move the code to get_vector_offset.
2731 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2733 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2736 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
2738 * config/rs6000/constraints.md: Improve documentation.
2740 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
2743 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2744 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2746 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
2748 * config.gcc: Remove "carrizo" support.
2749 * config/gcn/gcn-opts.h (processor_type): Likewise.
2750 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2751 * config/gcn/gcn.opt (gpu_type): Likewise.
2752 * config/gcn/t-omp-device: Likewise.
2754 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2757 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2758 * config/arm/arm.c (arm_gen_far_branch): New function
2760 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2762 2020-02-03 Julian Brown <julian@codesourcery.com>
2763 Tobias Burnus <tobias@codesourcery.com>
2765 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2767 2020-02-03 Jakub Jelinek <jakub@redhat.com>
2770 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2771 valid RTL to sum up the lowest and second lowest bytes of the popcnt
2774 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
2776 PR rtl-optimization/91333
2777 * ira-color.c (struct allocno_color_data): Add member
2779 (init_allocno_threads): Set the member up.
2780 (bucket_allocno_compare_func): Add compare hard reg
2783 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
2785 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2787 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2788 * config.in: Regenerated.
2789 * configure: Regenerated.
2790 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2791 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2792 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2794 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
2796 * configure: Regenerate.
2798 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
2800 PR rtl-optimization/91333
2801 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2802 reg preferences comparison up.
2804 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2806 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2807 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2808 aarch64-sve-builtins-base.h.
2809 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2810 aarch64-sve-builtins-base.cc.
2811 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2812 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2814 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2815 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2816 (svcvtnt): New functions.
2817 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2818 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2819 (svcvtnt): New functions.
2820 (svcvt): Add a form that converts f32 to bf16.
2821 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2822 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2824 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2825 Treat B as bfloat16_t.
2826 (ternary_bfloat_lane_base): New class.
2827 (ternary_bfloat_def): Likewise.
2828 (ternary_bfloat): New shape.
2829 (ternary_bfloat_lane_def): New class.
2830 (ternary_bfloat_lane): New shape.
2831 (ternary_bfloat_lanex2_def): New class.
2832 (ternary_bfloat_lanex2): New shape.
2833 (ternary_bfloat_opt_n_def): New class.
2834 (ternary_bfloat_opt_n): New shape.
2835 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2836 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2837 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2838 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2839 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2840 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2841 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2842 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2843 the pattern off the narrow mode instead of the wider one.
2844 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2845 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2846 (sve_fp_op): Handle them.
2847 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2848 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2850 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2852 * config/aarch64/arm_sve.h: Include arm_bf16.h.
2853 * config/aarch64/aarch64-modes.def (BF): Move definition before
2854 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
2855 (SVE_MODES): Handle BF modes.
2856 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2858 (aarch64_full_sve_mode): Likewise.
2859 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2861 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2862 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2863 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2864 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2866 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2868 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2869 (TYPES_all_data): Add bf16.
2870 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2871 (register_tuple_type): Increase buffer size.
2872 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2873 (bf16): New type suffix.
2874 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2875 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2876 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2877 Change type from all_data to all_arith.
2878 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2881 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
2882 Matthew Malcomson <matthew.malcomson@arm.com>
2883 Richard Sandiford <richard.sandiford@arm.com>
2885 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2886 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2887 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2888 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
2889 __ARM_FEATURE_MATMUL_FP64.
2890 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2891 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2892 be disabled at the same time.
2893 (f32mm): New extension.
2894 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2895 (AARCH64_FL_F64MM): Bump to the next bit up.
2896 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2897 (TARGET_SVE_F64MM): New macros.
2898 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2899 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2900 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2901 (UNSPEC_ZIP2Q): New unspeccs.
2902 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2903 (optab, sur, perm_insn): Handle the new unspecs.
2904 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
2905 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2906 TARGET_SVE_F64MM instead of separate tests.
2907 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2908 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2909 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2910 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2911 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2912 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2913 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2914 (TYPES_s_signed): New macro.
2915 (TYPES_s_integer): Use it.
2916 (TYPES_d_float): New macro.
2917 (TYPES_d_data): Use it.
2918 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2919 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2920 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2921 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2922 (svmmla): New shape.
2923 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2924 template parameters.
2925 (ternary_resize2_lane_base): Likewise.
2926 (ternary_resize2_base): New class.
2927 (ternary_qq_lane_base): Likewise.
2928 (ternary_intq_uintq_lane_def): Likewise.
2929 (ternary_intq_uintq_lane): New shape.
2930 (ternary_intq_uintq_opt_n_def): New class
2931 (ternary_intq_uintq_opt_n): New shape.
2932 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2933 (ternary_uintq_intq_def): New class.
2934 (ternary_uintq_intq): New shape.
2935 (ternary_uintq_intq_lane_def): New class.
2936 (ternary_uintq_intq_lane): New shape.
2937 (ternary_uintq_intq_opt_n_def): New class.
2938 (ternary_uintq_intq_opt_n): New shape.
2939 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2940 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2941 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2942 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2944 (svdotprod_lane_impl): ...this new class.
2945 (svmmla_impl, svusdot_impl): New classes.
2946 (svdot_lane): Update to use svdotprod_lane_impl.
2947 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2948 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2950 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2951 function, with no types defined.
2952 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2953 AARCH64_FL_I8MM functions.
2954 (svmmla): New AARCH64_FL_F32MM function.
2955 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2956 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2957 AARCH64_FL_F64MM function.
2958 (REQUIRED_EXTENSIONS):
2960 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2962 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2965 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
2967 * config/i386/i386.md (*movoi_internal_avx): Do not check for
2968 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
2969 (*movti_internal): Do not check for
2970 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2971 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2972 just after check for TARGET_AVX.
2973 (*movdf_internal): Ditto.
2974 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2975 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2976 * config/i386/sse.md (mov<mode>_internal): Only check
2977 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
2978 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2979 (<sse>_andnot<mode>3<mask_name>): Move check for
2980 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2981 (<code><mode>3<mask_name>): Ditto.
2982 (*andnot<mode>3): Ditto.
2983 (*andnottf3): Ditto.
2984 (*<code><mode>3): Ditto.
2985 (*<code>tf3): Ditto.
2986 (*andnot<VI:mode>3): Remove
2987 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2988 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2989 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2990 (sse4_1_blendv<ssemodesuffix>): Ditto.
2991 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2992 Explain that tune applies to 128bit instructions only.
2994 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
2996 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2997 to definition of hsa_kernel_description. Parse assembly to find SGPR
2998 and VGPR count of kernel and store in hsa_kernel_description.
3000 2020-01-31 Tamar Christina <tamar.christina@arm.com>
3002 PR rtl-optimization/91838
3003 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
3004 to truncate if allowed or reject combination.
3006 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
3008 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
3009 (find_inv_vars_cb): Likewise.
3011 2020-01-31 David Malcolm <dmalcolm@redhat.com>
3013 * calls.c (special_function_p): Split out the check for DECL_NAME
3014 being non-NULL and fndecl being extern at file scope into a
3015 new maybe_special_function_p and call it. Drop check for fndecl
3016 being non-NULL that was after a usage of DECL_NAME (fndecl).
3017 * tree.h (maybe_special_function_p): New inline function.
3019 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
3021 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
3022 (mask_gather_load<mode>): ... here, and zero-initialize the
3024 (maskload<mode>di): Zero-initialize the destination.
3027 2020-01-30 David Malcolm <dmalcolm@redhat.com>
3030 * doc/analyzer.texi (Limitations): Note that constraints on
3031 floating-point values are currently ignored.
3033 2020-01-30 Jakub Jelinek <jakub@redhat.com>
3036 * symtab.c (symtab_node::noninterposable_alias): If localalias
3037 already exists, but is not usable, append numbers after it until
3038 a unique name is found. Formatting fix.
3041 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
3044 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
3046 * config/gcn/gcn.c (print_operand): Handle LTGT.
3047 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
3049 2020-01-30 Richard Biener <rguenther@suse.de>
3051 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
3052 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
3054 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
3056 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
3057 without a DECL in .data.rel.ro.local.
3059 2020-01-30 Jakub Jelinek <jakub@redhat.com>
3062 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
3066 * config/i386/sse.md
3067 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
3068 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
3069 any_extend code iterator instead of always zero_extend.
3070 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
3071 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
3072 Use any_extend code iterator instead of always zero_extend.
3073 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
3074 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
3075 Use any_extend code iterator instead of always zero_extend.
3076 (*sse2_pmovmskb_ext): New define_insn.
3077 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
3080 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
3081 (*popcountsi2_zext_falsedep): New define_insn.
3083 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3085 * config.in: Regenerated.
3086 * configure: Regenerated.
3088 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
3091 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
3092 LLVM's assembler changed the default in version 9.
3094 2020-01-24 Jeff Law <law@redhat.com>
3096 PR tree-optimization/89689
3097 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
3099 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
3103 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3105 PR rtl-optimization/87763
3106 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
3107 simplification to handle subregs as well as bare regs.
3108 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
3110 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
3113 * ira.c (ira): Revert use of simplified LRA algorithm.
3115 2020-01-29 Martin Jambor <mjambor@suse.cz>
3117 PR tree-optimization/92706
3118 * tree-sra.c (struct access): Fields first_link, last_link,
3119 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
3120 next_rhs_queued and grp_rhs_queued respectively, new fields
3121 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
3122 (struct assign_link): Field next renamed to next_rhs, new field
3123 next_lhs. Updated comment.
3124 (work_queue_head): Renamed to rhs_work_queue_head.
3125 (lhs_work_queue_head): New variable.
3126 (add_link_to_lhs): New function.
3127 (relink_to_new_repr): Also relink LHS lists.
3128 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
3129 (add_access_to_lhs_work_queue): New function.
3130 (pop_access_from_work_queue): Renamed to
3131 pop_access_from_rhs_work_queue.
3132 (pop_access_from_lhs_work_queue): New function.
3133 (build_accesses_from_assign): Also add links to LHS lists and to LHS
3135 (child_would_conflict_in_lacc): Renamed to
3136 child_would_conflict_in_acc. Adjusted parameter names.
3137 (create_artificial_child_access): New parameter set_grp_read, use it.
3138 (subtree_mark_written_and_enqueue): Renamed to
3139 subtree_mark_written_and_rhs_enqueue.
3140 (propagate_subaccesses_across_link): Renamed to
3141 propagate_subaccesses_from_rhs.
3142 (propagate_subaccesses_from_lhs): New function.
3143 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
3146 2020-01-29 Martin Jambor <mjambor@suse.cz>
3148 PR tree-optimization/92706
3149 * tree-sra.c (struct access): Adjust comment of
3150 grp_total_scalarization.
3151 (find_access_in_subtree): Look for single children spanning an entire
3153 (scalarizable_type_p): Allow register accesses, adjust callers.
3154 (completely_scalarize): Remove function.
3155 (scalarize_elem): Likewise.
3156 (create_total_scalarization_access): Likewise.
3157 (sort_and_splice_var_accesses): Do not track total scalarization
3159 (analyze_access_subtree): New parameter totally, adjust to new meaning
3160 of grp_total_scalarization.
3161 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
3162 (can_totally_scalarize_forest_p): New function.
3163 (create_total_scalarization_access): Likewise.
3164 (create_total_access_and_reshape): Likewise.
3165 (total_should_skip_creating_access): Likewise.
3166 (totally_scalarize_subtree): Likewise.
3167 (analyze_all_variable_accesses): Perform total scalarization after
3168 subaccess propagation using the new functions above.
3169 (initialize_constant_pool_replacements): Output initializers by
3170 traversing the access tree.
3172 2020-01-29 Martin Jambor <mjambor@suse.cz>
3174 * tree-sra.c (verify_sra_access_forest): New function.
3175 (verify_all_sra_access_forests): Likewise.
3176 (create_artificial_child_access): Set parent.
3177 (analyze_all_variable_accesses): Call the verifier.
3179 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3181 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
3182 if called on indirect edge.
3183 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
3184 speculative call if needed.
3186 2020-01-29 Richard Biener <rguenther@suse.de>
3188 PR tree-optimization/93428
3189 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
3190 permutation when the load node is created.
3191 (vect_analyze_slp_instance): Re-use it here.
3193 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3195 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
3197 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
3199 PR rtl-optimization/93272
3200 * ira-lives.c (process_out_of_region_eh_regs): New function.
3201 (process_bb_node_lives): Call it.
3203 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3205 * coverage.c (read_counts_file): Make error message lowercase.
3207 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3209 * profile-count.c (profile_quality_display_names): Fix ordering.
3211 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3214 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
3215 hash only when edge is first within the sequence.
3216 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
3217 (symbol_table::create_edge): Do not set target_prob.
3218 (cgraph_edge::remove_caller): Watch for speculative calls when updating
3220 (cgraph_edge::make_speculative): Drop target_prob parameter.
3221 (cgraph_edge::speculative_call_info): Remove.
3222 (cgraph_edge::first_speculative_call_target): New member function.
3223 (update_call_stmt_hash_for_removing_direct_edge): New function.
3224 (cgraph_edge::resolve_speculation): Rewrite to new API.
3225 (cgraph_edge::speculative_call_for_target): New member function.
3226 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
3227 multiple speculation targets.
3228 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
3230 (verify_speculative_call): Verify that targets form an interval.
3231 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
3232 (cgraph_edge::first_speculative_call_target): New member function.
3233 (cgraph_edge::next_speculative_call_target): New member function.
3234 (cgraph_edge::speculative_call_target_ref): New member function.
3235 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
3236 (cgraph_edge): Remove target_prob.
3237 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3238 Fix handling of speculative calls.
3239 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
3240 * ipa-fnsummary.c (analyze_function_body): Likewise.
3241 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
3242 * ipa-profile.c (dump_histogram): Fix formating.
3243 (ipa_profile_generate_summary): Watch for overflows.
3244 (ipa_profile): Do not require probablity to be 1/2; update to new API.
3245 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
3246 (update_indirect_edges_after_inlining): Update to new API.
3247 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
3249 * profile-count.h: (profile_probability::adjusted): New.
3250 * tree-inline.c (copy_bb): Update to new speculative call API; fix
3251 updating of profile.
3252 * value-prof.c (gimple_ic_transform): Rename to ...
3253 (dump_ic_profile): ... this one; update dumping.
3254 (stream_in_histogram_value): Fix formating.
3255 (gimple_value_profile_transformations): Update.
3257 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
3260 * config/i386/i386.md (*movoi_internal_avx): Remove
3261 TARGET_SSE_TYPELESS_STORES check.
3262 (*movti_internal): Prefer TARGET_AVX over
3263 TARGET_SSE_TYPELESS_STORES.
3264 (*movtf_internal): Likewise.
3265 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
3266 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
3267 from TARGET_SSE_TYPELESS_STORES.
3269 2020-01-28 David Malcolm <dmalcolm@redhat.com>
3271 * diagnostic-core.h (warning_at): Rename overload to...
3272 (warning_meta): ...this.
3273 (emit_diagnostic_valist): Delete decl of overload taking
3274 diagnostic_metadata.
3275 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
3276 (warning_at): Rename overload taking diagnostic_metadata to...
3277 (warning_meta): ...this.
3279 2020-01-28 Richard Biener <rguenther@suse.de>
3281 PR tree-optimization/93439
3282 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
3283 * tree-cfg.c (move_sese_region_to_fn): ... here.
3284 (verify_types_in_gimple_reference): Verify used cliques are
3287 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
3290 * config/i386/i386-options.c (set_ix86_tune_features): Add an
3291 argument of a pointer to struct gcc_options and pass it to
3292 parse_mtune_ctrl_str.
3293 (ix86_function_specific_restore): Pass opts to
3294 set_ix86_tune_features.
3295 (ix86_option_override_internal): Likewise.
3296 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
3297 gcc_options and use it for x_ix86_tune_ctrl_string.
3299 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3301 PR rtl-optimization/87763
3302 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
3303 simplification to handle subregs as well as bare regs.
3304 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
3306 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3308 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
3309 for reduction chains that (now) include a call.
3311 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3313 PR tree-optimization/92822
3314 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
3315 out the don't-care elements of a vector whose significant elements
3316 are duplicates, make the don't-care elements duplicates too.
3318 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3320 PR tree-optimization/93434
3321 * tree-predcom.c (split_data_refs_to_components): Record which
3322 components have had aliasing loads removed. Prevent store-store
3323 commoning for all such components.
3325 2020-01-28 Jakub Jelinek <jakub@redhat.com>
3328 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
3329 -1 or is_vshift is true, use new_vector with number of elts npatterns
3330 rather than new_unary_operation.
3332 PR tree-optimization/93454
3333 * gimple-fold.c (fold_array_ctor_reference): Perform
3334 elt_size.to_uhwi () just once, instead of calling it in every
3335 iteration. Punt if that value is above size of the temporary
3336 buffer. Decrease third native_encode_expr argument when
3337 bufoff + elt_sz is above size of buf.
3339 2020-01-27 Joseph Myers <joseph@codesourcery.com>
3341 * config/mips/mips.c (mips_declare_object_name)
3342 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
3344 2020-01-27 Martin Liska <mliska@suse.cz>
3346 PR gcov-profile/93403
3347 * tree-profile.c (gimple_init_gcov_profiler): Generate
3348 both __gcov_indirect_call_profiler_v4 and
3349 __gcov_indirect_call_profiler_v4_atomic.
3351 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3354 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
3356 (@aarch64_split_simd_mov<mode>): Use it.
3357 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
3358 Leave the vec_extract patterns to handle 2-element vectors.
3359 (aarch64_simd_mov_from_<mode>high): Likewise.
3360 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
3361 (vec_extractv2dfv1df): Likewise.
3363 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3365 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
3366 jump conditions for *compare_condjump<GPI:mode>.
3368 2020-01-27 David Malcolm <dmalcolm@redhat.com>
3371 * digraph.cc (test_edge::test_edge): Specify template for base
3374 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
3376 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
3378 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
3380 * config/arc/arc-protos.h (gen_mlo): Remove.
3381 (gen_mhi): Likewise.
3382 * config/arc/arc.c (AUX_MULHI): Define.
3383 (arc_must_save_reister): Special handling for r58/59.
3384 (arc_compute_frame_size): Consider mlo/mhi registers.
3385 (arc_save_callee_saves): Emit fp/sp move only when emit_move
3387 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
3388 mlo/mhi name selection.
3389 (arc_restore_callee_saves): Don't early restore blink when ISR.
3390 (arc_expand_prologue): Add mlo/mhi saving.
3391 (arc_expand_epilogue): Add mlo/mhi restoring.
3394 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
3395 numbering when MUL64 option is used.
3396 (DWARF2_FRAME_REG_OUT): Define.
3397 * config/arc/arc.md (arc600_stall): New pattern.
3398 (VUNSPEC_ARC_ARC600_STALL): Define.
3399 (mulsi64): Use correct mlo/mhi registers.
3400 (mulsi_600): Clean it up.
3401 * config/arc/predicates.md (mlo_operand): Remove any dependency on
3403 (mhi_operand): Likewise.
3405 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
3406 Petro Karashchenko <petro.karashchenko@ring.com>
3408 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
3409 attributes if needed.
3410 (prepare_move_operands): Generate special unspec instruction for
3412 (arc_isuncached_mem_p): Propagate uncached attribute to each
3414 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
3415 (VUNSPEC_ARC_STDI): Likewise.
3416 (ALLI): New mode iterator.
3417 (mALLI): New mode attribute.
3418 (lddi): New instruction pattern.
3420 (stdidi_split): Split instruction for architectures which are not
3421 supporting ll64 option.
3422 (lddidi_split): Likewise.
3424 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3426 PR rtl-optimization/92989
3427 * lra-lives.c (process_bb_lives): Update the live-in set before
3428 processing additional clobbers.
3430 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3432 PR rtl-optimization/93170
3433 * cselib.c (cselib_invalidate_regno_val): New function, split out
3435 (cselib_invalidate_regno): ...here.
3436 (cselib_invalidated_by_call_p): New function.
3437 (cselib_process_insn): Iterate over all the hard-register entries in
3438 REG_VALUES and invalidate any that cross call-clobbered registers.
3440 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3442 * dojump.c (split_comparison): Use HONOR_NANS rather than
3443 HONOR_SNANS when splitting LTGT.
3445 2020-01-27 Martin Liska <mliska@suse.cz>
3448 * opts.c (print_filtered_help): Exclude language-specific
3449 options from --help=common unless enabled in all FEs.
3451 2020-01-27 Martin Liska <mliska@suse.cz>
3453 * opts.c (print_help): Exclude params from
3454 all except --help=param.
3456 2020-01-27 Martin Liska <mliska@suse.cz>
3459 * config/i386/i386-features.c (make_resolver_func):
3460 Align the code with ppc64 target implementation.
3461 Do not generate a unique name for resolver function.
3463 2020-01-27 Richard Biener <rguenther@suse.de>
3465 PR tree-optimization/93397
3466 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
3467 converted reduction chain SLP graph adjustment.
3469 2020-01-26 Marek Polacek <polacek@redhat.com>
3472 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
3475 2020-01-26 Jason Merrill <jason@redhat.com>
3478 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
3481 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
3483 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
3484 (rx_setmem): Likewise.
3486 2020-01-26 Jakub Jelinek <jakub@redhat.com>
3489 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
3490 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
3491 drop <di> from constraint of last operand.
3494 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
3495 TARGET_AVX2 and V4DFmode not in the split condition, but in the
3496 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
3498 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
3501 * ipa-cp.c (get_info_about_necessary_edges): Remove value
3504 2020-01-24 Jeff Law <law@redhat.com>
3506 PR tree-optimization/92788
3507 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
3510 2020-01-24 Jakub Jelinek <jakub@redhat.com>
3513 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
3514 *avx_vperm_broadcast_<mode>,
3515 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
3516 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
3517 Move before avx2_perm<mode>/avx512f_perm<mode>.
3520 * simplify-rtx.c (simplify_const_unary_operation,
3521 simplify_const_binary_operation): Punt for mode precision above
3522 MAX_BITSIZE_MODE_ANY_INT.
3524 2020-01-24 Andrew Pinski <apinski@marvell.com>
3526 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
3529 2020-01-24 Jeff Law <law@redhat.com>
3532 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
3533 for REGs. Call output_operand_lossage to get more reasonable
3536 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
3538 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
3539 gcn_fp_compare_operator.
3540 (vec_cmpu<mode>di): Use gcn_compare_operator.
3541 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
3542 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
3543 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
3544 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
3545 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
3546 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
3547 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
3548 gcn_fp_compare_operator.
3549 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
3550 gcn_fp_compare_operator.
3551 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
3552 gcn_fp_compare_operator.
3553 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
3554 gcn_fp_compare_operator.
3556 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
3558 * doc/install.texi (Cross-Compiler-Specific Options): Document
3559 `--with-toolexeclibdir' option.
3561 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
3563 * target.def (flags_regnum): Also mention effect on delay slot filling.
3564 * doc/tm.texi: Regenerate.
3566 2020-01-23 Jeff Law <law@redhat.com>
3568 PR translation/90162
3569 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
3571 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
3574 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
3577 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3579 PR rtl-optimization/93402
3580 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
3583 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3585 * config.in: Regenerated.
3586 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
3587 for TARGET_LIBC_GNUSTACK.
3588 * configure: Regenerated.
3589 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
3590 found to be 2.31 or greater.
3592 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3594 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
3596 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
3597 (mips_asm_file_end): New function. Delegate to
3598 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
3599 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3601 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3604 * config/i386/i386-modes.def (POImode): New mode.
3605 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
3606 * config/i386/i386.md (DPWI): New mode attribute.
3607 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
3609 (QPWI): ... this. Use POI instead of OI for TImode.
3610 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
3611 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
3614 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3617 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
3619 (speculation_tracker_rev): New pattern.
3620 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
3621 Use speculation_tracker_rev to track the inverse condition.
3623 2020-01-23 Richard Biener <rguenther@suse.de>
3625 PR tree-optimization/93381
3626 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
3627 alias-set of the def as argument and record the first one.
3628 (vn_walk_cb_data::first_set): New member.
3629 (vn_reference_lookup_3): Pass the alias-set of the current def
3630 to push_partial_def. Fix alias-set used in the aggregate copy
3632 (vn_reference_lookup): Consistently set *last_vuse_ptr.
3633 * real.c (clear_significand_below): Fix out-of-bound access.
3635 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3638 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
3639 New define_insn patterns.
3641 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3643 * doc/sourcebuild.texi (check-function-bodies): Add an
3644 optional target/xfail selector.
3646 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3648 PR rtl-optimization/93124
3649 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
3650 bare USE and CLOBBER insns.
3652 2020-01-22 Andrew Pinski <apinski@marvell.com>
3654 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
3656 2020-01-22 David Malcolm <dmalcolm@redhat.com>
3659 * gdbinit.in (break-on-saved-diagnostic): Update for move of
3660 diagnostic_manager into "ana" namespace.
3661 * selftest-run-tests.c (selftest::run_tests): Update for move of
3662 selftest::run_analyzer_selftests to
3663 ana::selftest::run_analyzer_selftests.
3665 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
3667 * cfgexpand.c (union_stack_vars): Update the size.
3669 2020-01-22 Richard Biener <rguenther@suse.de>
3671 PR tree-optimization/93381
3672 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
3673 throughout, handle all conversions the same.
3675 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3678 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
3679 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
3680 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
3681 Call force_reg on high_in2 unconditionally.
3683 2020-01-22 Martin Liska <mliska@suse.cz>
3685 PR tree-optimization/92924
3686 * profile.c (compute_value_histograms): Divide
3689 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3692 * output.h (assemble_name_resolve): Declare.
3693 * varasm.c (assemble_name_resolve): New function.
3694 (assemble_name): Use it.
3695 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
3697 2020-01-22 Joseph Myers <joseph@codesourcery.com>
3699 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
3700 update_web_docs_git instead of update_web_docs_svn.
3702 2020-01-21 Andrew Pinski <apinski@marvell.com>
3705 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
3706 as PTR mode. Have operand 1 as being modeless, it can be P mode.
3707 (*tlsgd_small_<mode>): Likewise.
3708 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
3709 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
3710 register. Convert that register back to dest using convert_mode.
3712 2020-01-21 Jim Wilson <jimw@sifive.com>
3714 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
3717 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
3718 Uros Bizjak <ubizjak@gmail.com>
3721 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3723 (legitimize_tls_address): Do GNU2 TLS address computation in
3724 ptr_mode and zero-extend result to Pmode.
3725 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3726 :P with :PTR and Pmode with ptr_mode.
3727 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3728 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3729 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3731 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3734 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3735 the last two operands are CONST_INT_P before using them as such.
3737 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3739 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3740 to get the integer element types.
3742 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3744 * config/aarch64/aarch64-sve-builtins.h
3745 (function_expander::convert_to_pmode): Declare.
3746 * config/aarch64/aarch64-sve-builtins.cc
3747 (function_expander::convert_to_pmode): New function.
3748 (function_expander::get_contiguous_base): Use it.
3749 (function_expander::prepare_gather_address_operands): Likewise.
3750 * config/aarch64/aarch64-sve-builtins-sve2.cc
3751 (svwhilerw_svwhilewr_impl::expand): Likewise.
3753 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
3756 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3757 cfun->machine->label_is_assembled.
3758 (aarch64_print_patchable_function_entry): New.
3759 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3760 * config/aarch64/aarch64.h (struct machine_function): New field,
3763 2020-01-21 David Malcolm <dmalcolm@redhat.com>
3766 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3769 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3772 * cgraph.c (cgraph_edge::resolve_speculation,
3773 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3774 call_stmt_site_hash.
3776 2020-01-21 Martin Liska <mliska@suse.cz>
3778 * config/rs6000/rs6000.c (common_mode_defined): Remove
3781 2020-01-21 Richard Biener <rguenther@suse.de>
3783 PR tree-optimization/92328
3784 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3785 type when value-numbering same-sized store by inserting a
3787 (eliminate_dom_walker::eliminate_stmt): When eliminating
3788 a redundant store handle bit-reinterpretation of the same value.
3790 2020-01-21 Andrew Pinski <apinski@marvel.com>
3793 * tree-into-ssa.c (prepare_block_for_update_1): Split out
3795 (prepare_block_for_update): This. Use a worklist instead of
3798 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3800 * gcc/config/arm/arm.c (clear_operation_p):
3801 Initialise last_regno, skip first iteration
3802 based on the first_set value and use ints instead
3803 of the unnecessary HOST_WIDE_INTs.
3805 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3808 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3809 compare_mode other than SFmode or DFmode.
3811 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
3814 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3815 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3816 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3818 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3820 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3822 2020-01-20 Andrew Pinski <apinski@marvell.com>
3825 * targhooks.c (default_print_patchable_function_entry): Use
3826 output_asm_insn to emit the nop instruction.
3828 2020-01-20 Fangrui Song <maskray@google.com>
3831 * targhooks.c (default_print_patchable_function_entry): Align to
3834 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
3837 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3838 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
3839 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3840 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
3841 (*tls_dynamic_gnu2_lea_64): Renamed to ...
3842 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
3843 Remove the {q} suffix from lea.
3844 (*tls_dynamic_gnu2_call_64): Renamed to ...
3845 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
3846 (*tls_dynamic_gnu2_combine_64): Renamed to ...
3847 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
3848 Pass Pmode to gen_tls_dynamic_gnu2_64.
3850 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3852 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3854 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
3856 * config/aarch64/aarch64-sve-builtins-base.cc
3857 (svld1ro_impl::memory_vector_mode): Remove parameter name.
3859 2020-01-20 Richard Biener <rguenther@suse.de>
3862 * dwarf2out.c (prune_unused_types): Unconditionally mark
3863 called function DIEs.
3865 2020-01-20 Martin Liska <mliska@suse.cz>
3867 PR tree-optimization/93199
3868 * tree-eh.c (struct leh_state): Add
3869 new field outer_non_cleanup.
3870 (cleanup_is_dead_in): Pass leh_state instead
3871 of eh_region. Add a checking that state->outer_non_cleanup
3872 points to outer non-clean up region.
3873 (lower_try_finally): Record outer_non_cleanup
3875 (lower_catch): Likewise.
3876 (lower_eh_filter): Likewise.
3877 (lower_eh_must_not_throw): Likewise.
3878 (lower_cleanup): Likewise.
3880 2020-01-20 Richard Biener <rguenther@suse.de>
3882 PR tree-optimization/93094
3883 * tree-vectorizer.h (vect_loop_versioning): Adjust.
3884 (vect_transform_loop): Likewise.
3885 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3886 loop_vectorized_call to vect_transform_loop.
3887 * tree-vect-loop.c (vect_transform_loop): Pass down
3888 loop_vectorized_call to vect_loop_versioning.
3889 * tree-vect-loop-manip.c (vect_loop_versioning): Use
3890 the earlier discovered loop_vectorized_call.
3892 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
3894 * doc/contribute.texi: Update for SVN -> Git transition.
3895 * doc/install.texi: Likewise.
3897 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3900 * cgraph.c (cgraph_edge::make_speculative): Increase number of
3901 speculative targets.
3902 (verify_speculative_call): New function
3903 (cgraph_node::verify_node): Use it.
3904 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3907 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3910 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3911 (cgraph_edge::make_direct): Remove all indirect targets.
3912 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3913 (cgraph_node::verify_node): Verify that only one call_stmt or
3914 lto_stmt_uid is set.
3915 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3917 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3918 (lto_output_ref): Simplify streaming of stmt.
3919 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3921 2020-01-18 Tamar Christina <tamar.christina@arm.com>
3923 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3924 Mark parameter unused.
3926 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
3928 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3930 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
3932 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3934 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
3936 * Makefile.in: Add coroutine-passes.o.
3937 * builtin-types.def (BT_CONST_SIZE): New.
3938 (BT_FN_BOOL_PTR): New.
3939 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3940 * builtins.def (DEF_COROUTINE_BUILTIN): New.
3941 * coroutine-builtins.def: New file.
3942 * coroutine-passes.cc: New file.
3943 * function.h (struct GTY function): Add a bit to indicate that the
3944 function is a coroutine component.
3945 * internal-fn.c (expand_CO_FRAME): New.
3946 (expand_CO_YIELD): New.
3947 (expand_CO_SUSPN): New.
3948 (expand_CO_ACTOR): New.
3949 * internal-fn.def (CO_ACTOR): New.
3953 * passes.def: Add pass_coroutine_lower_builtins,
3954 pass_coroutine_early_expand_ifns.
3955 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3956 (make_pass_coroutine_early_expand_ifns): New.
3957 * doc/invoke.texi: Document the fcoroutines command line
3960 2020-01-18 Jakub Jelinek <jakub@redhat.com>
3962 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3965 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3966 after checking the argument is a REG. Don't use REGNO (reg)
3967 again to set last_regno, reuse regno variable instead.
3969 2020-01-17 David Malcolm <dmalcolm@redhat.com>
3971 * doc/analyzer.texi (Limitations): Add note about NaN.
3973 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3974 Sudakshina Das <sudi.das@arm.com>
3976 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3977 and valid immediate.
3978 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3979 (lshrdi3): Generate thumb2_lsrl for valid immediates.
3980 * config/arm/constraints.md (Pg): New.
3981 * config/arm/predicates.md (long_shift_imm): New.
3982 (arm_reg_or_long_shift_imm): Likewise.
3983 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3984 (thumb2_lsll): Likewise.
3987 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3988 Sudakshina Das <sudi.das@arm.com>
3990 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3991 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3992 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3993 register pairs for doubleword quantities for ARMv8.1M-Mainline.
3994 * config/arm/thumb2.md (thumb2_asrl): New.
3995 (thumb2_lsll): Likewise.
3997 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3999 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
4002 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
4004 * gdbinit.in (help-gcc-hooks): New command.
4005 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
4006 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
4009 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
4011 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
4012 correct target macro.
4014 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
4016 * config/aarch64/aarch64-protos.h
4017 (aarch64_sve_ld1ro_operand_p): New.
4018 * config/aarch64/aarch64-sve-builtins-base.cc
4019 (class load_replicate): New.
4020 (class svld1ro_impl): New.
4021 (class svld1rq_impl): Change to inherit from load_replicate.
4022 (svld1ro): New sve intrinsic function base.
4023 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
4024 New DEF_SVE_FUNCTION.
4025 * config/aarch64/aarch64-sve-builtins-base.h
4026 (svld1ro): New decl.
4027 * config/aarch64/aarch64-sve-builtins.cc
4028 (function_expander::add_mem_operand): Modify assert to allow
4030 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
4032 * config/aarch64/aarch64.c
4033 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
4034 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
4035 (aarch64_sve_ld1ro_operand_p): New.
4036 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
4037 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
4038 * config/aarch64/predicates.md
4039 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
4041 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
4043 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
4044 Introduce this ACLE specified predefined macro.
4045 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
4046 (fp): Disabling this disables f64mm.
4047 (simd): Disabling this disables f64mm.
4048 (fp16): Disabling this disables f64mm.
4049 (sve): Disabling this disables f64mm.
4050 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
4051 (AARCH64_ISA_F64MM): New.
4052 (TARGET_F64MM): New.
4053 * doc/invoke.texi (f64mm): Document new option.
4055 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4057 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
4058 (neoversen1_tunings): Likewise.
4060 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
4063 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
4064 Add assert to ensure prolog has been emitted.
4065 (aarch64_split_atomic_op): Likewise.
4066 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
4067 Use epilogue_completed rather than reload_completed.
4068 (aarch64_atomic_exchange<mode>): Likewise.
4069 (aarch64_atomic_<atomic_optab><mode>): Likewise.
4070 (atomic_nand<mode>): Likewise.
4071 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
4072 (atomic_fetch_nand<mode>): Likewise.
4073 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
4074 (atomic_nand_fetch<mode>): Likewise.
4076 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
4079 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
4081 (REVERSE_CONDITION): Delete.
4082 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
4083 (CCFP_CCFPE): Likewise.
4084 (e): New mode attribute.
4085 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
4086 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
4087 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
4088 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
4089 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
4090 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
4091 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
4092 name of generator from gen_ccmpdi to gen_ccmpccdi.
4093 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
4094 the previous comparison but aren't able to, use the new ccmp_rev
4097 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
4099 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
4100 than testing directly for INTEGER_CST.
4101 (gimplify_target_expr, gimplify_omp_depend): Likewise.
4103 2020-01-17 Jakub Jelinek <jakub@redhat.com>
4105 PR tree-optimization/93292
4106 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
4107 get_vectype_for_scalar_type returns NULL.
4109 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
4111 * params.opt (-param=max-predicted-iterations): Increase range from 0.
4112 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
4114 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
4116 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
4118 * params.opt: (max-predicted-iterations): Set bounds.
4119 * predict.c (real_almost_one, real_br_prob_base,
4120 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
4121 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
4122 probabilities; do not truncate to reg_br_prob_bases.
4123 (estimate_loops_at_level): Pass max_cyclic_prob.
4124 (estimate_loops): Compute max_cyclic_prob.
4125 (estimate_bb_frequencies): Do not initialize real_*; update calculation
4127 * profile-count.c (profile_probability::to_sreal): New.
4128 * profile-count.h (class sreal): Move up in file.
4129 (profile_probability::to_sreal): Declare.
4131 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4134 (arm_invalid_conversion): New function for target hook.
4135 (arm_invalid_unary_op): New function for target hook.
4136 (arm_invalid_binary_op): New function for target hook.
4138 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4140 * config.gcc: Add arm_bf16.h.
4141 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
4142 (arm_simd_builtin_std_type): Add BFmode.
4143 (arm_init_simd_builtin_types): Define element types for vector types.
4144 (arm_init_bf16_types): New function.
4145 (arm_init_builtins): Add arm_init_bf16_types function call.
4146 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
4147 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
4148 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
4149 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
4150 (arm_vector_mode_supported_p): Add V4BF, V8BF.
4151 (arm_mangle_type): Add __bf16.
4152 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
4153 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
4154 arm_bf16_ptr_type_node.
4155 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
4156 define_split between ARM registers.
4157 * config/arm/arm_bf16.h: New file.
4158 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4159 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
4161 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
4162 * config/arm/vfp.md: Add BFmode to movhf patterns.
4164 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
4165 Andre Vieira <andre.simoesdiasvieira@arm.com>
4167 * config/arm/arm-cpus.in (mve, mve_float): New features.
4168 (dsp, mve, mve.fp): New options.
4169 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
4170 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
4171 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
4173 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4174 Thomas Preud'homme <thomas.preudhomme@arm.com>
4176 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
4178 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
4179 error for using -mcmse when targeting Armv8.1-M Mainline.
4181 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4182 Thomas Preud'homme <thomas.preudhomme@arm.com>
4184 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
4185 address in r4 when targeting Armv8.1-M Mainline.
4186 (nonsecure_call_value_internal): Likewise.
4187 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
4188 a register match_operand again. Emit BLXNS when targeting
4190 (nonsecure_call_value_reg_thumb2): Likewise.
4192 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4193 Thomas Preud'homme <thomas.preudhomme@arm.com>
4195 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
4196 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
4197 variable as true when floating-point ABI is not hard. Replace
4198 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
4199 Generate VLSTM and VLLDM instruction respectively before and
4200 after a function call to cmse_nonsecure_call function.
4201 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
4202 (VUNSPEC_VLLDM): Likewise.
4203 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
4204 (lazy_load_multiple_insn): Likewise.
4206 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4207 Thomas Preud'homme <thomas.preudhomme@arm.com>
4209 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
4210 (arm_emit_vfp_multi_reg_pop): Likewise.
4211 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
4212 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
4213 restore callee-saved VFP registers.
4215 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4216 Thomas Preud'homme <thomas.preudhomme@arm.com>
4218 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
4219 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
4220 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
4221 callee-saved GPRs as well as clear ip register before doing a nonsecure
4222 call then restore callee-saved GPRs after it when targeting
4224 (arm_reorg): Adapt to function rename.
4226 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4227 Thomas Preud'homme <thomas.preudhomme@arm.com>
4229 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
4230 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
4231 clear_vfp_multiple pattern based on a new vfp parameter.
4232 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
4233 targeting Armv8.1-M Mainline.
4234 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
4235 unconditionally when targeting Armv8.1-M Mainline architecture. Check
4236 whether VFP registers are available before looking call_used_regs for a
4238 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
4239 of prototype of clear_operation_p.
4240 (clear_vfp_multiple_operation): New predicate.
4241 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
4242 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
4244 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4245 Thomas Preud'homme <thomas.preudhomme@arm.com>
4247 * config/arm/arm-protos.h (clear_operation_p): Declare.
4248 * config/arm/arm.c (clear_operation_p): New function.
4249 (cmse_clear_registers): Generate clear_multiple instruction pattern if
4250 targeting Armv8.1-M Mainline or successor.
4251 (output_return_instruction): Only output APSR register clearing if
4252 Armv8.1-M Mainline instructions not available.
4253 (thumb_exit): Likewise.
4254 * config/arm/predicates.md (clear_multiple_operation): New predicate.
4255 * config/arm/thumb2.md (clear_apsr): New define_insn.
4256 (clear_multiple): Likewise.
4257 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
4259 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4260 Thomas Preud'homme <thomas.preudhomme@arm.com>
4262 * config/arm/arm.c (fp_sysreg_names): Declare and define.
4263 (use_return_insn): Also return false for Armv8.1-M Mainline.
4264 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
4265 Mainline instructions are available.
4266 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
4267 when targeting Armv8.1-M Mainline Security Extensions.
4268 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
4269 Mainline entry function.
4270 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
4271 targeting Armv8.1-M Mainline or successor.
4272 (arm_expand_epilogue): Fix indentation of caller-saved register
4273 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
4275 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
4276 (FP_SYSREGS): Likewise.
4277 (enum vfp_sysregs_encoding): Define enum.
4278 (fp_sysreg_names): Declare.
4279 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
4280 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
4281 (pop_fpsysreg_insn): Likewise.
4283 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4284 Thomas Preud'homme <thomas.preudhomme@arm.com>
4286 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
4287 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
4288 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
4289 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
4290 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
4291 (ARMv8_1m_main): New feature group.
4292 (armv8.1-m.main): New architecture.
4293 * config/arm/arm-tables.opt: Regenerate.
4294 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
4295 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
4296 (arm_options_perform_arch_sanity_checks): Error out when targeting
4297 Armv8.1-M Mainline Security Extensions.
4298 * config/arm/arm.h (arm_arch8_1m_main): Declare.
4300 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4302 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
4303 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
4304 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
4305 aarch64_bfdot_laneq): New.
4306 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
4307 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
4308 vbfdotq_laneq_f32): New.
4309 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
4310 VBFMLA_W, VBF): New.
4311 (isquadop): Add V4BF, V8BF.
4313 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4315 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
4316 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
4317 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
4318 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
4319 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
4320 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
4321 usdot_laneq, sudot_lane,sudot_laneq): New.
4322 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
4323 (aarch64_<sur>dot_lane): New.
4324 * config/aarch64/arm_neon.h (vusdot_s32): New.
4326 (vusdot_lane_s32): New.
4327 (vsudot_lane_s32): New.
4328 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
4329 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
4331 2020-01-16 Martin Liska <mliska@suse.cz>
4333 * value-prof.c (dump_histogram_value): Fix
4334 obvious spacing issue.
4336 2020-01-16 Andrew Pinski <apinski@marvell.com>
4338 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
4339 !storage_order_barrier_p.
4341 2020-01-16 Andrew Pinski <apinski@marvell.com>
4343 * sched-int.h (_dep): Add unused bit-field field for the padding.
4344 * sched-deps.c (init_dep_1): Init unused field.
4346 2020-01-16 Andrew Pinski <apinski@marvell.com>
4348 * optabs.h (create_expand_operand): Initialize target field also.
4350 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
4352 PR tree-optimization/92429
4353 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
4354 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
4356 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
4359 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
4361 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
4362 aarch64_sve_int_mode to each mode.
4364 2020-01-15 David Malcolm <dmalcolm@redhat.com>
4366 * doc/analyzer.texi (Overview): Add note about
4367 -fdump-ipa-analyzer.
4369 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
4371 PR tree-optimization/93231
4372 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
4373 input_type is unsigned. Use tree_to_shwi for shift constant.
4374 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
4375 (simplify_count_trailing_zeroes): Add test to handle known non-zero
4376 inputs more efficiently.
4378 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
4380 * config/i386/i386.md (*movsf_internal): Do not require
4381 SSE2 ISA for alternatives 14 and 15.
4383 2020-01-15 Richard Biener <rguenther@suse.de>
4386 * tree-eh.c (sink_clobbers): If we already visited the destination
4387 block do not defer insertion.
4388 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
4389 the purpose of defered insertion.
4391 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4393 * BASE-VER: Bump to 10.0.1.
4395 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
4397 PR tree-optimization/93247
4398 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
4399 type of the stmt that we're going to vectorize.
4401 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
4403 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
4404 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
4407 2020-01-15 Martin Liska <mliska@suse.cz>
4409 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
4410 2 calls of streamer_read_hwi in a function call.
4412 2020-01-15 Richard Biener <rguenther@suse.de>
4414 * alias.c (record_alias_subset): Avoid redundant work when
4415 subset is already recorded.
4417 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4419 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
4420 the analyzer options provide CWE identifiers.
4422 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4424 * tree-diagnostic-path.cc (path_summary::event_range::print):
4425 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
4426 using get_pure_location.
4428 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4430 PR tree-optimization/93262
4431 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
4432 perform head trimming only if the last argument is constant,
4433 either all ones, or larger or equal to head trim, in the latter
4434 case decrease the last argument by head_trim.
4436 PR tree-optimization/93249
4437 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
4438 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
4439 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
4440 perform head trim unless we can prove there are no '\0' chars
4441 from the source among the first head_trim chars.
4443 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4445 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
4447 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4450 * config/i386/sse.md
4451 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
4452 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
4453 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
4454 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
4455 just a single alternative instead of two, make operands 1 and 2
4458 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
4461 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
4464 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4466 * Makefile.in (lang_opt_files): Add analyzer.opt.
4467 (ANALYZER_OBJS): New.
4468 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
4469 tristate.o and ANALYZER_OBJS.
4470 (TEXI_GCCINT_FILES): Add analyzer.texi.
4471 * common.opt (-fanalyzer): New driver option.
4472 * config.in: Regenerate.
4473 * configure: Regenerate.
4474 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
4475 (gccdepdir): Also create depdir for "analyzer" subdir.
4476 * digraph.cc: New file.
4477 * digraph.h: New file.
4478 * doc/analyzer.texi: New file.
4479 * doc/gccint.texi ("Static Analyzer") New menu item.
4480 (analyzer.texi): Include it.
4481 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
4482 ("Warning Options"): Add static analysis warnings to the list.
4483 (-Wno-analyzer-double-fclose): New option.
4484 (-Wno-analyzer-double-free): New option.
4485 (-Wno-analyzer-exposure-through-output-file): New option.
4486 (-Wno-analyzer-file-leak): New option.
4487 (-Wno-analyzer-free-of-non-heap): New option.
4488 (-Wno-analyzer-malloc-leak): New option.
4489 (-Wno-analyzer-possible-null-argument): New option.
4490 (-Wno-analyzer-possible-null-dereference): New option.
4491 (-Wno-analyzer-null-argument): New option.
4492 (-Wno-analyzer-null-dereference): New option.
4493 (-Wno-analyzer-stale-setjmp-buffer): New option.
4494 (-Wno-analyzer-tainted-array-index): New option.
4495 (-Wno-analyzer-use-after-free): New option.
4496 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
4497 (-Wno-analyzer-use-of-uninitialized-value): New option.
4498 (-Wanalyzer-too-complex): New option.
4499 (-fanalyzer-call-summaries): New warning.
4500 (-fanalyzer-checker=): New warning.
4501 (-fanalyzer-fine-grained): New warning.
4502 (-fno-analyzer-state-merge): New warning.
4503 (-fno-analyzer-state-purge): New warning.
4504 (-fanalyzer-transitivity): New warning.
4505 (-fanalyzer-verbose-edges): New warning.
4506 (-fanalyzer-verbose-state-changes): New warning.
4507 (-fanalyzer-verbosity=): New warning.
4508 (-fdump-analyzer): New warning.
4509 (-fdump-analyzer-callgraph): New warning.
4510 (-fdump-analyzer-exploded-graph): New warning.
4511 (-fdump-analyzer-exploded-nodes): New warning.
4512 (-fdump-analyzer-exploded-nodes-2): New warning.
4513 (-fdump-analyzer-exploded-nodes-3): New warning.
4514 (-fdump-analyzer-supergraph): New warning.
4515 * doc/sourcebuild.texi (dg-require-dot): New.
4516 (dg-check-dot): New.
4517 * gdbinit.in (break-on-saved-diagnostic): New command.
4518 * graphviz.cc: New file.
4519 * graphviz.h: New file.
4520 * ordered-hash-map-tests.cc: New file.
4521 * ordered-hash-map.h: New file.
4522 * passes.def (pass_analyzer): Add before
4523 pass_ipa_whole_program_visibility.
4524 * selftest-run-tests.c (selftest::run_tests): Call
4525 selftest::ordered_hash_map_tests_cc_tests.
4526 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
4528 * shortest-paths.h: New file.
4529 * timevar.def (TV_ANALYZER): New timevar.
4530 (TV_ANALYZER_SUPERGRAPH): Likewise.
4531 (TV_ANALYZER_STATE_PURGE): Likewise.
4532 (TV_ANALYZER_PLAN): Likewise.
4533 (TV_ANALYZER_SCC): Likewise.
4534 (TV_ANALYZER_WORKLIST): Likewise.
4535 (TV_ANALYZER_DUMP): Likewise.
4536 (TV_ANALYZER_DIAGNOSTICS): Likewise.
4537 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
4538 * tree-pass.h (make_pass_analyzer): New decl.
4539 * tristate.cc: New file.
4540 * tristate.h: New file.
4542 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
4545 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
4546 alternatives 9 and 10.
4548 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4550 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
4551 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
4552 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
4553 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
4554 (selftest::hash_map_tests_c_tests): Call it.
4555 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
4556 New static constant, using the value of = H::empty_zero_p.
4557 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
4558 from default_hash_traits <Value>.
4559 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
4561 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
4562 * hash-table.h (hash_table::alloc_entries): Guard the loop of
4563 calls to mark_empty with !Descriptor::empty_zero_p.
4564 (hash_table::empty_slow): Conditionalize the memset call with a
4565 check that Descriptor::empty_zero_p; otherwise, loop through the
4566 entries calling mark_empty on them.
4567 * hash-traits.h (int_hash::empty_zero_p): New static constant.
4568 (pointer_hash::empty_zero_p): Likewise.
4569 (pair_hash::empty_zero_p): Likewise.
4570 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
4572 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
4573 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
4574 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
4575 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
4576 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
4577 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
4578 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
4579 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
4581 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
4584 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
4586 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
4587 fix typo on return value.
4589 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
4592 * cgraph.c (symbol_table::create_edge): Init speculative_id and
4594 (cgraph_edge::make_speculative): Add param for setting speculative_id
4596 (cgraph_edge::speculative_call_info): Update comments and find reference
4597 by speculative_id for multiple indirect targets.
4598 (cgraph_edge::resolve_speculation): Decrease the speculations
4599 for indirect edge, drop it's speculative if not direct target
4600 left. Update comments.
4601 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4602 (cgraph_node::dump): Print num_speculative_call_targets.
4603 (cgraph_node::verify_node): Don't report error if speculative
4604 edge not include statement.
4605 (cgraph_edge::num_speculative_call_targets_p): New function.
4606 * cgraph.h (int common_target_id): Remove.
4607 (int common_target_probability): Remove.
4608 (num_speculative_call_targets): New variable.
4609 (make_speculative): Add param for setting speculative_id.
4610 (cgraph_edge::num_speculative_call_targets_p): New declare.
4611 (target_prob): New variable.
4612 (speculative_id): New variable.
4613 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
4614 call summaries for multiple speculative call targets.
4615 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
4616 * ipa-profile.c (struct speculative_call_target): New struct.
4617 (class speculative_call_summary): New class.
4618 (class speculative_call_summaries): New class.
4619 (call_sums): New variable.
4620 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
4621 (ipa_profile_write_edge_summary): New function.
4622 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
4623 (ipa_profile_dump_all_summaries): New function.
4624 (ipa_profile_read_edge_summary): New function.
4625 (ipa_profile_read_summary_section): New function.
4626 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
4627 (ipa_profile): Generate num_speculative_call_targets from
4629 * ipa-ref.h (speculative_id): New variable.
4630 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
4631 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
4632 common_target_probability. Stream out speculative_id and
4633 num_speculative_call_targets.
4634 (input_edge): Likewise.
4635 * predict.c (dump_prediction): Remove edges count assert to be
4637 * symtab.c (symtab_node::create_reference): Init speculative_id.
4638 (symtab_node::clone_references): Clone speculative_id.
4639 (symtab_node::clone_referring): Clone speculative_id.
4640 (symtab_node::clone_reference): Clone speculative_id.
4641 (symtab_node::clear_stmts_in_references): Clear speculative_id.
4642 * tree-inline.c (copy_bb): Duplicate all the speculative edges
4643 if indirect call contains multiple speculative targets.
4644 * value-prof.h (check_ic_target): Remove.
4645 * value-prof.c (gimple_value_profile_transformations):
4646 Use void function gimple_ic_transform.
4647 * value-prof.c (gimple_ic_transform): Handle topn case.
4648 Fix comment typos. Change it to a void function.
4650 2020-01-13 Andrew Pinski <apinski@marvell.com>
4652 * config/aarch64/aarch64-cores.def (octeontx2): New define.
4653 (octeontx2t98): New define.
4654 (octeontx2t96): New define.
4655 (octeontx2t93): New define.
4656 (octeontx2f95): New define.
4657 (octeontx2f95n): New define.
4658 (octeontx2f95mm): New define.
4659 * config/aarch64/aarch64-tune.md: Regenerate.
4660 * doc/invoke.texi (-mcpu=): Document the new cpu types.
4662 2020-01-13 Jason Merrill <jason@redhat.com>
4664 PR c++/33799 - destroy return value if local cleanup throws.
4665 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
4667 2020-01-13 Martin Liska <mliska@suse.cz>
4669 * ipa-cp.c (get_max_overall_size): Use newly
4670 renamed param param_ipa_cp_unit_growth.
4671 * params.opt: Remove legacy param name.
4673 2020-01-13 Martin Sebor <msebor@redhat.com>
4675 PR tree-optimization/93213
4676 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
4677 stores to be eliminated.
4679 2020-01-13 Martin Liska <mliska@suse.cz>
4681 * opts.c (print_help): Do not print CL_PARAM
4682 and CL_WARNING for CL_OPTIMIZATION.
4684 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
4687 * doc/invoke.texi (Warning Options): Add caveat about some warnings
4688 depending on optimization settings.
4690 2020-01-13 Jakub Jelinek <jakub@redhat.com>
4692 PR tree-optimization/90838
4693 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4694 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
4695 argument rather than to initialize temporary for targets that
4696 don't use the mode argument at all. Initialize ctzval to avoid
4699 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
4701 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
4702 * tree-core.h: Document it.
4703 * gimplify.c (gimplify_omp_workshare): Set it.
4704 * omp-low.c (lower_omp_target): Use it.
4705 * tree-pretty-print.c (dump_omp_clause): Print it.
4707 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
4708 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
4710 2020-01-10 David Malcolm <dmalcolm@redhat.com>
4712 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
4713 * common.opt (fdiagnostics-path-format=): New option.
4714 (diagnostic_path_format): New enum.
4715 (fdiagnostics-show-path-depths): New option.
4716 * coretypes.h (diagnostic_event_id_t): New forward decl.
4717 * diagnostic-color.c (color_dict): Add "path".
4718 * diagnostic-event-id.h: New file.
4719 * diagnostic-format-json.cc (json_from_expanded_location): Make
4721 (json_end_diagnostic): Call context->make_json_for_path if it
4722 exists and the diagnostic has a path.
4723 (diagnostic_output_format_init): Clear context->print_path.
4724 * diagnostic-path.h: New file.
4725 * diagnostic-show-locus.c (colorizer::set_range): Special-case
4726 when printing a run of events in a diagnostic_path so that they
4727 all get the same color.
4728 (layout::m_diagnostic_path_p): New field.
4729 (layout::layout): Initialize it.
4730 (layout::print_any_labels): Don't colorize the label text for an
4731 event in a diagnostic_path.
4732 (gcc_rich_location::add_location_if_nearby): Add
4733 "restrict_to_current_line_spans" and "label" params. Pass the
4734 former to layout.maybe_add_location_range; pass the latter
4735 when calling add_range.
4736 * diagnostic.c: Include "diagnostic-path.h".
4737 (diagnostic_initialize): Initialize context->path_format and
4738 context->show_path_depths.
4739 (diagnostic_show_any_path): New function.
4740 (diagnostic_path::interprocedural_p): New function.
4741 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4742 (simple_diagnostic_path::num_events): New function.
4743 (simple_diagnostic_path::get_event): New function.
4744 (simple_diagnostic_path::add_event): New function.
4745 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4746 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4747 (debug): New overload taking a diagnostic_path *.
4748 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4749 * diagnostic.h (enum diagnostic_path_format): New enum.
4750 (json::value): New forward decl.
4751 (diagnostic_context::path_format): New field.
4752 (diagnostic_context::show_path_depths): New field.
4753 (diagnostic_context::print_path): New callback field.
4754 (diagnostic_context::make_json_for_path): New callback field.
4755 (diagnostic_show_any_path): New decl.
4756 (json_from_expanded_location): New decl.
4757 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4758 (-fdiagnostics-show-path-depths): New option.
4759 (-fdiagnostics-color): Add "path" to description of default
4760 GCC_COLORS; describe it.
4761 (-fdiagnostics-format=json): Document how diagnostic paths are
4762 represented in the JSON output format.
4763 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4764 Add optional params "restrict_to_current_line_spans" and "label".
4765 * opts.c (common_handle_option): Handle
4766 OPT_fdiagnostics_path_format_ and
4767 OPT_fdiagnostics_show_path_depths.
4768 * pretty-print.c: Include "diagnostic-event-id.h".
4769 (pp_format): Implement "%@" format code for printing
4770 diagnostic_event_id_t *.
4771 (selftest::test_pp_format): Add tests for "%@".
4772 * selftest-run-tests.c (selftest::run_tests): Call
4773 selftest::tree_diagnostic_path_cc_tests.
4774 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4775 * toplev.c (general_init): Initialize global_dc->path_format and
4776 global_dc->show_path_depths.
4777 * tree-diagnostic-path.cc: New file.
4778 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4779 non-static. Drop "diagnostic" param in favor of storing the
4780 original value of "where" and re-using it.
4781 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4782 maybe_unwind_expanded_macro_loc.
4783 (tree_diagnostics_defaults): Initialize context->print_path and
4784 context->make_json_for_path.
4785 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4787 (default_tree_make_json_for_path): New decl.
4788 (maybe_unwind_expanded_macro_loc): New decl.
4790 2020-01-10 Jakub Jelinek <jakub@redhat.com>
4792 PR tree-optimization/93210
4793 * fold-const.h (native_encode_initializer,
4794 can_native_interpret_type_p): Declare.
4795 * fold-const.c (native_encode_string): Fix up handling with off != -1,
4797 (native_encode_initializer): New function, moved from dwarf2out.c.
4798 Adjust to native_encode_expr compatible arguments, including dry-run
4799 and partial extraction modes. Don't handle STRING_CST.
4800 (can_native_interpret_type_p): No longer static.
4801 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4802 offset / BITS_PER_UNIT fits into int and don't call it if
4803 can_native_interpret_type_p fails. If suboff is NULL and for
4804 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4805 native_encode_initializer.
4806 (fold_const_aggregate_ref_1): Formatting fix.
4807 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4808 (tree_add_const_value_attribute): Adjust caller.
4810 PR tree-optimization/90838
4811 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4812 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4813 CTZ_DEFINED_VALUE_AT_ZERO.
4815 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
4818 * lra-constraints.c (match_reload): Permit input operands have the
4819 same mode as output while other input operands have a different
4822 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
4824 PR tree-optimization/90838
4825 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4826 (check_ctz_string): Likewise.
4827 (optimize_count_trailing_zeroes): Likewise.
4828 (simplify_count_trailing_zeroes): Likewise.
4829 (pass_forwprop::execute): Try ctz simplification.
4830 * match.pd: Add matching for ctz idioms.
4832 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4834 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4836 (aarch64_invalid_unary_op): New function for target hook.
4837 (aarch64_invalid_binary_op): New function for target hook.
4839 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4841 * config.gcc: Add arm_bf16.h.
4842 * config/aarch64/aarch64-builtins.c
4843 (aarch64_simd_builtin_std_type): Add BFmode.
4844 (aarch64_init_simd_builtin_types): Define element types for vector
4846 (aarch64_init_bf16_types): New function.
4847 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4848 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4850 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4851 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4853 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4854 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4855 * config/aarch64/aarch64.c
4856 (aarch64_classify_vector_mode): Add support for BF types.
4857 (aarch64_gimplify_va_arg_expr): Add support for BF types.
4858 (aarch64_vq_mode): Add support for BF types.
4859 (aarch64_simd_container_mode): Add support for BF types.
4860 (aarch64_mangle_type): Add support for BF scalar type.
4861 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4862 * config/aarch64/arm_bf16.h: New file.
4863 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4864 * config/aarch64/iterators.md: Add BF types to mode attributes.
4865 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4867 2020-01-10 Jason Merrill <jason@redhat.com>
4869 PR c++/93173 - incorrect tree sharing.
4870 * gimplify.c (copy_if_shared): No longer static.
4871 * gimplify.h: Declare it.
4873 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4875 * doc/invoke.texi (-msve-vector-bits=): Document that
4876 -msve-vector-bits=128 now generates VL-specific code for
4877 little-endian targets.
4878 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4879 build_vector_type_for_mode to construct the data vector types.
4880 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4881 VL-specific code for -msve-vector-bits=128 on little-endian targets.
4882 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4883 for 128-bit vectors.
4885 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4887 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4890 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4892 * config/aarch64/aarch64-builtins.c
4893 (aarch64_builtin_vectorized_function): Check for specific vector modes,
4894 rather than checking the number of elements and the element mode.
4896 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4898 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4899 get_related_vectype_for_scalar_type rather than build_vector_type
4900 to create the index type for a conditional reduction.
4902 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4904 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4905 for any type of gather or scatter, including strided accesses.
4907 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4909 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4912 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4914 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4916 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
4917 parameter and its use to reset DR_OFFSET's.
4918 (vect_transform_loop): Remove orig_drs_init argument.
4919 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4920 member of dr_vec_info rather than the offset of the associated
4921 data_reference's innermost_loop_behavior.
4922 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4923 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4924 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4925 get_dr_vinfo_offset.
4926 (vectorizable_store): Likewise.
4927 (vectorizable_load): Likewise.
4929 2020-01-10 Richard Biener <rguenther@suse.de>
4931 * gimple-ssa-store-merging
4932 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4934 2020-01-10 Martin Liska <mliska@suse.cz>
4937 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4938 encapsulation that was there before r280040.
4940 2020-01-10 Richard Biener <rguenther@suse.de>
4943 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4944 sequences to avoid walking them again for secondary opportunities.
4945 (pass_lower_eh_dispatch::execute): Instead actually insert
4948 2020-01-10 Richard Biener <rguenther@suse.de>
4951 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4952 (cleanup_all_empty_eh): Walk landing pads in reverse order to
4953 avoid quadraticness.
4955 2020-01-10 Martin Jambor <mjambor@suse.cz>
4957 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4958 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4959 to get param_ipa_sra_max_replacements.
4960 (param_splitting_across_edge): Pass the caller to
4961 pull_accesses_from_callee.
4963 2020-01-10 Martin Jambor <mjambor@suse.cz>
4965 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4966 * ipa-cp.c (max_new_size): Removed.
4967 (orig_overall_size): New variable.
4968 (get_max_overall_size): New function.
4969 (estimate_local_effects): Use it. Adjust dump.
4970 (decide_about_value): Likewise.
4971 (ipcp_propagate_stage): Do not calculate max_new_size, just store
4972 orig_overall_size. Adjust dump.
4973 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4975 2020-01-10 Martin Jambor <mjambor@suse.cz>
4977 * params.opt (param_ipa_max_agg_items): Mark as Optimization
4978 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4979 instead of param_ipa_max_agg_items.
4980 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4981 optimization info for the callee.
4983 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
4985 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4986 markers if debug_inline_points is false.
4988 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4990 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4992 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4993 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4994 aarch64-sve-builtins-sve2.h.
4995 (aarch64-sve-builtins-sve2.o): New rule.
4996 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4997 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4998 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4999 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
5000 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5001 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
5003 * config/aarch64/aarch64-sve.md: Update comments with SVE2
5004 instructions that are handled here.
5005 (@cond_asrd<mode>): Generalize to...
5006 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
5007 (*cond_asrd<mode>_2): Generalize to...
5008 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
5009 (*cond_asrd<mode>_z): Generalize to...
5010 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
5011 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
5012 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
5013 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
5014 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
5016 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
5017 (@aarch64_scatter_stnt<mode>): Likewise.
5018 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
5019 (@aarch64_mul_lane_<mode>): Likewise.
5020 (@aarch64_sve_suqadd<mode>_const): Likewise.
5021 (*<sur>h<addsub><mode>): Generalize to...
5022 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
5024 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
5025 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
5026 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
5027 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
5028 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
5029 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
5030 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
5031 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
5032 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
5033 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
5034 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
5035 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
5036 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
5037 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
5038 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
5039 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
5040 (@aarch64_sve2_xar<mode>): Likewise.
5041 (@aarch64_sve2_bcax<mode>): Likewise.
5042 (*aarch64_sve2_eor3<mode>): Rename to...
5043 (@aarch64_sve2_eor3<mode>): ...this.
5044 (@aarch64_sve2_bsl<mode>): New expander.
5045 (@aarch64_sve2_nbsl<mode>): Likewise.
5046 (@aarch64_sve2_bsl1n<mode>): Likewise.
5047 (@aarch64_sve2_bsl2n<mode>): Likewise.
5048 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
5049 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
5050 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
5051 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
5052 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
5053 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
5054 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
5055 (<su>mull<bt><Vwide>): Generalize to...
5056 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
5058 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
5059 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
5060 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
5061 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
5062 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
5063 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
5064 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
5065 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
5066 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
5067 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
5068 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
5069 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
5070 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
5071 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
5072 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
5073 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
5074 (<SHRNB:r>shrnb<mode>): Generalize to...
5075 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
5077 (<SHRNT:r>shrnt<mode>): Generalize to...
5078 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
5080 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
5081 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
5082 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
5083 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
5084 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
5085 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
5086 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
5087 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
5088 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
5089 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
5090 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
5091 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
5092 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
5093 (@aarch64_sve2_cvtnt<mode>): Likewise.
5094 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
5095 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
5096 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
5097 (@aarch64_sve2_cvtxnt<mode>): Likewise.
5098 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
5099 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
5100 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
5101 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
5102 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
5103 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
5104 (@aarch64_sve2_pmul<mode>): Likewise.
5105 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
5106 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
5107 (@aarch64_sve2_tbl2<mode>): Likewise.
5108 (@aarch64_sve2_tbx<mode>): Likewise.
5109 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
5110 (@aarch64_sve2_histcnt<mode>): Likewise.
5111 (@aarch64_sve2_histseg<mode>): Likewise.
5112 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
5113 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
5114 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
5115 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
5116 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
5117 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
5118 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
5119 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
5120 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
5121 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
5122 (SVE2_PMULL_PAIR_I): New mode iterators.
5123 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
5124 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
5125 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
5126 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
5127 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
5128 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
5129 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
5130 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
5131 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
5132 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
5133 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
5134 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
5135 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
5136 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
5137 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
5138 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
5139 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
5140 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
5141 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
5142 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
5143 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
5144 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
5145 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
5146 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
5147 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
5148 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
5149 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
5150 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
5151 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
5152 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
5153 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
5155 (VNARROW, Ventype): New mode attributes.
5156 (Vewtype): Handle VNx2DI. Fix typo in comment.
5157 (VDOUBLE): New mode attribute.
5158 (sve_lane_con): Handle VNx8HI.
5159 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
5160 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
5161 (sve_int_op, sve_int_op_rev): Handle the above codes.
5162 (sve_pred_int_rhs2_operand): Likewise.
5163 (MULLBT, SHRNB, SHRNT): Delete.
5164 (SVE_INT_SHIFT_IMM): New int iterator.
5165 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
5166 and UNSPEC_WHILEHS for TARGET_SVE2.
5167 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
5168 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
5169 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
5170 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
5171 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
5172 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
5173 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
5174 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
5175 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
5176 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
5177 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
5178 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
5179 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
5180 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
5181 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
5182 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
5183 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
5184 (optab): Handle the new unspecs.
5185 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
5187 (lr): Handle the new unspecs.
5189 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
5190 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
5191 (sve_int_qsub_op): New int attributes.
5192 (sve_fp_op, rot): Handle the new unspecs.
5193 * config/aarch64/aarch64-sve-builtins.h
5194 (function_resolver::require_matching_pointer_type): Declare.
5195 (function_resolver::resolve_unary): Add an optional boolean argument.
5196 (function_resolver::finish_opt_n_resolution): Add an optional
5197 type_suffix_index argument.
5198 (gimple_folder::redirect_call): Declare.
5199 (gimple_expander::prepare_gather_address_operands): Add an optional
5201 * config/aarch64/aarch64-sve-builtins.cc: Include
5202 aarch64-sve-builtins-sve2.h.
5203 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
5204 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
5205 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
5206 (TYPES_hsd_integer): Use TYPES_hsd_signed.
5207 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
5208 (TYPES_s_unsigned): Likewise.
5209 (TYPES_s_integer): Use TYPES_s_unsigned.
5210 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
5211 (TYPES_sd_integer): Use them.
5212 (TYPES_d_unsigned): New macro.
5213 (TYPES_d_integer): Use it.
5214 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
5215 (TYPES_cvt_narrow): Likewise.
5216 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
5217 (preds_mx): New variable.
5218 (function_builder::add_overloaded_function): Allow the new feature
5219 set to be more restrictive than the original one.
5220 (function_resolver::infer_pointer_type): Remove qualifiers from
5221 the pointer type before printing it.
5222 (function_resolver::require_matching_pointer_type): New function.
5223 (function_resolver::resolve_sv_displacement): Handle functions
5224 that don't support 32-bit vector indices or svint32_t vector offsets.
5225 (function_resolver::finish_opt_n_resolution): Take the inferred type
5226 as a separate argument.
5227 (function_resolver::resolve_unary): Optionally treat all forms in
5228 the same way as normal merging functions.
5229 (gimple_folder::redirect_call): New function.
5230 (function_expander::prepare_gather_address_operands): Add an argument
5231 that says whether scaled forms are available. If they aren't,
5232 handle scaling of vector indices and don't add the extension and
5234 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
5235 fall back to using cond_* instead.
5236 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
5237 Split out the member variables into...
5238 (rtx_code_function_base): ...this new base class.
5239 (rtx_code_function_rotated): Inherit rtx_code_function_base.
5240 (unspec_based_function): Split out the member variables into...
5241 (unspec_based_function_base): ...this new base class.
5242 (unspec_based_function_rotated): Inherit unspec_based_function_base.
5243 (unspec_based_function_exact_insn): New class.
5244 (unspec_based_add_function, unspec_based_add_lane_function)
5245 (unspec_based_lane_function, unspec_based_pred_function)
5246 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
5247 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
5248 (unspec_based_sub_function, unspec_based_sub_lane_function): New
5250 (unspec_based_fused_function): New class.
5251 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
5252 (unspec_based_fused_lane_function): New class.
5253 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
5255 (CODE_FOR_MODE1): New macro.
5256 (fixed_insn_function): New class.
5257 (while_comparison): Likewise.
5258 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
5259 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
5260 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
5261 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
5262 (load_gather_sv_restricted, shift_left_imm_long): Declare.
5263 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
5264 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
5265 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
5266 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
5267 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
5268 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
5269 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
5270 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
5271 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
5272 Also add an initial argument for unary_convert_narrowt, regardless
5273 of the predication type.
5274 (build_32_64): Allow loads and stores to specify MODE_none.
5275 (build_sv_index64, build_sv_uint_offset): New functions.
5276 (long_type_suffix): New function.
5277 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
5278 (binary_imm_long_base, load_gather_sv_base): Likewise.
5279 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
5280 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
5281 (unary_narrowb_base, unary_narrowt_base): Likewise.
5282 (binary_long_lane_def, binary_long_lane): New shape.
5283 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
5284 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
5285 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
5286 (binary_to_uint_def, binary_to_uint): Likewise.
5287 (binary_wide_def, binary_wide): Likewise.
5288 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
5289 (compare_def, compare): Likewise.
5290 (compare_ptr_def, compare_ptr): Likewise.
5291 (load_ext_gather_index_restricted_def,
5292 load_ext_gather_index_restricted): Likewise.
5293 (load_ext_gather_offset_restricted_def,
5294 load_ext_gather_offset_restricted): Likewise.
5295 (load_gather_sv_def): Inherit from load_gather_sv_base.
5296 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
5297 (shift_left_imm_def, shift_left_imm): Likewise.
5298 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
5299 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
5300 (store_scatter_index_restricted_def,
5301 store_scatter_index_restricted): Likewise.
5302 (store_scatter_offset_restricted_def,
5303 store_scatter_offset_restricted): Likewise.
5304 (tbl_tuple_def, tbl_tuple): Likewise.
5305 (ternary_long_lane_def, ternary_long_lane): Likewise.
5306 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
5307 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
5308 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
5309 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
5310 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
5311 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
5312 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
5313 (ternary_uint_def, ternary_uint): Likewise.
5314 (unary_convert): Fix typo in comment.
5315 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
5316 (unary_long_def, unary_long): Likewise.
5317 (unary_narrowb_def, unary_narrowb): Likewise.
5318 (unary_narrowt_def, unary_narrowt): Likewise.
5319 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
5320 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
5321 (unary_to_int_def, unary_to_int): Likewise.
5322 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
5323 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
5324 (svasrd_impl): Delete.
5325 (svcadd_impl::expand): Handle integer operations too.
5326 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
5327 new functions to derive the unspec numbers.
5328 (svmla_svmls_lane_impl): Replace with...
5329 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
5330 integer operations too.
5331 (svwhile_impl): Rename to...
5332 (svwhilelx_impl): ...this and inherit from while_comparison.
5333 (svasrd): Use unspec_based_function.
5334 (svmla_lane): Use svmla_lane_impl.
5335 (svmls_lane): Use svmls_lane_impl.
5336 (svrecpe, svrsqrte): Handle unsigned integer operations too.
5337 (svwhilele, svwhilelt): Use svwhilelx_impl.
5338 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
5339 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
5340 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
5341 * config/aarch64/aarch64-sve-builtins.def: Include
5342 aarch64-sve-builtins-sve2.def.
5344 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5346 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
5347 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
5348 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
5349 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
5350 immediates as well as vector ones.
5351 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
5352 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
5353 (aarch64_sve_qsub_immediate): Update calls accordingly.
5355 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5357 * config/aarch64/aarch64-sve2.md: Add banner comments.
5358 (<su>mulh<r>s<mode>3): Move further up file.
5359 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
5360 (*aarch64_sve2_sra<mode>): Move further down file.
5361 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
5363 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5365 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
5367 (while_optab_cmp): Handle them.
5368 * config/aarch64/aarch64-sve.md
5369 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
5370 and add a "@" marker.
5371 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
5372 instead of gen_aarch64_sve2_while_ptest.
5373 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
5375 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5377 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
5378 (UNSPEC_WHILELE): ...this.
5379 (UNSPEC_WHILE_LO): Rename to...
5380 (UNSPEC_WHILELO): ...this.
5381 (UNSPEC_WHILE_LS): Rename to...
5382 (UNSPEC_WHILELS): ...this.
5383 (UNSPEC_WHILE_LT): Rename to...
5384 (UNSPEC_WHILELT): ...this.
5385 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
5386 (cmp_op, while_optab_cmp): Likewise.
5387 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
5388 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
5389 (svwhilelt): Likewise.
5391 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5393 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
5394 (unary_to_uint): Define.
5395 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
5396 (unary_count): Rename to...
5397 (unary_to_uint_def, unary_to_uint): ...this.
5398 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
5400 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5402 * config/aarch64/aarch64-sve-builtins-functions.h
5403 (code_for_mode_function): New class.
5404 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
5405 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
5406 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
5407 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
5408 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
5410 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5412 * config/aarch64/iterators.md (addsub): New code attribute.
5413 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
5415 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
5416 in the asm string and attributes. Fix indentation.
5417 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
5419 (@aarch64_sve_<optab><mode>): ...this.
5420 * config/aarch64/aarch64-sve-builtins.h
5421 (function_expander::expand_signed_unpred_op): Delete.
5422 * config/aarch64/aarch64-sve-builtins.cc
5423 (function_expander::expand_signed_unpred_op): Likewise.
5424 (function_expander::map_to_rtx_codes): If the optab isn't defined,
5425 try using code_for_aarch64_sve instead.
5426 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
5427 (svqsub_impl): Likewise.
5428 (svqadd, svqsub): Use rtx_code_function instead.
5430 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5432 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
5433 (HADDSUB, sur, addsub): Remove them.
5435 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5437 * tree-nrv.c (pass_return_slot::execute): Handle all internal
5438 functions the same way, rather than singling out those that
5439 aren't mapped directly to optabs.
5441 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5443 * target.def (compatible_vector_types_p): New target hook.
5444 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
5445 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
5446 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
5447 * doc/tm.texi: Regenerate.
5448 * gimple-expr.c: Include target.h.
5449 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
5450 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
5452 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
5453 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
5454 Use the original predicate if it already has a suitable type.
5456 2020-01-09 Martin Jambor <mjambor@suse.cz>
5458 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
5459 resolve_speculation and redirect_call_stmt_to_callee static. Change
5460 return type of set_call_stmt to cgraph_edge *.
5461 * auto-profile.c (afdo_indirect_call): Adjust call to
5462 redirect_call_stmt_to_callee.
5463 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
5464 make the this pointer explicit, adjust self-recursive calls and the
5465 call top make_direct. Return the resulting edge.
5466 (cgraph_edge::remove): Make this pointer explicit.
5467 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
5468 (cgraph_edge::make_direct): Likewise, adjust call to
5469 resolve_speculation.
5470 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
5471 call to set_call_stmt.
5472 (cgraph_update_edges_for_call_stmt_node): Update call to
5473 set_call_stmt and remove.
5474 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
5475 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
5476 (cgraph_node::create_edge_including_clones): Moved "first" definition
5477 of edge to the block where it was used. Adjusted calls to
5479 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
5480 cgraph_edge::remove.
5481 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
5482 make_direct and redirect_call_stmt_to_callee.
5483 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
5484 resolve_speculation and make_direct.
5485 * ipa-inline-transform.c (inline_transform): Adjust call to
5486 redirect_call_stmt_to_callee.
5487 (check_speculations_1):: Adjust call to resolve_speculation.
5488 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
5489 resolve-speculation.
5490 (inline_small_functions): Adjust call to resolve_speculation.
5491 (ipa_inline): Likewise.
5492 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
5494 * ipa-visibility.c (function_and_variable_visibility): Make iteration
5495 safe with regards to edge removal, adjust calls to
5496 redirect_call_stmt_to_callee.
5497 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
5498 and redirect_call_stmt_to_callee.
5499 * multiple_target.c (create_dispatcher_calls): Adjust call to
5500 redirect_call_stmt_to_callee
5501 (redirect_to_specific_clone): Likewise.
5502 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
5503 Adjust calls to cgraph_edge::remove.
5504 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
5505 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
5506 (expand_call_inline): Adjust call to cgraph_edge::remove.
5508 2020-01-09 Martin Liska <mliska@suse.cz>
5510 * params.opt: Set Optimization for
5511 param_max_speculative_devirt_maydefs.
5513 2020-01-09 Martin Sebor <msebor@redhat.com>
5517 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
5519 2020-01-09 Martin Liska <mliska@suse.cz>
5521 * auto-profile.c (auto_profile): Use opt_for_fn
5523 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
5524 (propagate_vals_across_arith_jfunc): Likewise.
5525 (hint_time_bonus): Likewise.
5526 (incorporate_penalties): Likewise.
5527 (good_cloning_opportunity_p): Likewise.
5528 (perform_estimation_of_a_value): Likewise.
5529 (estimate_local_effects): Likewise.
5530 (ipcp_propagate_stage): Likewise.
5531 * ipa-fnsummary.c (decompose_param_expr): Likewise.
5532 (set_switch_stmt_execution_predicate): Likewise.
5533 (analyze_function_body): Likewise.
5534 * ipa-inline-analysis.c (offline_size): Likewise.
5535 * ipa-inline.c (early_inliner): Likewise.
5536 * ipa-prop.c (ipa_analyze_node): Likewise.
5537 (ipcp_transform_function): Likewise.
5538 * ipa-sra.c (process_scan_results): Likewise.
5539 (ipa_sra_summarize_function): Likewise.
5540 * params.opt: Rename ipcp-unit-growth to
5541 ipa-cp-unit-growth. Add Optimization for various
5542 IPA-related parameters.
5544 2020-01-09 Richard Biener <rguenther@suse.de>
5547 * gimplify.c (gimplify_expr): Deal with NOP definitions.
5549 2020-01-09 Richard Biener <rguenther@suse.de>
5551 PR tree-optimization/93040
5552 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
5554 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
5556 * common/config/avr/avr-common.c (avr_option_optimization_table)
5557 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
5559 2020-01-09 Martin Liska <mliska@suse.cz>
5561 * cgraphclones.c (symbol_table::materialize_all_clones):
5562 Use cgraph_node::dump_name.
5564 2020-01-09 Jakub Jelinek <jakub@redhat.com>
5567 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
5568 output_operand_lossage instead of gcc_unreachable.
5569 * doc/md.texi (riscv f constraint): Fix typo.
5572 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
5573 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
5574 CONST_SCALAR_INT_P instead of CONST_INT_P.
5575 (*subv<mode>4_1): Rename to ...
5576 (subv<mode>4_1): ... this.
5577 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5578 define_insn_and_split patterns.
5579 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5582 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5584 * vec.c (class selftest::count_dtor): New class.
5585 (selftest::test_auto_delete_vec): New test.
5586 (selftest::vec_c_tests): Call it.
5587 * vec.h (class auto_delete_vec): New class template.
5588 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
5590 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5592 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
5594 2020-01-08 Jim Wilson <jimw@sifive.com>
5596 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
5597 use of TLS_MODEL_LOCAL_EXEC when not pic.
5599 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5601 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
5604 2020-01-08 Jakub Jelinek <jakub@redhat.com>
5607 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
5608 *stack_protect_set_3 peephole2): Also check that the second
5609 insns source is general_operand.
5612 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
5613 predicate for output operand instead of register_operand.
5614 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
5615 memory destination and non-memory operands[2].
5617 2020-01-08 Martin Liska <mliska@suse.cz>
5619 * cgraph.c (cgraph_node::dump): Use ::dump_name or
5620 ::dump_asm_name instead of (::name or ::asm_name).
5621 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
5622 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
5623 (analyze_functions): Likewise.
5624 (expand_all_functions): Likewise.
5625 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5626 (propagate_bits_across_jump_function): Likewise.
5627 (dump_profile_updates): Likewise.
5628 (ipcp_store_bits_results): Likewise.
5629 (ipcp_store_vr_results): Likewise.
5630 * ipa-devirt.c (dump_targets): Likewise.
5631 * ipa-fnsummary.c (analyze_function_body): Likewise.
5632 * ipa-hsa.c (check_warn_node_versionable): Likewise.
5633 (process_hsa_functions): Likewise.
5634 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
5635 (set_alias_uids): Likewise.
5636 * ipa-inline-transform.c (save_inline_function_body): Likewise.
5637 * ipa-inline.c (recursive_inlining): Likewise.
5638 (inline_to_all_callers_1): Likewise.
5639 (ipa_inline): Likewise.
5640 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5641 (ipa_propagate_frequency): Likewise.
5642 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5643 (remove_described_reference): Likewise.
5644 * ipa-pure-const.c (worse_state): Likewise.
5645 (check_retval_uses): Likewise.
5646 (analyze_function): Likewise.
5647 (propagate_pure_const): Likewise.
5648 (propagate_nothrow): Likewise.
5649 (dump_malloc_lattice): Likewise.
5650 (propagate_malloc): Likewise.
5651 (pass_local_pure_const::execute): Likewise.
5652 * ipa-visibility.c (optimize_weakref): Likewise.
5653 (function_and_variable_visibility): Likewise.
5654 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
5655 (ipa_discover_variable_flags): Likewise.
5656 * lto-streamer-out.c (output_function): Likewise.
5657 (output_constructor): Likewise.
5658 * tree-inline.c (copy_bb): Likewise.
5659 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
5660 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
5662 2020-01-08 Richard Biener <rguenther@suse.de>
5665 * tree-eh.c (sink_clobbers): Update virtual operands for
5666 the first and last stmt only. Add a dry-run capability.
5667 (pass_lower_eh_dispatch::execute): Perform clobber sinking
5668 after CFG manipulations and in RPO order to catch all
5669 secondary opportunities reliably.
5671 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5674 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5676 2019-01-08 Richard Biener <rguenther@suse.de>
5679 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
5680 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
5681 virtual operand, also updating SSA use.
5682 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
5683 Update stmt after resetting virtual operand.
5684 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
5685 * gimple-iterator.c (gsi_remove): When not removing the stmt
5686 permanently do not delink immediate uses or mark the stmt modified.
5688 2020-01-08 Martin Liska <mliska@suse.cz>
5690 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
5691 (ipa_call_context::estimate_size_and_time): Likewise.
5692 (inline_analyze_function): Likewise.
5694 2020-01-08 Martin Liska <mliska@suse.cz>
5696 * cgraph.c (cgraph_node::dump): Use systematically
5699 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5701 Add -nodevicespecs option for avr.
5704 * config/avr/avr.opt (-nodevicespecs): New driver option.
5705 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
5706 "-specs=device-specs/..." if that option is not set.
5707 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5709 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5711 Implement 64-bit double functions for avr.
5714 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
5715 --with-double-comparison.
5716 * doc/install.texi: Document them.
5717 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
5718 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
5719 <WITH_DOUBLE_COMPARISON>: New built-in defines.
5720 * doc/invoke.texi (AVR Built-in Macros): Document them.
5721 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5722 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5723 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5725 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
5728 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5729 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5730 when only building rm-profile multilibs.
5732 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
5735 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5736 lattice for a value to check.
5737 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5738 finite propagation in self-recursive scc.
5740 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
5742 * ipa-inline.c (caller_growth_limits): Restore the AND.
5744 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5746 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5747 (VEC_ALLREG_ALT): New iterator.
5748 (VEC_ALLREG_INT_MODE): New iterator.
5749 (VCMP_MODE): New iterator.
5750 (VCMP_MODE_INT): New iterator.
5751 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5752 (vec_cmp<u>v64qidi): New define_expand.
5753 (vec_cmp<mode>di_exec): Use VCMP_MODE.
5754 (vec_cmpu<mode>di_exec): New define_expand.
5755 (vec_cmp<u>v64qidi_exec): New define_expand.
5756 (vec_cmp<mode>di_dup): Use VCMP_MODE.
5757 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5758 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5759 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5760 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5761 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5762 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5763 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5764 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5765 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5767 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5768 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5770 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5772 * config/gcn/constraints.md (DA): Update description and match.
5774 (Db): New constraint.
5775 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5777 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5778 Implement 'Db' mixed immediate type.
5779 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5780 (addcv64si3_dup<exec_vcc>): Delete.
5781 (subcv64si3<exec_vcc>): Rework constraints.
5782 (addv64di3): Rework constraints.
5783 (addv64di3_exec): Rework constraints.
5784 (subv64di3): Rework constraints.
5785 (addv64di3_dup): Delete.
5786 (addv64di3_dup_exec): Delete.
5787 (addv64di3_zext): Rework constraints.
5788 (addv64di3_zext_exec): Rework constraints.
5789 (addv64di3_zext_dup): Rework constraints.
5790 (addv64di3_zext_dup_exec): Rework constraints.
5791 (addv64di3_zext_dup2): Rework constraints.
5792 (addv64di3_zext_dup2_exec): Rework constraints.
5793 (addv64di3_sext_dup2): Rework constraints.
5794 (addv64di3_sext_dup2_exec): Rework constraints.
5796 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
5798 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5799 existing target checks.
5801 2020-01-07 Richard Biener <rguenther@suse.de>
5803 * doc/install.texi: Bump minimal supported MPC version.
5805 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5807 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5808 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5809 * langhooks.c: Include stor-layout.h.
5810 (lhd_simulate_enum_decl): New function.
5811 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5812 handle_arm_sve_h for the LTO frontend.
5813 (register_vector_type): Cope with null returns from pushdecl.
5815 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5817 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5818 (aarch64_sve::nvectors_if_data_type): Replace with...
5819 (aarch64_sve::builtin_type_p): ...this.
5820 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5821 (find_vector_type): Delete.
5822 (add_sve_type_attribute): New function.
5823 (lookup_sve_type_attribute): Likewise.
5824 (register_builtin_types): Add an "SVE type" attribute to each type.
5825 (register_tuple_type): Likewise.
5826 (svbool_type_p, nvectors_if_data_type): Delete.
5827 (mangle_builtin_type): Use lookup_sve_type_attribute.
5828 (builtin_type_p): Likewise. Add an overload that returns the
5829 number of constituent vector and predicate registers.
5830 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5831 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5832 instead of aarch64_sve_argument_p.
5833 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5834 (aarch64_pass_by_reference): Likewise.
5835 (aarch64_function_value_1): Likewise.
5836 (aarch64_return_in_memory): Likewise.
5837 (aarch64_layout_arg): Likewise.
5839 2020-01-07 Jakub Jelinek <jakub@redhat.com>
5841 PR tree-optimization/93156
5842 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5843 least significant bit is always clear.
5845 PR tree-optimization/93118
5846 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
5847 simplifier with two intermediate conversions.
5849 2020-01-07 Martin Liska <mliska@suse.cz>
5851 * params.opt: Add Optimization for various parameters.
5853 2020-01-07 Martin Liska <mliska@suse.cz>
5856 * doc/extend.texi: Explain cloning for target_clone
5859 2020-01-07 Martin Liska <mliska@suse.cz>
5861 PR tree-optimization/92860
5862 * common.opt: Make in Optimization option
5863 as it is affected by -O0, which is an Optimization
5865 * tree-inline.c (tree_inlinable_function_p):
5866 Use opt_for_fn for warn_inline.
5867 (expand_call_inline): Likewise.
5869 2020-01-07 Martin Liska <mliska@suse.cz>
5871 PR tree-optimization/92860
5872 * common.opt: Make flag_ree as optimization
5875 2020-01-07 Martin Liska <mliska@suse.cz>
5877 PR optimization/92860
5878 * params.opt: Mark param_min_crossjump_insns with Optimization
5881 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
5883 * ipa-inline-analysis.c (estimate_growth): Fix typo.
5884 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5886 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
5888 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5889 helper function to return the valid addressing formats for a given
5890 hard register and mode.
5891 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5893 * config/rs6000/constraints.md (Q constraint): Update
5895 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5898 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5899 Use 'Q' for doing vector extract from memory.
5900 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5902 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5903 doing vector extract from memory.
5904 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5905 extract from memory.
5907 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5908 for the offset being 34-bits when -mcpu=future is used.
5910 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
5912 * config/pa/pa.md: Revert change to use ordered_comparison_operator
5913 instead of cmpib_comparison_operator in cmpib patterns.
5914 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5915 of cmpib_comparison_operator. Revise comment.
5917 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5919 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5920 in an IFN_DIV_POW2 node to be equal.
5922 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5924 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5925 (vect_check_scalar_mask): ...this.
5926 (vectorizable_store, vectorizable_load): Update call accordingly.
5927 (vectorizable_call): Use vect_check_scalar_mask to check the mask
5928 argument in calls to conditional internal functions.
5930 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5932 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5933 '0' matching inputs.
5934 (subv64di3_exec): Likewise.
5936 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
5938 * config/mips/mips.c (vr4130_align_insns): Fix typo.
5939 * doc/md.texi (movstr): Likewise.
5941 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5943 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5946 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5948 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5950 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
5951 to a temporary file and use move-if-change to update the real
5952 file where necessary.
5954 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5956 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5957 rather than Upa for CPY /M.
5959 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5961 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5964 2020-01-06 Martin Liska <mliska@suse.cz>
5966 PR tree-optimization/92860
5967 * params.opt: Mark param_max_combine_insns with Optimization
5970 2020-01-05 Jakub Jelinek <jakub@redhat.com>
5973 * config/i386/i386.md (SWIDWI): New mode iterator.
5974 (DWI, dwi): Add TImode variants.
5975 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
5976 <general_hilo_operand> instead of <general_operand>. Use
5977 CONST_SCALAR_INT_P instead of CONST_INT_P.
5978 (*addv<mode>4_1): Rename to ...
5979 (addv<mode>4_1): ... this.
5980 (QWI): New mode attribute.
5981 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5982 define_insn_and_split patterns.
5983 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5985 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
5986 <general_hilo_operand> instead of <general_operand>.
5987 (*addcarry<mode>_1): New define_insn.
5988 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5990 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
5992 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5993 Use "call" instead of "set".
5995 2020-01-03 Martin Jambor <mjambor@suse.cz>
5998 * ipa-cp.c (print_all_lattices): Skip functions without info.
6000 2020-01-03 Jakub Jelinek <jakub@redhat.com>
6003 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
6004 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
6005 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
6006 for 'e' simd clones.
6009 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
6011 (mprefer-vector-width=): Add Save.
6012 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
6013 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
6014 (ix86_debug_options, ix86_function_specific_print): Adjust
6015 ix86_target_string callers.
6016 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
6017 (ix86_valid_target_attribute_tree): Likewise.
6018 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
6019 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
6020 ix86_target_string caller.
6023 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
6024 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
6025 instead of gen_int_shift_amount + convert_modes.
6027 PR rtl-optimization/93088
6028 * loop-iv.c (find_single_def_src): Punt after looking through
6029 128 reg copies for regs with single definitions. Move definitions
6032 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
6034 * config/arm/arm-c.c (arm_cpu_builtins): Define
6035 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
6036 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
6037 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
6038 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
6039 * config/arm/arm-tables.opt: Regenerated.
6040 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
6041 arm_arch_i8mm and arm_arch_bf16 when enabled.
6042 * config/arm/arm.h (TARGET_I8MM): New macro.
6043 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
6044 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
6045 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
6046 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
6047 (v8_6_a_simd_variants): New.
6048 (v8_*_a_simd_variants): Add i8mm and bf16.
6049 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
6051 2020-01-02 Jakub Jelinek <jakub@redhat.com>
6054 * predict.c (compute_function_frequency): Don't call
6055 warn_function_cold on functions that already have cold attribute.
6057 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
6060 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
6061 COMDAT group function labels in .data.rel.ro.local section.
6062 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
6065 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
6066 comparison_operator in B and S integer comparisons. Likewise, use
6067 ordered_comparison_operator instead of cmpib_comparison_operator in
6069 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
6071 2020-01-01 Jakub Jelinek <jakub@redhat.com>
6073 Update copyright years.
6075 * gcc.c (process_command): Update copyright notice dates.
6076 * gcov-dump.c (print_version): Ditto.
6077 * gcov.c (print_version): Ditto.
6078 * gcov-tool.c (print_version): Ditto.
6079 * gengtype.c (create_file): Ditto.
6080 * doc/cpp.texi: Bump @copying's copyright year.
6081 * doc/cppinternals.texi: Ditto.
6082 * doc/gcc.texi: Ditto.
6083 * doc/gccint.texi: Ditto.
6084 * doc/gcov.texi: Ditto.
6085 * doc/install.texi: Ditto.
6086 * doc/invoke.texi: Ditto.
6088 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
6090 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
6093 2020-01-01 Jakub Jelinek <jakub@redhat.com>
6095 PR tree-optimization/93098
6096 * match.pd (popcount): For shift amounts, use integer_onep
6097 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
6098 tests. Make sure that precision is power of two larger than or equal
6099 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
6100 instead of ULL suffixed constants. Formatting fixes.
6102 Copyright (C) 2020 Free Software Foundation, Inc.
6104 Copying and distribution of this file, with or without modification,
6105 are permitted in any medium without royalty provided the copyright
6106 notice and this notice are preserved.