1 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3 * doc/invoke.texi (-msve-vector-bits=): Document that
4 -msve-vector-bits=128 now generates VL-specific code for
6 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
7 build_vector_type_for_mode to construct the data vector types.
8 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
9 VL-specific code for -msve-vector-bits=128 on little-endian targets.
10 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
13 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
15 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
18 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
20 * config/aarch64/aarch64-builtins.c
21 (aarch64_builtin_vectorized_function): Check for specific vector modes,
22 rather than checking the number of elements and the element mode.
24 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
26 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27 get_related_vectype_for_scalar_type rather than build_vector_type
28 to create the index type for a conditional reduction.
30 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
32 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
33 for any type of gather or scatter, including strided accesses.
35 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
37 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
40 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
42 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
44 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
45 parameter and its use to reset DR_OFFSET's.
46 (vect_transform_loop): Remove orig_drs_init argument.
47 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
48 member of dr_vec_info rather than the offset of the associated
49 data_reference's innermost_loop_behavior.
50 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
51 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
52 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
54 (vectorizable_store): Likewise.
55 (vectorizable_load): Likewise.
57 2020-01-10 Richard Biener <rguenther@suse.de>
59 * gimple-ssa-store-merging
60 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
62 2020-01-10 Martin Liska <mliska@suse.cz>
65 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
66 encapsulation that was there before r280040.
68 2020-01-10 Richard Biener <rguenther@suse.de>
71 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
72 sequences to avoid walking them again for secondary opportunities.
73 (pass_lower_eh_dispatch::execute): Instead actually insert
76 2020-01-10 Richard Biener <rguenther@suse.de>
79 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
80 (cleanup_all_empty_eh): Walk landing pads in reverse order to
83 2020-01-10 Martin Jambor <mjambor@suse.cz>
85 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
86 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
87 to get param_ipa_sra_max_replacements.
88 (param_splitting_across_edge): Pass the caller to
89 pull_accesses_from_callee.
91 2020-01-10 Martin Jambor <mjambor@suse.cz>
93 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
94 * ipa-cp.c (max_new_size): Removed.
95 (orig_overall_size): New variable.
96 (get_max_overall_size): New function.
97 (estimate_local_effects): Use it. Adjust dump.
98 (decide_about_value): Likewise.
99 (ipcp_propagate_stage): Do not calculate max_new_size, just store
100 orig_overall_size. Adjust dump.
101 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
103 2020-01-10 Martin Jambor <mjambor@suse.cz>
105 * params.opt (param_ipa_max_agg_items): Mark as Optimization
106 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
107 instead of param_ipa_max_agg_items.
108 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
109 optimization info for the callee.
111 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
113 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
114 markers if debug_inline_points is false.
116 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
118 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
120 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
121 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
122 aarch64-sve-builtins-sve2.h.
123 (aarch64-sve-builtins-sve2.o): New rule.
124 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
125 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
126 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
127 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
128 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
129 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
131 * config/aarch64/aarch64-sve.md: Update comments with SVE2
132 instructions that are handled here.
133 (@cond_asrd<mode>): Generalize to...
134 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
135 (*cond_asrd<mode>_2): Generalize to...
136 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
137 (*cond_asrd<mode>_z): Generalize to...
138 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
139 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
140 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
141 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
142 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
144 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
145 (@aarch64_scatter_stnt<mode>): Likewise.
146 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
147 (@aarch64_mul_lane_<mode>): Likewise.
148 (@aarch64_sve_suqadd<mode>_const): Likewise.
149 (*<sur>h<addsub><mode>): Generalize to...
150 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
152 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
153 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
154 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
155 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
156 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
157 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
158 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
159 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
160 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
161 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
162 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
163 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
164 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
165 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
166 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
167 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
168 (@aarch64_sve2_xar<mode>): Likewise.
169 (@aarch64_sve2_bcax<mode>): Likewise.
170 (*aarch64_sve2_eor3<mode>): Rename to...
171 (@aarch64_sve2_eor3<mode>): ...this.
172 (@aarch64_sve2_bsl<mode>): New expander.
173 (@aarch64_sve2_nbsl<mode>): Likewise.
174 (@aarch64_sve2_bsl1n<mode>): Likewise.
175 (@aarch64_sve2_bsl2n<mode>): Likewise.
176 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
177 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
178 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
179 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
180 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
181 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
182 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
183 (<su>mull<bt><Vwide>): Generalize to...
184 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
186 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
187 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
188 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
189 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
190 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
191 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
192 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
193 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
194 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
195 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
196 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
197 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
198 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
199 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
200 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
201 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
202 (<SHRNB:r>shrnb<mode>): Generalize to...
203 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
205 (<SHRNT:r>shrnt<mode>): Generalize to...
206 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
208 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
209 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
210 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
211 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
212 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
213 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
214 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
215 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
216 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
217 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
218 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
219 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
220 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
221 (@aarch64_sve2_cvtnt<mode>): Likewise.
222 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
223 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
224 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
225 (@aarch64_sve2_cvtxnt<mode>): Likewise.
226 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
227 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
228 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
229 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
230 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
231 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
232 (@aarch64_sve2_pmul<mode>): Likewise.
233 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
234 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
235 (@aarch64_sve2_tbl2<mode>): Likewise.
236 (@aarch64_sve2_tbx<mode>): Likewise.
237 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
238 (@aarch64_sve2_histcnt<mode>): Likewise.
239 (@aarch64_sve2_histseg<mode>): Likewise.
240 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
241 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
242 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
243 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
244 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
245 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
246 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
247 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
248 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
249 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
250 (SVE2_PMULL_PAIR_I): New mode iterators.
251 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
252 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
253 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
254 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
255 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
256 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
257 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
258 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
259 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
260 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
261 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
262 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
263 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
264 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
265 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
266 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
267 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
268 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
269 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
270 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
271 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
272 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
273 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
274 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
275 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
276 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
277 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
278 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
279 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
280 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
281 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
283 (VNARROW, Ventype): New mode attributes.
284 (Vewtype): Handle VNx2DI. Fix typo in comment.
285 (VDOUBLE): New mode attribute.
286 (sve_lane_con): Handle VNx8HI.
287 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
288 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
289 (sve_int_op, sve_int_op_rev): Handle the above codes.
290 (sve_pred_int_rhs2_operand): Likewise.
291 (MULLBT, SHRNB, SHRNT): Delete.
292 (SVE_INT_SHIFT_IMM): New int iterator.
293 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
294 and UNSPEC_WHILEHS for TARGET_SVE2.
295 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
296 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
297 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
298 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
299 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
300 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
301 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
302 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
303 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
304 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
305 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
306 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
307 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
308 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
309 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
310 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
311 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
312 (optab): Handle the new unspecs.
313 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
315 (lr): Handle the new unspecs.
317 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
318 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
319 (sve_int_qsub_op): New int attributes.
320 (sve_fp_op, rot): Handle the new unspecs.
321 * config/aarch64/aarch64-sve-builtins.h
322 (function_resolver::require_matching_pointer_type): Declare.
323 (function_resolver::resolve_unary): Add an optional boolean argument.
324 (function_resolver::finish_opt_n_resolution): Add an optional
325 type_suffix_index argument.
326 (gimple_folder::redirect_call): Declare.
327 (gimple_expander::prepare_gather_address_operands): Add an optional
329 * config/aarch64/aarch64-sve-builtins.cc: Include
330 aarch64-sve-builtins-sve2.h.
331 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
332 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
333 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
334 (TYPES_hsd_integer): Use TYPES_hsd_signed.
335 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
336 (TYPES_s_unsigned): Likewise.
337 (TYPES_s_integer): Use TYPES_s_unsigned.
338 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
339 (TYPES_sd_integer): Use them.
340 (TYPES_d_unsigned): New macro.
341 (TYPES_d_integer): Use it.
342 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
343 (TYPES_cvt_narrow): Likewise.
344 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
345 (preds_mx): New variable.
346 (function_builder::add_overloaded_function): Allow the new feature
347 set to be more restrictive than the original one.
348 (function_resolver::infer_pointer_type): Remove qualifiers from
349 the pointer type before printing it.
350 (function_resolver::require_matching_pointer_type): New function.
351 (function_resolver::resolve_sv_displacement): Handle functions
352 that don't support 32-bit vector indices or svint32_t vector offsets.
353 (function_resolver::finish_opt_n_resolution): Take the inferred type
354 as a separate argument.
355 (function_resolver::resolve_unary): Optionally treat all forms in
356 the same way as normal merging functions.
357 (gimple_folder::redirect_call): New function.
358 (function_expander::prepare_gather_address_operands): Add an argument
359 that says whether scaled forms are available. If they aren't,
360 handle scaling of vector indices and don't add the extension and
362 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
363 fall back to using cond_* instead.
364 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
365 Split out the member variables into...
366 (rtx_code_function_base): ...this new base class.
367 (rtx_code_function_rotated): Inherit rtx_code_function_base.
368 (unspec_based_function): Split out the member variables into...
369 (unspec_based_function_base): ...this new base class.
370 (unspec_based_function_rotated): Inherit unspec_based_function_base.
371 (unspec_based_function_exact_insn): New class.
372 (unspec_based_add_function, unspec_based_add_lane_function)
373 (unspec_based_lane_function, unspec_based_pred_function)
374 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
375 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
376 (unspec_based_sub_function, unspec_based_sub_lane_function): New
378 (unspec_based_fused_function): New class.
379 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
380 (unspec_based_fused_lane_function): New class.
381 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
383 (CODE_FOR_MODE1): New macro.
384 (fixed_insn_function): New class.
385 (while_comparison): Likewise.
386 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
387 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
388 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
389 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
390 (load_gather_sv_restricted, shift_left_imm_long): Declare.
391 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
392 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
393 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
394 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
395 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
396 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
397 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
398 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
399 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
400 Also add an initial argument for unary_convert_narrowt, regardless
401 of the predication type.
402 (build_32_64): Allow loads and stores to specify MODE_none.
403 (build_sv_index64, build_sv_uint_offset): New functions.
404 (long_type_suffix): New function.
405 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
406 (binary_imm_long_base, load_gather_sv_base): Likewise.
407 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
408 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
409 (unary_narrowb_base, unary_narrowt_base): Likewise.
410 (binary_long_lane_def, binary_long_lane): New shape.
411 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
412 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
413 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
414 (binary_to_uint_def, binary_to_uint): Likewise.
415 (binary_wide_def, binary_wide): Likewise.
416 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
417 (compare_def, compare): Likewise.
418 (compare_ptr_def, compare_ptr): Likewise.
419 (load_ext_gather_index_restricted_def,
420 load_ext_gather_index_restricted): Likewise.
421 (load_ext_gather_offset_restricted_def,
422 load_ext_gather_offset_restricted): Likewise.
423 (load_gather_sv_def): Inherit from load_gather_sv_base.
424 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
425 (shift_left_imm_def, shift_left_imm): Likewise.
426 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
427 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
428 (store_scatter_index_restricted_def,
429 store_scatter_index_restricted): Likewise.
430 (store_scatter_offset_restricted_def,
431 store_scatter_offset_restricted): Likewise.
432 (tbl_tuple_def, tbl_tuple): Likewise.
433 (ternary_long_lane_def, ternary_long_lane): Likewise.
434 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
435 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
436 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
437 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
438 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
439 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
440 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
441 (ternary_uint_def, ternary_uint): Likewise.
442 (unary_convert): Fix typo in comment.
443 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
444 (unary_long_def, unary_long): Likewise.
445 (unary_narrowb_def, unary_narrowb): Likewise.
446 (unary_narrowt_def, unary_narrowt): Likewise.
447 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
448 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
449 (unary_to_int_def, unary_to_int): Likewise.
450 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
451 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
452 (svasrd_impl): Delete.
453 (svcadd_impl::expand): Handle integer operations too.
454 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
455 new functions to derive the unspec numbers.
456 (svmla_svmls_lane_impl): Replace with...
457 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
458 integer operations too.
459 (svwhile_impl): Rename to...
460 (svwhilelx_impl): ...this and inherit from while_comparison.
461 (svasrd): Use unspec_based_function.
462 (svmla_lane): Use svmla_lane_impl.
463 (svmls_lane): Use svmls_lane_impl.
464 (svrecpe, svrsqrte): Handle unsigned integer operations too.
465 (svwhilele, svwhilelt): Use svwhilelx_impl.
466 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
467 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
468 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
469 * config/aarch64/aarch64-sve-builtins.def: Include
470 aarch64-sve-builtins-sve2.def.
472 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
474 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
475 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
476 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
477 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
478 immediates as well as vector ones.
479 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
480 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
481 (aarch64_sve_qsub_immediate): Update calls accordingly.
483 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
485 * config/aarch64/aarch64-sve2.md: Add banner comments.
486 (<su>mulh<r>s<mode>3): Move further up file.
487 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
488 (*aarch64_sve2_sra<mode>): Move further down file.
489 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
491 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
493 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
495 (while_optab_cmp): Handle them.
496 * config/aarch64/aarch64-sve.md
497 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
498 and add a "@" marker.
499 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
500 instead of gen_aarch64_sve2_while_ptest.
501 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
503 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
505 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
506 (UNSPEC_WHILELE): ...this.
507 (UNSPEC_WHILE_LO): Rename to...
508 (UNSPEC_WHILELO): ...this.
509 (UNSPEC_WHILE_LS): Rename to...
510 (UNSPEC_WHILELS): ...this.
511 (UNSPEC_WHILE_LT): Rename to...
512 (UNSPEC_WHILELT): ...this.
513 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
514 (cmp_op, while_optab_cmp): Likewise.
515 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
516 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
517 (svwhilelt): Likewise.
519 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
521 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
522 (unary_to_uint): Define.
523 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
524 (unary_count): Rename to...
525 (unary_to_uint_def, unary_to_uint): ...this.
526 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
528 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
530 * config/aarch64/aarch64-sve-builtins-functions.h
531 (code_for_mode_function): New class.
532 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
533 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
534 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
535 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
536 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
538 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
540 * config/aarch64/iterators.md (addsub): New code attribute.
541 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
543 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
544 in the asm string and attributes. Fix indentation.
545 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
547 (@aarch64_sve_<optab><mode>): ...this.
548 * config/aarch64/aarch64-sve-builtins.h
549 (function_expander::expand_signed_unpred_op): Delete.
550 * config/aarch64/aarch64-sve-builtins.cc
551 (function_expander::expand_signed_unpred_op): Likewise.
552 (function_expander::map_to_rtx_codes): If the optab isn't defined,
553 try using code_for_aarch64_sve instead.
554 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
555 (svqsub_impl): Likewise.
556 (svqadd, svqsub): Use rtx_code_function instead.
558 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
560 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
561 (HADDSUB, sur, addsub): Remove them.
563 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
565 * tree-nrv.c (pass_return_slot::execute): Handle all internal
566 functions the same way, rather than singling out those that
567 aren't mapped directly to optabs.
569 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
571 * target.def (compatible_vector_types_p): New target hook.
572 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
573 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
574 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
575 * doc/tm.texi: Regenerate.
576 * gimple-expr.c: Include target.h.
577 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
578 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
580 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
581 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
582 Use the original predicate if it already has a suitable type.
584 2020-01-09 Martin Jambor <mjambor@suse.cz>
586 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
587 resolve_speculation and redirect_call_stmt_to_callee static. Change
588 return type of set_call_stmt to cgraph_edge *.
589 * auto-profile.c (afdo_indirect_call): Adjust call to
590 redirect_call_stmt_to_callee.
591 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
592 make the this pointer explicit, adjust self-recursive calls and the
593 call top make_direct. Return the resulting edge.
594 (cgraph_edge::remove): Make this pointer explicit.
595 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
596 (cgraph_edge::make_direct): Likewise, adjust call to
598 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
599 call to set_call_stmt.
600 (cgraph_update_edges_for_call_stmt_node): Update call to
601 set_call_stmt and remove.
602 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
603 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
604 (cgraph_node::create_edge_including_clones): Moved "first" definition
605 of edge to the block where it was used. Adjusted calls to
607 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
609 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
610 make_direct and redirect_call_stmt_to_callee.
611 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
612 resolve_speculation and make_direct.
613 * ipa-inline-transform.c (inline_transform): Adjust call to
614 redirect_call_stmt_to_callee.
615 (check_speculations_1):: Adjust call to resolve_speculation.
616 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
618 (inline_small_functions): Adjust call to resolve_speculation.
619 (ipa_inline): Likewise.
620 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
622 * ipa-visibility.c (function_and_variable_visibility): Make iteration
623 safe with regards to edge removal, adjust calls to
624 redirect_call_stmt_to_callee.
625 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
626 and redirect_call_stmt_to_callee.
627 * multiple_target.c (create_dispatcher_calls): Adjust call to
628 redirect_call_stmt_to_callee
629 (redirect_to_specific_clone): Likewise.
630 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
631 Adjust calls to cgraph_edge::remove.
632 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
633 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
634 (expand_call_inline): Adjust call to cgraph_edge::remove.
636 2020-01-09 Martin Liska <mliska@suse.cz>
638 * params.opt: Set Optimization for
639 param_max_speculative_devirt_maydefs.
641 2020-01-09 Martin Sebor <msebor@redhat.com>
645 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
647 2020-01-09 Martin Liska <mliska@suse.cz>
649 * auto-profile.c (auto_profile): Use opt_for_fn
651 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
652 (propagate_vals_across_arith_jfunc): Likewise.
653 (hint_time_bonus): Likewise.
654 (incorporate_penalties): Likewise.
655 (good_cloning_opportunity_p): Likewise.
656 (perform_estimation_of_a_value): Likewise.
657 (estimate_local_effects): Likewise.
658 (ipcp_propagate_stage): Likewise.
659 * ipa-fnsummary.c (decompose_param_expr): Likewise.
660 (set_switch_stmt_execution_predicate): Likewise.
661 (analyze_function_body): Likewise.
662 * ipa-inline-analysis.c (offline_size): Likewise.
663 * ipa-inline.c (early_inliner): Likewise.
664 * ipa-prop.c (ipa_analyze_node): Likewise.
665 (ipcp_transform_function): Likewise.
666 * ipa-sra.c (process_scan_results): Likewise.
667 (ipa_sra_summarize_function): Likewise.
668 * params.opt: Rename ipcp-unit-growth to
669 ipa-cp-unit-growth. Add Optimization for various
670 IPA-related parameters.
672 2020-01-09 Richard Biener <rguenther@suse.de>
675 * gimplify.c (gimplify_expr): Deal with NOP definitions.
677 2020-01-09 Richard Biener <rguenther@suse.de>
679 PR tree-optimization/93040
680 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
682 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
684 * common/config/avr/avr-common.c (avr_option_optimization_table)
685 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
687 2020-01-09 Martin Liska <mliska@suse.cz>
689 * cgraphclones.c (symbol_table::materialize_all_clones):
690 Use cgraph_node::dump_name.
692 2020-01-09 Jakub Jelinek <jakub@redhat.com>
695 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
696 output_operand_lossage instead of gcc_unreachable.
697 * doc/md.texi (riscv f constraint): Fix typo.
700 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
701 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
702 CONST_SCALAR_INT_P instead of CONST_INT_P.
703 (*subv<mode>4_1): Rename to ...
704 (subv<mode>4_1): ... this.
705 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
706 define_insn_and_split patterns.
707 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
710 2020-01-08 David Malcolm <dmalcolm@redhat.com>
712 * vec.c (class selftest::count_dtor): New class.
713 (selftest::test_auto_delete_vec): New test.
714 (selftest::vec_c_tests): Call it.
715 * vec.h (class auto_delete_vec): New class template.
716 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
718 2020-01-08 David Malcolm <dmalcolm@redhat.com>
720 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
722 2020-01-08 Jim Wilson <jimw@sifive.com>
724 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
725 use of TLS_MODEL_LOCAL_EXEC when not pic.
727 2020-01-08 David Malcolm <dmalcolm@redhat.com>
729 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
732 2020-01-08 Jakub Jelinek <jakub@redhat.com>
735 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
736 *stack_protect_set_3 peephole2): Also check that the second
737 insns source is general_operand.
740 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
741 predicate for output operand instead of register_operand.
742 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
743 memory destination and non-memory operands[2].
745 2020-01-08 Martin Liska <mliska@suse.cz>
747 * cgraph.c (cgraph_node::dump): Use ::dump_name or
748 ::dump_asm_name instead of (::name or ::asm_name).
749 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
750 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
751 (analyze_functions): Likewise.
752 (expand_all_functions): Likewise.
753 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
754 (propagate_bits_across_jump_function): Likewise.
755 (dump_profile_updates): Likewise.
756 (ipcp_store_bits_results): Likewise.
757 (ipcp_store_vr_results): Likewise.
758 * ipa-devirt.c (dump_targets): Likewise.
759 * ipa-fnsummary.c (analyze_function_body): Likewise.
760 * ipa-hsa.c (check_warn_node_versionable): Likewise.
761 (process_hsa_functions): Likewise.
762 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
763 (set_alias_uids): Likewise.
764 * ipa-inline-transform.c (save_inline_function_body): Likewise.
765 * ipa-inline.c (recursive_inlining): Likewise.
766 (inline_to_all_callers_1): Likewise.
767 (ipa_inline): Likewise.
768 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
769 (ipa_propagate_frequency): Likewise.
770 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
771 (remove_described_reference): Likewise.
772 * ipa-pure-const.c (worse_state): Likewise.
773 (check_retval_uses): Likewise.
774 (analyze_function): Likewise.
775 (propagate_pure_const): Likewise.
776 (propagate_nothrow): Likewise.
777 (dump_malloc_lattice): Likewise.
778 (propagate_malloc): Likewise.
779 (pass_local_pure_const::execute): Likewise.
780 * ipa-visibility.c (optimize_weakref): Likewise.
781 (function_and_variable_visibility): Likewise.
782 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
783 (ipa_discover_variable_flags): Likewise.
784 * lto-streamer-out.c (output_function): Likewise.
785 (output_constructor): Likewise.
786 * tree-inline.c (copy_bb): Likewise.
787 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
788 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
790 2020-01-08 Richard Biener <rguenther@suse.de>
793 * tree-eh.c (sink_clobbers): Update virtual operands for
794 the first and last stmt only. Add a dry-run capability.
795 (pass_lower_eh_dispatch::execute): Perform clobber sinking
796 after CFG manipulations and in RPO order to catch all
797 secondary opportunities reliably.
799 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
802 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
804 2019-01-08 Richard Biener <rguenther@suse.de>
807 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
808 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
809 virtual operand, also updating SSA use.
810 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
811 Update stmt after resetting virtual operand.
812 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
813 * gimple-iterator.c (gsi_remove): When not removing the stmt
814 permanently do not delink immediate uses or mark the stmt modified.
816 2020-01-08 Martin Liska <mliska@suse.cz>
818 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
819 (ipa_call_context::estimate_size_and_time): Likewise.
820 (inline_analyze_function): Likewise.
822 2020-01-08 Martin Liska <mliska@suse.cz>
824 * cgraph.c (cgraph_node::dump): Use systematically
827 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
829 Add -nodevicespecs option for avr.
832 * config/avr/avr.opt (-nodevicespecs): New driver option.
833 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
834 "-specs=device-specs/..." if that option is not set.
835 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
837 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
839 Implement 64-bit double functions for avr.
842 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
843 --with-double-comparison.
844 * doc/install.texi: Document them.
845 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
846 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
847 <WITH_DOUBLE_COMPARISON>: New built-in defines.
848 * doc/invoke.texi (AVR Built-in Macros): Document them.
849 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
850 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
851 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
853 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
856 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
857 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
858 when only building rm-profile multilibs.
860 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
863 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
864 lattice for a value to check.
865 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
866 finite propagation in self-recursive scc.
868 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
870 * ipa-inline.c (caller_growth_limits): Restore the AND.
872 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
874 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
875 (VEC_ALLREG_ALT): New iterator.
876 (VEC_ALLREG_INT_MODE): New iterator.
877 (VCMP_MODE): New iterator.
878 (VCMP_MODE_INT): New iterator.
879 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
880 (vec_cmp<u>v64qidi): New define_expand.
881 (vec_cmp<mode>di_exec): Use VCMP_MODE.
882 (vec_cmpu<mode>di_exec): New define_expand.
883 (vec_cmp<u>v64qidi_exec): New define_expand.
884 (vec_cmp<mode>di_dup): Use VCMP_MODE.
885 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
886 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
887 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
888 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
889 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
890 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
891 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
892 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
893 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
895 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
896 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
898 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
900 * config/gcn/constraints.md (DA): Update description and match.
902 (Db): New constraint.
903 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
905 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
906 Implement 'Db' mixed immediate type.
907 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
908 (addcv64si3_dup<exec_vcc>): Delete.
909 (subcv64si3<exec_vcc>): Rework constraints.
910 (addv64di3): Rework constraints.
911 (addv64di3_exec): Rework constraints.
912 (subv64di3): Rework constraints.
913 (addv64di3_dup): Delete.
914 (addv64di3_dup_exec): Delete.
915 (addv64di3_zext): Rework constraints.
916 (addv64di3_zext_exec): Rework constraints.
917 (addv64di3_zext_dup): Rework constraints.
918 (addv64di3_zext_dup_exec): Rework constraints.
919 (addv64di3_zext_dup2): Rework constraints.
920 (addv64di3_zext_dup2_exec): Rework constraints.
921 (addv64di3_sext_dup2): Rework constraints.
922 (addv64di3_sext_dup2_exec): Rework constraints.
924 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
926 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
927 existing target checks.
929 2020-01-07 Richard Biener <rguenther@suse.de>
931 * doc/install.texi: Bump minimal supported MPC version.
933 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
935 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
936 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
937 * langhooks.c: Include stor-layout.h.
938 (lhd_simulate_enum_decl): New function.
939 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
940 handle_arm_sve_h for the LTO frontend.
941 (register_vector_type): Cope with null returns from pushdecl.
943 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
945 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
946 (aarch64_sve::nvectors_if_data_type): Replace with...
947 (aarch64_sve::builtin_type_p): ...this.
948 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
949 (find_vector_type): Delete.
950 (add_sve_type_attribute): New function.
951 (lookup_sve_type_attribute): Likewise.
952 (register_builtin_types): Add an "SVE type" attribute to each type.
953 (register_tuple_type): Likewise.
954 (svbool_type_p, nvectors_if_data_type): Delete.
955 (mangle_builtin_type): Use lookup_sve_type_attribute.
956 (builtin_type_p): Likewise. Add an overload that returns the
957 number of constituent vector and predicate registers.
958 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
959 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
960 instead of aarch64_sve_argument_p.
961 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
962 (aarch64_pass_by_reference): Likewise.
963 (aarch64_function_value_1): Likewise.
964 (aarch64_return_in_memory): Likewise.
965 (aarch64_layout_arg): Likewise.
967 2020-01-07 Jakub Jelinek <jakub@redhat.com>
969 PR tree-optimization/93156
970 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
971 least significant bit is always clear.
973 PR tree-optimization/93118
974 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
975 simplifier with two intermediate conversions.
977 2020-01-07 Martin Liska <mliska@suse.cz>
979 * params.opt: Add Optimization for various parameters.
981 2020-01-07 Martin Liska <mliska@suse.cz>
984 * doc/extend.texi: Explain cloning for target_clone
987 2020-01-07 Martin Liska <mliska@suse.cz>
989 PR tree-optimization/92860
990 * common.opt: Make in Optimization option
991 as it is affected by -O0, which is an Optimization
993 * tree-inline.c (tree_inlinable_function_p):
994 Use opt_for_fn for warn_inline.
995 (expand_call_inline): Likewise.
997 2020-01-07 Martin Liska <mliska@suse.cz>
999 PR tree-optimization/92860
1000 * common.opt: Make flag_ree as optimization
1003 2020-01-07 Martin Liska <mliska@suse.cz>
1005 PR optimization/92860
1006 * params.opt: Mark param_min_crossjump_insns with Optimization
1009 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
1011 * ipa-inline-analysis.c (estimate_growth): Fix typo.
1012 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
1014 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
1016 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
1017 helper function to return the valid addressing formats for a given
1018 hard register and mode.
1019 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
1021 * config/rs6000/constraints.md (Q constraint): Update
1023 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
1026 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
1027 Use 'Q' for doing vector extract from memory.
1028 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
1030 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
1031 doing vector extract from memory.
1032 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
1033 extract from memory.
1035 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
1036 for the offset being 34-bits when -mcpu=future is used.
1038 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
1040 * config/pa/pa.md: Revert change to use ordered_comparison_operator
1041 instead of cmpib_comparison_operator in cmpib patterns.
1042 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
1043 of cmpib_comparison_operator. Revise comment.
1045 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1047 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
1048 in an IFN_DIV_POW2 node to be equal.
1050 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1052 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
1053 (vect_check_scalar_mask): ...this.
1054 (vectorizable_store, vectorizable_load): Update call accordingly.
1055 (vectorizable_call): Use vect_check_scalar_mask to check the mask
1056 argument in calls to conditional internal functions.
1058 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1060 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
1061 '0' matching inputs.
1062 (subv64di3_exec): Likewise.
1064 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
1066 * config/mips/mips.c (vr4130_align_insns): Fix typo.
1067 * doc/md.texi (movstr): Likewise.
1069 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1071 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
1074 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1076 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
1078 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
1079 to a temporary file and use move-if-change to update the real
1080 file where necessary.
1082 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1084 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
1085 rather than Upa for CPY /M.
1087 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1089 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
1092 2020-01-06 Martin Liska <mliska@suse.cz>
1094 PR tree-optimization/92860
1095 * params.opt: Mark param_max_combine_insns with Optimization
1098 2020-01-05 Jakub Jelinek <jakub@redhat.com>
1101 * config/i386/i386.md (SWIDWI): New mode iterator.
1102 (DWI, dwi): Add TImode variants.
1103 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
1104 <general_hilo_operand> instead of <general_operand>. Use
1105 CONST_SCALAR_INT_P instead of CONST_INT_P.
1106 (*addv<mode>4_1): Rename to ...
1107 (addv<mode>4_1): ... this.
1108 (QWI): New mode attribute.
1109 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
1110 define_insn_and_split patterns.
1111 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
1113 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
1114 <general_hilo_operand> instead of <general_operand>.
1115 (*addcarry<mode>_1): New define_insn.
1116 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
1118 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
1120 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
1121 Use "call" instead of "set".
1123 2020-01-03 Martin Jambor <mjambor@suse.cz>
1126 * ipa-cp.c (print_all_lattices): Skip functions without info.
1128 2020-01-03 Jakub Jelinek <jakub@redhat.com>
1131 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
1132 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
1133 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
1134 for 'e' simd clones.
1137 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
1139 (mprefer-vector-width=): Add Save.
1140 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
1141 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
1142 (ix86_debug_options, ix86_function_specific_print): Adjust
1143 ix86_target_string callers.
1144 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
1145 (ix86_valid_target_attribute_tree): Likewise.
1146 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
1147 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
1148 ix86_target_string caller.
1151 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
1152 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
1153 instead of gen_int_shift_amount + convert_modes.
1155 PR rtl-optimization/93088
1156 * loop-iv.c (find_single_def_src): Punt after looking through
1157 128 reg copies for regs with single definitions. Move definitions
1160 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
1162 * config/arm/arm-c.c (arm_cpu_builtins): Define
1163 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
1164 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
1165 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
1166 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
1167 * config/arm/arm-tables.opt: Regenerated.
1168 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
1169 arm_arch_i8mm and arm_arch_bf16 when enabled.
1170 * config/arm/arm.h (TARGET_I8MM): New macro.
1171 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
1172 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
1173 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
1174 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
1175 (v8_6_a_simd_variants): New.
1176 (v8_*_a_simd_variants): Add i8mm and bf16.
1177 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
1179 2020-01-02 Jakub Jelinek <jakub@redhat.com>
1182 * predict.c (compute_function_frequency): Don't call
1183 warn_function_cold on functions that already have cold attribute.
1185 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
1188 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
1189 COMDAT group function labels in .data.rel.ro.local section.
1190 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
1193 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
1194 comparison_operator in B and S integer comparisons. Likewise, use
1195 ordered_comparison_operator instead of cmpib_comparison_operator in
1197 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
1199 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1201 Update copyright years.
1203 * gcc.c (process_command): Update copyright notice dates.
1204 * gcov-dump.c (print_version): Ditto.
1205 * gcov.c (print_version): Ditto.
1206 * gcov-tool.c (print_version): Ditto.
1207 * gengtype.c (create_file): Ditto.
1208 * doc/cpp.texi: Bump @copying's copyright year.
1209 * doc/cppinternals.texi: Ditto.
1210 * doc/gcc.texi: Ditto.
1211 * doc/gccint.texi: Ditto.
1212 * doc/gcov.texi: Ditto.
1213 * doc/install.texi: Ditto.
1214 * doc/invoke.texi: Ditto.
1216 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
1218 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
1221 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1223 PR tree-optimization/93098
1224 * match.pd (popcount): For shift amounts, use integer_onep
1225 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
1226 tests. Make sure that precision is power of two larger than or equal
1227 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
1228 instead of ULL suffixed constants. Formatting fixes.
1230 Copyright (C) 2020 Free Software Foundation, Inc.
1232 Copying and distribution of this file, with or without modification,
1233 are permitted in any medium without royalty provided the copyright
1234 notice and this notice are preserved.