1 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
6 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
8 * config/aarch64/aarch64-builtins.c
9 (aarch64_builtin_vectorized_function): Check for specific vector modes,
10 rather than checking the number of elements and the element mode.
12 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
14 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
15 get_related_vectype_for_scalar_type rather than build_vector_type
16 to create the index type for a conditional reduction.
18 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
20 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
21 for any type of gather or scatter, including strided accesses.
23 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
25 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
28 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
30 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
32 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
33 parameter and its use to reset DR_OFFSET's.
34 (vect_transform_loop): Remove orig_drs_init argument.
35 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
36 member of dr_vec_info rather than the offset of the associated
37 data_reference's innermost_loop_behavior.
38 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
39 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
40 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
42 (vectorizable_store): Likewise.
43 (vectorizable_load): Likewise.
45 2020-01-10 Richard Biener <rguenther@suse.de>
47 * gimple-ssa-store-merging
48 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
50 2020-01-10 Martin Liska <mliska@suse.cz>
53 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
54 encapsulation that was there before r280040.
56 2020-01-10 Richard Biener <rguenther@suse.de>
59 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
60 sequences to avoid walking them again for secondary opportunities.
61 (pass_lower_eh_dispatch::execute): Instead actually insert
64 2020-01-10 Richard Biener <rguenther@suse.de>
67 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
68 (cleanup_all_empty_eh): Walk landing pads in reverse order to
71 2020-01-10 Martin Jambor <mjambor@suse.cz>
73 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
74 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
75 to get param_ipa_sra_max_replacements.
76 (param_splitting_across_edge): Pass the caller to
77 pull_accesses_from_callee.
79 2020-01-10 Martin Jambor <mjambor@suse.cz>
81 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
82 * ipa-cp.c (max_new_size): Removed.
83 (orig_overall_size): New variable.
84 (get_max_overall_size): New function.
85 (estimate_local_effects): Use it. Adjust dump.
86 (decide_about_value): Likewise.
87 (ipcp_propagate_stage): Do not calculate max_new_size, just store
88 orig_overall_size. Adjust dump.
89 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
91 2020-01-10 Martin Jambor <mjambor@suse.cz>
93 * params.opt (param_ipa_max_agg_items): Mark as Optimization
94 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
95 instead of param_ipa_max_agg_items.
96 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
97 optimization info for the callee.
99 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
101 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
102 markers if debug_inline_points is false.
104 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
106 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
108 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
109 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
110 aarch64-sve-builtins-sve2.h.
111 (aarch64-sve-builtins-sve2.o): New rule.
112 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
113 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
114 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
115 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
116 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
117 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
119 * config/aarch64/aarch64-sve.md: Update comments with SVE2
120 instructions that are handled here.
121 (@cond_asrd<mode>): Generalize to...
122 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
123 (*cond_asrd<mode>_2): Generalize to...
124 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
125 (*cond_asrd<mode>_z): Generalize to...
126 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
127 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
128 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
129 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
130 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
132 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
133 (@aarch64_scatter_stnt<mode>): Likewise.
134 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
135 (@aarch64_mul_lane_<mode>): Likewise.
136 (@aarch64_sve_suqadd<mode>_const): Likewise.
137 (*<sur>h<addsub><mode>): Generalize to...
138 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
140 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
141 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
142 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
143 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
144 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
145 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
146 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
147 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
148 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
149 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
150 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
151 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
152 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
153 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
154 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
155 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
156 (@aarch64_sve2_xar<mode>): Likewise.
157 (@aarch64_sve2_bcax<mode>): Likewise.
158 (*aarch64_sve2_eor3<mode>): Rename to...
159 (@aarch64_sve2_eor3<mode>): ...this.
160 (@aarch64_sve2_bsl<mode>): New expander.
161 (@aarch64_sve2_nbsl<mode>): Likewise.
162 (@aarch64_sve2_bsl1n<mode>): Likewise.
163 (@aarch64_sve2_bsl2n<mode>): Likewise.
164 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
165 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
166 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
167 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
168 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
169 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
170 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
171 (<su>mull<bt><Vwide>): Generalize to...
172 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
174 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
175 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
176 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
177 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
178 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
179 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
180 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
181 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
182 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
183 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
184 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
185 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
186 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
187 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
188 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
189 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
190 (<SHRNB:r>shrnb<mode>): Generalize to...
191 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
193 (<SHRNT:r>shrnt<mode>): Generalize to...
194 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
196 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
197 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
198 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
199 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
200 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
201 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
202 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
203 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
204 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
205 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
206 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
207 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
208 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
209 (@aarch64_sve2_cvtnt<mode>): Likewise.
210 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
211 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
212 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
213 (@aarch64_sve2_cvtxnt<mode>): Likewise.
214 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
215 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
216 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
217 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
218 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
219 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
220 (@aarch64_sve2_pmul<mode>): Likewise.
221 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
222 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
223 (@aarch64_sve2_tbl2<mode>): Likewise.
224 (@aarch64_sve2_tbx<mode>): Likewise.
225 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
226 (@aarch64_sve2_histcnt<mode>): Likewise.
227 (@aarch64_sve2_histseg<mode>): Likewise.
228 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
229 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
230 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
231 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
232 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
233 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
234 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
235 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
236 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
237 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
238 (SVE2_PMULL_PAIR_I): New mode iterators.
239 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
240 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
241 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
242 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
243 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
244 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
245 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
246 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
247 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
248 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
249 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
250 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
251 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
252 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
253 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
254 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
255 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
256 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
257 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
258 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
259 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
260 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
261 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
262 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
263 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
264 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
265 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
266 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
267 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
268 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
269 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
271 (VNARROW, Ventype): New mode attributes.
272 (Vewtype): Handle VNx2DI. Fix typo in comment.
273 (VDOUBLE): New mode attribute.
274 (sve_lane_con): Handle VNx8HI.
275 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
276 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
277 (sve_int_op, sve_int_op_rev): Handle the above codes.
278 (sve_pred_int_rhs2_operand): Likewise.
279 (MULLBT, SHRNB, SHRNT): Delete.
280 (SVE_INT_SHIFT_IMM): New int iterator.
281 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
282 and UNSPEC_WHILEHS for TARGET_SVE2.
283 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
284 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
285 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
286 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
287 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
288 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
289 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
290 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
291 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
292 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
293 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
294 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
295 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
296 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
297 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
298 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
299 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
300 (optab): Handle the new unspecs.
301 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
303 (lr): Handle the new unspecs.
305 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
306 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
307 (sve_int_qsub_op): New int attributes.
308 (sve_fp_op, rot): Handle the new unspecs.
309 * config/aarch64/aarch64-sve-builtins.h
310 (function_resolver::require_matching_pointer_type): Declare.
311 (function_resolver::resolve_unary): Add an optional boolean argument.
312 (function_resolver::finish_opt_n_resolution): Add an optional
313 type_suffix_index argument.
314 (gimple_folder::redirect_call): Declare.
315 (gimple_expander::prepare_gather_address_operands): Add an optional
317 * config/aarch64/aarch64-sve-builtins.cc: Include
318 aarch64-sve-builtins-sve2.h.
319 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
320 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
321 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
322 (TYPES_hsd_integer): Use TYPES_hsd_signed.
323 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
324 (TYPES_s_unsigned): Likewise.
325 (TYPES_s_integer): Use TYPES_s_unsigned.
326 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
327 (TYPES_sd_integer): Use them.
328 (TYPES_d_unsigned): New macro.
329 (TYPES_d_integer): Use it.
330 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
331 (TYPES_cvt_narrow): Likewise.
332 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
333 (preds_mx): New variable.
334 (function_builder::add_overloaded_function): Allow the new feature
335 set to be more restrictive than the original one.
336 (function_resolver::infer_pointer_type): Remove qualifiers from
337 the pointer type before printing it.
338 (function_resolver::require_matching_pointer_type): New function.
339 (function_resolver::resolve_sv_displacement): Handle functions
340 that don't support 32-bit vector indices or svint32_t vector offsets.
341 (function_resolver::finish_opt_n_resolution): Take the inferred type
342 as a separate argument.
343 (function_resolver::resolve_unary): Optionally treat all forms in
344 the same way as normal merging functions.
345 (gimple_folder::redirect_call): New function.
346 (function_expander::prepare_gather_address_operands): Add an argument
347 that says whether scaled forms are available. If they aren't,
348 handle scaling of vector indices and don't add the extension and
350 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
351 fall back to using cond_* instead.
352 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
353 Split out the member variables into...
354 (rtx_code_function_base): ...this new base class.
355 (rtx_code_function_rotated): Inherit rtx_code_function_base.
356 (unspec_based_function): Split out the member variables into...
357 (unspec_based_function_base): ...this new base class.
358 (unspec_based_function_rotated): Inherit unspec_based_function_base.
359 (unspec_based_function_exact_insn): New class.
360 (unspec_based_add_function, unspec_based_add_lane_function)
361 (unspec_based_lane_function, unspec_based_pred_function)
362 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
363 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
364 (unspec_based_sub_function, unspec_based_sub_lane_function): New
366 (unspec_based_fused_function): New class.
367 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
368 (unspec_based_fused_lane_function): New class.
369 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
371 (CODE_FOR_MODE1): New macro.
372 (fixed_insn_function): New class.
373 (while_comparison): Likewise.
374 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
375 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
376 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
377 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
378 (load_gather_sv_restricted, shift_left_imm_long): Declare.
379 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
380 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
381 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
382 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
383 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
384 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
385 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
386 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
387 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
388 Also add an initial argument for unary_convert_narrowt, regardless
389 of the predication type.
390 (build_32_64): Allow loads and stores to specify MODE_none.
391 (build_sv_index64, build_sv_uint_offset): New functions.
392 (long_type_suffix): New function.
393 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
394 (binary_imm_long_base, load_gather_sv_base): Likewise.
395 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
396 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
397 (unary_narrowb_base, unary_narrowt_base): Likewise.
398 (binary_long_lane_def, binary_long_lane): New shape.
399 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
400 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
401 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
402 (binary_to_uint_def, binary_to_uint): Likewise.
403 (binary_wide_def, binary_wide): Likewise.
404 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
405 (compare_def, compare): Likewise.
406 (compare_ptr_def, compare_ptr): Likewise.
407 (load_ext_gather_index_restricted_def,
408 load_ext_gather_index_restricted): Likewise.
409 (load_ext_gather_offset_restricted_def,
410 load_ext_gather_offset_restricted): Likewise.
411 (load_gather_sv_def): Inherit from load_gather_sv_base.
412 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
413 (shift_left_imm_def, shift_left_imm): Likewise.
414 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
415 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
416 (store_scatter_index_restricted_def,
417 store_scatter_index_restricted): Likewise.
418 (store_scatter_offset_restricted_def,
419 store_scatter_offset_restricted): Likewise.
420 (tbl_tuple_def, tbl_tuple): Likewise.
421 (ternary_long_lane_def, ternary_long_lane): Likewise.
422 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
423 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
424 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
425 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
426 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
427 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
428 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
429 (ternary_uint_def, ternary_uint): Likewise.
430 (unary_convert): Fix typo in comment.
431 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
432 (unary_long_def, unary_long): Likewise.
433 (unary_narrowb_def, unary_narrowb): Likewise.
434 (unary_narrowt_def, unary_narrowt): Likewise.
435 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
436 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
437 (unary_to_int_def, unary_to_int): Likewise.
438 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
439 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
440 (svasrd_impl): Delete.
441 (svcadd_impl::expand): Handle integer operations too.
442 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
443 new functions to derive the unspec numbers.
444 (svmla_svmls_lane_impl): Replace with...
445 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
446 integer operations too.
447 (svwhile_impl): Rename to...
448 (svwhilelx_impl): ...this and inherit from while_comparison.
449 (svasrd): Use unspec_based_function.
450 (svmla_lane): Use svmla_lane_impl.
451 (svmls_lane): Use svmls_lane_impl.
452 (svrecpe, svrsqrte): Handle unsigned integer operations too.
453 (svwhilele, svwhilelt): Use svwhilelx_impl.
454 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
455 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
456 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
457 * config/aarch64/aarch64-sve-builtins.def: Include
458 aarch64-sve-builtins-sve2.def.
460 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
462 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
463 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
464 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
465 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
466 immediates as well as vector ones.
467 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
468 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
469 (aarch64_sve_qsub_immediate): Update calls accordingly.
471 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
473 * config/aarch64/aarch64-sve2.md: Add banner comments.
474 (<su>mulh<r>s<mode>3): Move further up file.
475 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
476 (*aarch64_sve2_sra<mode>): Move further down file.
477 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
479 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
481 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
483 (while_optab_cmp): Handle them.
484 * config/aarch64/aarch64-sve.md
485 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
486 and add a "@" marker.
487 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
488 instead of gen_aarch64_sve2_while_ptest.
489 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
491 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
493 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
494 (UNSPEC_WHILELE): ...this.
495 (UNSPEC_WHILE_LO): Rename to...
496 (UNSPEC_WHILELO): ...this.
497 (UNSPEC_WHILE_LS): Rename to...
498 (UNSPEC_WHILELS): ...this.
499 (UNSPEC_WHILE_LT): Rename to...
500 (UNSPEC_WHILELT): ...this.
501 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
502 (cmp_op, while_optab_cmp): Likewise.
503 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
504 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
505 (svwhilelt): Likewise.
507 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
509 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
510 (unary_to_uint): Define.
511 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
512 (unary_count): Rename to...
513 (unary_to_uint_def, unary_to_uint): ...this.
514 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
516 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
518 * config/aarch64/aarch64-sve-builtins-functions.h
519 (code_for_mode_function): New class.
520 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
521 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
522 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
523 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
524 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
526 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
528 * config/aarch64/iterators.md (addsub): New code attribute.
529 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
531 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
532 in the asm string and attributes. Fix indentation.
533 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
535 (@aarch64_sve_<optab><mode>): ...this.
536 * config/aarch64/aarch64-sve-builtins.h
537 (function_expander::expand_signed_unpred_op): Delete.
538 * config/aarch64/aarch64-sve-builtins.cc
539 (function_expander::expand_signed_unpred_op): Likewise.
540 (function_expander::map_to_rtx_codes): If the optab isn't defined,
541 try using code_for_aarch64_sve instead.
542 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
543 (svqsub_impl): Likewise.
544 (svqadd, svqsub): Use rtx_code_function instead.
546 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
548 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
549 (HADDSUB, sur, addsub): Remove them.
551 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
553 * tree-nrv.c (pass_return_slot::execute): Handle all internal
554 functions the same way, rather than singling out those that
555 aren't mapped directly to optabs.
557 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
559 * target.def (compatible_vector_types_p): New target hook.
560 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
561 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
562 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
563 * doc/tm.texi: Regenerate.
564 * gimple-expr.c: Include target.h.
565 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
566 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
568 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
569 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
570 Use the original predicate if it already has a suitable type.
572 2020-01-09 Martin Jambor <mjambor@suse.cz>
574 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
575 resolve_speculation and redirect_call_stmt_to_callee static. Change
576 return type of set_call_stmt to cgraph_edge *.
577 * auto-profile.c (afdo_indirect_call): Adjust call to
578 redirect_call_stmt_to_callee.
579 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
580 make the this pointer explicit, adjust self-recursive calls and the
581 call top make_direct. Return the resulting edge.
582 (cgraph_edge::remove): Make this pointer explicit.
583 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
584 (cgraph_edge::make_direct): Likewise, adjust call to
586 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
587 call to set_call_stmt.
588 (cgraph_update_edges_for_call_stmt_node): Update call to
589 set_call_stmt and remove.
590 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
591 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
592 (cgraph_node::create_edge_including_clones): Moved "first" definition
593 of edge to the block where it was used. Adjusted calls to
595 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
597 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
598 make_direct and redirect_call_stmt_to_callee.
599 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
600 resolve_speculation and make_direct.
601 * ipa-inline-transform.c (inline_transform): Adjust call to
602 redirect_call_stmt_to_callee.
603 (check_speculations_1):: Adjust call to resolve_speculation.
604 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
606 (inline_small_functions): Adjust call to resolve_speculation.
607 (ipa_inline): Likewise.
608 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
610 * ipa-visibility.c (function_and_variable_visibility): Make iteration
611 safe with regards to edge removal, adjust calls to
612 redirect_call_stmt_to_callee.
613 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
614 and redirect_call_stmt_to_callee.
615 * multiple_target.c (create_dispatcher_calls): Adjust call to
616 redirect_call_stmt_to_callee
617 (redirect_to_specific_clone): Likewise.
618 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
619 Adjust calls to cgraph_edge::remove.
620 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
621 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
622 (expand_call_inline): Adjust call to cgraph_edge::remove.
624 2020-01-09 Martin Liska <mliska@suse.cz>
626 * params.opt: Set Optimization for
627 param_max_speculative_devirt_maydefs.
629 2020-01-09 Martin Sebor <msebor@redhat.com>
633 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
635 2020-01-09 Martin Liska <mliska@suse.cz>
637 * auto-profile.c (auto_profile): Use opt_for_fn
639 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
640 (propagate_vals_across_arith_jfunc): Likewise.
641 (hint_time_bonus): Likewise.
642 (incorporate_penalties): Likewise.
643 (good_cloning_opportunity_p): Likewise.
644 (perform_estimation_of_a_value): Likewise.
645 (estimate_local_effects): Likewise.
646 (ipcp_propagate_stage): Likewise.
647 * ipa-fnsummary.c (decompose_param_expr): Likewise.
648 (set_switch_stmt_execution_predicate): Likewise.
649 (analyze_function_body): Likewise.
650 * ipa-inline-analysis.c (offline_size): Likewise.
651 * ipa-inline.c (early_inliner): Likewise.
652 * ipa-prop.c (ipa_analyze_node): Likewise.
653 (ipcp_transform_function): Likewise.
654 * ipa-sra.c (process_scan_results): Likewise.
655 (ipa_sra_summarize_function): Likewise.
656 * params.opt: Rename ipcp-unit-growth to
657 ipa-cp-unit-growth. Add Optimization for various
658 IPA-related parameters.
660 2020-01-09 Richard Biener <rguenther@suse.de>
663 * gimplify.c (gimplify_expr): Deal with NOP definitions.
665 2020-01-09 Richard Biener <rguenther@suse.de>
667 PR tree-optimization/93040
668 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
670 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
672 * common/config/avr/avr-common.c (avr_option_optimization_table)
673 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
675 2020-01-09 Martin Liska <mliska@suse.cz>
677 * cgraphclones.c (symbol_table::materialize_all_clones):
678 Use cgraph_node::dump_name.
680 2020-01-09 Jakub Jelinek <jakub@redhat.com>
683 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
684 output_operand_lossage instead of gcc_unreachable.
685 * doc/md.texi (riscv f constraint): Fix typo.
688 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
689 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
690 CONST_SCALAR_INT_P instead of CONST_INT_P.
691 (*subv<mode>4_1): Rename to ...
692 (subv<mode>4_1): ... this.
693 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
694 define_insn_and_split patterns.
695 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
698 2020-01-08 David Malcolm <dmalcolm@redhat.com>
700 * vec.c (class selftest::count_dtor): New class.
701 (selftest::test_auto_delete_vec): New test.
702 (selftest::vec_c_tests): Call it.
703 * vec.h (class auto_delete_vec): New class template.
704 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
706 2020-01-08 David Malcolm <dmalcolm@redhat.com>
708 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
710 2020-01-08 Jim Wilson <jimw@sifive.com>
712 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
713 use of TLS_MODEL_LOCAL_EXEC when not pic.
715 2020-01-08 David Malcolm <dmalcolm@redhat.com>
717 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
720 2020-01-08 Jakub Jelinek <jakub@redhat.com>
723 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
724 *stack_protect_set_3 peephole2): Also check that the second
725 insns source is general_operand.
728 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
729 predicate for output operand instead of register_operand.
730 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
731 memory destination and non-memory operands[2].
733 2020-01-08 Martin Liska <mliska@suse.cz>
735 * cgraph.c (cgraph_node::dump): Use ::dump_name or
736 ::dump_asm_name instead of (::name or ::asm_name).
737 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
738 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
739 (analyze_functions): Likewise.
740 (expand_all_functions): Likewise.
741 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
742 (propagate_bits_across_jump_function): Likewise.
743 (dump_profile_updates): Likewise.
744 (ipcp_store_bits_results): Likewise.
745 (ipcp_store_vr_results): Likewise.
746 * ipa-devirt.c (dump_targets): Likewise.
747 * ipa-fnsummary.c (analyze_function_body): Likewise.
748 * ipa-hsa.c (check_warn_node_versionable): Likewise.
749 (process_hsa_functions): Likewise.
750 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
751 (set_alias_uids): Likewise.
752 * ipa-inline-transform.c (save_inline_function_body): Likewise.
753 * ipa-inline.c (recursive_inlining): Likewise.
754 (inline_to_all_callers_1): Likewise.
755 (ipa_inline): Likewise.
756 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
757 (ipa_propagate_frequency): Likewise.
758 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
759 (remove_described_reference): Likewise.
760 * ipa-pure-const.c (worse_state): Likewise.
761 (check_retval_uses): Likewise.
762 (analyze_function): Likewise.
763 (propagate_pure_const): Likewise.
764 (propagate_nothrow): Likewise.
765 (dump_malloc_lattice): Likewise.
766 (propagate_malloc): Likewise.
767 (pass_local_pure_const::execute): Likewise.
768 * ipa-visibility.c (optimize_weakref): Likewise.
769 (function_and_variable_visibility): Likewise.
770 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
771 (ipa_discover_variable_flags): Likewise.
772 * lto-streamer-out.c (output_function): Likewise.
773 (output_constructor): Likewise.
774 * tree-inline.c (copy_bb): Likewise.
775 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
776 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
778 2020-01-08 Richard Biener <rguenther@suse.de>
781 * tree-eh.c (sink_clobbers): Update virtual operands for
782 the first and last stmt only. Add a dry-run capability.
783 (pass_lower_eh_dispatch::execute): Perform clobber sinking
784 after CFG manipulations and in RPO order to catch all
785 secondary opportunities reliably.
787 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
790 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
792 2019-01-08 Richard Biener <rguenther@suse.de>
795 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
796 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
797 virtual operand, also updating SSA use.
798 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
799 Update stmt after resetting virtual operand.
800 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
801 * gimple-iterator.c (gsi_remove): When not removing the stmt
802 permanently do not delink immediate uses or mark the stmt modified.
804 2020-01-08 Martin Liska <mliska@suse.cz>
806 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
807 (ipa_call_context::estimate_size_and_time): Likewise.
808 (inline_analyze_function): Likewise.
810 2020-01-08 Martin Liska <mliska@suse.cz>
812 * cgraph.c (cgraph_node::dump): Use systematically
815 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
817 Add -nodevicespecs option for avr.
820 * config/avr/avr.opt (-nodevicespecs): New driver option.
821 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
822 "-specs=device-specs/..." if that option is not set.
823 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
825 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
827 Implement 64-bit double functions for avr.
830 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
831 --with-double-comparison.
832 * doc/install.texi: Document them.
833 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
834 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
835 <WITH_DOUBLE_COMPARISON>: New built-in defines.
836 * doc/invoke.texi (AVR Built-in Macros): Document them.
837 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
838 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
839 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
841 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
844 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
845 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
846 when only building rm-profile multilibs.
848 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
851 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
852 lattice for a value to check.
853 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
854 finite propagation in self-recursive scc.
856 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
858 * ipa-inline.c (caller_growth_limits): Restore the AND.
860 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
862 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
863 (VEC_ALLREG_ALT): New iterator.
864 (VEC_ALLREG_INT_MODE): New iterator.
865 (VCMP_MODE): New iterator.
866 (VCMP_MODE_INT): New iterator.
867 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
868 (vec_cmp<u>v64qidi): New define_expand.
869 (vec_cmp<mode>di_exec): Use VCMP_MODE.
870 (vec_cmpu<mode>di_exec): New define_expand.
871 (vec_cmp<u>v64qidi_exec): New define_expand.
872 (vec_cmp<mode>di_dup): Use VCMP_MODE.
873 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
874 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
875 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
876 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
877 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
878 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
879 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
880 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
881 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
883 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
884 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
886 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
888 * config/gcn/constraints.md (DA): Update description and match.
890 (Db): New constraint.
891 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
893 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
894 Implement 'Db' mixed immediate type.
895 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
896 (addcv64si3_dup<exec_vcc>): Delete.
897 (subcv64si3<exec_vcc>): Rework constraints.
898 (addv64di3): Rework constraints.
899 (addv64di3_exec): Rework constraints.
900 (subv64di3): Rework constraints.
901 (addv64di3_dup): Delete.
902 (addv64di3_dup_exec): Delete.
903 (addv64di3_zext): Rework constraints.
904 (addv64di3_zext_exec): Rework constraints.
905 (addv64di3_zext_dup): Rework constraints.
906 (addv64di3_zext_dup_exec): Rework constraints.
907 (addv64di3_zext_dup2): Rework constraints.
908 (addv64di3_zext_dup2_exec): Rework constraints.
909 (addv64di3_sext_dup2): Rework constraints.
910 (addv64di3_sext_dup2_exec): Rework constraints.
912 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
914 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
915 existing target checks.
917 2020-01-07 Richard Biener <rguenther@suse.de>
919 * doc/install.texi: Bump minimal supported MPC version.
921 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
923 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
924 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
925 * langhooks.c: Include stor-layout.h.
926 (lhd_simulate_enum_decl): New function.
927 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
928 handle_arm_sve_h for the LTO frontend.
929 (register_vector_type): Cope with null returns from pushdecl.
931 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
933 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
934 (aarch64_sve::nvectors_if_data_type): Replace with...
935 (aarch64_sve::builtin_type_p): ...this.
936 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
937 (find_vector_type): Delete.
938 (add_sve_type_attribute): New function.
939 (lookup_sve_type_attribute): Likewise.
940 (register_builtin_types): Add an "SVE type" attribute to each type.
941 (register_tuple_type): Likewise.
942 (svbool_type_p, nvectors_if_data_type): Delete.
943 (mangle_builtin_type): Use lookup_sve_type_attribute.
944 (builtin_type_p): Likewise. Add an overload that returns the
945 number of constituent vector and predicate registers.
946 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
947 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
948 instead of aarch64_sve_argument_p.
949 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
950 (aarch64_pass_by_reference): Likewise.
951 (aarch64_function_value_1): Likewise.
952 (aarch64_return_in_memory): Likewise.
953 (aarch64_layout_arg): Likewise.
955 2020-01-07 Jakub Jelinek <jakub@redhat.com>
957 PR tree-optimization/93156
958 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
959 least significant bit is always clear.
961 PR tree-optimization/93118
962 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
963 simplifier with two intermediate conversions.
965 2020-01-07 Martin Liska <mliska@suse.cz>
967 * params.opt: Add Optimization for various parameters.
969 2020-01-07 Martin Liska <mliska@suse.cz>
972 * doc/extend.texi: Explain cloning for target_clone
975 2020-01-07 Martin Liska <mliska@suse.cz>
977 PR tree-optimization/92860
978 * common.opt: Make in Optimization option
979 as it is affected by -O0, which is an Optimization
981 * tree-inline.c (tree_inlinable_function_p):
982 Use opt_for_fn for warn_inline.
983 (expand_call_inline): Likewise.
985 2020-01-07 Martin Liska <mliska@suse.cz>
987 PR tree-optimization/92860
988 * common.opt: Make flag_ree as optimization
991 2020-01-07 Martin Liska <mliska@suse.cz>
993 PR optimization/92860
994 * params.opt: Mark param_min_crossjump_insns with Optimization
997 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
999 * ipa-inline-analysis.c (estimate_growth): Fix typo.
1000 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
1002 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
1004 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
1005 helper function to return the valid addressing formats for a given
1006 hard register and mode.
1007 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
1009 * config/rs6000/constraints.md (Q constraint): Update
1011 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
1014 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
1015 Use 'Q' for doing vector extract from memory.
1016 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
1018 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
1019 doing vector extract from memory.
1020 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
1021 extract from memory.
1023 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
1024 for the offset being 34-bits when -mcpu=future is used.
1026 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
1028 * config/pa/pa.md: Revert change to use ordered_comparison_operator
1029 instead of cmpib_comparison_operator in cmpib patterns.
1030 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
1031 of cmpib_comparison_operator. Revise comment.
1033 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1035 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
1036 in an IFN_DIV_POW2 node to be equal.
1038 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1040 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
1041 (vect_check_scalar_mask): ...this.
1042 (vectorizable_store, vectorizable_load): Update call accordingly.
1043 (vectorizable_call): Use vect_check_scalar_mask to check the mask
1044 argument in calls to conditional internal functions.
1046 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1048 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
1049 '0' matching inputs.
1050 (subv64di3_exec): Likewise.
1052 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
1054 * config/mips/mips.c (vr4130_align_insns): Fix typo.
1055 * doc/md.texi (movstr): Likewise.
1057 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1059 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
1062 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1064 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
1066 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
1067 to a temporary file and use move-if-change to update the real
1068 file where necessary.
1070 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1072 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
1073 rather than Upa for CPY /M.
1075 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1077 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
1080 2020-01-06 Martin Liska <mliska@suse.cz>
1082 PR tree-optimization/92860
1083 * params.opt: Mark param_max_combine_insns with Optimization
1086 2020-01-05 Jakub Jelinek <jakub@redhat.com>
1089 * config/i386/i386.md (SWIDWI): New mode iterator.
1090 (DWI, dwi): Add TImode variants.
1091 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
1092 <general_hilo_operand> instead of <general_operand>. Use
1093 CONST_SCALAR_INT_P instead of CONST_INT_P.
1094 (*addv<mode>4_1): Rename to ...
1095 (addv<mode>4_1): ... this.
1096 (QWI): New mode attribute.
1097 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
1098 define_insn_and_split patterns.
1099 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
1101 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
1102 <general_hilo_operand> instead of <general_operand>.
1103 (*addcarry<mode>_1): New define_insn.
1104 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
1106 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
1108 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
1109 Use "call" instead of "set".
1111 2020-01-03 Martin Jambor <mjambor@suse.cz>
1114 * ipa-cp.c (print_all_lattices): Skip functions without info.
1116 2020-01-03 Jakub Jelinek <jakub@redhat.com>
1119 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
1120 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
1121 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
1122 for 'e' simd clones.
1125 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
1127 (mprefer-vector-width=): Add Save.
1128 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
1129 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
1130 (ix86_debug_options, ix86_function_specific_print): Adjust
1131 ix86_target_string callers.
1132 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
1133 (ix86_valid_target_attribute_tree): Likewise.
1134 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
1135 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
1136 ix86_target_string caller.
1139 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
1140 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
1141 instead of gen_int_shift_amount + convert_modes.
1143 PR rtl-optimization/93088
1144 * loop-iv.c (find_single_def_src): Punt after looking through
1145 128 reg copies for regs with single definitions. Move definitions
1148 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
1150 * config/arm/arm-c.c (arm_cpu_builtins): Define
1151 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
1152 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
1153 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
1154 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
1155 * config/arm/arm-tables.opt: Regenerated.
1156 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
1157 arm_arch_i8mm and arm_arch_bf16 when enabled.
1158 * config/arm/arm.h (TARGET_I8MM): New macro.
1159 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
1160 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
1161 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
1162 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
1163 (v8_6_a_simd_variants): New.
1164 (v8_*_a_simd_variants): Add i8mm and bf16.
1165 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
1167 2020-01-02 Jakub Jelinek <jakub@redhat.com>
1170 * predict.c (compute_function_frequency): Don't call
1171 warn_function_cold on functions that already have cold attribute.
1173 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
1176 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
1177 COMDAT group function labels in .data.rel.ro.local section.
1178 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
1181 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
1182 comparison_operator in B and S integer comparisons. Likewise, use
1183 ordered_comparison_operator instead of cmpib_comparison_operator in
1185 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
1187 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1189 Update copyright years.
1191 * gcc.c (process_command): Update copyright notice dates.
1192 * gcov-dump.c (print_version): Ditto.
1193 * gcov.c (print_version): Ditto.
1194 * gcov-tool.c (print_version): Ditto.
1195 * gengtype.c (create_file): Ditto.
1196 * doc/cpp.texi: Bump @copying's copyright year.
1197 * doc/cppinternals.texi: Ditto.
1198 * doc/gcc.texi: Ditto.
1199 * doc/gccint.texi: Ditto.
1200 * doc/gcov.texi: Ditto.
1201 * doc/install.texi: Ditto.
1202 * doc/invoke.texi: Ditto.
1204 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
1206 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
1209 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1211 PR tree-optimization/93098
1212 * match.pd (popcount): For shift amounts, use integer_onep
1213 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
1214 tests. Make sure that precision is power of two larger than or equal
1215 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
1216 instead of ULL suffixed constants. Formatting fixes.
1218 Copyright (C) 2020 Free Software Foundation, Inc.
1220 Copying and distribution of this file, with or without modification,
1221 are permitted in any medium without royalty provided the copyright
1222 notice and this notice are preserved.