1 2020-03-16 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 (CF): Define mve_builtin_data.
7 (ARM_BUILTIN_MVE_PATTERN_START): Define.
8 (arm_init_mve_builtins): Define function.
9 (arm_init_builtins): Add TARGET_HAVE_MVE check.
10 (arm_expand_builtin_1): Check the range of fcode.
11 (arm_expand_mve_builtin): Define function to expand MVE builtins.
12 (arm_expand_builtin): Check the range of fcode.
13 * config/arm/arm_mve.h (__ARM_FEATURE_MVE): Define MVE floating point
15 (__ARM_MVE_PRESERVE_USER_NAMESPACE): Define to protect user namespace.
16 (vst4q_s8): Define macro.
17 (vst4q_s16): Likewise.
18 (vst4q_s32): Likewise.
20 (vst4q_u16): Likewise.
21 (vst4q_u32): Likewise.
22 (vst4q_f16): Likewise.
23 (vst4q_f32): Likewise.
24 (__arm_vst4q_s8): Define inline builtin.
25 (__arm_vst4q_s16): Likewise.
26 (__arm_vst4q_s32): Likewise.
27 (__arm_vst4q_u8): Likewise.
28 (__arm_vst4q_u16): Likewise.
29 (__arm_vst4q_u32): Likewise.
30 (__arm_vst4q_f16): Likewise.
31 (__arm_vst4q_f32): Likewise.
32 (__ARM_mve_typeid): Define macro with MVE types.
33 (__ARM_mve_coerce): Define macro with _Generic feature.
34 (vst4q): Define polymorphic variant for different vst4q builtins.
35 * config/arm/arm_mve_builtins.def: New file.
36 * config/arm/iterators.md (VSTRUCT): Modify to allow XI and OI
38 * config/arm/mve.md (MVE_VLD_ST): Define iterator.
39 (unspec): Define unspec.
40 (mve_vst4q<mode>): Define RTL pattern.
41 * config/arm/neon.md (mov<mode>): Modify expand to allow XI and OI
43 (neon_mov<mode>): Modify RTL define_insn to allow XI and OI modes
45 (define_split): Allow OI mode split for MVE after reload.
46 (define_split): Allow XI mode split for MVE after reload.
47 * config/arm/t-arm (arm.o): Add entry for arm_mve_builtins.def.
48 (arm-builtins.o): Likewise.
50 2020-03-17 Christophe Lyon <christophe.lyon@linaro.org>
52 * c-typeck.c (process_init_element): Handle constructor_type with
53 type size represented by POLY_INT_CST.
55 2020-03-17 Jakub Jelinek <jakub@redhat.com>
57 PR tree-optimization/94187
58 * tree-ssa-strlen.c (count_nonzero_bytes): Punt if
59 nchars - offset < nbytes.
62 * builtins.c (expand_builtin_strnlen): Do return NULL_RTX if we would
63 emit a warning if it was enabled and don't depend on TREE_NO_WARNING
66 2020-03-16 Vladimir Makarov <vmakarov@redhat.com>
69 * lra-spills.c (remove_pseudos): Do not reuse insn alternative
70 after changing memory subreg.
72 2020-03-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
73 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
75 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Modify function to add
76 emulator calls for dobule precision arithmetic operations for MVE.
78 2020-03-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
79 Mihail Ionescu <mihail.ionescu@arm.com>
80 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
82 * common/config/arm/arm-common.c (arm_asm_auto_mfpu): When vfp_base
83 feature bit is on and -mfpu=auto is passed as compiler option, do not
84 generate error on not finding any matching fpu. Because in this case
86 * config/arm/arm-cpus.in (vfp_base): Define feature bit, this bit is
87 enabled for MVE and also for all VFP extensions.
88 (VFPv2): Modify fgroup to enable vfp_base feature bit when ever VFPv2
90 (MVE): Define fgroup to enable feature bits mve, vfp_base and armv7em.
91 (MVE_FP): Define fgroup to enable feature bits is fgroup MVE and FPv5
92 along with feature bits mve_float.
93 (mve): Modify add options in armv8.1-m.main arch for MVE.
94 (mve.fp): Modify add options in armv8.1-m.main arch for MVE with
96 * config/arm/arm.c (use_return_insn): Replace the
97 check with TARGET_VFP_BASE.
98 (thumb2_legitimate_index_p): Replace TARGET_HARD_FLOAT with
100 (arm_rtx_costs_internal): Replace "TARGET_HARD_FLOAT || TARGET_HAVE_MVE"
101 with TARGET_VFP_BASE, to allow cost calculations for copies in MVE as
103 (arm_get_vfp_saved_size): Replace TARGET_HARD_FLOAT with
104 TARGET_VFP_BASE, to allow space calculation for VFP registers in MVE
106 (arm_compute_frame_layout): Likewise.
107 (arm_save_coproc_regs): Likewise.
108 (arm_fixed_condition_code_regs): Modify to enable using VFPCC_REGNUM
110 (arm_hard_regno_mode_ok): Replace "TARGET_HARD_FLOAT || TARGET_HAVE_MVE"
111 with equivalent macro TARGET_VFP_BASE.
112 (arm_expand_epilogue_apcs_frame): Likewise.
113 (arm_expand_epilogue): Likewise.
114 (arm_conditional_register_usage): Likewise.
115 (arm_declare_function_name): Add check to skip printing .fpu directive
116 in assembly file when TARGET_VFP_BASE is enabled and fpu_to_print is
118 * config/arm/arm.h (TARGET_VFP_BASE): Define.
119 * config/arm/arm.md (arch): Add "mve" to arch.
120 (eq_attr "arch" "mve"): Enable on TARGET_HAVE_MVE is true.
121 (vfp_pop_multiple_with_writeback): Replace "TARGET_HARD_FLOAT
122 || TARGET_HAVE_MVE" with equivalent macro TARGET_VFP_BASE.
123 * config/arm/constraints.md (Uf): Define to allow modification to FPCCR
125 * config/arm/thumb2.md (thumb2_movsfcc_soft_insn): Modify target guard
126 to not allow for MVE.
127 * config/arm/unspecs.md (UNSPEC_GET_FPSCR): Move to volatile unspecs
129 (VUNSPEC_GET_FPSCR): Define.
130 * config/arm/vfp.md (thumb2_movhi_vfp): Add support for VMSR and VMRS
131 instructions which move to general-purpose Register from Floating-point
132 Special register and vice-versa.
133 (thumb2_movhi_fp16): Likewise.
134 (thumb2_movsi_vfp): Add support for VMSR and VMRS instructions along
135 with MCR and MRC instructions which set and get Floating-point Status
136 and Control Register (FPSCR).
137 (movdi_vfp): Modify pattern to enable Single-precision scalar float move
139 (thumb2_movdf_vfp): Modify pattern to enable Double-precision scalar
140 float move patterns in MVE.
141 (thumb2_movsfcc_vfp): Modify pattern to enable single float conditional
142 code move patterns of VFP also in MVE by adding TARGET_VFP_BASE check.
143 (thumb2_movdfcc_vfp): Modify pattern to enable double float conditional
144 code move patterns of VFP also in MVE by adding TARGET_VFP_BASE check.
145 (push_multi_vfp): Add support to use VFP VPUSH pattern for MVE by adding
146 TARGET_VFP_BASE check.
147 (set_fpscr): Add support to set FPSCR register for MVE. Modify pattern
148 using VFPCC_REGNUM as few MVE intrinsics use carry bit of FPSCR
150 (get_fpscr): Add support to get FPSCR register for MVE. Modify pattern
151 using VFPCC_REGNUM as few MVE intrinsics use carry bit of FPSCR
155 2020-03-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
156 Mihail Ionescu <mihail.ionescu@arm.com>
157 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
159 * config.gcc (arm_mve.h): Include mve intrinsics header file.
160 * config/arm/aout.h (p0): Add new register name for MVE predicated
162 * config/arm-builtins.c (ARM_BUILTIN_SIMD_LANE_CHECK): Define macro
163 common to Neon and MVE.
164 (ARM_BUILTIN_NEON_LANE_CHECK): Renamed to ARM_BUILTIN_SIMD_LANE_CHECK.
165 (arm_init_simd_builtin_types): Disable poly types for MVE.
166 (arm_init_neon_builtins): Move a check to arm_init_builtins function.
167 (arm_init_builtins): Use ARM_BUILTIN_SIMD_LANE_CHECK instead of
168 ARM_BUILTIN_NEON_LANE_CHECK.
169 (mve_dereference_pointer): Add function.
170 (arm_expand_builtin_args): Call to mve_dereference_pointer when MVE is
172 (arm_expand_neon_builtin): Moved to arm_expand_builtin function.
173 (arm_expand_builtin): Moved from arm_expand_neon_builtin function.
174 * config/arm/arm-c.c (__ARM_FEATURE_MVE): Define macro for MVE and MVE
175 with floating point enabled.
176 * config/arm/arm-protos.h (neon_immediate_valid_for_move): Renamed to
177 simd_immediate_valid_for_move.
178 (simd_immediate_valid_for_move): Renamed from
179 neon_immediate_valid_for_move function.
180 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Generate
181 error if vfpv2 feature bit is disabled and mve feature bit is also
182 disabled for HARD_FLOAT_ABI.
183 (use_return_insn): Check to not push VFP regs for MVE.
184 (aapcs_vfp_allocate): Add MVE check to have same Procedure Call Standard
186 (aapcs_vfp_allocate_return_reg): Likewise.
187 (thumb2_legitimate_address_p): Check to return 0 on valid Thumb-2
188 address operand for MVE.
189 (arm_rtx_costs_internal): MVE check to determine cost of rtx.
190 (neon_valid_immediate): Rename to simd_valid_immediate.
191 (simd_valid_immediate): Rename from neon_valid_immediate.
192 (simd_valid_immediate): MVE check on size of vector is 128 bits.
193 (neon_immediate_valid_for_move): Rename to
194 simd_immediate_valid_for_move.
195 (simd_immediate_valid_for_move): Rename from
196 neon_immediate_valid_for_move.
197 (neon_immediate_valid_for_logic): Modify call to neon_valid_immediate
199 (neon_make_constant): Modify call to neon_valid_immediate function.
200 (neon_vector_mem_operand): Return VFP register for POST_INC or PRE_DEC
202 (output_move_neon): Add MVE check to generate vldm/vstm instrcutions.
203 (arm_compute_frame_layout): Calculate space for saved VFP registers for
205 (arm_save_coproc_regs): Save coproc registers for MVE.
206 (arm_print_operand): Add case 'E' to print memory operands for MVE.
207 (arm_print_operand_address): Check to print register number for MVE.
208 (arm_hard_regno_mode_ok): Check for arm hard regno mode ok for MVE.
209 (arm_modes_tieable_p): Check to allow structure mode for MVE.
210 (arm_regno_class): Add VPR_REGNUM check.
211 (arm_expand_epilogue_apcs_frame): MVE check to calculate epilogue code
213 (arm_expand_epilogue): MVE check for enabling pop instructions in
215 (arm_print_asm_arch_directives): Modify function to disable print of
216 .arch_extension "mve" and "fp" for cases where MVE is enabled with
218 (arm_vector_mode_supported_p): Check for modes available in MVE interger
219 and MVE floating point.
220 (arm_array_mode_supported_p): Add TARGET_HAVE_MVE check for array mode
222 (arm_conditional_register_usage): Enable usage of conditional regsiter
224 (fixed_regs[VPR_REGNUM]): Enable VPR_REG for MVE.
225 (arm_declare_function_name): Modify function to disable print of
226 .arch_extension "mve" and "fp" for cases where MVE is enabled with
228 * config/arm/arm.h (TARGET_HAVE_MVE): Disable for soft float abi and
229 when target general registers are required.
230 (TARGET_HAVE_MVE_FLOAT): Likewise.
231 (FIXED_REGISTERS): Add bit for VFP_REG class which is enabled in arm.c
233 (CALL_USED_REGISTERS): Set bit for VFP_REG class in CALL_USED_REGISTERS
234 which indicate this is not available for across function calls.
235 (FIRST_PSEUDO_REGISTER): Modify.
236 (VALID_MVE_MODE): Define valid MVE mode.
237 (VALID_MVE_SI_MODE): Define valid MVE SI mode.
238 (VALID_MVE_SF_MODE): Define valid MVE SF mode.
239 (VALID_MVE_STRUCT_MODE): Define valid MVE struct mode.
240 (VPR_REGNUM): Add Vector Predication Register in arm_regs_in_sequence
242 (IS_VPR_REGNUM): Macro to check for VPR_REG register.
243 (REG_ALLOC_ORDER): Add VPR_REGNUM entry.
244 (enum reg_class): Add VPR_REG entry.
245 (REG_CLASS_NAMES): Add VPR_REG entry.
246 * config/arm/arm.md (VPR_REGNUM): Define.
247 (conds): Check is_mve_type attrbiute to differentiate "conditional" and
248 "unconditional" instructions.
249 (arm_movsf_soft_insn): Modify RTL to not allow for MVE.
250 (movdf_soft_insn): Modify RTL to not allow for MVE.
251 (vfp_pop_multiple_with_writeback): Enable for MVE.
252 (include "mve.md"): Include mve.md file.
253 * config/arm/arm_mve.h: Add MVE intrinsics head file.
254 * config/arm/constraints.md (Up): Constraint to enable "p0" register in MVE
255 for vector predicated operands.
256 * config/arm/iterators.md (VNIM1): Define.
258 (VHFBF_split): Define
259 * config/arm/mve.md: New file.
260 (mve_mov<mode>): Define RTL for move, store and load in MVE.
261 (mve_mov<mode>): Define move RTL pattern with vec_duplicate operator for
263 * config/arm/neon.md (neon_immediate_valid_for_move): Rename with
264 simd_immediate_valid_for_move.
265 (neon_mov<mode>): Split pattern and move expand pattern "movv8hf" which
266 is common to MVE and NEON to vec-common.md file.
267 (vec_init<mode><V_elem_l>): Add TARGET_HAVE_MVE check.
268 * config/arm/predicates.md (vpr_register_operand): Define.
269 * config/arm/t-arm: Add mve.md file.
270 * config/arm/types.md (mve_move): Add MVE instructions mve_move to
272 (mve_store): Add MVE instructions mve_store to attribute "type".
273 (mve_load): Add MVE instructions mve_load to attribute "type".
274 (is_mve_type): Define attribute.
275 * config/arm/vec-common.md (mov<mode>): Modify RTL expand to support
276 standard move patterns in MVE along with NEON and IWMMXT with mode
278 (mov<mode>): Modify RTL expand to support standard move patterns in NEON
279 and IWMMXT with mode iterator V8HF.
280 (movv8hf): Define RTL expand to support standard "movv8hf" pattern in
282 * config/arm/vfp.md (neon_immediate_valid_for_move): Rename to
283 simd_immediate_valid_for_move.
286 2020-03-16 H.J. Lu <hongjiu.lu@intel.com>
289 * config/i386/i386.md (*movsi_internal): Call ix86_output_ssemov
290 for TYPE_SSEMOV. Remove ext_sse_reg_operand and TARGET_AVX512VL
292 * config/i386/predicates.md (ext_sse_reg_operand): Removed.
294 2020-03-16 Jakub Jelinek <jakub@redhat.com>
297 * tree-inline.c (insert_init_stmt): Don't gimple_regimplify_operands
300 PR tree-optimization/94166
301 * tree-ssa-reassoc.c (sort_by_mach_mode): Use SSA_NAME_VERSION
302 as secondary comparison key.
304 2020-03-16 Bin Cheng <bin.cheng@linux.alibaba.com>
306 PR tree-optimization/94125
307 * tree-loop-distribution.c
308 (loop_distribution::break_alias_scc_partitions): Update post order
309 number for merged scc.
311 2020-03-15 H.J. Lu <hongjiu.lu@intel.com>
314 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_SI and
316 * config/i386/i386.md (*movsf_internal): Call ix86_output_ssemov
317 for TYPE_SSEMOV. Remove TARGET_PREFER_AVX256, TARGET_AVX512VL
318 and ext_sse_reg_operand check.
320 2020-03-15 Lewis Hyatt <lhyatt@gmail.com>
322 * common.opt: Avoid redundancy in the help text.
323 * config/arc/arc.opt: Likewise.
324 * config/cr16/cr16.opt: Likewise.
326 2020-03-14 Jakub Jelinek <jakub@redhat.com>
329 * tree-nested.c (convert_nonlocal_omp_clauses,
330 convert_local_omp_clauses): Handle {,in_,task_}reduction clauses
331 with C/C++ array sections.
333 2020-03-14 H.J. Lu <hongjiu.lu@intel.com>
336 * config/i386/i386.md (*movdi_internal): Call ix86_output_ssemov
337 for TYPE_SSEMOV. Remove ext_sse_reg_operand and TARGET_AVX512VL
340 2020-03-14 Jakub Jelinek <jakub@redhat.com>
342 * gimple-fold.c (gimple_fold_builtin_strncpy): Change
343 "a an" to "an" in a comment.
344 * hsa-common.h (is_a_helper): Likewise.
345 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Likewise.
346 * config/arc/arc.c (arc600_corereg_hazard): Likewise.
347 * config/s390/s390.c (s390_indirect_branch_via_thunk): Likewise.
349 2020-03-13 Aaron Sawdey <acsawdey@linux.ibm.com>
352 * config/rs6000/rs6000.c (num_insns_constant_multi): Don't shift a
353 64-bit value by 64 bits (UB).
355 2020-03-13 Vladimir Makarov <vmakarov@redhat.com>
357 PR rtl-optimization/92303
358 * lra-spills.c (remove_pseudos): Try to simplify memory subreg.
360 2020-03-13 Segher Boessenkool <segher@kernel.crashing.org>
362 PR rtl-optimization/94148
363 PR rtl-optimization/94042
364 * df-core.c (BB_LAST_CHANGE_AGE): Delete.
365 (df_worklist_propagate_forward): New parameter last_change_age, use
366 that instead of bb->aux.
367 (df_worklist_propagate_backward): Ditto.
368 (df_worklist_dataflow_doublequeue): Use a local array last_change_age.
370 2020-03-13 Richard Biener <rguenther@suse.de>
372 PR tree-optimization/94163
373 * tree-ssa-pre.c (create_expression_by_pieces): Check
374 whether alignment would be zero.
376 2020-03-13 Martin Liska <mliska@suse.cz>
379 * lto-wrapper.c (run_gcc): Use concat for appending
380 to collect_gcc_options.
382 2020-03-13 Jakub Jelinek <jakub@redhat.com>
385 * config/aarch64/aarch64.c (aarch64_add_offset_1): Use gen_int_mode
388 2020-03-13 H.J. Lu <hongjiu.lu@intel.com>
391 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_DF.
392 * config/i386/i386.md (*movdf_internal): Call ix86_output_ssemov
393 for TYPE_SSEMOV. Remove TARGET_AVX512F, TARGET_PREFER_AVX256,
394 TARGET_AVX512VL and ext_sse_reg_operand check.
396 2020-03-13 Bu Le <bule1@huawei.com>
399 * config/aarch64/aarch64.opt (-param=aarch64-float-recp-precision=)
400 (-param=aarch64-double-recp-precision=): New options.
401 * doc/invoke.texi: Document them.
402 * config/aarch64/aarch64.c (aarch64_emit_approx_div): Use them
403 instead of hard-coding the choice of 1 for float and 2 for double.
405 2020-03-13 Eric Botcazou <ebotcazou@adacore.com>
407 PR rtl-optimization/94119
408 * resource.h (clear_hashed_info_until_next_barrier): Declare.
409 * resource.c (clear_hashed_info_until_next_barrier): New function.
410 * reorg.c (add_to_delay_list): Fix formatting.
411 (relax_delay_slots): Call clear_hashed_info_until_next_barrier on
412 the next instruction after removing a BARRIER.
414 2020-03-13 Eric Botcazou <ebotcazou@adacore.com>
417 * expmed.c (store_integral_bit_field): For fields larger than a word,
418 call extract_bit_field on the value if the mode is BLKmode. Remove
419 specific path for big-endian targets and tidy things up a little bit.
421 2020-03-12 Richard Sandiford <richard.sandiford@arm.com>
423 PR rtl-optimization/90275
424 * cse.c (cse_insn): Delete no-op register moves too.
426 2020-03-12 Darius Galis <darius.galis@cyberthorstudios.com>
428 * config/rx/rx.md (CTRLREG_CPEN): Remove.
429 * config/rx/rx.c (rx_print_operand): Remove CTRLREG_CPEN support.
431 2020-03-12 Richard Biener <rguenther@suse.de>
433 PR tree-optimization/94103
434 * tree-ssa-sccvn.c (visit_reference_op_load): Avoid type
435 punning when the mode precision is not sufficient.
437 2020-03-12 H.J. Lu <hongjiu.lu@intel.com>
440 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_DI,
441 MODE_V1DF and MODE_V2SF.
442 * config/i386/mmx.md (MMXMODE:*mov<mode>_internal): Call
443 ix86_output_ssemov for TYPE_SSEMOV. Remove ext_sse_reg_operand
446 2020-03-12 Jakub Jelinek <jakub@redhat.com>
448 * doc/tm.texi.in (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Change
449 ASM_OUTPUT_ALIGNED_DECL in description to ASM_OUTPUT_ALIGNED_LOCAL
450 and ASM_OUTPUT_DECL to ASM_OUTPUT_LOCAL.
451 * doc/tm.texi: Regenerated.
453 PR tree-optimization/94130
454 * tree-ssa-dse.c: Include gimplify.h.
455 (increment_start_addr): If stmt has lhs, drop the lhs from call and
456 set it after the call to the original value of the first argument.
458 (decrement_count): Formatting fix.
460 2020-03-11 Delia Burduv <delia.burduv@arm.com>
462 * config/arm/arm-builtins.c
463 (arm_init_simd_builtin_scalar_types): New.
464 * config/arm/arm_neon.h (vld2_bf16): Used new builtin type.
465 (vld2q_bf16): Used new builtin type.
466 (vld3_bf16): Used new builtin type.
467 (vld3q_bf16): Used new builtin type.
468 (vld4_bf16): Used new builtin type.
469 (vld4q_bf16): Used new builtin type.
470 (vld2_dup_bf16): Used new builtin type.
471 (vld2q_dup_bf16): Used new builtin type.
472 (vld3_dup_bf16): Used new builtin type.
473 (vld3q_dup_bf16): Used new builtin type.
474 (vld4_dup_bf16): Used new builtin type.
475 (vld4q_dup_bf16): Used new builtin type.
477 2020-03-11 Jakub Jelinek <jakub@redhat.com>
480 * config/pdp11/pdp11.c (pdp11_asm_output_var): Call switch_to_section
481 at the start to switch to data section. Don't print extra newline if
482 .globl directive has not been emitted.
484 2020-03-11 Richard Biener <rguenther@suse.de>
486 * match.pd ((T *)(ptr - ptr-cst) -> &MEM[ptr + -ptr-cst]):
489 2020-03-11 Eric Botcazou <ebotcazou@adacore.com>
492 * tree.c (variably_modified_type_p) <RECORD_TYPE>: Recurse into fields
493 whose type is a qualified union.
495 2020-03-11 Jakub Jelinek <jakub@redhat.com>
498 * config/aarch64/aarch64.c (aarch64_add_offset_1): Use absu_hwi
499 instead of abs_hwi, change moffset type to unsigned HOST_WIDE_INT.
502 * value-prof.c (dump_histogram_value): Use abs_hwi instead of
504 (get_nth_most_common_value): Use abs_hwi instead of abs.
507 * dfp.c (decimal_to_binary): Only use decimal128ToString if from->cl
508 is rvc_normal, otherwise use real_to_decimal to print the number to
511 PR tree-optimization/94114
512 * tree-loop-distribution.c (generate_memset_builtin): Call
513 rewrite_to_non_trapping_overflow even on mem.
514 (generate_memcpy_builtin): Call rewrite_to_non_trapping_overflow even
517 2020-03-10 Jeff Law <law@redhat.com>
519 * config/bfin/bfin.md (movsi_insv): Add length attribute.
521 2020-03-10 Jiufu Guo <guojiufu@linux.ibm.com>
524 * gcc/config/rs6000/rs6000.c (rs6000_emit_p9_fp_minmax): Check
525 NAN and SIGNED_ZEROR for smax/smin.
527 2020-03-10 Will Schmidt <will_schmidt@vnet.ibm.com>
530 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
531 clause to handle P9V_BUILTIN_VEC_LXVL with const arguments.
533 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
535 * loop-iv.c (find_simple_exit): Make it static.
536 * cfgloop.h: Remove the corresponding prototype.
538 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
540 * ddg.c (create_ddg): Fix intendation.
541 (set_recurrence_length): Likewise.
542 (create_ddg_all_sccs): Likewise.
544 2020-03-10 Jakub Jelinek <jakub@redhat.com>
547 * config/i386/i386.md (*testqi_ext_3): Call ix86_match_ccmode with
548 CCZmode instead of CCNOmode if operands[2] has DImode and pos + len
551 2020-03-09 Jason Merrill <jason@redhat.com>
553 * gdbinit.in (pgs): Fix typo in documentation.
555 2020-03-09 Vladimir Makarov <vmakarov@redhat.com>
559 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
561 PR rtl-optimization/93564
562 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
563 do not honor reg alloc order.
565 2020-03-09 Andrew Pinski <apinski@marvell.com>
568 * doc/extend.texi (x86 Operand Modifiers): Fix column
571 2020-03-09 Martin Liska <mliska@suse.cz>
574 * config/rs6000/rs6000.c (rs6000_option_override_internal):
575 Remove set of str_align_loops and str_align_jumps as these
576 should be set in previous 2 conditions in the function.
578 2020-03-09 Jakub Jelinek <jakub@redhat.com>
580 PR rtl-optimization/94045
581 * params.opt (-param=max-find-base-term-values=): New option.
582 * alias.c (find_base_term): Add cut-off for number of visited VALUEs
583 in a single toplevel find_base_term call.
585 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
588 * config/aarch64/aarch64-builtins.c (TYPES_TERNOPU_LANE): Add define.
589 * config/aarch64/aarch64-simd.md
590 (aarch64_vec_<su>mult_lane<Qlane>): Add new insn for widening lane mul.
591 (aarch64_vec_<su>mlal_lane<Qlane>): Likewise.
592 * config/aarch64/aarch64-simd-builtins.def: Add intrinsics.
593 * config/aarch64/arm_neon.h:
594 (vmlal_lane_s16): Expand using intrinsics rather than inline asm.
595 (vmlal_lane_u16): Likewise.
596 (vmlal_lane_s32): Likewise.
597 (vmlal_lane_u32): Likewise.
598 (vmlal_laneq_s16): Likewise.
599 (vmlal_laneq_u16): Likewise.
600 (vmlal_laneq_s32): Likewise.
601 (vmlal_laneq_u32): Likewise.
602 (vmull_lane_s16): Likewise.
603 (vmull_lane_u16): Likewise.
604 (vmull_lane_s32): Likewise.
605 (vmull_lane_u32): Likewise.
606 (vmull_laneq_s16): Likewise.
607 (vmull_laneq_u16): Likewise.
608 (vmull_laneq_s32): Likewise.
609 (vmull_laneq_u32): Likewise.
610 * config/aarch64/iterators.md (Vcondtype): New iterator for lane mul.
613 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
615 * aarch64/aarch64-simd.md (aarch64_mla_elt<mode>): Correct lane syntax.
616 (aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
617 (aarch64_mls_elt<mode>): Likewise.
618 (aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
619 (aarch64_fma4_elt<mode>): Likewise.
620 (aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
621 (aarch64_fma4_elt_to_64v2df): Likewise.
622 (aarch64_fnma4_elt<mode>): Likewise.
623 (aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
624 (aarch64_fnma4_elt_to_64v2df): Likewise.
626 2020-03-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
628 * config/aarch64/aarch64-sve2.md (@aarch64_sve_<sve_int_op><mode>:
629 Specify movprfx attribute.
630 (@aarch64_sve_<sve_int_op>_lane_<mode>): Likewise.
632 2020-03-06 David Edelsohn <dje.gcc@gmail.com>
635 * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
637 (TARGET_NO_FP_IN_TOC): Same.
638 * config/rs6000/aix71.h: Same.
639 * config/rs6000/aix72.h: Same.
641 2020-03-06 Andrew Pinski <apinski@marvell.com>
642 Jeff Law <law@redhat.com>
644 PR rtl-optimization/93996
645 * haifa-sched.c (remove_notes): Be more careful when adding
648 2020-03-06 Delia Burduv <delia.burduv@arm.com>
650 * config/arm/arm_neon.h (vld2_bf16): New.
656 (vld2_dup_bf16): New.
657 (vld2q_dup_bf16): New.
658 (vld3_dup_bf16): New.
659 (vld3q_dup_bf16): New.
660 (vld4_dup_bf16): New.
661 (vld4q_dup_bf16): New.
662 * config/arm/arm_neon_builtins.def
663 (vld2): Changed to VAR13 and added v4bf, v8bf
664 (vld2_dup): Changed to VAR8 and added v4bf, v8bf
665 (vld3): Changed to VAR13 and added v4bf, v8bf
666 (vld3_dup): Changed to VAR8 and added v4bf, v8bf
667 (vld4): Changed to VAR13 and added v4bf, v8bf
668 (vld4_dup): Changed to VAR8 and added v4bf, v8bf
669 * config/arm/iterators.md (VDXBF2): New iterator.
670 *config/arm/neon.md (neon_vld2): Use new iterators.
671 (neon_vld2_dup<mode): Use new iterators.
672 (neon_vld3<mode>): Likewise.
673 (neon_vld3qa<mode>): Likewise.
674 (neon_vld3qb<mode>): Likewise.
675 (neon_vld3_dup<mode>): Likewise.
676 (neon_vld4<mode>): Likewise.
677 (neon_vld4qa<mode>): Likewise.
678 (neon_vld4qb<mode>): Likewise.
679 (neon_vld4_dup<mode>): Likewise.
680 (neon_vld2_dupv8bf): New.
681 (neon_vld3_dupv8bf): Likewise.
682 (neon_vld4_dupv8bf): Likewise.
684 2020-03-06 Delia Burduv <delia.burduv@arm.com>
686 * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
687 (bfloat16x8x2_t): New typedef.
688 (bfloat16x4x3_t): New typedef.
689 (bfloat16x8x3_t): New typedef.
690 (bfloat16x4x4_t): New typedef.
691 (bfloat16x8x4_t): New typedef.
698 * config/arm/arm-builtins.c (v2bf_UP): Define.
700 (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
701 * config/arm/arm-modes.def (V2BF): New mode.
702 * config/arm/arm-simd-builtin-types.def
703 (Bfloat16x2_t): New entry.
704 * config/arm/arm_neon_builtins.def
705 (vst2): Changed to VAR13 and added v4bf, v8bf
706 (vst3): Changed to VAR13 and added v4bf, v8bf
707 (vst4): Changed to VAR13 and added v4bf, v8bf
708 * config/arm/iterators.md (VDXBF): New iterator.
709 (VQ2BF): New iterator.
710 *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
711 (neon_vst2<mode>): Used new iterators.
712 (neon_vst3<mode>): Used new iterators.
713 (neon_vst3<mode>): Used new iterators.
714 (neon_vst3qa<mode>): Used new iterators.
715 (neon_vst3qb<mode>): Used new iterators.
716 (neon_vst4<mode>): Used new iterators.
717 (neon_vst4<mode>): Used new iterators.
718 (neon_vst4qa<mode>): Used new iterators.
719 (neon_vst4qb<mode>): Used new iterators.
721 2020-03-06 Delia Burduv <delia.burduv@arm.com>
723 * config/aarch64/aarch64-simd-builtins.def
724 (bfcvtn): New built-in function.
725 (bfcvtn_q): New built-in function.
726 (bfcvtn2): New built-in function.
727 (bfcvt): New built-in function.
728 * config/aarch64/aarch64-simd.md
729 (aarch64_bfcvtn<q><mode>): New pattern.
730 (aarch64_bfcvtn2v8bf): New pattern.
731 (aarch64_bfcvtbf): New pattern.
732 * config/aarch64/arm_bf16.h (float32_t): New typedef.
733 (vcvth_bf16_f32): New intrinsic.
734 * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
735 (vcvtq_low_bf16_f32): New intrinsic.
736 (vcvtq_high_bf16_f32): New intrinsic.
737 * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
738 (UNSPEC_BFCVTN): New UNSPEC.
739 (UNSPEC_BFCVTN2): New UNSPEC.
740 (UNSPEC_BFCVT): New UNSPEC.
741 * config/arm/types.md (bf_cvt): New type.
743 2020-03-06 Andreas Krebbel <krebbel@linux.ibm.com>
745 * config/s390/s390.md ("tabort"): Get rid of two consecutive
746 blanks in format string.
748 2020-03-05 H.J. Lu <hongjiu.lu@intel.com>
752 * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
753 * config/i386/i386.c (ix86_get_ssemov): New function.
754 (ix86_output_ssemov): Likewise.
755 * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
756 ix86_output_ssemov for TYPE_SSEMOV. Remove TARGET_AVX512VL
758 (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
759 (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
760 Remove ext_sse_reg_operand and TARGET_AVX512VL check.
761 (*movti_internal): Likewise.
762 (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
764 2020-03-05 Jeff Law <law@redhat.com>
766 PR tree-optimization/91890
767 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
768 Use gimple_or_expr_nonartificial_location.
769 (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
770 Use gimple_or_expr_nonartificial_location.
771 * gimple.c (gimple_or_expr_nonartificial_location): New function.
772 * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
773 * tree-ssa-strlen.c (maybe_warn_overflow): Use
774 gimple_or_expr_nonartificial_location.
775 (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
776 (maybe_warn_pointless_strcmp): Likewise.
778 2020-03-05 Jakub Jelinek <jakub@redhat.com>
781 * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
782 SRC and MASK arguments to __m128 from __m128d.
783 (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
785 (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
787 * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
788 argument to __m128i from __m128d.
789 (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
791 (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
792 (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
795 2020-03-05 Delia Burduv <delia.burduv@arm.com>
797 * config/arm/arm_neon.h (vbfmmlaq_f32): New.
798 (vbfmlalbq_f32): New.
799 (vbfmlaltq_f32): New.
800 (vbfmlalbq_lane_f32): New.
801 (vbfmlaltq_lane_f32): New.
802 (vbfmlalbq_laneq_f32): New.
803 (vbfmlaltq_laneq_f32): New.
804 * config/arm/arm_neon_builtins.def (vmmla): New.
811 * config/arm/iterators.md (BF_MA): New int iterator.
812 (bt): New int attribute.
813 (VQXBF): Copy of VQX with V8BF.
814 * config/arm/neon.md (neon_vmmlav8bf): New insn.
815 (neon_vfma<bt>v8bf): New insn.
816 (neon_vfma<bt>_lanev8bf): New insn.
817 (neon_vfma<bt>_laneqv8bf): New expand.
818 (neon_vget_high<mode>): Changed iterator to VQXBF.
819 * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
820 (UNSPEC_BFMAB): New UNSPEC.
821 (UNSPEC_BFMAT): New UNSPEC.
823 2020-03-05 Jakub Jelinek <jakub@redhat.com>
826 * tree-pretty-print.h (pretty_print_string): Declare.
827 * tree-pretty-print.c (pretty_print_string): Remove forward
828 declaration, no longer static. Change nbytes parameter type
829 from unsigned to size_t.
830 * print-rtl.c (print_value) <case CONST_STRING>: Use
831 pretty_print_string and for shrink way too long strings.
833 2020-03-05 Richard Biener <rguenther@suse.de>
834 Jakub Jelinek <jakub@redhat.com>
836 PR tree-optimization/93582
837 * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
838 last operand as signed when looking for memset offset. Formatting
841 2020-03-04 Andrew Pinski <apinski@marvell.com>
844 * value-prof.c (dump_histogram_value): Use std::abs.
846 2020-03-04 Martin Sebor <msebor@redhat.com>
848 PR tree-optimization/93986
849 * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
850 operands to the same precision widest_int to avoid ICEs.
852 2020-03-04 Bill Schmidt <wschmidt@linux.ibm.com>
855 * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
856 * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
857 for OPTION_MASK_ALTIVEC.
859 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
861 * config.gcc: Include the glibc-stdint.h header for zTPF.
863 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
865 * config/s390/s390.c (s390_secondary_memory_needed): Disallow
866 direct FPR-GPR copies.
867 (s390_register_info_gprtofpr): Disallow GPR content to be saved in
870 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
872 * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
873 operands to the prologue_tpf expander.
874 (s390_emit_epilogue): Likewise.
875 (s390_option_override_internal): Do error checking and setup for
877 * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
878 (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
879 (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
880 (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
881 * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
882 operands for the check flag and the branch target.
883 * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
884 ("mtpf-trace-hook-prologue-target")
885 ("mtpf-trace-hook-epilogue-check")
886 ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
888 * doc/invoke.texi: Document -mtpf-trace-skip option. The other
889 options are for debugging purposes and will not be documented
892 2020-03-04 Jakub Jelinek <jakub@redhat.com>
895 * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
897 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
898 argument. Change pd argument so that it can be modified. Turn
899 constant non-CONSTRUCTOR store into non-constant if it is too large.
900 Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
902 (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
905 2020-02-04 Richard Biener <rguenther@suse.de>
907 PR tree-optimization/93964
908 * graphite-isl-ast-to-gimple.c
909 (gcc_expression_from_isl_ast_expr_id): Add intermediate
910 conversion for pointer to integer converts.
911 * graphite-scop-detection.c (assign_parameter_index_in_region):
914 2020-03-04 Martin Liska <mliska@suse.cz>
918 * doc/invoke.texi: Clarify --help=language and --help=common
921 2020-03-04 Jakub Jelinek <jakub@redhat.com>
923 PR tree-optimization/94001
924 * tree-tailcall.c (process_assignment): Before comparing op1 to
925 *ass_var, verify *ass_var is non-NULL.
927 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
930 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
933 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
935 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
936 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
937 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
938 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
939 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
940 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
941 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
942 (V_bf_low, V_bf_cvt_m): New mode attributes.
943 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
944 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
945 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
946 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
947 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
949 2020-03-03 Jakub Jelinek <jakub@redhat.com>
951 PR tree-optimization/93582
952 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
953 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
954 members, initialize them in the constructor and if mask is non-NULL,
955 artificially push_partial_def {} for the portions of the mask that
957 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
958 val and return (void *)-1. Formatting fix.
959 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
961 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
962 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
964 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
966 (visit_stmt): Formatting fix.
968 2020-03-03 Richard Biener <rguenther@suse.de>
970 PR tree-optimization/93946
971 * alias.h (refs_same_for_tbaa_p): Declare.
972 * alias.c (refs_same_for_tbaa_p): New function.
973 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
975 * tree-ssa-scopedtables.h
976 (avail_exprs_stack::lookup_avail_expr): Add output argument
977 giving access to the hashtable entry.
978 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
980 * tree-ssa-dom.c: Include alias.h.
981 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
982 removing redundant store.
983 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
984 (ao_ref_init_from_vn_reference): Adjust prototype.
985 (vn_reference_lookup_pieces): Likewise.
986 (vn_reference_insert_pieces): Likewise.
987 * tree-ssa-sccvn.c: Track base alias set in addition to alias
989 (eliminate_dom_walker::eliminate_stmt): Also check base alias
990 set when removing redundant stores.
991 (visit_reference_op_store): Likewise.
992 * dse.c (record_store): Adjust valdity check for redundant
995 2020-03-03 Jakub Jelinek <jakub@redhat.com>
998 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
1000 PR rtl-optimization/94002
1001 * explow.c (plus_constant): Punt if cst has VOIDmode and
1002 get_pool_mode is different from mode.
1004 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
1006 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
1007 address has an offset which fits the scalling constraint for a
1008 load/store operation.
1009 (legitimate_scaled_address_p): Update use
1010 leigitimate_small_data_address_p.
1011 (arc_print_operand): Likewise.
1012 (arc_legitimate_address_p): Likewise.
1013 (legitimate_small_data_address_p): Likewise.
1015 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
1017 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
1018 (fnmasf4_fpu): Likewise.
1020 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
1022 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
1025 (adddi3_i): Remove pattern.
1026 (subdi3_i): Likewise.
1028 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
1030 * config/arc/arc.md (eh_return): Add length info.
1032 2020-03-02 David Malcolm <dmalcolm@redhat.com>
1034 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
1036 2020-03-02 David Malcolm <dmalcolm@redhat.com>
1038 * doc/invoke.texi (Static Analyzer Options): Add
1039 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
1042 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
1045 * config/i386/i386.md (movstrict<mode>): Allow only
1046 registers with VALID_INT_MODE_P modes.
1048 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
1050 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
1051 (reduc_insn): Use 'U' and 'B' operand codes.
1052 (reduc_<reduc_op>_scal_<mode>): Allow all types.
1053 (reduc_<reduc_op>_scal_v64di): Delete.
1054 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
1055 (*plus_carry_dpp_shr_v64si): Change to ...
1056 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
1057 (mov_from_lane63_v64di): Change to ...
1058 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
1059 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
1060 Support UNSPEC_MOV_DPP_SHR output formats.
1061 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
1062 Add "use_extends" reductions.
1063 (print_operand_address): Add 'I' and 'U' codes.
1064 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
1066 2020-03-02 Martin Liska <mliska@suse.cz>
1068 * lto-wrapper.c: Fix typo in comment about
1069 C++ standard version.
1071 2020-03-01 Martin Sebor <msebor@redhat.com>
1074 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
1076 2020-03-01 Martin Sebor <msebor@redhat.com>
1079 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
1080 of a pointer in the outermost ADDR_EXPRs.
1082 2020-02-28 Jeff Law <law@redhat.com>
1084 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
1085 * config/v850/v850.c (v850_asm_trampoline_template): Update
1088 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
1091 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
1094 2020-02-28 Martin Liska <mliska@suse.cz>
1097 * configure.ac: Improve detection of ld_date by requiring
1098 either two dashes or none.
1099 * configure: Regenerate.
1101 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
1103 PR rtl-optimization/93564
1104 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
1105 do not honor reg alloc order.
1107 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
1110 * config/aarch64/aarch64.c (aarch64_override_options): Fix
1111 misleading warning string.
1113 2020-02-27 Martin Sebor <msebor@redhat.com>
1115 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
1117 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
1120 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
1121 Split the insn into two parts. This insn only does variable
1122 extract from a register.
1123 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
1124 variable extract from memory.
1125 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
1126 only does variable extract from a register.
1127 (vsx_extract_v4sf_var_load): New insn, do variable extract from
1129 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
1130 into two parts. This insn only does variable extract from a
1132 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
1133 do variable extract from memory.
1135 2020-02-27 Martin Jambor <mjambor@suse.cz>
1136 Feng Xue <fxue@os.amperecomputing.com>
1139 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
1140 new function calls_same_node_or_its_all_contexts_clone_p.
1141 (cgraph_edge_brings_value_p): Use it.
1142 (cgraph_edge_brings_value_p): Likewise.
1143 (self_recursive_pass_through_p): Return false if caller is a clone.
1144 (self_recursive_agg_pass_through_p): Likewise.
1146 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
1149 * alias.c (ends_tbaa_access_path_p): Break out from ...
1150 (component_uses_parent_alias_set_from): ... here.
1151 * alias.h (ends_tbaa_access_path_p): Declare.
1152 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
1153 handle trailing arrays past end of tbaa access path.
1154 (aliasing_component_refs_p): ... here; likewise.
1155 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
1156 path; disambiguate also past end of it.
1157 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
1160 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
1162 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
1163 beginning of the file.
1164 (vcreate_bf16, vcombine_bf16): New.
1165 (vdup_n_bf16, vdupq_n_bf16): New.
1166 (vdup_lane_bf16, vdup_laneq_bf16): New.
1167 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
1168 (vduph_lane_bf16, vduph_laneq_bf16): New.
1169 (vset_lane_bf16, vsetq_lane_bf16): New.
1170 (vget_lane_bf16, vgetq_lane_bf16): New.
1171 (vget_high_bf16, vget_low_bf16): New.
1172 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
1173 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
1174 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
1175 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
1176 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
1177 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
1178 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
1179 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
1180 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
1181 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
1182 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
1183 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
1184 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
1185 (vreinterpretq_bf16_p128): New.
1186 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
1187 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
1188 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
1189 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
1190 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
1191 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
1192 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
1193 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
1194 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
1195 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
1196 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
1197 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
1198 (vreinterpretq_p128_bf16): New.
1199 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
1201 (V_elem_l): Likewise.
1202 (VD_LANE): Likewise.
1204 (V_DOUBLE): Likewise.
1205 (VDQX): Add V4BF and V8BF.
1206 (V_two_elem, V_three_elem, V_four_elem): Likewise.
1209 (V_double_vector_mode): Likewise.
1210 (V_cmp_result): Likewise.
1211 (V_uf_sclr): Likewise.
1212 (V_sz_elem): Likewise.
1213 (Is_d_reg): Likewise.
1214 (V_mode_nunits): Likewise.
1215 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
1217 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
1219 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
1220 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
1221 (<expander><mode>3<exec>): Likewise.
1222 (<expander><mode>3): New.
1223 (v<expander><mode>3): New.
1224 (<expander><mode>3): New.
1225 (<expander><mode>3<exec>): Rename to ...
1226 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
1227 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
1229 2020-02-27 Alexandre Oliva <oliva@adacore.com>
1231 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
1234 2020-02-27 Richard Biener <rguenther@suse.de>
1236 PR tree-optimization/93508
1237 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
1238 non-_CHK variants. Valueize their length arguments.
1240 2020-02-27 Richard Biener <rguenther@suse.de>
1242 PR tree-optimization/93953
1243 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
1244 to the hash-map entry.
1246 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
1248 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
1250 2020-02-27 Mark Williams <mwilliams@fb.com>
1252 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
1253 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
1254 -ffile-prefix-map and -fmacro-prefix-map.
1255 * lto-streamer-out.c: Include file-prefix-map.h.
1256 (lto_output_location): Remap the file part of locations.
1258 2020-02-27 Jakub Jelinek <jakub@redhat.com>
1261 * gimplify.c (gimplify_init_constructor): Don't promote readonly
1262 DECL_REGISTER variables to TREE_STATIC.
1264 PR tree-optimization/93582
1265 PR tree-optimization/93945
1266 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
1267 non-zero INTEGER_CST second argument and ref->offset or ref->size
1268 not a multiple of BITS_PER_UNIT.
1270 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
1272 * doc/install.texi (Binaries): Update description of BullFreeware.
1274 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
1278 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
1279 C++ Language Options, Warning Options, and Static Analyzer
1280 Options lists. Document negative form of options enabled by
1281 default. Move some things around to more accurately sort
1282 warnings by category.
1283 (C++ Dialect Options, Warning Options, Static Analyzer
1284 Options): Document negative form of options when enabled by
1285 default. Move some things around to more accurately sort
1286 warnings by category. Add some missing index entries.
1289 2020-02-26 Carl Love <cel@us.ibm.com>
1292 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
1293 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
1294 for the vector unsigned short arguments. It is also listed as the
1295 name of the built-in for arguments vector unsigned short,
1296 vector unsigned int and vector unsigned long long built-ins. The
1297 name of the builtins for these arguments should be:
1298 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
1299 __builtin_crypto_vpmsumd respectively.
1301 2020-02-26 Richard Biener <rguenther@suse.de>
1303 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
1304 and load permutation.
1306 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
1309 * optabs-tree.c (supportable_convert_operation): Reject types with
1312 2020-02-26 David Malcolm <dmalcolm@redhat.com>
1314 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
1316 2020-02-26 Jakub Jelinek <jakub@redhat.com>
1318 PR tree-optimization/93820
1319 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
1320 argument to ALL_INTEGER_CST_P boolean.
1321 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
1322 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
1323 adjacent INTEGER_CST store into merged_store->only_constants like
1326 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1329 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
1331 * cfghooks.c (verify_flow_info): Likewise.
1332 * predict.c (combine_predictions_for_bb): Likewise.
1333 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
1334 sucessor -> successor.
1335 (find_traces_1_round): Fix comment typo, destinarion -> destination.
1336 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
1338 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
1339 message typo, sucessors -> successors.
1341 2020-02-25 Martin Sebor <msebor@redhat.com>
1343 * doc/extend.texi (attribute access): Correct an example.
1345 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
1347 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
1349 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
1350 (VAR15, VAR16): New.
1351 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
1352 (VD): Enable for V4BF.
1354 (VQ): Enable for V8BF.
1356 (VQ_NO2E): Likewise.
1357 (VDBL, Vdbl): Add V4BF.
1358 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
1359 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
1360 (bfloat16x8x2_t): Likewise.
1361 (bfloat16x4x3_t): Likewise.
1362 (bfloat16x8x3_t): Likewise.
1363 (bfloat16x4x4_t): Likewise.
1364 (bfloat16x8x4_t): Likewise.
1365 (vcombine_bf16): New.
1366 (vld1_bf16, vld1_bf16_x2): New.
1367 (vld1_bf16_x3, vld1_bf16_x4): New.
1368 (vld1q_bf16, vld1q_bf16_x2): New.
1369 (vld1q_bf16_x3, vld1q_bf16_x4): New.
1370 (vld1_lane_bf16): New.
1371 (vld1q_lane_bf16): New.
1372 (vld1_dup_bf16): New.
1373 (vld1q_dup_bf16): New.
1376 (vld2_dup_bf16): New.
1377 (vld2q_dup_bf16): New.
1380 (vld3_dup_bf16): New.
1381 (vld3q_dup_bf16): New.
1384 (vld4_dup_bf16): New.
1385 (vld4q_dup_bf16): New.
1386 (vst1_bf16, vst1_bf16_x2): New.
1387 (vst1_bf16_x3, vst1_bf16_x4): New.
1388 (vst1q_bf16, vst1q_bf16_x2): New.
1389 (vst1q_bf16_x3, vst1q_bf16_x4): New.
1390 (vst1_lane_bf16): New.
1391 (vst1q_lane_bf16): New.
1399 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
1401 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
1402 (VALL_F16): Likewise.
1403 (VALLDI_F16): Likewise.
1406 (vswap_width_name): Likewise.
1407 (VSWAP_WIDTH): Likewise.
1411 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
1412 (vget_lane_bf16, vgetq_lane_bf16): New.
1413 (vcreate_bf16): New.
1414 (vdup_n_bf16, vdupq_n_bf16): New.
1415 (vdup_lane_bf16, vdup_laneq_bf16): New.
1416 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
1417 (vduph_lane_bf16, vduph_laneq_bf16): New.
1418 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
1419 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
1420 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
1421 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
1422 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
1423 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
1424 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
1425 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
1426 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
1427 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
1428 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
1429 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
1430 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
1431 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
1432 (vreinterpretq_bf16_p128): New.
1433 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
1434 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
1435 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
1436 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
1437 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
1438 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
1439 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
1440 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
1441 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
1442 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
1443 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
1444 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
1445 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
1446 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
1447 (vreinterpretq_p128_bf16): New.
1449 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
1451 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
1452 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
1453 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
1454 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
1455 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
1456 * config/arm/iterators.md (VSF2BF): New attribute.
1457 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
1458 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
1459 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
1461 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
1463 * config/arm/arm.md (required_for_purecode): New attribute.
1464 (enabled): Handle required_for_purecode.
1465 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
1466 work with -mpure-code.
1468 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1470 PR rtl-optimization/93908
1471 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
1474 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
1476 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
1478 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
1480 * doc/install.texi (--enable-checking): Adjust wording.
1482 2020-02-25 Richard Biener <rguenther@suse.de>
1484 PR tree-optimization/93868
1485 * tree-vect-slp.c (slp_copy_subtree): New function.
1486 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
1487 re-arranging stmts in it.
1489 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1492 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
1493 dummy function and remove it at the end.
1495 PR translation/93864
1496 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
1497 paramter -> parameter.
1498 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
1499 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
1501 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
1503 * doc/install.texi (--enable-checking): Properly document current
1505 (--enable-stage1-checking): Minor clarification about bootstrap.
1507 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1510 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
1511 -fanalyzer-checker=taint is also required.
1512 (-fanalyzer-checker=): Note that providing this option enables the
1513 given checker, and doing so may be required for checkers that are
1514 disabled by default.
1516 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1518 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
1519 significant control flow events; add a "3" which shows all
1520 control flow events; the old "3" becomes "4".
1522 2020-02-24 Jakub Jelinek <jakub@redhat.com>
1524 PR tree-optimization/93582
1525 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
1526 pd.offset and pd.size to be counted in bits rather than bytes, add
1527 support for maxsizei that is not a multiple of BITS_PER_UNIT and
1528 handle bitfield stores and loads.
1529 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
1530 uncomparable quantities - bytes vs. bits. Allow push_partial_def
1531 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
1532 pd.offset/pd.size to be counted in bits rather than bytes.
1533 Formatting fix. Rename shadowed len variable to buflen.
1535 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1536 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
1539 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
1540 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
1541 * opts-common.c (parse_options_from_collect_gcc_options): New function.
1542 (prepend_xassembler_to_collect_as_options): Likewise.
1543 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
1544 (prepend_xassembler_to_collect_as_options): Likewise.
1545 * lto-opts.c (lto_write_options): Stream assembler options
1546 in COLLECT_AS_OPTIONS.
1547 * lto-wrapper.c (xassembler_options_error): New static variable.
1548 (get_options_from_collect_gcc_options): Move parsing options code to
1549 parse_options_from_collect_gcc_options and call it.
1550 (merge_and_complain): Validate -Xassembler options.
1551 (append_compiler_options): Handle OPT_Xassembler.
1552 (run_gcc): Append command line -Xassembler options to
1553 collect_gcc_options.
1554 * doc/invoke.texi: Add documentation about using Xassembler
1557 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
1559 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
1561 (riscv_rtx_costs): Update cost model for LTGT.
1563 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
1565 PR rtl-optimization/93564
1566 * ira-color.c (struct update_cost_queue_elem): New member start.
1567 (queue_update_cost, get_next_update_cost): Add new arg start.
1568 (allocnos_conflict_p): New function.
1569 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
1570 Add checking conflicts with allocnos_conflict_p.
1571 (update_costs_from_prefs, restore_costs_from_copies): Adjust
1572 update_costs_from_allocno calls.
1573 (update_conflict_hard_regno_costs): Add checking conflicts with
1574 allocnos_conflict_p. Adjust calls of queue_update_cost and
1575 get_next_update_cost.
1576 (assign_hard_reg): Adjust calls of queue_update_cost. Add
1578 (bucket_allocno_compare_func): Restore previous version.
1580 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
1582 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
1583 double-word size when handling aggregate return values.
1584 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
1585 that homogeneous SFmode and DFmode aggregates are passed and returned
1586 in general registers.
1588 2020-02-21 Jakub Jelinek <jakub@redhat.com>
1590 PR translation/93759
1591 * opts.c (print_filtered_help): Translate help before appending
1592 messages to it rather than after that.
1594 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1596 PR rtl-optimization/PR92989
1597 * lra-lives.c (process_bb_lives): Restore the original order
1598 of the bb liveness update. Call make_hard_regno_dead for each
1599 register clobbered at the start of an EH receiver.
1601 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
1604 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
1605 self-recursively generated.
1607 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
1610 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
1613 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
1615 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
1616 Document new target supports option.
1618 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
1620 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
1621 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
1622 * config/arm/iterators.md (MATMUL): New iterator.
1623 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
1624 (mmla_sfx): New attribute.
1625 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
1626 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
1627 (UNSPEC_MATMUL_US): New.
1629 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1631 * config/arm/arm.md: Prevent scalar shifts from being used when big
1634 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
1635 Richard Biener <rguenther@suse.de>
1637 PR tree-optimization/93586
1638 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1639 after mismatched array refs; do not sure type size information to
1640 recover from unmatched referneces with !flag_strict_aliasing_p.
1642 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1644 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1645 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1646 (scatter_store<mode>): Rename to ...
1647 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1648 (scatter<mode>_exec): Delete. Move contents ...
1649 (mask_scatter_store<mode>): ... here, and rename that to ...
1650 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1651 Remove mode conversion.
1652 (mask_gather_load<mode>): Rename to ...
1653 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1654 Remove mode conversion.
1655 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1657 2020-02-21 Martin Jambor <mjambor@suse.cz>
1659 PR tree-optimization/93845
1660 * tree-sra.c (verify_sra_access_forest): Only test access size of
1663 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1665 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1666 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1667 (addv64di3_exec): Likewise.
1668 (subv64di3): Likewise.
1669 (subv64di3_exec): Likewise.
1670 (addv64di3_zext): Likewise.
1671 (addv64di3_zext_exec): Likewise.
1672 (addv64di3_zext_dup): Likewise.
1673 (addv64di3_zext_dup_exec): Likewise.
1674 (addv64di3_zext_dup2): Likewise.
1675 (addv64di3_zext_dup2_exec): Likewise.
1676 (addv64di3_sext_dup2): Likewise.
1677 (addv64di3_sext_dup2_exec): Likewise.
1678 (<expander>v64di3): Likewise.
1679 (<expander>v64di3_exec): Likewise.
1680 (*<reduc_op>_dpp_shr_v64di): Likewise.
1681 (*plus_carry_dpp_shr_v64di): Likewise.
1682 * config/gcn/gcn.md (adddi3): Likewise.
1683 (addptrdi3): Likewise.
1684 (<expander>di3): Likewise.
1686 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1688 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1690 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1692 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1693 support. Use aarch64_emit_mult instead of emitting multiplication
1694 instructions directly.
1695 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1696 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1698 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1700 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1701 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
1702 instead of emitting multiplication instructions directly.
1703 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1704 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1705 (@aarch64_frecps<mode>): New expanders.
1707 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1709 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1710 on and produce uint64_ts rather than ints.
1711 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1712 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1714 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1716 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1717 an unused xmsk register when handling approximate rsqrt.
1719 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1721 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1722 flag_finite_math_only condition.
1724 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
1727 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1728 to destination operand for shufps alternative.
1729 (*vec_extractv2si_1): Ditto.
1731 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
1734 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1737 2020-02-20 Martin Liska <mliska@suse.cz>
1739 PR translation/93831
1740 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1742 2020-02-20 Martin Liska <mliska@suse.cz>
1744 PR translation/93830
1745 * common/config/avr/avr-common.c: Remote trailing "|".
1747 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1749 * collect2.c (maybe_run_lto_and_relink): Fix typo in
1752 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1754 PR tree-optimization/93767
1755 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1756 access-size bias from the offset calculations for negative strides.
1758 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1760 * collect2.c (c_file, o_file): Make const again.
1761 (ldout,lderrout, dump_ld_file): Remove.
1762 (tool_cleanup): Avoid calling not signal-safe functions.
1763 (maybe_run_lto_and_relink): Avoid possible signal handler
1764 access to unintialzed memory (lto_o_files).
1765 (main): Avoid leaking temp files in $TMPDIR.
1766 Initialize c_file/o_file with concat, which avoids exposing
1767 uninitialized memory to signal handler, which calls unlink(!).
1768 Avoid calling maybe_unlink when the main function returns,
1769 since the atexit handler is already doing this.
1770 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1772 2020-02-19 Martin Jambor <mjambor@suse.cz>
1774 PR tree-optimization/93776
1775 * tree-sra.c (create_access): Do not create zero size accesses.
1776 (get_access_for_expr): Do not search for zero sized accesses.
1778 2020-02-19 Martin Jambor <mjambor@suse.cz>
1780 PR tree-optimization/93667
1781 * tree-sra.c (scalarizable_type_p): Return false if record fields
1782 do not follow wach other.
1784 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1786 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1787 rather than fmv.x.s/fmv.s.x.
1789 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
1791 * config/aarch64/aarch64-simd-builtins.def
1792 (intrinsic_vec_smult_lo_): New.
1793 (intrinsic_vec_umult_lo_): Likewise.
1794 (vec_widen_smult_hi_): Likewise.
1795 (vec_widen_umult_hi_): Likewise.
1796 * config/aarch64/aarch64-simd.md
1797 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1798 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1799 (vmull_high_s16): Likewise.
1800 (vmull_high_s32): Likewise.
1801 (vmull_high_u8): Likewise.
1802 (vmull_high_u16): Likewise.
1803 (vmull_high_u32): Likewise.
1804 (vmull_s8): Likewise.
1805 (vmull_s16): Likewise.
1806 (vmull_s32): Likewise.
1807 (vmull_u8): Likewise.
1808 (vmull_u16): Likewise.
1809 (vmull_u32): Likewise.
1811 2020-02-18 Martin Liska <mliska@suse.cz>
1813 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1814 bootstrap by missing removal of invalid sanity check.
1816 2020-02-18 Martin Liska <mliska@suse.cz>
1819 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1820 Always compare LHS of gimple_assign.
1822 2020-02-18 Martin Liska <mliska@suse.cz>
1825 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1826 and return type of functions.
1827 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1828 Drop MALLOC attribute for void functions.
1829 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1830 malloc_state for a new VOID clone.
1832 2020-02-18 Martin Liska <mliska@suse.cz>
1835 * common.opt: Add -fprofile-reproducibility.
1836 * doc/invoke.texi: Document it.
1837 * value-prof.c (dump_histogram_value):
1838 Document and support behavior for counters[0]
1839 being a negative value.
1840 (get_nth_most_common_value): Handle negative
1841 counters[0] in respect to flag_profile_reproducible.
1843 2020-02-18 Jakub Jelinek <jakub@redhat.com>
1846 * cgraph.c (verify_speculative_call): Use speculative_id instead of
1847 speculative_uid in messages. Remove trailing whitespace from error
1848 message. Use num_speculative_call_targets instead of
1849 num_speculative_targets in a message.
1850 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1851 edge messages and stmt instead of cal_stmt in reference message.
1853 PR tree-optimization/93780
1854 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1855 before calling build_vector_type.
1856 (execute_update_addresses_taken): Likewise.
1859 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1860 typo, functoin -> function.
1861 * tree.c (free_lang_data_in_decl): Fix comment typo,
1862 functoin -> function.
1863 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1865 2020-02-17 David Malcolm <dmalcolm@redhat.com>
1867 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1869 (print_option_information): Don't call get_option_url if URLs
1872 2020-02-17 Alexandre Oliva <oliva@adacore.com>
1874 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1875 handling of register_common-less targets.
1877 2020-02-17 Martin Liska <mliska@suse.cz>
1880 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1882 2020-02-17 Martin Liska <mliska@suse.cz>
1884 PR translation/93755
1885 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1888 2020-02-17 Martin Liska <mliska@suse.cz>
1891 * config/rx/elf.opt: Fix typo.
1893 2020-02-17 Richard Biener <rguenther@suse.de>
1896 * opts-global.c (print_ignored_options): Use inform and
1899 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
1902 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1904 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
1907 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1908 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
1910 2020-02-15 Jason Merrill <jason@redhat.com>
1912 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
1914 2020-02-15 Jakub Jelinek <jakub@redhat.com>
1916 PR tree-optimization/93744
1917 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
1918 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
1919 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1920 sure @2 in the first and @1 in the other patterns has no side-effects.
1922 2020-02-15 David Malcolm <dmalcolm@redhat.com>
1923 Bernd Edlinger <bernd.edlinger@hotmail.de>
1927 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1928 * configure.ac (--with-diagnostics-urls): New configuration
1929 option, based on --with-diagnostics-color.
1930 (DIAGNOSTICS_URLS_DEFAULT): New define.
1931 * config.h: Regenerate.
1932 * configure: Regenerate.
1933 * diagnostic.c (diagnostic_urls_init): Handle -1 for
1934 DIAGNOSTICS_URLS_DEFAULT from configure-time
1935 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1936 and TERM_URLS environment variable.
1937 * diagnostic-url.h (diagnostic_url_format): New enum type.
1938 (diagnostic_urls_enabled_p): rename to...
1939 (determine_url_format): ... this, and change return type.
1940 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1941 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1942 the linux console, and mingw.
1943 (diagnostic_urls_enabled_p): rename to...
1944 (determine_url_format): ... this, and adjust.
1945 * pretty-print.h (pretty_printer::show_urls): rename to...
1946 (pretty_printer::url_format): ... this, and change to enum.
1947 * pretty-print.c (pretty_printer::pretty_printer,
1948 pp_begin_url, pp_end_url, test_urls): Adjust.
1949 * doc/install.texi (--with-diagnostics-urls): Document the new
1950 configuration option.
1951 (--with-diagnostics-color): Document the existing interaction
1952 with GCC_COLORS better.
1953 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1954 vindex reference. Update description of defaults based on the above.
1955 (-fdiagnostics-color): Update description of how -fdiagnostics-color
1956 interacts with GCC_COLORS.
1958 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
1961 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1962 conjunction with TARGET_GNU_TLS in early return.
1964 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
1966 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1967 the mode is not wider than UNITS_PER_WORD.
1969 2020-02-14 Martin Jambor <mjambor@suse.cz>
1971 PR tree-optimization/93516
1972 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1973 access of the same type as the parent.
1974 (propagate_subaccesses_from_lhs): Likewise.
1976 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
1979 * config/i386/avx512vbmi2intrin.h
1980 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1981 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1982 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1983 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1984 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1985 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1986 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1987 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1988 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1989 of lacking a closing parenthesis.
1990 * config/i386/avx512vbmi2vlintrin.h
1991 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1992 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1993 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1994 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1995 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1996 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1997 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1998 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1999 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
2000 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
2001 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
2002 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
2003 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
2004 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
2005 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
2006 _mm_shldi_epi32, _mm_mask_shldi_epi32,
2007 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
2008 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
2010 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
2013 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
2014 the target function entry.
2016 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
2018 * common/config/arc/arc-common.c (arc_option_optimization_table):
2019 Disable if-conversion step when optimized for size.
2021 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
2023 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
2024 R12-R15 are always in ARCOMPACT16_REGS register class.
2025 * config/arc/arc.opt (mq-class): Deprecate.
2026 * config/arc/constraint.md ("q"): Remove dependency on mq-class
2028 * doc/invoke.texi (mq-class): Update text.
2029 * common/config/arc/arc-common.c (arc_option_optimization_table):
2032 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
2034 * config/arc/arc.c (arc_insn_cost): New function.
2035 (TARGET_INSN_COST): Define.
2036 * config/arc/arc.md (cost): New attribute.
2037 (add_n): Use arc_nonmemory_operand.
2038 (ashlsi3_insn): Likewise, also update constraints.
2039 (ashrsi3_insn): Likewise.
2040 (rotrsi3): Likewise.
2041 (add_shift): Likewise.
2042 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
2044 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
2046 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
2048 (umulsidi_600): Likewise.
2050 2020-02-13 Jakub Jelinek <jakub@redhat.com>
2053 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
2054 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
2055 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
2056 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
2057 pass __A to the builtin followed by __W instead of __A followed by
2059 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
2060 _mm512_mask_popcnt_epi64): Likewise.
2061 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
2062 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
2063 _mm256_mask_popcnt_epi64): Likewise.
2065 PR tree-optimization/93582
2066 * fold-const.h (shift_bytes_in_array_left,
2067 shift_bytes_in_array_right): Declare.
2068 * fold-const.c (shift_bytes_in_array_left,
2069 shift_bytes_in_array_right): New function, moved from
2070 gimple-ssa-store-merging.c, no longer static.
2071 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
2072 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
2073 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
2074 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
2075 shift_bytes_in_array.
2076 (verify_shift_bytes_in_array): Rename to ...
2077 (verify_shift_bytes_in_array_left): ... this. Use
2078 shift_bytes_in_array_left instead of shift_bytes_in_array.
2079 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
2080 instead of verify_shift_bytes_in_array.
2081 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
2082 / native_interpret_expr where the store covers all needed bits,
2083 punt on PDP-endian, otherwise allow all involved offsets and sizes
2084 not to be byte-aligned.
2087 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
2088 use const_0_to_255_operand predicate instead of immediate_operand.
2089 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
2090 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
2091 vgf2p8affineinvqb_<mode><mask_name>,
2092 vgf2p8affineqb_<mode><mask_name>): Drop mode from
2093 const_0_to_255_operand predicated operands.
2095 2020-02-12 Jeff Law <law@redhat.com>
2097 * config/h8300/h8300.md (comparison shortening peepholes): Use
2098 a mode iterator to merge the HImode and SImode peepholes.
2100 2020-02-12 Jakub Jelinek <jakub@redhat.com>
2103 * real.c (is_even): Make static. Function comment fix.
2104 (is_halfway_below): Make static, don't assert R is not inf/nan,
2105 instead return false for those. Small formatting fixes.
2107 2020-02-12 Martin Sebor <msebor@redhat.com>
2110 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
2111 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
2112 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
2113 (strlen_check_and_optimize_call): Adjust callee name.
2115 2020-02-12 Jeff Law <law@redhat.com>
2117 * config/h8300/h8300.md (comparison shortening peepholes): Drop
2118 (and (xor)) variant. Combine other two into single peephole.
2120 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
2122 PR rtl-optimization/93565
2123 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
2125 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
2127 * config/aarch64/aarch64-simd.md
2128 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
2129 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
2130 generating separate ADDV and zero_extend patterns.
2131 * config/aarch64/iterators.md (VDQV_E): New iterator.
2133 2020-02-12 Jeff Law <law@redhat.com>
2135 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
2136 expanders, splits, etc.
2137 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
2138 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
2139 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
2140 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
2141 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
2142 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
2144 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
2146 2020-02-12 Jakub Jelinek <jakub@redhat.com>
2149 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
2150 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
2151 TARGET_AVX512DQ from condition.
2152 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
2153 instead of <mask_mode512bit_condition> in condition. If
2154 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
2156 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
2159 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
2162 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
2164 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
2166 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
2167 where strlen is more legible.
2168 (rs6000_builtin_vectorized_libmass): Ditto.
2169 (rs6000_print_options_internal): Ditto.
2171 2020-02-11 Martin Sebor <msebor@redhat.com>
2173 PR tree-optimization/93683
2174 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
2176 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
2178 * config/rs6000/predicates.md (cint34_operand): Rename the
2179 -mprefixed-addr option to be -mprefixed.
2180 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
2181 the -mprefixed-addr option to be -mprefixed.
2182 (OTHER_FUTURE_MASKS): Likewise.
2183 (POWERPC_MASKS): Likewise.
2184 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
2185 the -mprefixed-addr option to be -mprefixed. Change error
2186 messages to refer to -mprefixed.
2187 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
2189 (rs6000_legitimate_offset_address_p): Likewise.
2190 (rs6000_mode_dependent_address): Likewise.
2191 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
2192 "-mprefixed" for target attributes and pragmas.
2193 (address_to_insn_form): Rename the -mprefixed-addr option to be
2195 (rs6000_adjust_insn_length): Likewise.
2196 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
2197 -mprefixed-addr option to be -mprefixed.
2198 (ASM_OUTPUT_OPCODE): Likewise.
2199 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
2200 -mprefixed-addr option to be -mprefixed.
2201 * config/rs6000/rs6000.opt (-mprefixed): Rename the
2202 -mprefixed-addr option to be prefixed. Change the option from
2203 being undocumented to being documented.
2204 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
2205 -mprefixed option. Update the -mpcrel documentation to mention
2208 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
2210 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
2211 including FIRST_PSEUDO_REGISTER - 1.
2212 * ira-color.c (print_hard_reg_set): Ditto.
2214 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2216 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
2217 (USTERNOP_QUALIFIERS): New define.
2218 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
2219 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
2220 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
2221 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
2222 * config/arm/arm_neon.h (vusdot_s32): New.
2223 (vusdot_lane_s32): New.
2224 (vusdotq_lane_s32): New.
2225 (vsudot_lane_s32): New.
2226 (vsudotq_lane_s32): New.
2227 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
2228 * config/arm/iterators.md (DOTPROD_I8MM): New.
2229 (sup, opsuffix): Add <us/su>.
2230 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
2231 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
2233 2020-02-11 Richard Biener <rguenther@suse.de>
2235 PR tree-optimization/93661
2236 PR tree-optimization/93662
2237 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
2239 * tree-sra.c (get_access_for_expr): Likewise.
2241 2020-02-10 Jakub Jelinek <jakub@redhat.com>
2244 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
2245 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
2246 Change condition from TARGET_AVX2 to TARGET_AVX.
2248 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
2251 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
2252 argument of strncmp.
2254 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
2256 Try to generate zero-based comparisons.
2257 * config/cris/cris.c (cris_reduce_compare): New function.
2258 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
2259 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
2260 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
2262 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
2265 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
2266 in Thumb state and also as a destination in Arm state. Add T16
2269 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
2271 * md.texi (Define Subst): Match closing paren in example.
2273 2020-02-10 Jakub Jelinek <jakub@redhat.com>
2277 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
2278 arguments of strncmp.
2280 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
2283 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
2284 but different source value.
2285 (adjust_callers_for_value_intersection): New function.
2286 (gather_edges_for_value): Adjust order of callers to let a
2287 non-self-recursive caller be the first element.
2288 (self_recursive_pass_through_p): Add a new parameter "simple", and
2289 check generalized self-recursive pass-through jump function.
2290 (self_recursive_agg_pass_through_p): Likewise.
2291 (find_more_scalar_values_for_callers_subset): Compute value from
2292 pass-through jump function for self-recursive.
2293 (intersect_with_plats): Cleanup previous implementation code for value
2294 itersection with self-recursive call edge.
2295 (intersect_with_agg_replacements): Likewise.
2296 (intersect_aggregates_with_edge): Deduce value from pass-through jump
2297 function for self-recursive call edge. Cleanup previous implementation
2298 code for value intersection with self-recursive call edge.
2299 (decide_whether_version_node): Remove dead callers and adjust order
2300 to let a non-self-recursive caller be the first element.
2302 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
2304 * recog.c: Move pass_split_before_sched2 code in front of
2305 pass_split_before_regstack.
2306 (pass_data_split_before_sched2): Rename pass to split3 from split4.
2307 (pass_data_split_before_regstack): Rename pass to split4 from split3.
2308 (rest_of_handle_split_before_sched2): Remove.
2309 (pass_split_before_sched2::execute): Unconditionally call
2311 (enable_split_before_sched2): New function.
2312 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
2313 (pass_split_before_regstack::gate): Ditto.
2314 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
2315 Update name check for renamed split4 pass.
2316 * config/sh/sh.c (register_sh_passes): Update pass insertion
2317 point for renamed split4 pass.
2319 2020-02-09 Jakub Jelinek <jakub@redhat.com>
2321 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
2322 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
2323 copying them around between host and target.
2325 2020-02-08 Andrew Pinski <apinski@marvell.com>
2328 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
2329 STRICT_ALIGNMENT also.
2331 2020-02-08 Jim Wilson <jimw@sifive.com>
2334 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
2336 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
2337 Jakub Jelinek <jakub@redhat.com>
2340 * config/i386/i386.h (CALL_USED_REGISTERS): Make
2341 xmm16-xmm31 call-used even in 64-bit ms-abi.
2343 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
2345 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
2346 (simd_ummla, simd_usmmla): Likewise.
2347 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
2348 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
2349 (vusmmlaq_s32): New.
2351 2020-02-07 Richard Biener <rguenther@suse.de>
2354 * tree-inline.c (fold_marked_statements): Do a PRE walk,
2355 skipping unreachable regions.
2356 (optimize_inline_calls): Skip folding stmts when we didn't
2359 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
2362 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
2363 Don't return aggregates with only SFmode and DFmode in SSE
2365 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
2367 2020-02-07 Jakub Jelinek <jakub@redhat.com>
2370 * config/rs6000/rs6000-logue.c
2371 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
2372 if it fails, move rs into end_addr and retry. Add
2373 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
2374 the insn pattern doesn't describe well what exactly happens to
2378 * config/i386/predicates.md (avx_identity_operand): Remove.
2379 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
2380 (avx_<castmode><avxsizesuffix>_<castmode>,
2381 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
2382 a VEC_CONCAT of the operand and UNSPEC_CAST.
2383 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
2384 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
2388 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
2389 recog_data.insn if distance_non_agu_define changed it.
2391 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
2394 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
2395 we only had X-FORM (reg+reg) addressing for vectors. Also before
2396 ISA 3.0, we only had X-FORM addressing for scalars in the
2397 traditional Altivec registers.
2399 2020-02-06 <zhongyunde@huawei.com>
2400 Vladimir Makarov <vmakarov@redhat.com>
2402 PR rtl-optimization/93561
2403 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
2404 hard register range.
2406 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2408 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
2411 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
2413 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
2414 where the low and the high 32 bits are equal to each other specially,
2415 with an rldimi instruction.
2417 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
2419 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
2421 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
2423 * config/arm/arm-tables.opt: Regenerate.
2425 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2428 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
2429 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
2430 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
2432 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2434 PR rtl-optimization/87763
2435 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
2437 2020-02-06 Delia Burduv <delia.burduv@arm.com>
2439 * config/aarch64/aarch64-simd-builtins.def
2440 (bfmlaq): New built-in function.
2441 (bfmlalb): New built-in function.
2442 (bfmlalt): New built-in function.
2443 (bfmlalb_lane): New built-in function.
2444 (bfmlalt_lane): New built-in function.
2445 * config/aarch64/aarch64-simd.md
2446 (aarch64_bfmmlaqv4sf): New pattern.
2447 (aarch64_bfmlal<bt>v4sf): New pattern.
2448 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
2449 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
2450 (vbfmlalbq_f32): New intrinsic.
2451 (vbfmlaltq_f32): New intrinsic.
2452 (vbfmlalbq_lane_f32): New intrinsic.
2453 (vbfmlaltq_lane_f32): New intrinsic.
2454 (vbfmlalbq_laneq_f32): New intrinsic.
2455 (vbfmlaltq_laneq_f32): New intrinsic.
2456 * config/aarch64/iterators.md (BF_MLA): New int iterator.
2457 (bt): New int attribute.
2459 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
2461 * config/i386/i386.md (*pushtf): Emit "#" instead of
2462 calling gcc_unreachable in insn output.
2465 (*pushsf_rex64): Ditto for alternatives other than 1.
2466 (*pushsf): Ditto for alternatives other than 1.
2468 2020-02-06 Martin Liska <mliska@suse.cz>
2470 PR gcov-profile/91971
2471 PR gcov-profile/93466
2472 * coverage.c (coverage_init): Revert mangling of
2473 path into filename. It can lead to huge filename length.
2474 Creation of subfolders seem more natural.
2476 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2479 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
2480 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
2481 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
2483 2020-02-06 Jakub Jelinek <jakub@redhat.com>
2486 * config/i386/predicates.md (avx_identity_operand): New predicate.
2487 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
2488 define_insn_and_split.
2491 * omp-low.c (use_pointer_for_field): For nested constructs, also
2492 look for map clauses on target construct.
2493 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
2494 taskreg_nesting_level.
2497 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
2498 shared clause, call omp_notice_variable on outer context if any.
2500 2020-02-05 Jason Merrill <jason@redhat.com>
2503 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
2504 non-zero address even if weak and not yet defined.
2506 2020-02-05 Martin Sebor <msebor@redhat.com>
2508 PR tree-optimization/92765
2509 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
2510 * tree-ssa-strlen.c (compute_string_length): Remove.
2511 (determine_min_objsize): Remove.
2512 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
2513 Avoid using type size as the upper bound on string length.
2514 (handle_builtin_string_cmp): Add an argument. Adjust.
2515 (strlen_check_and_optimize_call): Pass additional argument to
2516 handle_builtin_string_cmp.
2518 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
2520 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
2521 (*pushdi2_rex64 peephole2): Unconditionally split after
2523 (*ashl<mode>3_doubleword): Ditto.
2524 (*<shift_insn><mode>3_doubleword): Ditto.
2526 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
2529 * config/rs6000/rs6000.c (get_vector_offset): Fix
2531 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
2533 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
2535 2020-02-05 David Malcolm <dmalcolm@redhat.com>
2538 (Special Functions for Debugging the Analyzer): Update description
2539 of __analyzer_dump_exploded_nodes.
2541 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2544 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
2545 include sets and not clobbers in the vzeroupper pattern.
2546 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
2547 the parallel has 17 (64-bit) or 9 (32-bit) elts.
2548 (*avx_vzeroupper_1): New define_insn_and_split.
2551 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
2552 don't run when !optimize.
2553 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
2556 2020-02-05 Richard Biener <rguenther@suse.de>
2559 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
2560 checks before matching calls.
2562 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2564 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
2565 function comment typo.
2568 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
2569 simd_clone_create failed when i == 0, adjust clone->nargs by
2572 2020-02-05 Martin Liska <mliska@suse.cz>
2575 * doc/invoke.texi: Document that one should
2576 not combine ASLR and -fpch.
2578 2020-02-04 Richard Biener <rguenther@suse.de>
2580 PR tree-optimization/93538
2581 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
2583 2020-02-04 Richard Biener <rguenther@suse.de>
2585 PR tree-optimization/91123
2586 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
2587 (vn_walk_cb_data::last_vuse): New member.
2588 (vn_walk_cb_data::saved_operands): Likewsie.
2589 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
2590 (vn_walk_cb_data::push_partial_def): Use finish.
2591 (vn_reference_lookup_2): Update last_vuse and use finish if
2592 we've saved operands.
2593 (vn_reference_lookup_3): Use finish and update calls to
2594 push_partial_defs everywhere. When translating through
2595 memcpy or aggregate copies save off operands and alias-set.
2596 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
2597 operation for redundant store removal.
2599 2020-02-04 Richard Biener <rguenther@suse.de>
2601 PR tree-optimization/92819
2602 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
2603 generating more stmts than before.
2605 2020-02-04 Martin Liska <mliska@suse.cz>
2607 * config/arm/arm.c (arm_gen_far_branch): Move the function
2608 outside of selftests.
2610 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2612 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
2613 function to adjust PC-relative vector addresses.
2614 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
2615 handle vectors with PC-relative addresses.
2617 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2619 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
2621 (hard_reg_and_mode_to_addr_mask): Delete.
2622 (rs6000_adjust_vec_address): If the original vector address
2623 was REG+REG or REG+OFFSET and the element is not zero, do the add
2624 of the elements in the original address before adding the offset
2625 for the vector element. Use address_to_insn_form to validate the
2626 address using the register being loaded, rather than guessing
2627 whether the address is a DS-FORM or DQ-FORM address.
2629 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2631 * config/rs6000/rs6000.c (get_vector_offset): New helper function
2632 to calculate the offset in memory from the start of a vector of a
2633 particular element. Add code to keep the element number in
2634 bounds if the element number is variable.
2635 (rs6000_adjust_vec_address): Move calculation of offset of the
2636 vector element to get_vector_offset.
2637 (rs6000_split_vec_extract_var): Do not do the initial AND of
2638 element here, move the code to get_vector_offset.
2640 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2642 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2645 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
2647 * config/rs6000/constraints.md: Improve documentation.
2649 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
2652 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2653 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2655 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
2657 * config.gcc: Remove "carrizo" support.
2658 * config/gcn/gcn-opts.h (processor_type): Likewise.
2659 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2660 * config/gcn/gcn.opt (gpu_type): Likewise.
2661 * config/gcn/t-omp-device: Likewise.
2663 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2666 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2667 * config/arm/arm.c (arm_gen_far_branch): New function
2669 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2671 2020-02-03 Julian Brown <julian@codesourcery.com>
2672 Tobias Burnus <tobias@codesourcery.com>
2674 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2676 2020-02-03 Jakub Jelinek <jakub@redhat.com>
2679 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2680 valid RTL to sum up the lowest and second lowest bytes of the popcnt
2683 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
2685 PR rtl-optimization/91333
2686 * ira-color.c (struct allocno_color_data): Add member
2688 (init_allocno_threads): Set the member up.
2689 (bucket_allocno_compare_func): Add compare hard reg
2692 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
2694 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2696 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2697 * config.in: Regenerated.
2698 * configure: Regenerated.
2699 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2700 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2701 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2703 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
2705 * configure: Regenerate.
2707 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
2709 PR rtl-optimization/91333
2710 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2711 reg preferences comparison up.
2713 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2715 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2716 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2717 aarch64-sve-builtins-base.h.
2718 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2719 aarch64-sve-builtins-base.cc.
2720 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2721 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2723 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2724 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2725 (svcvtnt): New functions.
2726 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2727 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2728 (svcvtnt): New functions.
2729 (svcvt): Add a form that converts f32 to bf16.
2730 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2731 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2733 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2734 Treat B as bfloat16_t.
2735 (ternary_bfloat_lane_base): New class.
2736 (ternary_bfloat_def): Likewise.
2737 (ternary_bfloat): New shape.
2738 (ternary_bfloat_lane_def): New class.
2739 (ternary_bfloat_lane): New shape.
2740 (ternary_bfloat_lanex2_def): New class.
2741 (ternary_bfloat_lanex2): New shape.
2742 (ternary_bfloat_opt_n_def): New class.
2743 (ternary_bfloat_opt_n): New shape.
2744 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2745 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2746 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2747 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2748 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2749 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2750 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2751 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2752 the pattern off the narrow mode instead of the wider one.
2753 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2754 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2755 (sve_fp_op): Handle them.
2756 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2757 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2759 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2761 * config/aarch64/arm_sve.h: Include arm_bf16.h.
2762 * config/aarch64/aarch64-modes.def (BF): Move definition before
2763 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
2764 (SVE_MODES): Handle BF modes.
2765 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2767 (aarch64_full_sve_mode): Likewise.
2768 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2770 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2771 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2772 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2773 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2775 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2777 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2778 (TYPES_all_data): Add bf16.
2779 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2780 (register_tuple_type): Increase buffer size.
2781 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2782 (bf16): New type suffix.
2783 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2784 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2785 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2786 Change type from all_data to all_arith.
2787 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2790 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
2791 Matthew Malcomson <matthew.malcomson@arm.com>
2792 Richard Sandiford <richard.sandiford@arm.com>
2794 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2795 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2796 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2797 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
2798 __ARM_FEATURE_MATMUL_FP64.
2799 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2800 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2801 be disabled at the same time.
2802 (f32mm): New extension.
2803 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2804 (AARCH64_FL_F64MM): Bump to the next bit up.
2805 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2806 (TARGET_SVE_F64MM): New macros.
2807 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2808 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2809 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2810 (UNSPEC_ZIP2Q): New unspeccs.
2811 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2812 (optab, sur, perm_insn): Handle the new unspecs.
2813 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
2814 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2815 TARGET_SVE_F64MM instead of separate tests.
2816 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2817 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2818 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2819 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2820 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2821 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2822 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2823 (TYPES_s_signed): New macro.
2824 (TYPES_s_integer): Use it.
2825 (TYPES_d_float): New macro.
2826 (TYPES_d_data): Use it.
2827 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2828 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2829 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2830 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2831 (svmmla): New shape.
2832 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2833 template parameters.
2834 (ternary_resize2_lane_base): Likewise.
2835 (ternary_resize2_base): New class.
2836 (ternary_qq_lane_base): Likewise.
2837 (ternary_intq_uintq_lane_def): Likewise.
2838 (ternary_intq_uintq_lane): New shape.
2839 (ternary_intq_uintq_opt_n_def): New class
2840 (ternary_intq_uintq_opt_n): New shape.
2841 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2842 (ternary_uintq_intq_def): New class.
2843 (ternary_uintq_intq): New shape.
2844 (ternary_uintq_intq_lane_def): New class.
2845 (ternary_uintq_intq_lane): New shape.
2846 (ternary_uintq_intq_opt_n_def): New class.
2847 (ternary_uintq_intq_opt_n): New shape.
2848 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2849 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2850 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2851 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2853 (svdotprod_lane_impl): ...this new class.
2854 (svmmla_impl, svusdot_impl): New classes.
2855 (svdot_lane): Update to use svdotprod_lane_impl.
2856 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2857 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2859 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2860 function, with no types defined.
2861 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2862 AARCH64_FL_I8MM functions.
2863 (svmmla): New AARCH64_FL_F32MM function.
2864 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2865 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2866 AARCH64_FL_F64MM function.
2867 (REQUIRED_EXTENSIONS):
2869 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2871 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2874 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
2876 * config/i386/i386.md (*movoi_internal_avx): Do not check for
2877 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
2878 (*movti_internal): Do not check for
2879 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2880 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2881 just after check for TARGET_AVX.
2882 (*movdf_internal): Ditto.
2883 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2884 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2885 * config/i386/sse.md (mov<mode>_internal): Only check
2886 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
2887 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2888 (<sse>_andnot<mode>3<mask_name>): Move check for
2889 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2890 (<code><mode>3<mask_name>): Ditto.
2891 (*andnot<mode>3): Ditto.
2892 (*andnottf3): Ditto.
2893 (*<code><mode>3): Ditto.
2894 (*<code>tf3): Ditto.
2895 (*andnot<VI:mode>3): Remove
2896 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2897 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2898 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2899 (sse4_1_blendv<ssemodesuffix>): Ditto.
2900 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2901 Explain that tune applies to 128bit instructions only.
2903 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
2905 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2906 to definition of hsa_kernel_description. Parse assembly to find SGPR
2907 and VGPR count of kernel and store in hsa_kernel_description.
2909 2020-01-31 Tamar Christina <tamar.christina@arm.com>
2911 PR rtl-optimization/91838
2912 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
2913 to truncate if allowed or reject combination.
2915 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2917 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
2918 (find_inv_vars_cb): Likewise.
2920 2020-01-31 David Malcolm <dmalcolm@redhat.com>
2922 * calls.c (special_function_p): Split out the check for DECL_NAME
2923 being non-NULL and fndecl being extern at file scope into a
2924 new maybe_special_function_p and call it. Drop check for fndecl
2925 being non-NULL that was after a usage of DECL_NAME (fndecl).
2926 * tree.h (maybe_special_function_p): New inline function.
2928 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2930 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2931 (mask_gather_load<mode>): ... here, and zero-initialize the
2933 (maskload<mode>di): Zero-initialize the destination.
2936 2020-01-30 David Malcolm <dmalcolm@redhat.com>
2939 * doc/analyzer.texi (Limitations): Note that constraints on
2940 floating-point values are currently ignored.
2942 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2945 * symtab.c (symtab_node::noninterposable_alias): If localalias
2946 already exists, but is not usable, append numbers after it until
2947 a unique name is found. Formatting fix.
2950 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2953 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2955 * config/gcn/gcn.c (print_operand): Handle LTGT.
2956 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2958 2020-01-30 Richard Biener <rguenther@suse.de>
2960 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2961 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2963 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
2965 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2966 without a DECL in .data.rel.ro.local.
2968 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2971 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2975 * config/i386/sse.md
2976 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2977 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
2978 any_extend code iterator instead of always zero_extend.
2979 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2980 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2981 Use any_extend code iterator instead of always zero_extend.
2982 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2983 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2984 Use any_extend code iterator instead of always zero_extend.
2985 (*sse2_pmovmskb_ext): New define_insn.
2986 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2989 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2990 (*popcountsi2_zext_falsedep): New define_insn.
2992 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2994 * config.in: Regenerated.
2995 * configure: Regenerated.
2997 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
3000 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
3001 LLVM's assembler changed the default in version 9.
3003 2020-01-24 Jeff Law <law@redhat.com>
3005 PR tree-optimization/89689
3006 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
3008 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
3012 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3014 PR rtl-optimization/87763
3015 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
3016 simplification to handle subregs as well as bare regs.
3017 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
3019 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
3022 * ira.c (ira): Revert use of simplified LRA algorithm.
3024 2020-01-29 Martin Jambor <mjambor@suse.cz>
3026 PR tree-optimization/92706
3027 * tree-sra.c (struct access): Fields first_link, last_link,
3028 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
3029 next_rhs_queued and grp_rhs_queued respectively, new fields
3030 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
3031 (struct assign_link): Field next renamed to next_rhs, new field
3032 next_lhs. Updated comment.
3033 (work_queue_head): Renamed to rhs_work_queue_head.
3034 (lhs_work_queue_head): New variable.
3035 (add_link_to_lhs): New function.
3036 (relink_to_new_repr): Also relink LHS lists.
3037 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
3038 (add_access_to_lhs_work_queue): New function.
3039 (pop_access_from_work_queue): Renamed to
3040 pop_access_from_rhs_work_queue.
3041 (pop_access_from_lhs_work_queue): New function.
3042 (build_accesses_from_assign): Also add links to LHS lists and to LHS
3044 (child_would_conflict_in_lacc): Renamed to
3045 child_would_conflict_in_acc. Adjusted parameter names.
3046 (create_artificial_child_access): New parameter set_grp_read, use it.
3047 (subtree_mark_written_and_enqueue): Renamed to
3048 subtree_mark_written_and_rhs_enqueue.
3049 (propagate_subaccesses_across_link): Renamed to
3050 propagate_subaccesses_from_rhs.
3051 (propagate_subaccesses_from_lhs): New function.
3052 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
3055 2020-01-29 Martin Jambor <mjambor@suse.cz>
3057 PR tree-optimization/92706
3058 * tree-sra.c (struct access): Adjust comment of
3059 grp_total_scalarization.
3060 (find_access_in_subtree): Look for single children spanning an entire
3062 (scalarizable_type_p): Allow register accesses, adjust callers.
3063 (completely_scalarize): Remove function.
3064 (scalarize_elem): Likewise.
3065 (create_total_scalarization_access): Likewise.
3066 (sort_and_splice_var_accesses): Do not track total scalarization
3068 (analyze_access_subtree): New parameter totally, adjust to new meaning
3069 of grp_total_scalarization.
3070 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
3071 (can_totally_scalarize_forest_p): New function.
3072 (create_total_scalarization_access): Likewise.
3073 (create_total_access_and_reshape): Likewise.
3074 (total_should_skip_creating_access): Likewise.
3075 (totally_scalarize_subtree): Likewise.
3076 (analyze_all_variable_accesses): Perform total scalarization after
3077 subaccess propagation using the new functions above.
3078 (initialize_constant_pool_replacements): Output initializers by
3079 traversing the access tree.
3081 2020-01-29 Martin Jambor <mjambor@suse.cz>
3083 * tree-sra.c (verify_sra_access_forest): New function.
3084 (verify_all_sra_access_forests): Likewise.
3085 (create_artificial_child_access): Set parent.
3086 (analyze_all_variable_accesses): Call the verifier.
3088 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3090 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
3091 if called on indirect edge.
3092 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
3093 speculative call if needed.
3095 2020-01-29 Richard Biener <rguenther@suse.de>
3097 PR tree-optimization/93428
3098 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
3099 permutation when the load node is created.
3100 (vect_analyze_slp_instance): Re-use it here.
3102 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3104 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
3106 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
3108 PR rtl-optimization/93272
3109 * ira-lives.c (process_out_of_region_eh_regs): New function.
3110 (process_bb_node_lives): Call it.
3112 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3114 * coverage.c (read_counts_file): Make error message lowercase.
3116 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3118 * profile-count.c (profile_quality_display_names): Fix ordering.
3120 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
3123 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
3124 hash only when edge is first within the sequence.
3125 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
3126 (symbol_table::create_edge): Do not set target_prob.
3127 (cgraph_edge::remove_caller): Watch for speculative calls when updating
3129 (cgraph_edge::make_speculative): Drop target_prob parameter.
3130 (cgraph_edge::speculative_call_info): Remove.
3131 (cgraph_edge::first_speculative_call_target): New member function.
3132 (update_call_stmt_hash_for_removing_direct_edge): New function.
3133 (cgraph_edge::resolve_speculation): Rewrite to new API.
3134 (cgraph_edge::speculative_call_for_target): New member function.
3135 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
3136 multiple speculation targets.
3137 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
3139 (verify_speculative_call): Verify that targets form an interval.
3140 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
3141 (cgraph_edge::first_speculative_call_target): New member function.
3142 (cgraph_edge::next_speculative_call_target): New member function.
3143 (cgraph_edge::speculative_call_target_ref): New member function.
3144 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
3145 (cgraph_edge): Remove target_prob.
3146 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3147 Fix handling of speculative calls.
3148 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
3149 * ipa-fnsummary.c (analyze_function_body): Likewise.
3150 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
3151 * ipa-profile.c (dump_histogram): Fix formating.
3152 (ipa_profile_generate_summary): Watch for overflows.
3153 (ipa_profile): Do not require probablity to be 1/2; update to new API.
3154 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
3155 (update_indirect_edges_after_inlining): Update to new API.
3156 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
3158 * profile-count.h: (profile_probability::adjusted): New.
3159 * tree-inline.c (copy_bb): Update to new speculative call API; fix
3160 updating of profile.
3161 * value-prof.c (gimple_ic_transform): Rename to ...
3162 (dump_ic_profile): ... this one; update dumping.
3163 (stream_in_histogram_value): Fix formating.
3164 (gimple_value_profile_transformations): Update.
3166 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
3169 * config/i386/i386.md (*movoi_internal_avx): Remove
3170 TARGET_SSE_TYPELESS_STORES check.
3171 (*movti_internal): Prefer TARGET_AVX over
3172 TARGET_SSE_TYPELESS_STORES.
3173 (*movtf_internal): Likewise.
3174 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
3175 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
3176 from TARGET_SSE_TYPELESS_STORES.
3178 2020-01-28 David Malcolm <dmalcolm@redhat.com>
3180 * diagnostic-core.h (warning_at): Rename overload to...
3181 (warning_meta): ...this.
3182 (emit_diagnostic_valist): Delete decl of overload taking
3183 diagnostic_metadata.
3184 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
3185 (warning_at): Rename overload taking diagnostic_metadata to...
3186 (warning_meta): ...this.
3188 2020-01-28 Richard Biener <rguenther@suse.de>
3190 PR tree-optimization/93439
3191 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
3192 * tree-cfg.c (move_sese_region_to_fn): ... here.
3193 (verify_types_in_gimple_reference): Verify used cliques are
3196 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
3199 * config/i386/i386-options.c (set_ix86_tune_features): Add an
3200 argument of a pointer to struct gcc_options and pass it to
3201 parse_mtune_ctrl_str.
3202 (ix86_function_specific_restore): Pass opts to
3203 set_ix86_tune_features.
3204 (ix86_option_override_internal): Likewise.
3205 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
3206 gcc_options and use it for x_ix86_tune_ctrl_string.
3208 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3210 PR rtl-optimization/87763
3211 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
3212 simplification to handle subregs as well as bare regs.
3213 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
3215 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3217 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
3218 for reduction chains that (now) include a call.
3220 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3222 PR tree-optimization/92822
3223 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
3224 out the don't-care elements of a vector whose significant elements
3225 are duplicates, make the don't-care elements duplicates too.
3227 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
3229 PR tree-optimization/93434
3230 * tree-predcom.c (split_data_refs_to_components): Record which
3231 components have had aliasing loads removed. Prevent store-store
3232 commoning for all such components.
3234 2020-01-28 Jakub Jelinek <jakub@redhat.com>
3237 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
3238 -1 or is_vshift is true, use new_vector with number of elts npatterns
3239 rather than new_unary_operation.
3241 PR tree-optimization/93454
3242 * gimple-fold.c (fold_array_ctor_reference): Perform
3243 elt_size.to_uhwi () just once, instead of calling it in every
3244 iteration. Punt if that value is above size of the temporary
3245 buffer. Decrease third native_encode_expr argument when
3246 bufoff + elt_sz is above size of buf.
3248 2020-01-27 Joseph Myers <joseph@codesourcery.com>
3250 * config/mips/mips.c (mips_declare_object_name)
3251 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
3253 2020-01-27 Martin Liska <mliska@suse.cz>
3255 PR gcov-profile/93403
3256 * tree-profile.c (gimple_init_gcov_profiler): Generate
3257 both __gcov_indirect_call_profiler_v4 and
3258 __gcov_indirect_call_profiler_v4_atomic.
3260 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3263 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
3265 (@aarch64_split_simd_mov<mode>): Use it.
3266 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
3267 Leave the vec_extract patterns to handle 2-element vectors.
3268 (aarch64_simd_mov_from_<mode>high): Likewise.
3269 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
3270 (vec_extractv2dfv1df): Likewise.
3272 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3274 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
3275 jump conditions for *compare_condjump<GPI:mode>.
3277 2020-01-27 David Malcolm <dmalcolm@redhat.com>
3280 * digraph.cc (test_edge::test_edge): Specify template for base
3283 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
3285 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
3287 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
3289 * config/arc/arc-protos.h (gen_mlo): Remove.
3290 (gen_mhi): Likewise.
3291 * config/arc/arc.c (AUX_MULHI): Define.
3292 (arc_must_save_reister): Special handling for r58/59.
3293 (arc_compute_frame_size): Consider mlo/mhi registers.
3294 (arc_save_callee_saves): Emit fp/sp move only when emit_move
3296 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
3297 mlo/mhi name selection.
3298 (arc_restore_callee_saves): Don't early restore blink when ISR.
3299 (arc_expand_prologue): Add mlo/mhi saving.
3300 (arc_expand_epilogue): Add mlo/mhi restoring.
3303 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
3304 numbering when MUL64 option is used.
3305 (DWARF2_FRAME_REG_OUT): Define.
3306 * config/arc/arc.md (arc600_stall): New pattern.
3307 (VUNSPEC_ARC_ARC600_STALL): Define.
3308 (mulsi64): Use correct mlo/mhi registers.
3309 (mulsi_600): Clean it up.
3310 * config/arc/predicates.md (mlo_operand): Remove any dependency on
3312 (mhi_operand): Likewise.
3314 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
3315 Petro Karashchenko <petro.karashchenko@ring.com>
3317 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
3318 attributes if needed.
3319 (prepare_move_operands): Generate special unspec instruction for
3321 (arc_isuncached_mem_p): Propagate uncached attribute to each
3323 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
3324 (VUNSPEC_ARC_STDI): Likewise.
3325 (ALLI): New mode iterator.
3326 (mALLI): New mode attribute.
3327 (lddi): New instruction pattern.
3329 (stdidi_split): Split instruction for architectures which are not
3330 supporting ll64 option.
3331 (lddidi_split): Likewise.
3333 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3335 PR rtl-optimization/92989
3336 * lra-lives.c (process_bb_lives): Update the live-in set before
3337 processing additional clobbers.
3339 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3341 PR rtl-optimization/93170
3342 * cselib.c (cselib_invalidate_regno_val): New function, split out
3344 (cselib_invalidate_regno): ...here.
3345 (cselib_invalidated_by_call_p): New function.
3346 (cselib_process_insn): Iterate over all the hard-register entries in
3347 REG_VALUES and invalidate any that cross call-clobbered registers.
3349 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
3351 * dojump.c (split_comparison): Use HONOR_NANS rather than
3352 HONOR_SNANS when splitting LTGT.
3354 2020-01-27 Martin Liska <mliska@suse.cz>
3357 * opts.c (print_filtered_help): Exclude language-specific
3358 options from --help=common unless enabled in all FEs.
3360 2020-01-27 Martin Liska <mliska@suse.cz>
3362 * opts.c (print_help): Exclude params from
3363 all except --help=param.
3365 2020-01-27 Martin Liska <mliska@suse.cz>
3368 * config/i386/i386-features.c (make_resolver_func):
3369 Align the code with ppc64 target implementation.
3370 Do not generate a unique name for resolver function.
3372 2020-01-27 Richard Biener <rguenther@suse.de>
3374 PR tree-optimization/93397
3375 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
3376 converted reduction chain SLP graph adjustment.
3378 2020-01-26 Marek Polacek <polacek@redhat.com>
3381 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
3384 2020-01-26 Jason Merrill <jason@redhat.com>
3387 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
3390 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
3392 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
3393 (rx_setmem): Likewise.
3395 2020-01-26 Jakub Jelinek <jakub@redhat.com>
3398 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
3399 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
3400 drop <di> from constraint of last operand.
3403 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
3404 TARGET_AVX2 and V4DFmode not in the split condition, but in the
3405 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
3407 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
3410 * ipa-cp.c (get_info_about_necessary_edges): Remove value
3413 2020-01-24 Jeff Law <law@redhat.com>
3415 PR tree-optimization/92788
3416 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
3419 2020-01-24 Jakub Jelinek <jakub@redhat.com>
3422 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
3423 *avx_vperm_broadcast_<mode>,
3424 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
3425 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
3426 Move before avx2_perm<mode>/avx512f_perm<mode>.
3429 * simplify-rtx.c (simplify_const_unary_operation,
3430 simplify_const_binary_operation): Punt for mode precision above
3431 MAX_BITSIZE_MODE_ANY_INT.
3433 2020-01-24 Andrew Pinski <apinski@marvell.com>
3435 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
3438 2020-01-24 Jeff Law <law@redhat.com>
3441 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
3442 for REGs. Call output_operand_lossage to get more reasonable
3445 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
3447 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
3448 gcn_fp_compare_operator.
3449 (vec_cmpu<mode>di): Use gcn_compare_operator.
3450 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
3451 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
3452 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
3453 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
3454 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
3455 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
3456 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
3457 gcn_fp_compare_operator.
3458 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
3459 gcn_fp_compare_operator.
3460 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
3461 gcn_fp_compare_operator.
3462 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
3463 gcn_fp_compare_operator.
3465 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
3467 * doc/install.texi (Cross-Compiler-Specific Options): Document
3468 `--with-toolexeclibdir' option.
3470 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
3472 * target.def (flags_regnum): Also mention effect on delay slot filling.
3473 * doc/tm.texi: Regenerate.
3475 2020-01-23 Jeff Law <law@redhat.com>
3477 PR translation/90162
3478 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
3480 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
3483 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
3486 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3488 PR rtl-optimization/93402
3489 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
3492 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3494 * config.in: Regenerated.
3495 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
3496 for TARGET_LIBC_GNUSTACK.
3497 * configure: Regenerated.
3498 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
3499 found to be 2.31 or greater.
3501 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3503 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
3505 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
3506 (mips_asm_file_end): New function. Delegate to
3507 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
3508 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3510 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3513 * config/i386/i386-modes.def (POImode): New mode.
3514 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
3515 * config/i386/i386.md (DPWI): New mode attribute.
3516 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
3518 (QPWI): ... this. Use POI instead of OI for TImode.
3519 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
3520 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
3523 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3526 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
3528 (speculation_tracker_rev): New pattern.
3529 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
3530 Use speculation_tracker_rev to track the inverse condition.
3532 2020-01-23 Richard Biener <rguenther@suse.de>
3534 PR tree-optimization/93381
3535 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
3536 alias-set of the def as argument and record the first one.
3537 (vn_walk_cb_data::first_set): New member.
3538 (vn_reference_lookup_3): Pass the alias-set of the current def
3539 to push_partial_def. Fix alias-set used in the aggregate copy
3541 (vn_reference_lookup): Consistently set *last_vuse_ptr.
3542 * real.c (clear_significand_below): Fix out-of-bound access.
3544 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3547 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
3548 New define_insn patterns.
3550 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3552 * doc/sourcebuild.texi (check-function-bodies): Add an
3553 optional target/xfail selector.
3555 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3557 PR rtl-optimization/93124
3558 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
3559 bare USE and CLOBBER insns.
3561 2020-01-22 Andrew Pinski <apinski@marvell.com>
3563 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
3565 2020-01-22 David Malcolm <dmalcolm@redhat.com>
3568 * gdbinit.in (break-on-saved-diagnostic): Update for move of
3569 diagnostic_manager into "ana" namespace.
3570 * selftest-run-tests.c (selftest::run_tests): Update for move of
3571 selftest::run_analyzer_selftests to
3572 ana::selftest::run_analyzer_selftests.
3574 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
3576 * cfgexpand.c (union_stack_vars): Update the size.
3578 2020-01-22 Richard Biener <rguenther@suse.de>
3580 PR tree-optimization/93381
3581 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
3582 throughout, handle all conversions the same.
3584 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3587 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
3588 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
3589 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
3590 Call force_reg on high_in2 unconditionally.
3592 2020-01-22 Martin Liska <mliska@suse.cz>
3594 PR tree-optimization/92924
3595 * profile.c (compute_value_histograms): Divide
3598 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3601 * output.h (assemble_name_resolve): Declare.
3602 * varasm.c (assemble_name_resolve): New function.
3603 (assemble_name): Use it.
3604 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
3606 2020-01-22 Joseph Myers <joseph@codesourcery.com>
3608 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
3609 update_web_docs_git instead of update_web_docs_svn.
3611 2020-01-21 Andrew Pinski <apinski@marvell.com>
3614 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
3615 as PTR mode. Have operand 1 as being modeless, it can be P mode.
3616 (*tlsgd_small_<mode>): Likewise.
3617 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
3618 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
3619 register. Convert that register back to dest using convert_mode.
3621 2020-01-21 Jim Wilson <jimw@sifive.com>
3623 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
3626 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
3627 Uros Bizjak <ubizjak@gmail.com>
3630 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3632 (legitimize_tls_address): Do GNU2 TLS address computation in
3633 ptr_mode and zero-extend result to Pmode.
3634 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3635 :P with :PTR and Pmode with ptr_mode.
3636 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3637 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3638 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3640 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3643 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3644 the last two operands are CONST_INT_P before using them as such.
3646 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3648 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3649 to get the integer element types.
3651 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3653 * config/aarch64/aarch64-sve-builtins.h
3654 (function_expander::convert_to_pmode): Declare.
3655 * config/aarch64/aarch64-sve-builtins.cc
3656 (function_expander::convert_to_pmode): New function.
3657 (function_expander::get_contiguous_base): Use it.
3658 (function_expander::prepare_gather_address_operands): Likewise.
3659 * config/aarch64/aarch64-sve-builtins-sve2.cc
3660 (svwhilerw_svwhilewr_impl::expand): Likewise.
3662 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
3665 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3666 cfun->machine->label_is_assembled.
3667 (aarch64_print_patchable_function_entry): New.
3668 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3669 * config/aarch64/aarch64.h (struct machine_function): New field,
3672 2020-01-21 David Malcolm <dmalcolm@redhat.com>
3675 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3678 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3681 * cgraph.c (cgraph_edge::resolve_speculation,
3682 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3683 call_stmt_site_hash.
3685 2020-01-21 Martin Liska <mliska@suse.cz>
3687 * config/rs6000/rs6000.c (common_mode_defined): Remove
3690 2020-01-21 Richard Biener <rguenther@suse.de>
3692 PR tree-optimization/92328
3693 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3694 type when value-numbering same-sized store by inserting a
3696 (eliminate_dom_walker::eliminate_stmt): When eliminating
3697 a redundant store handle bit-reinterpretation of the same value.
3699 2020-01-21 Andrew Pinski <apinski@marvel.com>
3702 * tree-into-ssa.c (prepare_block_for_update_1): Split out
3704 (prepare_block_for_update): This. Use a worklist instead of
3707 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3709 * gcc/config/arm/arm.c (clear_operation_p):
3710 Initialise last_regno, skip first iteration
3711 based on the first_set value and use ints instead
3712 of the unnecessary HOST_WIDE_INTs.
3714 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3717 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3718 compare_mode other than SFmode or DFmode.
3720 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
3723 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3724 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3725 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3727 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3729 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3731 2020-01-20 Andrew Pinski <apinski@marvell.com>
3734 * targhooks.c (default_print_patchable_function_entry): Use
3735 output_asm_insn to emit the nop instruction.
3737 2020-01-20 Fangrui Song <maskray@google.com>
3740 * targhooks.c (default_print_patchable_function_entry): Align to
3743 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
3746 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3747 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
3748 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3749 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
3750 (*tls_dynamic_gnu2_lea_64): Renamed to ...
3751 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
3752 Remove the {q} suffix from lea.
3753 (*tls_dynamic_gnu2_call_64): Renamed to ...
3754 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
3755 (*tls_dynamic_gnu2_combine_64): Renamed to ...
3756 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
3757 Pass Pmode to gen_tls_dynamic_gnu2_64.
3759 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3761 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3763 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
3765 * config/aarch64/aarch64-sve-builtins-base.cc
3766 (svld1ro_impl::memory_vector_mode): Remove parameter name.
3768 2020-01-20 Richard Biener <rguenther@suse.de>
3771 * dwarf2out.c (prune_unused_types): Unconditionally mark
3772 called function DIEs.
3774 2020-01-20 Martin Liska <mliska@suse.cz>
3776 PR tree-optimization/93199
3777 * tree-eh.c (struct leh_state): Add
3778 new field outer_non_cleanup.
3779 (cleanup_is_dead_in): Pass leh_state instead
3780 of eh_region. Add a checking that state->outer_non_cleanup
3781 points to outer non-clean up region.
3782 (lower_try_finally): Record outer_non_cleanup
3784 (lower_catch): Likewise.
3785 (lower_eh_filter): Likewise.
3786 (lower_eh_must_not_throw): Likewise.
3787 (lower_cleanup): Likewise.
3789 2020-01-20 Richard Biener <rguenther@suse.de>
3791 PR tree-optimization/93094
3792 * tree-vectorizer.h (vect_loop_versioning): Adjust.
3793 (vect_transform_loop): Likewise.
3794 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3795 loop_vectorized_call to vect_transform_loop.
3796 * tree-vect-loop.c (vect_transform_loop): Pass down
3797 loop_vectorized_call to vect_loop_versioning.
3798 * tree-vect-loop-manip.c (vect_loop_versioning): Use
3799 the earlier discovered loop_vectorized_call.
3801 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
3803 * doc/contribute.texi: Update for SVN -> Git transition.
3804 * doc/install.texi: Likewise.
3806 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3809 * cgraph.c (cgraph_edge::make_speculative): Increase number of
3810 speculative targets.
3811 (verify_speculative_call): New function
3812 (cgraph_node::verify_node): Use it.
3813 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3816 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3819 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3820 (cgraph_edge::make_direct): Remove all indirect targets.
3821 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3822 (cgraph_node::verify_node): Verify that only one call_stmt or
3823 lto_stmt_uid is set.
3824 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3826 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3827 (lto_output_ref): Simplify streaming of stmt.
3828 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3830 2020-01-18 Tamar Christina <tamar.christina@arm.com>
3832 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3833 Mark parameter unused.
3835 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
3837 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3839 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
3841 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3843 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
3845 * Makefile.in: Add coroutine-passes.o.
3846 * builtin-types.def (BT_CONST_SIZE): New.
3847 (BT_FN_BOOL_PTR): New.
3848 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3849 * builtins.def (DEF_COROUTINE_BUILTIN): New.
3850 * coroutine-builtins.def: New file.
3851 * coroutine-passes.cc: New file.
3852 * function.h (struct GTY function): Add a bit to indicate that the
3853 function is a coroutine component.
3854 * internal-fn.c (expand_CO_FRAME): New.
3855 (expand_CO_YIELD): New.
3856 (expand_CO_SUSPN): New.
3857 (expand_CO_ACTOR): New.
3858 * internal-fn.def (CO_ACTOR): New.
3862 * passes.def: Add pass_coroutine_lower_builtins,
3863 pass_coroutine_early_expand_ifns.
3864 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3865 (make_pass_coroutine_early_expand_ifns): New.
3866 * doc/invoke.texi: Document the fcoroutines command line
3869 2020-01-18 Jakub Jelinek <jakub@redhat.com>
3871 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3874 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3875 after checking the argument is a REG. Don't use REGNO (reg)
3876 again to set last_regno, reuse regno variable instead.
3878 2020-01-17 David Malcolm <dmalcolm@redhat.com>
3880 * doc/analyzer.texi (Limitations): Add note about NaN.
3882 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3883 Sudakshina Das <sudi.das@arm.com>
3885 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3886 and valid immediate.
3887 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3888 (lshrdi3): Generate thumb2_lsrl for valid immediates.
3889 * config/arm/constraints.md (Pg): New.
3890 * config/arm/predicates.md (long_shift_imm): New.
3891 (arm_reg_or_long_shift_imm): Likewise.
3892 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3893 (thumb2_lsll): Likewise.
3896 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3897 Sudakshina Das <sudi.das@arm.com>
3899 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3900 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3901 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3902 register pairs for doubleword quantities for ARMv8.1M-Mainline.
3903 * config/arm/thumb2.md (thumb2_asrl): New.
3904 (thumb2_lsll): Likewise.
3906 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3908 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
3911 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
3913 * gdbinit.in (help-gcc-hooks): New command.
3914 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
3915 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
3918 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3920 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3921 correct target macro.
3923 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3925 * config/aarch64/aarch64-protos.h
3926 (aarch64_sve_ld1ro_operand_p): New.
3927 * config/aarch64/aarch64-sve-builtins-base.cc
3928 (class load_replicate): New.
3929 (class svld1ro_impl): New.
3930 (class svld1rq_impl): Change to inherit from load_replicate.
3931 (svld1ro): New sve intrinsic function base.
3932 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3933 New DEF_SVE_FUNCTION.
3934 * config/aarch64/aarch64-sve-builtins-base.h
3935 (svld1ro): New decl.
3936 * config/aarch64/aarch64-sve-builtins.cc
3937 (function_expander::add_mem_operand): Modify assert to allow
3939 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3941 * config/aarch64/aarch64.c
3942 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3943 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3944 (aarch64_sve_ld1ro_operand_p): New.
3945 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3946 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3947 * config/aarch64/predicates.md
3948 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3950 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3952 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3953 Introduce this ACLE specified predefined macro.
3954 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3955 (fp): Disabling this disables f64mm.
3956 (simd): Disabling this disables f64mm.
3957 (fp16): Disabling this disables f64mm.
3958 (sve): Disabling this disables f64mm.
3959 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3960 (AARCH64_ISA_F64MM): New.
3961 (TARGET_F64MM): New.
3962 * doc/invoke.texi (f64mm): Document new option.
3964 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3966 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3967 (neoversen1_tunings): Likewise.
3969 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3972 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3973 Add assert to ensure prolog has been emitted.
3974 (aarch64_split_atomic_op): Likewise.
3975 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3976 Use epilogue_completed rather than reload_completed.
3977 (aarch64_atomic_exchange<mode>): Likewise.
3978 (aarch64_atomic_<atomic_optab><mode>): Likewise.
3979 (atomic_nand<mode>): Likewise.
3980 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3981 (atomic_fetch_nand<mode>): Likewise.
3982 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3983 (atomic_nand_fetch<mode>): Likewise.
3985 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3988 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3990 (REVERSE_CONDITION): Delete.
3991 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3992 (CCFP_CCFPE): Likewise.
3993 (e): New mode attribute.
3994 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3995 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3996 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3997 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3998 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3999 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
4000 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
4001 name of generator from gen_ccmpdi to gen_ccmpccdi.
4002 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
4003 the previous comparison but aren't able to, use the new ccmp_rev
4006 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
4008 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
4009 than testing directly for INTEGER_CST.
4010 (gimplify_target_expr, gimplify_omp_depend): Likewise.
4012 2020-01-17 Jakub Jelinek <jakub@redhat.com>
4014 PR tree-optimization/93292
4015 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
4016 get_vectype_for_scalar_type returns NULL.
4018 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
4020 * params.opt (-param=max-predicted-iterations): Increase range from 0.
4021 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
4023 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
4025 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
4027 * params.opt: (max-predicted-iterations): Set bounds.
4028 * predict.c (real_almost_one, real_br_prob_base,
4029 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
4030 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
4031 probabilities; do not truncate to reg_br_prob_bases.
4032 (estimate_loops_at_level): Pass max_cyclic_prob.
4033 (estimate_loops): Compute max_cyclic_prob.
4034 (estimate_bb_frequencies): Do not initialize real_*; update calculation
4036 * profile-count.c (profile_probability::to_sreal): New.
4037 * profile-count.h (class sreal): Move up in file.
4038 (profile_probability::to_sreal): Declare.
4040 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4043 (arm_invalid_conversion): New function for target hook.
4044 (arm_invalid_unary_op): New function for target hook.
4045 (arm_invalid_binary_op): New function for target hook.
4047 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4049 * config.gcc: Add arm_bf16.h.
4050 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
4051 (arm_simd_builtin_std_type): Add BFmode.
4052 (arm_init_simd_builtin_types): Define element types for vector types.
4053 (arm_init_bf16_types): New function.
4054 (arm_init_builtins): Add arm_init_bf16_types function call.
4055 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
4056 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
4057 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
4058 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
4059 (arm_vector_mode_supported_p): Add V4BF, V8BF.
4060 (arm_mangle_type): Add __bf16.
4061 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
4062 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
4063 arm_bf16_ptr_type_node.
4064 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
4065 define_split between ARM registers.
4066 * config/arm/arm_bf16.h: New file.
4067 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4068 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
4070 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
4071 * config/arm/vfp.md: Add BFmode to movhf patterns.
4073 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
4074 Andre Vieira <andre.simoesdiasvieira@arm.com>
4076 * config/arm/arm-cpus.in (mve, mve_float): New features.
4077 (dsp, mve, mve.fp): New options.
4078 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
4079 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
4080 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
4082 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4083 Thomas Preud'homme <thomas.preudhomme@arm.com>
4085 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
4087 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
4088 error for using -mcmse when targeting Armv8.1-M Mainline.
4090 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4091 Thomas Preud'homme <thomas.preudhomme@arm.com>
4093 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
4094 address in r4 when targeting Armv8.1-M Mainline.
4095 (nonsecure_call_value_internal): Likewise.
4096 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
4097 a register match_operand again. Emit BLXNS when targeting
4099 (nonsecure_call_value_reg_thumb2): Likewise.
4101 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4102 Thomas Preud'homme <thomas.preudhomme@arm.com>
4104 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
4105 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
4106 variable as true when floating-point ABI is not hard. Replace
4107 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
4108 Generate VLSTM and VLLDM instruction respectively before and
4109 after a function call to cmse_nonsecure_call function.
4110 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
4111 (VUNSPEC_VLLDM): Likewise.
4112 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
4113 (lazy_load_multiple_insn): Likewise.
4115 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4116 Thomas Preud'homme <thomas.preudhomme@arm.com>
4118 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
4119 (arm_emit_vfp_multi_reg_pop): Likewise.
4120 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
4121 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
4122 restore callee-saved VFP registers.
4124 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4125 Thomas Preud'homme <thomas.preudhomme@arm.com>
4127 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
4128 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
4129 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
4130 callee-saved GPRs as well as clear ip register before doing a nonsecure
4131 call then restore callee-saved GPRs after it when targeting
4133 (arm_reorg): Adapt to function rename.
4135 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4136 Thomas Preud'homme <thomas.preudhomme@arm.com>
4138 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
4139 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
4140 clear_vfp_multiple pattern based on a new vfp parameter.
4141 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
4142 targeting Armv8.1-M Mainline.
4143 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
4144 unconditionally when targeting Armv8.1-M Mainline architecture. Check
4145 whether VFP registers are available before looking call_used_regs for a
4147 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
4148 of prototype of clear_operation_p.
4149 (clear_vfp_multiple_operation): New predicate.
4150 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
4151 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
4153 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4154 Thomas Preud'homme <thomas.preudhomme@arm.com>
4156 * config/arm/arm-protos.h (clear_operation_p): Declare.
4157 * config/arm/arm.c (clear_operation_p): New function.
4158 (cmse_clear_registers): Generate clear_multiple instruction pattern if
4159 targeting Armv8.1-M Mainline or successor.
4160 (output_return_instruction): Only output APSR register clearing if
4161 Armv8.1-M Mainline instructions not available.
4162 (thumb_exit): Likewise.
4163 * config/arm/predicates.md (clear_multiple_operation): New predicate.
4164 * config/arm/thumb2.md (clear_apsr): New define_insn.
4165 (clear_multiple): Likewise.
4166 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
4168 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4169 Thomas Preud'homme <thomas.preudhomme@arm.com>
4171 * config/arm/arm.c (fp_sysreg_names): Declare and define.
4172 (use_return_insn): Also return false for Armv8.1-M Mainline.
4173 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
4174 Mainline instructions are available.
4175 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
4176 when targeting Armv8.1-M Mainline Security Extensions.
4177 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
4178 Mainline entry function.
4179 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
4180 targeting Armv8.1-M Mainline or successor.
4181 (arm_expand_epilogue): Fix indentation of caller-saved register
4182 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
4184 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
4185 (FP_SYSREGS): Likewise.
4186 (enum vfp_sysregs_encoding): Define enum.
4187 (fp_sysreg_names): Declare.
4188 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
4189 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
4190 (pop_fpsysreg_insn): Likewise.
4192 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
4193 Thomas Preud'homme <thomas.preudhomme@arm.com>
4195 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
4196 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
4197 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
4198 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
4199 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
4200 (ARMv8_1m_main): New feature group.
4201 (armv8.1-m.main): New architecture.
4202 * config/arm/arm-tables.opt: Regenerate.
4203 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
4204 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
4205 (arm_options_perform_arch_sanity_checks): Error out when targeting
4206 Armv8.1-M Mainline Security Extensions.
4207 * config/arm/arm.h (arm_arch8_1m_main): Declare.
4209 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4211 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
4212 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
4213 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
4214 aarch64_bfdot_laneq): New.
4215 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
4216 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
4217 vbfdotq_laneq_f32): New.
4218 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
4219 VBFMLA_W, VBF): New.
4220 (isquadop): Add V4BF, V8BF.
4222 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4224 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
4225 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
4226 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
4227 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
4228 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
4229 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
4230 usdot_laneq, sudot_lane,sudot_laneq): New.
4231 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
4232 (aarch64_<sur>dot_lane): New.
4233 * config/aarch64/arm_neon.h (vusdot_s32): New.
4235 (vusdot_lane_s32): New.
4236 (vsudot_lane_s32): New.
4237 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
4238 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
4240 2020-01-16 Martin Liska <mliska@suse.cz>
4242 * value-prof.c (dump_histogram_value): Fix
4243 obvious spacing issue.
4245 2020-01-16 Andrew Pinski <apinski@marvell.com>
4247 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
4248 !storage_order_barrier_p.
4250 2020-01-16 Andrew Pinski <apinski@marvell.com>
4252 * sched-int.h (_dep): Add unused bit-field field for the padding.
4253 * sched-deps.c (init_dep_1): Init unused field.
4255 2020-01-16 Andrew Pinski <apinski@marvell.com>
4257 * optabs.h (create_expand_operand): Initialize target field also.
4259 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
4261 PR tree-optimization/92429
4262 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
4263 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
4265 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
4268 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
4270 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
4271 aarch64_sve_int_mode to each mode.
4273 2020-01-15 David Malcolm <dmalcolm@redhat.com>
4275 * doc/analyzer.texi (Overview): Add note about
4276 -fdump-ipa-analyzer.
4278 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
4280 PR tree-optimization/93231
4281 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
4282 input_type is unsigned. Use tree_to_shwi for shift constant.
4283 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
4284 (simplify_count_trailing_zeroes): Add test to handle known non-zero
4285 inputs more efficiently.
4287 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
4289 * config/i386/i386.md (*movsf_internal): Do not require
4290 SSE2 ISA for alternatives 14 and 15.
4292 2020-01-15 Richard Biener <rguenther@suse.de>
4295 * tree-eh.c (sink_clobbers): If we already visited the destination
4296 block do not defer insertion.
4297 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
4298 the purpose of defered insertion.
4300 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4302 * BASE-VER: Bump to 10.0.1.
4304 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
4306 PR tree-optimization/93247
4307 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
4308 type of the stmt that we're going to vectorize.
4310 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
4312 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
4313 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
4316 2020-01-15 Martin Liska <mliska@suse.cz>
4318 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
4319 2 calls of streamer_read_hwi in a function call.
4321 2020-01-15 Richard Biener <rguenther@suse.de>
4323 * alias.c (record_alias_subset): Avoid redundant work when
4324 subset is already recorded.
4326 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4328 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
4329 the analyzer options provide CWE identifiers.
4331 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4333 * tree-diagnostic-path.cc (path_summary::event_range::print):
4334 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
4335 using get_pure_location.
4337 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4339 PR tree-optimization/93262
4340 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
4341 perform head trimming only if the last argument is constant,
4342 either all ones, or larger or equal to head trim, in the latter
4343 case decrease the last argument by head_trim.
4345 PR tree-optimization/93249
4346 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
4347 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
4348 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
4349 perform head trim unless we can prove there are no '\0' chars
4350 from the source among the first head_trim chars.
4352 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4354 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
4356 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4359 * config/i386/sse.md
4360 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
4361 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
4362 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
4363 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
4364 just a single alternative instead of two, make operands 1 and 2
4367 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
4370 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
4373 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4375 * Makefile.in (lang_opt_files): Add analyzer.opt.
4376 (ANALYZER_OBJS): New.
4377 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
4378 tristate.o and ANALYZER_OBJS.
4379 (TEXI_GCCINT_FILES): Add analyzer.texi.
4380 * common.opt (-fanalyzer): New driver option.
4381 * config.in: Regenerate.
4382 * configure: Regenerate.
4383 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
4384 (gccdepdir): Also create depdir for "analyzer" subdir.
4385 * digraph.cc: New file.
4386 * digraph.h: New file.
4387 * doc/analyzer.texi: New file.
4388 * doc/gccint.texi ("Static Analyzer") New menu item.
4389 (analyzer.texi): Include it.
4390 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
4391 ("Warning Options"): Add static analysis warnings to the list.
4392 (-Wno-analyzer-double-fclose): New option.
4393 (-Wno-analyzer-double-free): New option.
4394 (-Wno-analyzer-exposure-through-output-file): New option.
4395 (-Wno-analyzer-file-leak): New option.
4396 (-Wno-analyzer-free-of-non-heap): New option.
4397 (-Wno-analyzer-malloc-leak): New option.
4398 (-Wno-analyzer-possible-null-argument): New option.
4399 (-Wno-analyzer-possible-null-dereference): New option.
4400 (-Wno-analyzer-null-argument): New option.
4401 (-Wno-analyzer-null-dereference): New option.
4402 (-Wno-analyzer-stale-setjmp-buffer): New option.
4403 (-Wno-analyzer-tainted-array-index): New option.
4404 (-Wno-analyzer-use-after-free): New option.
4405 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
4406 (-Wno-analyzer-use-of-uninitialized-value): New option.
4407 (-Wanalyzer-too-complex): New option.
4408 (-fanalyzer-call-summaries): New warning.
4409 (-fanalyzer-checker=): New warning.
4410 (-fanalyzer-fine-grained): New warning.
4411 (-fno-analyzer-state-merge): New warning.
4412 (-fno-analyzer-state-purge): New warning.
4413 (-fanalyzer-transitivity): New warning.
4414 (-fanalyzer-verbose-edges): New warning.
4415 (-fanalyzer-verbose-state-changes): New warning.
4416 (-fanalyzer-verbosity=): New warning.
4417 (-fdump-analyzer): New warning.
4418 (-fdump-analyzer-callgraph): New warning.
4419 (-fdump-analyzer-exploded-graph): New warning.
4420 (-fdump-analyzer-exploded-nodes): New warning.
4421 (-fdump-analyzer-exploded-nodes-2): New warning.
4422 (-fdump-analyzer-exploded-nodes-3): New warning.
4423 (-fdump-analyzer-supergraph): New warning.
4424 * doc/sourcebuild.texi (dg-require-dot): New.
4425 (dg-check-dot): New.
4426 * gdbinit.in (break-on-saved-diagnostic): New command.
4427 * graphviz.cc: New file.
4428 * graphviz.h: New file.
4429 * ordered-hash-map-tests.cc: New file.
4430 * ordered-hash-map.h: New file.
4431 * passes.def (pass_analyzer): Add before
4432 pass_ipa_whole_program_visibility.
4433 * selftest-run-tests.c (selftest::run_tests): Call
4434 selftest::ordered_hash_map_tests_cc_tests.
4435 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
4437 * shortest-paths.h: New file.
4438 * timevar.def (TV_ANALYZER): New timevar.
4439 (TV_ANALYZER_SUPERGRAPH): Likewise.
4440 (TV_ANALYZER_STATE_PURGE): Likewise.
4441 (TV_ANALYZER_PLAN): Likewise.
4442 (TV_ANALYZER_SCC): Likewise.
4443 (TV_ANALYZER_WORKLIST): Likewise.
4444 (TV_ANALYZER_DUMP): Likewise.
4445 (TV_ANALYZER_DIAGNOSTICS): Likewise.
4446 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
4447 * tree-pass.h (make_pass_analyzer): New decl.
4448 * tristate.cc: New file.
4449 * tristate.h: New file.
4451 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
4454 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
4455 alternatives 9 and 10.
4457 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4459 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
4460 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
4461 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
4462 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
4463 (selftest::hash_map_tests_c_tests): Call it.
4464 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
4465 New static constant, using the value of = H::empty_zero_p.
4466 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
4467 from default_hash_traits <Value>.
4468 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
4470 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
4471 * hash-table.h (hash_table::alloc_entries): Guard the loop of
4472 calls to mark_empty with !Descriptor::empty_zero_p.
4473 (hash_table::empty_slow): Conditionalize the memset call with a
4474 check that Descriptor::empty_zero_p; otherwise, loop through the
4475 entries calling mark_empty on them.
4476 * hash-traits.h (int_hash::empty_zero_p): New static constant.
4477 (pointer_hash::empty_zero_p): Likewise.
4478 (pair_hash::empty_zero_p): Likewise.
4479 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
4481 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
4482 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
4483 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
4484 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
4485 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
4486 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
4487 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
4488 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
4490 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
4493 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
4495 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
4496 fix typo on return value.
4498 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
4501 * cgraph.c (symbol_table::create_edge): Init speculative_id and
4503 (cgraph_edge::make_speculative): Add param for setting speculative_id
4505 (cgraph_edge::speculative_call_info): Update comments and find reference
4506 by speculative_id for multiple indirect targets.
4507 (cgraph_edge::resolve_speculation): Decrease the speculations
4508 for indirect edge, drop it's speculative if not direct target
4509 left. Update comments.
4510 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4511 (cgraph_node::dump): Print num_speculative_call_targets.
4512 (cgraph_node::verify_node): Don't report error if speculative
4513 edge not include statement.
4514 (cgraph_edge::num_speculative_call_targets_p): New function.
4515 * cgraph.h (int common_target_id): Remove.
4516 (int common_target_probability): Remove.
4517 (num_speculative_call_targets): New variable.
4518 (make_speculative): Add param for setting speculative_id.
4519 (cgraph_edge::num_speculative_call_targets_p): New declare.
4520 (target_prob): New variable.
4521 (speculative_id): New variable.
4522 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
4523 call summaries for multiple speculative call targets.
4524 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
4525 * ipa-profile.c (struct speculative_call_target): New struct.
4526 (class speculative_call_summary): New class.
4527 (class speculative_call_summaries): New class.
4528 (call_sums): New variable.
4529 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
4530 (ipa_profile_write_edge_summary): New function.
4531 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
4532 (ipa_profile_dump_all_summaries): New function.
4533 (ipa_profile_read_edge_summary): New function.
4534 (ipa_profile_read_summary_section): New function.
4535 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
4536 (ipa_profile): Generate num_speculative_call_targets from
4538 * ipa-ref.h (speculative_id): New variable.
4539 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
4540 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
4541 common_target_probability. Stream out speculative_id and
4542 num_speculative_call_targets.
4543 (input_edge): Likewise.
4544 * predict.c (dump_prediction): Remove edges count assert to be
4546 * symtab.c (symtab_node::create_reference): Init speculative_id.
4547 (symtab_node::clone_references): Clone speculative_id.
4548 (symtab_node::clone_referring): Clone speculative_id.
4549 (symtab_node::clone_reference): Clone speculative_id.
4550 (symtab_node::clear_stmts_in_references): Clear speculative_id.
4551 * tree-inline.c (copy_bb): Duplicate all the speculative edges
4552 if indirect call contains multiple speculative targets.
4553 * value-prof.h (check_ic_target): Remove.
4554 * value-prof.c (gimple_value_profile_transformations):
4555 Use void function gimple_ic_transform.
4556 * value-prof.c (gimple_ic_transform): Handle topn case.
4557 Fix comment typos. Change it to a void function.
4559 2020-01-13 Andrew Pinski <apinski@marvell.com>
4561 * config/aarch64/aarch64-cores.def (octeontx2): New define.
4562 (octeontx2t98): New define.
4563 (octeontx2t96): New define.
4564 (octeontx2t93): New define.
4565 (octeontx2f95): New define.
4566 (octeontx2f95n): New define.
4567 (octeontx2f95mm): New define.
4568 * config/aarch64/aarch64-tune.md: Regenerate.
4569 * doc/invoke.texi (-mcpu=): Document the new cpu types.
4571 2020-01-13 Jason Merrill <jason@redhat.com>
4573 PR c++/33799 - destroy return value if local cleanup throws.
4574 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
4576 2020-01-13 Martin Liska <mliska@suse.cz>
4578 * ipa-cp.c (get_max_overall_size): Use newly
4579 renamed param param_ipa_cp_unit_growth.
4580 * params.opt: Remove legacy param name.
4582 2020-01-13 Martin Sebor <msebor@redhat.com>
4584 PR tree-optimization/93213
4585 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
4586 stores to be eliminated.
4588 2020-01-13 Martin Liska <mliska@suse.cz>
4590 * opts.c (print_help): Do not print CL_PARAM
4591 and CL_WARNING for CL_OPTIMIZATION.
4593 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
4596 * doc/invoke.texi (Warning Options): Add caveat about some warnings
4597 depending on optimization settings.
4599 2020-01-13 Jakub Jelinek <jakub@redhat.com>
4601 PR tree-optimization/90838
4602 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4603 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
4604 argument rather than to initialize temporary for targets that
4605 don't use the mode argument at all. Initialize ctzval to avoid
4608 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
4610 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
4611 * tree-core.h: Document it.
4612 * gimplify.c (gimplify_omp_workshare): Set it.
4613 * omp-low.c (lower_omp_target): Use it.
4614 * tree-pretty-print.c (dump_omp_clause): Print it.
4616 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
4617 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
4619 2020-01-10 David Malcolm <dmalcolm@redhat.com>
4621 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
4622 * common.opt (fdiagnostics-path-format=): New option.
4623 (diagnostic_path_format): New enum.
4624 (fdiagnostics-show-path-depths): New option.
4625 * coretypes.h (diagnostic_event_id_t): New forward decl.
4626 * diagnostic-color.c (color_dict): Add "path".
4627 * diagnostic-event-id.h: New file.
4628 * diagnostic-format-json.cc (json_from_expanded_location): Make
4630 (json_end_diagnostic): Call context->make_json_for_path if it
4631 exists and the diagnostic has a path.
4632 (diagnostic_output_format_init): Clear context->print_path.
4633 * diagnostic-path.h: New file.
4634 * diagnostic-show-locus.c (colorizer::set_range): Special-case
4635 when printing a run of events in a diagnostic_path so that they
4636 all get the same color.
4637 (layout::m_diagnostic_path_p): New field.
4638 (layout::layout): Initialize it.
4639 (layout::print_any_labels): Don't colorize the label text for an
4640 event in a diagnostic_path.
4641 (gcc_rich_location::add_location_if_nearby): Add
4642 "restrict_to_current_line_spans" and "label" params. Pass the
4643 former to layout.maybe_add_location_range; pass the latter
4644 when calling add_range.
4645 * diagnostic.c: Include "diagnostic-path.h".
4646 (diagnostic_initialize): Initialize context->path_format and
4647 context->show_path_depths.
4648 (diagnostic_show_any_path): New function.
4649 (diagnostic_path::interprocedural_p): New function.
4650 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4651 (simple_diagnostic_path::num_events): New function.
4652 (simple_diagnostic_path::get_event): New function.
4653 (simple_diagnostic_path::add_event): New function.
4654 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4655 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4656 (debug): New overload taking a diagnostic_path *.
4657 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4658 * diagnostic.h (enum diagnostic_path_format): New enum.
4659 (json::value): New forward decl.
4660 (diagnostic_context::path_format): New field.
4661 (diagnostic_context::show_path_depths): New field.
4662 (diagnostic_context::print_path): New callback field.
4663 (diagnostic_context::make_json_for_path): New callback field.
4664 (diagnostic_show_any_path): New decl.
4665 (json_from_expanded_location): New decl.
4666 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4667 (-fdiagnostics-show-path-depths): New option.
4668 (-fdiagnostics-color): Add "path" to description of default
4669 GCC_COLORS; describe it.
4670 (-fdiagnostics-format=json): Document how diagnostic paths are
4671 represented in the JSON output format.
4672 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4673 Add optional params "restrict_to_current_line_spans" and "label".
4674 * opts.c (common_handle_option): Handle
4675 OPT_fdiagnostics_path_format_ and
4676 OPT_fdiagnostics_show_path_depths.
4677 * pretty-print.c: Include "diagnostic-event-id.h".
4678 (pp_format): Implement "%@" format code for printing
4679 diagnostic_event_id_t *.
4680 (selftest::test_pp_format): Add tests for "%@".
4681 * selftest-run-tests.c (selftest::run_tests): Call
4682 selftest::tree_diagnostic_path_cc_tests.
4683 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4684 * toplev.c (general_init): Initialize global_dc->path_format and
4685 global_dc->show_path_depths.
4686 * tree-diagnostic-path.cc: New file.
4687 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4688 non-static. Drop "diagnostic" param in favor of storing the
4689 original value of "where" and re-using it.
4690 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4691 maybe_unwind_expanded_macro_loc.
4692 (tree_diagnostics_defaults): Initialize context->print_path and
4693 context->make_json_for_path.
4694 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4696 (default_tree_make_json_for_path): New decl.
4697 (maybe_unwind_expanded_macro_loc): New decl.
4699 2020-01-10 Jakub Jelinek <jakub@redhat.com>
4701 PR tree-optimization/93210
4702 * fold-const.h (native_encode_initializer,
4703 can_native_interpret_type_p): Declare.
4704 * fold-const.c (native_encode_string): Fix up handling with off != -1,
4706 (native_encode_initializer): New function, moved from dwarf2out.c.
4707 Adjust to native_encode_expr compatible arguments, including dry-run
4708 and partial extraction modes. Don't handle STRING_CST.
4709 (can_native_interpret_type_p): No longer static.
4710 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4711 offset / BITS_PER_UNIT fits into int and don't call it if
4712 can_native_interpret_type_p fails. If suboff is NULL and for
4713 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4714 native_encode_initializer.
4715 (fold_const_aggregate_ref_1): Formatting fix.
4716 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4717 (tree_add_const_value_attribute): Adjust caller.
4719 PR tree-optimization/90838
4720 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4721 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4722 CTZ_DEFINED_VALUE_AT_ZERO.
4724 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
4727 * lra-constraints.c (match_reload): Permit input operands have the
4728 same mode as output while other input operands have a different
4731 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
4733 PR tree-optimization/90838
4734 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4735 (check_ctz_string): Likewise.
4736 (optimize_count_trailing_zeroes): Likewise.
4737 (simplify_count_trailing_zeroes): Likewise.
4738 (pass_forwprop::execute): Try ctz simplification.
4739 * match.pd: Add matching for ctz idioms.
4741 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4743 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4745 (aarch64_invalid_unary_op): New function for target hook.
4746 (aarch64_invalid_binary_op): New function for target hook.
4748 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4750 * config.gcc: Add arm_bf16.h.
4751 * config/aarch64/aarch64-builtins.c
4752 (aarch64_simd_builtin_std_type): Add BFmode.
4753 (aarch64_init_simd_builtin_types): Define element types for vector
4755 (aarch64_init_bf16_types): New function.
4756 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4757 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4759 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4760 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4762 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4763 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4764 * config/aarch64/aarch64.c
4765 (aarch64_classify_vector_mode): Add support for BF types.
4766 (aarch64_gimplify_va_arg_expr): Add support for BF types.
4767 (aarch64_vq_mode): Add support for BF types.
4768 (aarch64_simd_container_mode): Add support for BF types.
4769 (aarch64_mangle_type): Add support for BF scalar type.
4770 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4771 * config/aarch64/arm_bf16.h: New file.
4772 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4773 * config/aarch64/iterators.md: Add BF types to mode attributes.
4774 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4776 2020-01-10 Jason Merrill <jason@redhat.com>
4778 PR c++/93173 - incorrect tree sharing.
4779 * gimplify.c (copy_if_shared): No longer static.
4780 * gimplify.h: Declare it.
4782 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4784 * doc/invoke.texi (-msve-vector-bits=): Document that
4785 -msve-vector-bits=128 now generates VL-specific code for
4786 little-endian targets.
4787 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4788 build_vector_type_for_mode to construct the data vector types.
4789 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4790 VL-specific code for -msve-vector-bits=128 on little-endian targets.
4791 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4792 for 128-bit vectors.
4794 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4796 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4799 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4801 * config/aarch64/aarch64-builtins.c
4802 (aarch64_builtin_vectorized_function): Check for specific vector modes,
4803 rather than checking the number of elements and the element mode.
4805 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4807 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4808 get_related_vectype_for_scalar_type rather than build_vector_type
4809 to create the index type for a conditional reduction.
4811 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4813 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4814 for any type of gather or scatter, including strided accesses.
4816 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4818 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4821 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4823 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4825 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
4826 parameter and its use to reset DR_OFFSET's.
4827 (vect_transform_loop): Remove orig_drs_init argument.
4828 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4829 member of dr_vec_info rather than the offset of the associated
4830 data_reference's innermost_loop_behavior.
4831 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4832 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4833 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4834 get_dr_vinfo_offset.
4835 (vectorizable_store): Likewise.
4836 (vectorizable_load): Likewise.
4838 2020-01-10 Richard Biener <rguenther@suse.de>
4840 * gimple-ssa-store-merging
4841 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4843 2020-01-10 Martin Liska <mliska@suse.cz>
4846 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4847 encapsulation that was there before r280040.
4849 2020-01-10 Richard Biener <rguenther@suse.de>
4852 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4853 sequences to avoid walking them again for secondary opportunities.
4854 (pass_lower_eh_dispatch::execute): Instead actually insert
4857 2020-01-10 Richard Biener <rguenther@suse.de>
4860 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4861 (cleanup_all_empty_eh): Walk landing pads in reverse order to
4862 avoid quadraticness.
4864 2020-01-10 Martin Jambor <mjambor@suse.cz>
4866 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4867 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4868 to get param_ipa_sra_max_replacements.
4869 (param_splitting_across_edge): Pass the caller to
4870 pull_accesses_from_callee.
4872 2020-01-10 Martin Jambor <mjambor@suse.cz>
4874 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4875 * ipa-cp.c (max_new_size): Removed.
4876 (orig_overall_size): New variable.
4877 (get_max_overall_size): New function.
4878 (estimate_local_effects): Use it. Adjust dump.
4879 (decide_about_value): Likewise.
4880 (ipcp_propagate_stage): Do not calculate max_new_size, just store
4881 orig_overall_size. Adjust dump.
4882 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4884 2020-01-10 Martin Jambor <mjambor@suse.cz>
4886 * params.opt (param_ipa_max_agg_items): Mark as Optimization
4887 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4888 instead of param_ipa_max_agg_items.
4889 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4890 optimization info for the callee.
4892 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
4894 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4895 markers if debug_inline_points is false.
4897 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4899 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4901 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4902 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4903 aarch64-sve-builtins-sve2.h.
4904 (aarch64-sve-builtins-sve2.o): New rule.
4905 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4906 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4907 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4908 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
4909 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4910 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
4912 * config/aarch64/aarch64-sve.md: Update comments with SVE2
4913 instructions that are handled here.
4914 (@cond_asrd<mode>): Generalize to...
4915 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
4916 (*cond_asrd<mode>_2): Generalize to...
4917 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
4918 (*cond_asrd<mode>_z): Generalize to...
4919 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4920 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4921 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4922 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4923 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4925 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4926 (@aarch64_scatter_stnt<mode>): Likewise.
4927 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4928 (@aarch64_mul_lane_<mode>): Likewise.
4929 (@aarch64_sve_suqadd<mode>_const): Likewise.
4930 (*<sur>h<addsub><mode>): Generalize to...
4931 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4933 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4934 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4935 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4936 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4937 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4938 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4939 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4940 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4941 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4942 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4943 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4944 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4945 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4946 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4947 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4948 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4949 (@aarch64_sve2_xar<mode>): Likewise.
4950 (@aarch64_sve2_bcax<mode>): Likewise.
4951 (*aarch64_sve2_eor3<mode>): Rename to...
4952 (@aarch64_sve2_eor3<mode>): ...this.
4953 (@aarch64_sve2_bsl<mode>): New expander.
4954 (@aarch64_sve2_nbsl<mode>): Likewise.
4955 (@aarch64_sve2_bsl1n<mode>): Likewise.
4956 (@aarch64_sve2_bsl2n<mode>): Likewise.
4957 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4958 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4959 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4960 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4961 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4962 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4963 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4964 (<su>mull<bt><Vwide>): Generalize to...
4965 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4967 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4968 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4969 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4970 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4971 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4972 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4973 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4974 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4975 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4976 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4977 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4978 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4979 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4980 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4981 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4982 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4983 (<SHRNB:r>shrnb<mode>): Generalize to...
4984 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4986 (<SHRNT:r>shrnt<mode>): Generalize to...
4987 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4989 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4990 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4991 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4992 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4993 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4994 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4995 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4996 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4997 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4998 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4999 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
5000 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
5001 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
5002 (@aarch64_sve2_cvtnt<mode>): Likewise.
5003 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
5004 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
5005 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
5006 (@aarch64_sve2_cvtxnt<mode>): Likewise.
5007 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
5008 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
5009 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
5010 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
5011 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
5012 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
5013 (@aarch64_sve2_pmul<mode>): Likewise.
5014 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
5015 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
5016 (@aarch64_sve2_tbl2<mode>): Likewise.
5017 (@aarch64_sve2_tbx<mode>): Likewise.
5018 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
5019 (@aarch64_sve2_histcnt<mode>): Likewise.
5020 (@aarch64_sve2_histseg<mode>): Likewise.
5021 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
5022 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
5023 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
5024 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
5025 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
5026 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
5027 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
5028 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
5029 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
5030 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
5031 (SVE2_PMULL_PAIR_I): New mode iterators.
5032 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
5033 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
5034 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
5035 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
5036 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
5037 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
5038 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
5039 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
5040 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
5041 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
5042 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
5043 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
5044 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
5045 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
5046 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
5047 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
5048 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
5049 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
5050 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
5051 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
5052 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
5053 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
5054 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
5055 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
5056 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
5057 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
5058 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
5059 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
5060 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
5061 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
5062 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
5064 (VNARROW, Ventype): New mode attributes.
5065 (Vewtype): Handle VNx2DI. Fix typo in comment.
5066 (VDOUBLE): New mode attribute.
5067 (sve_lane_con): Handle VNx8HI.
5068 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
5069 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
5070 (sve_int_op, sve_int_op_rev): Handle the above codes.
5071 (sve_pred_int_rhs2_operand): Likewise.
5072 (MULLBT, SHRNB, SHRNT): Delete.
5073 (SVE_INT_SHIFT_IMM): New int iterator.
5074 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
5075 and UNSPEC_WHILEHS for TARGET_SVE2.
5076 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
5077 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
5078 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
5079 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
5080 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
5081 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
5082 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
5083 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
5084 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
5085 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
5086 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
5087 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
5088 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
5089 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
5090 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
5091 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
5092 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
5093 (optab): Handle the new unspecs.
5094 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
5096 (lr): Handle the new unspecs.
5098 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
5099 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
5100 (sve_int_qsub_op): New int attributes.
5101 (sve_fp_op, rot): Handle the new unspecs.
5102 * config/aarch64/aarch64-sve-builtins.h
5103 (function_resolver::require_matching_pointer_type): Declare.
5104 (function_resolver::resolve_unary): Add an optional boolean argument.
5105 (function_resolver::finish_opt_n_resolution): Add an optional
5106 type_suffix_index argument.
5107 (gimple_folder::redirect_call): Declare.
5108 (gimple_expander::prepare_gather_address_operands): Add an optional
5110 * config/aarch64/aarch64-sve-builtins.cc: Include
5111 aarch64-sve-builtins-sve2.h.
5112 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
5113 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
5114 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
5115 (TYPES_hsd_integer): Use TYPES_hsd_signed.
5116 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
5117 (TYPES_s_unsigned): Likewise.
5118 (TYPES_s_integer): Use TYPES_s_unsigned.
5119 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
5120 (TYPES_sd_integer): Use them.
5121 (TYPES_d_unsigned): New macro.
5122 (TYPES_d_integer): Use it.
5123 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
5124 (TYPES_cvt_narrow): Likewise.
5125 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
5126 (preds_mx): New variable.
5127 (function_builder::add_overloaded_function): Allow the new feature
5128 set to be more restrictive than the original one.
5129 (function_resolver::infer_pointer_type): Remove qualifiers from
5130 the pointer type before printing it.
5131 (function_resolver::require_matching_pointer_type): New function.
5132 (function_resolver::resolve_sv_displacement): Handle functions
5133 that don't support 32-bit vector indices or svint32_t vector offsets.
5134 (function_resolver::finish_opt_n_resolution): Take the inferred type
5135 as a separate argument.
5136 (function_resolver::resolve_unary): Optionally treat all forms in
5137 the same way as normal merging functions.
5138 (gimple_folder::redirect_call): New function.
5139 (function_expander::prepare_gather_address_operands): Add an argument
5140 that says whether scaled forms are available. If they aren't,
5141 handle scaling of vector indices and don't add the extension and
5143 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
5144 fall back to using cond_* instead.
5145 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
5146 Split out the member variables into...
5147 (rtx_code_function_base): ...this new base class.
5148 (rtx_code_function_rotated): Inherit rtx_code_function_base.
5149 (unspec_based_function): Split out the member variables into...
5150 (unspec_based_function_base): ...this new base class.
5151 (unspec_based_function_rotated): Inherit unspec_based_function_base.
5152 (unspec_based_function_exact_insn): New class.
5153 (unspec_based_add_function, unspec_based_add_lane_function)
5154 (unspec_based_lane_function, unspec_based_pred_function)
5155 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
5156 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
5157 (unspec_based_sub_function, unspec_based_sub_lane_function): New
5159 (unspec_based_fused_function): New class.
5160 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
5161 (unspec_based_fused_lane_function): New class.
5162 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
5164 (CODE_FOR_MODE1): New macro.
5165 (fixed_insn_function): New class.
5166 (while_comparison): Likewise.
5167 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
5168 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
5169 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
5170 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
5171 (load_gather_sv_restricted, shift_left_imm_long): Declare.
5172 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
5173 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
5174 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
5175 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
5176 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
5177 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
5178 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
5179 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
5180 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
5181 Also add an initial argument for unary_convert_narrowt, regardless
5182 of the predication type.
5183 (build_32_64): Allow loads and stores to specify MODE_none.
5184 (build_sv_index64, build_sv_uint_offset): New functions.
5185 (long_type_suffix): New function.
5186 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
5187 (binary_imm_long_base, load_gather_sv_base): Likewise.
5188 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
5189 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
5190 (unary_narrowb_base, unary_narrowt_base): Likewise.
5191 (binary_long_lane_def, binary_long_lane): New shape.
5192 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
5193 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
5194 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
5195 (binary_to_uint_def, binary_to_uint): Likewise.
5196 (binary_wide_def, binary_wide): Likewise.
5197 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
5198 (compare_def, compare): Likewise.
5199 (compare_ptr_def, compare_ptr): Likewise.
5200 (load_ext_gather_index_restricted_def,
5201 load_ext_gather_index_restricted): Likewise.
5202 (load_ext_gather_offset_restricted_def,
5203 load_ext_gather_offset_restricted): Likewise.
5204 (load_gather_sv_def): Inherit from load_gather_sv_base.
5205 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
5206 (shift_left_imm_def, shift_left_imm): Likewise.
5207 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
5208 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
5209 (store_scatter_index_restricted_def,
5210 store_scatter_index_restricted): Likewise.
5211 (store_scatter_offset_restricted_def,
5212 store_scatter_offset_restricted): Likewise.
5213 (tbl_tuple_def, tbl_tuple): Likewise.
5214 (ternary_long_lane_def, ternary_long_lane): Likewise.
5215 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
5216 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
5217 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
5218 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
5219 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
5220 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
5221 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
5222 (ternary_uint_def, ternary_uint): Likewise.
5223 (unary_convert): Fix typo in comment.
5224 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
5225 (unary_long_def, unary_long): Likewise.
5226 (unary_narrowb_def, unary_narrowb): Likewise.
5227 (unary_narrowt_def, unary_narrowt): Likewise.
5228 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
5229 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
5230 (unary_to_int_def, unary_to_int): Likewise.
5231 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
5232 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
5233 (svasrd_impl): Delete.
5234 (svcadd_impl::expand): Handle integer operations too.
5235 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
5236 new functions to derive the unspec numbers.
5237 (svmla_svmls_lane_impl): Replace with...
5238 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
5239 integer operations too.
5240 (svwhile_impl): Rename to...
5241 (svwhilelx_impl): ...this and inherit from while_comparison.
5242 (svasrd): Use unspec_based_function.
5243 (svmla_lane): Use svmla_lane_impl.
5244 (svmls_lane): Use svmls_lane_impl.
5245 (svrecpe, svrsqrte): Handle unsigned integer operations too.
5246 (svwhilele, svwhilelt): Use svwhilelx_impl.
5247 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
5248 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
5249 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
5250 * config/aarch64/aarch64-sve-builtins.def: Include
5251 aarch64-sve-builtins-sve2.def.
5253 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5255 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
5256 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
5257 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
5258 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
5259 immediates as well as vector ones.
5260 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
5261 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
5262 (aarch64_sve_qsub_immediate): Update calls accordingly.
5264 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5266 * config/aarch64/aarch64-sve2.md: Add banner comments.
5267 (<su>mulh<r>s<mode>3): Move further up file.
5268 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
5269 (*aarch64_sve2_sra<mode>): Move further down file.
5270 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
5272 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5274 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
5276 (while_optab_cmp): Handle them.
5277 * config/aarch64/aarch64-sve.md
5278 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
5279 and add a "@" marker.
5280 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
5281 instead of gen_aarch64_sve2_while_ptest.
5282 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
5284 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5286 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
5287 (UNSPEC_WHILELE): ...this.
5288 (UNSPEC_WHILE_LO): Rename to...
5289 (UNSPEC_WHILELO): ...this.
5290 (UNSPEC_WHILE_LS): Rename to...
5291 (UNSPEC_WHILELS): ...this.
5292 (UNSPEC_WHILE_LT): Rename to...
5293 (UNSPEC_WHILELT): ...this.
5294 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
5295 (cmp_op, while_optab_cmp): Likewise.
5296 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
5297 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
5298 (svwhilelt): Likewise.
5300 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5302 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
5303 (unary_to_uint): Define.
5304 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
5305 (unary_count): Rename to...
5306 (unary_to_uint_def, unary_to_uint): ...this.
5307 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
5309 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5311 * config/aarch64/aarch64-sve-builtins-functions.h
5312 (code_for_mode_function): New class.
5313 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
5314 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
5315 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
5316 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
5317 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
5319 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5321 * config/aarch64/iterators.md (addsub): New code attribute.
5322 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
5324 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
5325 in the asm string and attributes. Fix indentation.
5326 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
5328 (@aarch64_sve_<optab><mode>): ...this.
5329 * config/aarch64/aarch64-sve-builtins.h
5330 (function_expander::expand_signed_unpred_op): Delete.
5331 * config/aarch64/aarch64-sve-builtins.cc
5332 (function_expander::expand_signed_unpred_op): Likewise.
5333 (function_expander::map_to_rtx_codes): If the optab isn't defined,
5334 try using code_for_aarch64_sve instead.
5335 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
5336 (svqsub_impl): Likewise.
5337 (svqadd, svqsub): Use rtx_code_function instead.
5339 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5341 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
5342 (HADDSUB, sur, addsub): Remove them.
5344 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5346 * tree-nrv.c (pass_return_slot::execute): Handle all internal
5347 functions the same way, rather than singling out those that
5348 aren't mapped directly to optabs.
5350 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
5352 * target.def (compatible_vector_types_p): New target hook.
5353 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
5354 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
5355 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
5356 * doc/tm.texi: Regenerate.
5357 * gimple-expr.c: Include target.h.
5358 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
5359 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
5361 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
5362 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
5363 Use the original predicate if it already has a suitable type.
5365 2020-01-09 Martin Jambor <mjambor@suse.cz>
5367 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
5368 resolve_speculation and redirect_call_stmt_to_callee static. Change
5369 return type of set_call_stmt to cgraph_edge *.
5370 * auto-profile.c (afdo_indirect_call): Adjust call to
5371 redirect_call_stmt_to_callee.
5372 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
5373 make the this pointer explicit, adjust self-recursive calls and the
5374 call top make_direct. Return the resulting edge.
5375 (cgraph_edge::remove): Make this pointer explicit.
5376 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
5377 (cgraph_edge::make_direct): Likewise, adjust call to
5378 resolve_speculation.
5379 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
5380 call to set_call_stmt.
5381 (cgraph_update_edges_for_call_stmt_node): Update call to
5382 set_call_stmt and remove.
5383 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
5384 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
5385 (cgraph_node::create_edge_including_clones): Moved "first" definition
5386 of edge to the block where it was used. Adjusted calls to
5388 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
5389 cgraph_edge::remove.
5390 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
5391 make_direct and redirect_call_stmt_to_callee.
5392 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
5393 resolve_speculation and make_direct.
5394 * ipa-inline-transform.c (inline_transform): Adjust call to
5395 redirect_call_stmt_to_callee.
5396 (check_speculations_1):: Adjust call to resolve_speculation.
5397 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
5398 resolve-speculation.
5399 (inline_small_functions): Adjust call to resolve_speculation.
5400 (ipa_inline): Likewise.
5401 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
5403 * ipa-visibility.c (function_and_variable_visibility): Make iteration
5404 safe with regards to edge removal, adjust calls to
5405 redirect_call_stmt_to_callee.
5406 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
5407 and redirect_call_stmt_to_callee.
5408 * multiple_target.c (create_dispatcher_calls): Adjust call to
5409 redirect_call_stmt_to_callee
5410 (redirect_to_specific_clone): Likewise.
5411 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
5412 Adjust calls to cgraph_edge::remove.
5413 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
5414 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
5415 (expand_call_inline): Adjust call to cgraph_edge::remove.
5417 2020-01-09 Martin Liska <mliska@suse.cz>
5419 * params.opt: Set Optimization for
5420 param_max_speculative_devirt_maydefs.
5422 2020-01-09 Martin Sebor <msebor@redhat.com>
5426 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
5428 2020-01-09 Martin Liska <mliska@suse.cz>
5430 * auto-profile.c (auto_profile): Use opt_for_fn
5432 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
5433 (propagate_vals_across_arith_jfunc): Likewise.
5434 (hint_time_bonus): Likewise.
5435 (incorporate_penalties): Likewise.
5436 (good_cloning_opportunity_p): Likewise.
5437 (perform_estimation_of_a_value): Likewise.
5438 (estimate_local_effects): Likewise.
5439 (ipcp_propagate_stage): Likewise.
5440 * ipa-fnsummary.c (decompose_param_expr): Likewise.
5441 (set_switch_stmt_execution_predicate): Likewise.
5442 (analyze_function_body): Likewise.
5443 * ipa-inline-analysis.c (offline_size): Likewise.
5444 * ipa-inline.c (early_inliner): Likewise.
5445 * ipa-prop.c (ipa_analyze_node): Likewise.
5446 (ipcp_transform_function): Likewise.
5447 * ipa-sra.c (process_scan_results): Likewise.
5448 (ipa_sra_summarize_function): Likewise.
5449 * params.opt: Rename ipcp-unit-growth to
5450 ipa-cp-unit-growth. Add Optimization for various
5451 IPA-related parameters.
5453 2020-01-09 Richard Biener <rguenther@suse.de>
5456 * gimplify.c (gimplify_expr): Deal with NOP definitions.
5458 2020-01-09 Richard Biener <rguenther@suse.de>
5460 PR tree-optimization/93040
5461 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
5463 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
5465 * common/config/avr/avr-common.c (avr_option_optimization_table)
5466 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
5468 2020-01-09 Martin Liska <mliska@suse.cz>
5470 * cgraphclones.c (symbol_table::materialize_all_clones):
5471 Use cgraph_node::dump_name.
5473 2020-01-09 Jakub Jelinek <jakub@redhat.com>
5476 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
5477 output_operand_lossage instead of gcc_unreachable.
5478 * doc/md.texi (riscv f constraint): Fix typo.
5481 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
5482 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
5483 CONST_SCALAR_INT_P instead of CONST_INT_P.
5484 (*subv<mode>4_1): Rename to ...
5485 (subv<mode>4_1): ... this.
5486 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5487 define_insn_and_split patterns.
5488 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5491 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5493 * vec.c (class selftest::count_dtor): New class.
5494 (selftest::test_auto_delete_vec): New test.
5495 (selftest::vec_c_tests): Call it.
5496 * vec.h (class auto_delete_vec): New class template.
5497 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
5499 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5501 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
5503 2020-01-08 Jim Wilson <jimw@sifive.com>
5505 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
5506 use of TLS_MODEL_LOCAL_EXEC when not pic.
5508 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5510 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
5513 2020-01-08 Jakub Jelinek <jakub@redhat.com>
5516 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
5517 *stack_protect_set_3 peephole2): Also check that the second
5518 insns source is general_operand.
5521 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
5522 predicate for output operand instead of register_operand.
5523 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
5524 memory destination and non-memory operands[2].
5526 2020-01-08 Martin Liska <mliska@suse.cz>
5528 * cgraph.c (cgraph_node::dump): Use ::dump_name or
5529 ::dump_asm_name instead of (::name or ::asm_name).
5530 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
5531 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
5532 (analyze_functions): Likewise.
5533 (expand_all_functions): Likewise.
5534 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5535 (propagate_bits_across_jump_function): Likewise.
5536 (dump_profile_updates): Likewise.
5537 (ipcp_store_bits_results): Likewise.
5538 (ipcp_store_vr_results): Likewise.
5539 * ipa-devirt.c (dump_targets): Likewise.
5540 * ipa-fnsummary.c (analyze_function_body): Likewise.
5541 * ipa-hsa.c (check_warn_node_versionable): Likewise.
5542 (process_hsa_functions): Likewise.
5543 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
5544 (set_alias_uids): Likewise.
5545 * ipa-inline-transform.c (save_inline_function_body): Likewise.
5546 * ipa-inline.c (recursive_inlining): Likewise.
5547 (inline_to_all_callers_1): Likewise.
5548 (ipa_inline): Likewise.
5549 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5550 (ipa_propagate_frequency): Likewise.
5551 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5552 (remove_described_reference): Likewise.
5553 * ipa-pure-const.c (worse_state): Likewise.
5554 (check_retval_uses): Likewise.
5555 (analyze_function): Likewise.
5556 (propagate_pure_const): Likewise.
5557 (propagate_nothrow): Likewise.
5558 (dump_malloc_lattice): Likewise.
5559 (propagate_malloc): Likewise.
5560 (pass_local_pure_const::execute): Likewise.
5561 * ipa-visibility.c (optimize_weakref): Likewise.
5562 (function_and_variable_visibility): Likewise.
5563 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
5564 (ipa_discover_variable_flags): Likewise.
5565 * lto-streamer-out.c (output_function): Likewise.
5566 (output_constructor): Likewise.
5567 * tree-inline.c (copy_bb): Likewise.
5568 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
5569 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
5571 2020-01-08 Richard Biener <rguenther@suse.de>
5574 * tree-eh.c (sink_clobbers): Update virtual operands for
5575 the first and last stmt only. Add a dry-run capability.
5576 (pass_lower_eh_dispatch::execute): Perform clobber sinking
5577 after CFG manipulations and in RPO order to catch all
5578 secondary opportunities reliably.
5580 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5583 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5585 2019-01-08 Richard Biener <rguenther@suse.de>
5588 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
5589 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
5590 virtual operand, also updating SSA use.
5591 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
5592 Update stmt after resetting virtual operand.
5593 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
5594 * gimple-iterator.c (gsi_remove): When not removing the stmt
5595 permanently do not delink immediate uses or mark the stmt modified.
5597 2020-01-08 Martin Liska <mliska@suse.cz>
5599 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
5600 (ipa_call_context::estimate_size_and_time): Likewise.
5601 (inline_analyze_function): Likewise.
5603 2020-01-08 Martin Liska <mliska@suse.cz>
5605 * cgraph.c (cgraph_node::dump): Use systematically
5608 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5610 Add -nodevicespecs option for avr.
5613 * config/avr/avr.opt (-nodevicespecs): New driver option.
5614 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
5615 "-specs=device-specs/..." if that option is not set.
5616 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5618 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5620 Implement 64-bit double functions for avr.
5623 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
5624 --with-double-comparison.
5625 * doc/install.texi: Document them.
5626 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
5627 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
5628 <WITH_DOUBLE_COMPARISON>: New built-in defines.
5629 * doc/invoke.texi (AVR Built-in Macros): Document them.
5630 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5631 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5632 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5634 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
5637 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5638 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5639 when only building rm-profile multilibs.
5641 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
5644 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5645 lattice for a value to check.
5646 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5647 finite propagation in self-recursive scc.
5649 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
5651 * ipa-inline.c (caller_growth_limits): Restore the AND.
5653 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5655 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5656 (VEC_ALLREG_ALT): New iterator.
5657 (VEC_ALLREG_INT_MODE): New iterator.
5658 (VCMP_MODE): New iterator.
5659 (VCMP_MODE_INT): New iterator.
5660 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5661 (vec_cmp<u>v64qidi): New define_expand.
5662 (vec_cmp<mode>di_exec): Use VCMP_MODE.
5663 (vec_cmpu<mode>di_exec): New define_expand.
5664 (vec_cmp<u>v64qidi_exec): New define_expand.
5665 (vec_cmp<mode>di_dup): Use VCMP_MODE.
5666 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5667 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5668 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5669 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5670 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5671 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5672 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5673 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5674 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5676 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5677 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5679 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5681 * config/gcn/constraints.md (DA): Update description and match.
5683 (Db): New constraint.
5684 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5686 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5687 Implement 'Db' mixed immediate type.
5688 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5689 (addcv64si3_dup<exec_vcc>): Delete.
5690 (subcv64si3<exec_vcc>): Rework constraints.
5691 (addv64di3): Rework constraints.
5692 (addv64di3_exec): Rework constraints.
5693 (subv64di3): Rework constraints.
5694 (addv64di3_dup): Delete.
5695 (addv64di3_dup_exec): Delete.
5696 (addv64di3_zext): Rework constraints.
5697 (addv64di3_zext_exec): Rework constraints.
5698 (addv64di3_zext_dup): Rework constraints.
5699 (addv64di3_zext_dup_exec): Rework constraints.
5700 (addv64di3_zext_dup2): Rework constraints.
5701 (addv64di3_zext_dup2_exec): Rework constraints.
5702 (addv64di3_sext_dup2): Rework constraints.
5703 (addv64di3_sext_dup2_exec): Rework constraints.
5705 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
5707 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5708 existing target checks.
5710 2020-01-07 Richard Biener <rguenther@suse.de>
5712 * doc/install.texi: Bump minimal supported MPC version.
5714 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5716 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5717 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5718 * langhooks.c: Include stor-layout.h.
5719 (lhd_simulate_enum_decl): New function.
5720 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5721 handle_arm_sve_h for the LTO frontend.
5722 (register_vector_type): Cope with null returns from pushdecl.
5724 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5726 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5727 (aarch64_sve::nvectors_if_data_type): Replace with...
5728 (aarch64_sve::builtin_type_p): ...this.
5729 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5730 (find_vector_type): Delete.
5731 (add_sve_type_attribute): New function.
5732 (lookup_sve_type_attribute): Likewise.
5733 (register_builtin_types): Add an "SVE type" attribute to each type.
5734 (register_tuple_type): Likewise.
5735 (svbool_type_p, nvectors_if_data_type): Delete.
5736 (mangle_builtin_type): Use lookup_sve_type_attribute.
5737 (builtin_type_p): Likewise. Add an overload that returns the
5738 number of constituent vector and predicate registers.
5739 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5740 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5741 instead of aarch64_sve_argument_p.
5742 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5743 (aarch64_pass_by_reference): Likewise.
5744 (aarch64_function_value_1): Likewise.
5745 (aarch64_return_in_memory): Likewise.
5746 (aarch64_layout_arg): Likewise.
5748 2020-01-07 Jakub Jelinek <jakub@redhat.com>
5750 PR tree-optimization/93156
5751 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5752 least significant bit is always clear.
5754 PR tree-optimization/93118
5755 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
5756 simplifier with two intermediate conversions.
5758 2020-01-07 Martin Liska <mliska@suse.cz>
5760 * params.opt: Add Optimization for various parameters.
5762 2020-01-07 Martin Liska <mliska@suse.cz>
5765 * doc/extend.texi: Explain cloning for target_clone
5768 2020-01-07 Martin Liska <mliska@suse.cz>
5770 PR tree-optimization/92860
5771 * common.opt: Make in Optimization option
5772 as it is affected by -O0, which is an Optimization
5774 * tree-inline.c (tree_inlinable_function_p):
5775 Use opt_for_fn for warn_inline.
5776 (expand_call_inline): Likewise.
5778 2020-01-07 Martin Liska <mliska@suse.cz>
5780 PR tree-optimization/92860
5781 * common.opt: Make flag_ree as optimization
5784 2020-01-07 Martin Liska <mliska@suse.cz>
5786 PR optimization/92860
5787 * params.opt: Mark param_min_crossjump_insns with Optimization
5790 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
5792 * ipa-inline-analysis.c (estimate_growth): Fix typo.
5793 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5795 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
5797 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5798 helper function to return the valid addressing formats for a given
5799 hard register and mode.
5800 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5802 * config/rs6000/constraints.md (Q constraint): Update
5804 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5807 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5808 Use 'Q' for doing vector extract from memory.
5809 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5811 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5812 doing vector extract from memory.
5813 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5814 extract from memory.
5816 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5817 for the offset being 34-bits when -mcpu=future is used.
5819 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
5821 * config/pa/pa.md: Revert change to use ordered_comparison_operator
5822 instead of cmpib_comparison_operator in cmpib patterns.
5823 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5824 of cmpib_comparison_operator. Revise comment.
5826 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5828 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5829 in an IFN_DIV_POW2 node to be equal.
5831 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5833 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5834 (vect_check_scalar_mask): ...this.
5835 (vectorizable_store, vectorizable_load): Update call accordingly.
5836 (vectorizable_call): Use vect_check_scalar_mask to check the mask
5837 argument in calls to conditional internal functions.
5839 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5841 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5842 '0' matching inputs.
5843 (subv64di3_exec): Likewise.
5845 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
5847 * config/mips/mips.c (vr4130_align_insns): Fix typo.
5848 * doc/md.texi (movstr): Likewise.
5850 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5852 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5855 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5857 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5859 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
5860 to a temporary file and use move-if-change to update the real
5861 file where necessary.
5863 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5865 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5866 rather than Upa for CPY /M.
5868 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5870 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5873 2020-01-06 Martin Liska <mliska@suse.cz>
5875 PR tree-optimization/92860
5876 * params.opt: Mark param_max_combine_insns with Optimization
5879 2020-01-05 Jakub Jelinek <jakub@redhat.com>
5882 * config/i386/i386.md (SWIDWI): New mode iterator.
5883 (DWI, dwi): Add TImode variants.
5884 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
5885 <general_hilo_operand> instead of <general_operand>. Use
5886 CONST_SCALAR_INT_P instead of CONST_INT_P.
5887 (*addv<mode>4_1): Rename to ...
5888 (addv<mode>4_1): ... this.
5889 (QWI): New mode attribute.
5890 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5891 define_insn_and_split patterns.
5892 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5894 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
5895 <general_hilo_operand> instead of <general_operand>.
5896 (*addcarry<mode>_1): New define_insn.
5897 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5899 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
5901 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5902 Use "call" instead of "set".
5904 2020-01-03 Martin Jambor <mjambor@suse.cz>
5907 * ipa-cp.c (print_all_lattices): Skip functions without info.
5909 2020-01-03 Jakub Jelinek <jakub@redhat.com>
5912 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
5913 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
5914 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
5915 for 'e' simd clones.
5918 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5920 (mprefer-vector-width=): Add Save.
5921 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5922 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
5923 (ix86_debug_options, ix86_function_specific_print): Adjust
5924 ix86_target_string callers.
5925 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5926 (ix86_valid_target_attribute_tree): Likewise.
5927 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5928 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5929 ix86_target_string caller.
5932 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5933 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
5934 instead of gen_int_shift_amount + convert_modes.
5936 PR rtl-optimization/93088
5937 * loop-iv.c (find_single_def_src): Punt after looking through
5938 128 reg copies for regs with single definitions. Move definitions
5941 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
5943 * config/arm/arm-c.c (arm_cpu_builtins): Define
5944 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5945 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5946 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5947 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5948 * config/arm/arm-tables.opt: Regenerated.
5949 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5950 arm_arch_i8mm and arm_arch_bf16 when enabled.
5951 * config/arm/arm.h (TARGET_I8MM): New macro.
5952 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5953 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5954 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5955 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5956 (v8_6_a_simd_variants): New.
5957 (v8_*_a_simd_variants): Add i8mm and bf16.
5958 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5960 2020-01-02 Jakub Jelinek <jakub@redhat.com>
5963 * predict.c (compute_function_frequency): Don't call
5964 warn_function_cold on functions that already have cold attribute.
5966 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
5969 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
5970 COMDAT group function labels in .data.rel.ro.local section.
5971 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5974 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5975 comparison_operator in B and S integer comparisons. Likewise, use
5976 ordered_comparison_operator instead of cmpib_comparison_operator in
5978 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5980 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5982 Update copyright years.
5984 * gcc.c (process_command): Update copyright notice dates.
5985 * gcov-dump.c (print_version): Ditto.
5986 * gcov.c (print_version): Ditto.
5987 * gcov-tool.c (print_version): Ditto.
5988 * gengtype.c (create_file): Ditto.
5989 * doc/cpp.texi: Bump @copying's copyright year.
5990 * doc/cppinternals.texi: Ditto.
5991 * doc/gcc.texi: Ditto.
5992 * doc/gccint.texi: Ditto.
5993 * doc/gcov.texi: Ditto.
5994 * doc/install.texi: Ditto.
5995 * doc/invoke.texi: Ditto.
5997 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5999 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
6002 2020-01-01 Jakub Jelinek <jakub@redhat.com>
6004 PR tree-optimization/93098
6005 * match.pd (popcount): For shift amounts, use integer_onep
6006 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
6007 tests. Make sure that precision is power of two larger than or equal
6008 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
6009 instead of ULL suffixed constants. Formatting fixes.
6011 Copyright (C) 2020 Free Software Foundation, Inc.
6013 Copying and distribution of this file, with or without modification,
6014 are permitted in any medium without royalty provided the copyright
6015 notice and this notice are preserved.