1 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
6 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
8 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
10 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
11 parameter and its use to reset DR_OFFSET's.
12 (vect_transform_loop): Remove orig_drs_init argument.
13 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
14 member of dr_vec_info rather than the offset of the associated
15 data_reference's innermost_loop_behavior.
16 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
17 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
18 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
20 (vectorizable_store): Likewise.
21 (vectorizable_load): Likewise.
23 2020-01-10 Richard Biener <rguenther@suse.de>
25 * gimple-ssa-store-merging
26 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
28 2020-01-10 Martin Liska <mliska@suse.cz>
31 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
32 encapsulation that was there before r280040.
34 2020-01-10 Richard Biener <rguenther@suse.de>
37 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
38 sequences to avoid walking them again for secondary opportunities.
39 (pass_lower_eh_dispatch::execute): Instead actually insert
42 2020-01-10 Richard Biener <rguenther@suse.de>
45 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
46 (cleanup_all_empty_eh): Walk landing pads in reverse order to
49 2020-01-10 Martin Jambor <mjambor@suse.cz>
51 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
52 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
53 to get param_ipa_sra_max_replacements.
54 (param_splitting_across_edge): Pass the caller to
55 pull_accesses_from_callee.
57 2020-01-10 Martin Jambor <mjambor@suse.cz>
59 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
60 * ipa-cp.c (max_new_size): Removed.
61 (orig_overall_size): New variable.
62 (get_max_overall_size): New function.
63 (estimate_local_effects): Use it. Adjust dump.
64 (decide_about_value): Likewise.
65 (ipcp_propagate_stage): Do not calculate max_new_size, just store
66 orig_overall_size. Adjust dump.
67 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
69 2020-01-10 Martin Jambor <mjambor@suse.cz>
71 * params.opt (param_ipa_max_agg_items): Mark as Optimization
72 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
73 instead of param_ipa_max_agg_items.
74 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
75 optimization info for the callee.
77 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
79 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
80 markers if debug_inline_points is false.
82 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
84 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
86 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
87 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
88 aarch64-sve-builtins-sve2.h.
89 (aarch64-sve-builtins-sve2.o): New rule.
90 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
91 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
92 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
93 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
94 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
95 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
97 * config/aarch64/aarch64-sve.md: Update comments with SVE2
98 instructions that are handled here.
99 (@cond_asrd<mode>): Generalize to...
100 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
101 (*cond_asrd<mode>_2): Generalize to...
102 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
103 (*cond_asrd<mode>_z): Generalize to...
104 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
105 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
106 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
107 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
108 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
110 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
111 (@aarch64_scatter_stnt<mode>): Likewise.
112 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
113 (@aarch64_mul_lane_<mode>): Likewise.
114 (@aarch64_sve_suqadd<mode>_const): Likewise.
115 (*<sur>h<addsub><mode>): Generalize to...
116 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
118 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
119 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
120 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
121 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
122 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
123 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
124 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
125 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
126 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
127 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
128 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
129 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
130 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
131 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
132 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
133 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
134 (@aarch64_sve2_xar<mode>): Likewise.
135 (@aarch64_sve2_bcax<mode>): Likewise.
136 (*aarch64_sve2_eor3<mode>): Rename to...
137 (@aarch64_sve2_eor3<mode>): ...this.
138 (@aarch64_sve2_bsl<mode>): New expander.
139 (@aarch64_sve2_nbsl<mode>): Likewise.
140 (@aarch64_sve2_bsl1n<mode>): Likewise.
141 (@aarch64_sve2_bsl2n<mode>): Likewise.
142 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
143 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
144 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
145 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
146 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
147 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
148 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
149 (<su>mull<bt><Vwide>): Generalize to...
150 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
152 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
153 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
154 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
155 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
156 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
157 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
158 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
159 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
160 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
161 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
162 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
163 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
164 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
165 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
166 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
167 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
168 (<SHRNB:r>shrnb<mode>): Generalize to...
169 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
171 (<SHRNT:r>shrnt<mode>): Generalize to...
172 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
174 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
175 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
176 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
177 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
178 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
179 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
180 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
181 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
182 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
183 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
184 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
185 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
186 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
187 (@aarch64_sve2_cvtnt<mode>): Likewise.
188 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
189 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
190 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
191 (@aarch64_sve2_cvtxnt<mode>): Likewise.
192 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
193 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
194 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
195 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
196 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
197 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
198 (@aarch64_sve2_pmul<mode>): Likewise.
199 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
200 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
201 (@aarch64_sve2_tbl2<mode>): Likewise.
202 (@aarch64_sve2_tbx<mode>): Likewise.
203 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
204 (@aarch64_sve2_histcnt<mode>): Likewise.
205 (@aarch64_sve2_histseg<mode>): Likewise.
206 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
207 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
208 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
209 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
210 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
211 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
212 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
213 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
214 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
215 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
216 (SVE2_PMULL_PAIR_I): New mode iterators.
217 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
218 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
219 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
220 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
221 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
222 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
223 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
224 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
225 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
226 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
227 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
228 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
229 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
230 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
231 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
232 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
233 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
234 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
235 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
236 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
237 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
238 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
239 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
240 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
241 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
242 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
243 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
244 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
245 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
246 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
247 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
249 (VNARROW, Ventype): New mode attributes.
250 (Vewtype): Handle VNx2DI. Fix typo in comment.
251 (VDOUBLE): New mode attribute.
252 (sve_lane_con): Handle VNx8HI.
253 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
254 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
255 (sve_int_op, sve_int_op_rev): Handle the above codes.
256 (sve_pred_int_rhs2_operand): Likewise.
257 (MULLBT, SHRNB, SHRNT): Delete.
258 (SVE_INT_SHIFT_IMM): New int iterator.
259 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
260 and UNSPEC_WHILEHS for TARGET_SVE2.
261 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
262 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
263 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
264 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
265 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
266 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
267 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
268 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
269 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
270 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
271 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
272 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
273 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
274 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
275 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
276 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
277 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
278 (optab): Handle the new unspecs.
279 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
281 (lr): Handle the new unspecs.
283 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
284 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
285 (sve_int_qsub_op): New int attributes.
286 (sve_fp_op, rot): Handle the new unspecs.
287 * config/aarch64/aarch64-sve-builtins.h
288 (function_resolver::require_matching_pointer_type): Declare.
289 (function_resolver::resolve_unary): Add an optional boolean argument.
290 (function_resolver::finish_opt_n_resolution): Add an optional
291 type_suffix_index argument.
292 (gimple_folder::redirect_call): Declare.
293 (gimple_expander::prepare_gather_address_operands): Add an optional
295 * config/aarch64/aarch64-sve-builtins.cc: Include
296 aarch64-sve-builtins-sve2.h.
297 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
298 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
299 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
300 (TYPES_hsd_integer): Use TYPES_hsd_signed.
301 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
302 (TYPES_s_unsigned): Likewise.
303 (TYPES_s_integer): Use TYPES_s_unsigned.
304 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
305 (TYPES_sd_integer): Use them.
306 (TYPES_d_unsigned): New macro.
307 (TYPES_d_integer): Use it.
308 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
309 (TYPES_cvt_narrow): Likewise.
310 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
311 (preds_mx): New variable.
312 (function_builder::add_overloaded_function): Allow the new feature
313 set to be more restrictive than the original one.
314 (function_resolver::infer_pointer_type): Remove qualifiers from
315 the pointer type before printing it.
316 (function_resolver::require_matching_pointer_type): New function.
317 (function_resolver::resolve_sv_displacement): Handle functions
318 that don't support 32-bit vector indices or svint32_t vector offsets.
319 (function_resolver::finish_opt_n_resolution): Take the inferred type
320 as a separate argument.
321 (function_resolver::resolve_unary): Optionally treat all forms in
322 the same way as normal merging functions.
323 (gimple_folder::redirect_call): New function.
324 (function_expander::prepare_gather_address_operands): Add an argument
325 that says whether scaled forms are available. If they aren't,
326 handle scaling of vector indices and don't add the extension and
328 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
329 fall back to using cond_* instead.
330 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
331 Split out the member variables into...
332 (rtx_code_function_base): ...this new base class.
333 (rtx_code_function_rotated): Inherit rtx_code_function_base.
334 (unspec_based_function): Split out the member variables into...
335 (unspec_based_function_base): ...this new base class.
336 (unspec_based_function_rotated): Inherit unspec_based_function_base.
337 (unspec_based_function_exact_insn): New class.
338 (unspec_based_add_function, unspec_based_add_lane_function)
339 (unspec_based_lane_function, unspec_based_pred_function)
340 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
341 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
342 (unspec_based_sub_function, unspec_based_sub_lane_function): New
344 (unspec_based_fused_function): New class.
345 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
346 (unspec_based_fused_lane_function): New class.
347 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
349 (CODE_FOR_MODE1): New macro.
350 (fixed_insn_function): New class.
351 (while_comparison): Likewise.
352 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
353 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
354 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
355 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
356 (load_gather_sv_restricted, shift_left_imm_long): Declare.
357 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
358 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
359 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
360 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
361 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
362 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
363 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
364 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
365 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
366 Also add an initial argument for unary_convert_narrowt, regardless
367 of the predication type.
368 (build_32_64): Allow loads and stores to specify MODE_none.
369 (build_sv_index64, build_sv_uint_offset): New functions.
370 (long_type_suffix): New function.
371 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
372 (binary_imm_long_base, load_gather_sv_base): Likewise.
373 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
374 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
375 (unary_narrowb_base, unary_narrowt_base): Likewise.
376 (binary_long_lane_def, binary_long_lane): New shape.
377 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
378 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
379 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
380 (binary_to_uint_def, binary_to_uint): Likewise.
381 (binary_wide_def, binary_wide): Likewise.
382 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
383 (compare_def, compare): Likewise.
384 (compare_ptr_def, compare_ptr): Likewise.
385 (load_ext_gather_index_restricted_def,
386 load_ext_gather_index_restricted): Likewise.
387 (load_ext_gather_offset_restricted_def,
388 load_ext_gather_offset_restricted): Likewise.
389 (load_gather_sv_def): Inherit from load_gather_sv_base.
390 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
391 (shift_left_imm_def, shift_left_imm): Likewise.
392 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
393 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
394 (store_scatter_index_restricted_def,
395 store_scatter_index_restricted): Likewise.
396 (store_scatter_offset_restricted_def,
397 store_scatter_offset_restricted): Likewise.
398 (tbl_tuple_def, tbl_tuple): Likewise.
399 (ternary_long_lane_def, ternary_long_lane): Likewise.
400 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
401 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
402 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
403 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
404 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
405 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
406 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
407 (ternary_uint_def, ternary_uint): Likewise.
408 (unary_convert): Fix typo in comment.
409 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
410 (unary_long_def, unary_long): Likewise.
411 (unary_narrowb_def, unary_narrowb): Likewise.
412 (unary_narrowt_def, unary_narrowt): Likewise.
413 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
414 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
415 (unary_to_int_def, unary_to_int): Likewise.
416 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
417 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
418 (svasrd_impl): Delete.
419 (svcadd_impl::expand): Handle integer operations too.
420 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
421 new functions to derive the unspec numbers.
422 (svmla_svmls_lane_impl): Replace with...
423 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
424 integer operations too.
425 (svwhile_impl): Rename to...
426 (svwhilelx_impl): ...this and inherit from while_comparison.
427 (svasrd): Use unspec_based_function.
428 (svmla_lane): Use svmla_lane_impl.
429 (svmls_lane): Use svmls_lane_impl.
430 (svrecpe, svrsqrte): Handle unsigned integer operations too.
431 (svwhilele, svwhilelt): Use svwhilelx_impl.
432 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
433 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
434 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
435 * config/aarch64/aarch64-sve-builtins.def: Include
436 aarch64-sve-builtins-sve2.def.
438 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
440 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
441 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
442 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
443 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
444 immediates as well as vector ones.
445 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
446 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
447 (aarch64_sve_qsub_immediate): Update calls accordingly.
449 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
451 * config/aarch64/aarch64-sve2.md: Add banner comments.
452 (<su>mulh<r>s<mode>3): Move further up file.
453 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
454 (*aarch64_sve2_sra<mode>): Move further down file.
455 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
457 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
459 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
461 (while_optab_cmp): Handle them.
462 * config/aarch64/aarch64-sve.md
463 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
464 and add a "@" marker.
465 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
466 instead of gen_aarch64_sve2_while_ptest.
467 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
469 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
471 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
472 (UNSPEC_WHILELE): ...this.
473 (UNSPEC_WHILE_LO): Rename to...
474 (UNSPEC_WHILELO): ...this.
475 (UNSPEC_WHILE_LS): Rename to...
476 (UNSPEC_WHILELS): ...this.
477 (UNSPEC_WHILE_LT): Rename to...
478 (UNSPEC_WHILELT): ...this.
479 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
480 (cmp_op, while_optab_cmp): Likewise.
481 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
482 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
483 (svwhilelt): Likewise.
485 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
487 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
488 (unary_to_uint): Define.
489 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
490 (unary_count): Rename to...
491 (unary_to_uint_def, unary_to_uint): ...this.
492 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
494 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
496 * config/aarch64/aarch64-sve-builtins-functions.h
497 (code_for_mode_function): New class.
498 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
499 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
500 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
501 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
502 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
504 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
506 * config/aarch64/iterators.md (addsub): New code attribute.
507 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
509 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
510 in the asm string and attributes. Fix indentation.
511 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
513 (@aarch64_sve_<optab><mode>): ...this.
514 * config/aarch64/aarch64-sve-builtins.h
515 (function_expander::expand_signed_unpred_op): Delete.
516 * config/aarch64/aarch64-sve-builtins.cc
517 (function_expander::expand_signed_unpred_op): Likewise.
518 (function_expander::map_to_rtx_codes): If the optab isn't defined,
519 try using code_for_aarch64_sve instead.
520 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
521 (svqsub_impl): Likewise.
522 (svqadd, svqsub): Use rtx_code_function instead.
524 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
526 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
527 (HADDSUB, sur, addsub): Remove them.
529 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
531 * tree-nrv.c (pass_return_slot::execute): Handle all internal
532 functions the same way, rather than singling out those that
533 aren't mapped directly to optabs.
535 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
537 * target.def (compatible_vector_types_p): New target hook.
538 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
539 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
540 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
541 * doc/tm.texi: Regenerate.
542 * gimple-expr.c: Include target.h.
543 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
544 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
546 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
547 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
548 Use the original predicate if it already has a suitable type.
550 2020-01-09 Martin Jambor <mjambor@suse.cz>
552 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
553 resolve_speculation and redirect_call_stmt_to_callee static. Change
554 return type of set_call_stmt to cgraph_edge *.
555 * auto-profile.c (afdo_indirect_call): Adjust call to
556 redirect_call_stmt_to_callee.
557 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
558 make the this pointer explicit, adjust self-recursive calls and the
559 call top make_direct. Return the resulting edge.
560 (cgraph_edge::remove): Make this pointer explicit.
561 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
562 (cgraph_edge::make_direct): Likewise, adjust call to
564 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
565 call to set_call_stmt.
566 (cgraph_update_edges_for_call_stmt_node): Update call to
567 set_call_stmt and remove.
568 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
569 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
570 (cgraph_node::create_edge_including_clones): Moved "first" definition
571 of edge to the block where it was used. Adjusted calls to
573 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
575 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
576 make_direct and redirect_call_stmt_to_callee.
577 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
578 resolve_speculation and make_direct.
579 * ipa-inline-transform.c (inline_transform): Adjust call to
580 redirect_call_stmt_to_callee.
581 (check_speculations_1):: Adjust call to resolve_speculation.
582 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
584 (inline_small_functions): Adjust call to resolve_speculation.
585 (ipa_inline): Likewise.
586 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
588 * ipa-visibility.c (function_and_variable_visibility): Make iteration
589 safe with regards to edge removal, adjust calls to
590 redirect_call_stmt_to_callee.
591 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
592 and redirect_call_stmt_to_callee.
593 * multiple_target.c (create_dispatcher_calls): Adjust call to
594 redirect_call_stmt_to_callee
595 (redirect_to_specific_clone): Likewise.
596 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
597 Adjust calls to cgraph_edge::remove.
598 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
599 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
600 (expand_call_inline): Adjust call to cgraph_edge::remove.
602 2020-01-09 Martin Liska <mliska@suse.cz>
604 * params.opt: Set Optimization for
605 param_max_speculative_devirt_maydefs.
607 2020-01-09 Martin Sebor <msebor@redhat.com>
611 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
613 2020-01-09 Martin Liska <mliska@suse.cz>
615 * auto-profile.c (auto_profile): Use opt_for_fn
617 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
618 (propagate_vals_across_arith_jfunc): Likewise.
619 (hint_time_bonus): Likewise.
620 (incorporate_penalties): Likewise.
621 (good_cloning_opportunity_p): Likewise.
622 (perform_estimation_of_a_value): Likewise.
623 (estimate_local_effects): Likewise.
624 (ipcp_propagate_stage): Likewise.
625 * ipa-fnsummary.c (decompose_param_expr): Likewise.
626 (set_switch_stmt_execution_predicate): Likewise.
627 (analyze_function_body): Likewise.
628 * ipa-inline-analysis.c (offline_size): Likewise.
629 * ipa-inline.c (early_inliner): Likewise.
630 * ipa-prop.c (ipa_analyze_node): Likewise.
631 (ipcp_transform_function): Likewise.
632 * ipa-sra.c (process_scan_results): Likewise.
633 (ipa_sra_summarize_function): Likewise.
634 * params.opt: Rename ipcp-unit-growth to
635 ipa-cp-unit-growth. Add Optimization for various
636 IPA-related parameters.
638 2020-01-09 Richard Biener <rguenther@suse.de>
641 * gimplify.c (gimplify_expr): Deal with NOP definitions.
643 2020-01-09 Richard Biener <rguenther@suse.de>
645 PR tree-optimization/93040
646 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
648 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
650 * common/config/avr/avr-common.c (avr_option_optimization_table)
651 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
653 2020-01-09 Martin Liska <mliska@suse.cz>
655 * cgraphclones.c (symbol_table::materialize_all_clones):
656 Use cgraph_node::dump_name.
658 2020-01-09 Jakub Jelinek <jakub@redhat.com>
661 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
662 output_operand_lossage instead of gcc_unreachable.
663 * doc/md.texi (riscv f constraint): Fix typo.
666 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
667 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
668 CONST_SCALAR_INT_P instead of CONST_INT_P.
669 (*subv<mode>4_1): Rename to ...
670 (subv<mode>4_1): ... this.
671 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
672 define_insn_and_split patterns.
673 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
676 2020-01-08 David Malcolm <dmalcolm@redhat.com>
678 * vec.c (class selftest::count_dtor): New class.
679 (selftest::test_auto_delete_vec): New test.
680 (selftest::vec_c_tests): Call it.
681 * vec.h (class auto_delete_vec): New class template.
682 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
684 2020-01-08 David Malcolm <dmalcolm@redhat.com>
686 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
688 2020-01-08 Jim Wilson <jimw@sifive.com>
690 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
691 use of TLS_MODEL_LOCAL_EXEC when not pic.
693 2020-01-08 David Malcolm <dmalcolm@redhat.com>
695 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
698 2020-01-08 Jakub Jelinek <jakub@redhat.com>
701 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
702 *stack_protect_set_3 peephole2): Also check that the second
703 insns source is general_operand.
706 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
707 predicate for output operand instead of register_operand.
708 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
709 memory destination and non-memory operands[2].
711 2020-01-08 Martin Liska <mliska@suse.cz>
713 * cgraph.c (cgraph_node::dump): Use ::dump_name or
714 ::dump_asm_name instead of (::name or ::asm_name).
715 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
716 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
717 (analyze_functions): Likewise.
718 (expand_all_functions): Likewise.
719 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
720 (propagate_bits_across_jump_function): Likewise.
721 (dump_profile_updates): Likewise.
722 (ipcp_store_bits_results): Likewise.
723 (ipcp_store_vr_results): Likewise.
724 * ipa-devirt.c (dump_targets): Likewise.
725 * ipa-fnsummary.c (analyze_function_body): Likewise.
726 * ipa-hsa.c (check_warn_node_versionable): Likewise.
727 (process_hsa_functions): Likewise.
728 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
729 (set_alias_uids): Likewise.
730 * ipa-inline-transform.c (save_inline_function_body): Likewise.
731 * ipa-inline.c (recursive_inlining): Likewise.
732 (inline_to_all_callers_1): Likewise.
733 (ipa_inline): Likewise.
734 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
735 (ipa_propagate_frequency): Likewise.
736 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
737 (remove_described_reference): Likewise.
738 * ipa-pure-const.c (worse_state): Likewise.
739 (check_retval_uses): Likewise.
740 (analyze_function): Likewise.
741 (propagate_pure_const): Likewise.
742 (propagate_nothrow): Likewise.
743 (dump_malloc_lattice): Likewise.
744 (propagate_malloc): Likewise.
745 (pass_local_pure_const::execute): Likewise.
746 * ipa-visibility.c (optimize_weakref): Likewise.
747 (function_and_variable_visibility): Likewise.
748 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
749 (ipa_discover_variable_flags): Likewise.
750 * lto-streamer-out.c (output_function): Likewise.
751 (output_constructor): Likewise.
752 * tree-inline.c (copy_bb): Likewise.
753 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
754 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
756 2020-01-08 Richard Biener <rguenther@suse.de>
759 * tree-eh.c (sink_clobbers): Update virtual operands for
760 the first and last stmt only. Add a dry-run capability.
761 (pass_lower_eh_dispatch::execute): Perform clobber sinking
762 after CFG manipulations and in RPO order to catch all
763 secondary opportunities reliably.
765 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
768 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
770 2019-01-08 Richard Biener <rguenther@suse.de>
773 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
774 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
775 virtual operand, also updating SSA use.
776 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
777 Update stmt after resetting virtual operand.
778 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
779 * gimple-iterator.c (gsi_remove): When not removing the stmt
780 permanently do not delink immediate uses or mark the stmt modified.
782 2020-01-08 Martin Liska <mliska@suse.cz>
784 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
785 (ipa_call_context::estimate_size_and_time): Likewise.
786 (inline_analyze_function): Likewise.
788 2020-01-08 Martin Liska <mliska@suse.cz>
790 * cgraph.c (cgraph_node::dump): Use systematically
793 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
795 Add -nodevicespecs option for avr.
798 * config/avr/avr.opt (-nodevicespecs): New driver option.
799 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
800 "-specs=device-specs/..." if that option is not set.
801 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
803 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
805 Implement 64-bit double functions for avr.
808 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
809 --with-double-comparison.
810 * doc/install.texi: Document them.
811 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
812 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
813 <WITH_DOUBLE_COMPARISON>: New built-in defines.
814 * doc/invoke.texi (AVR Built-in Macros): Document them.
815 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
816 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
817 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
819 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
822 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
823 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
824 when only building rm-profile multilibs.
826 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
829 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
830 lattice for a value to check.
831 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
832 finite propagation in self-recursive scc.
834 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
836 * ipa-inline.c (caller_growth_limits): Restore the AND.
838 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
840 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
841 (VEC_ALLREG_ALT): New iterator.
842 (VEC_ALLREG_INT_MODE): New iterator.
843 (VCMP_MODE): New iterator.
844 (VCMP_MODE_INT): New iterator.
845 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
846 (vec_cmp<u>v64qidi): New define_expand.
847 (vec_cmp<mode>di_exec): Use VCMP_MODE.
848 (vec_cmpu<mode>di_exec): New define_expand.
849 (vec_cmp<u>v64qidi_exec): New define_expand.
850 (vec_cmp<mode>di_dup): Use VCMP_MODE.
851 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
852 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
853 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
854 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
855 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
856 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
857 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
858 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
859 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
861 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
862 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
864 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
866 * config/gcn/constraints.md (DA): Update description and match.
868 (Db): New constraint.
869 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
871 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
872 Implement 'Db' mixed immediate type.
873 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
874 (addcv64si3_dup<exec_vcc>): Delete.
875 (subcv64si3<exec_vcc>): Rework constraints.
876 (addv64di3): Rework constraints.
877 (addv64di3_exec): Rework constraints.
878 (subv64di3): Rework constraints.
879 (addv64di3_dup): Delete.
880 (addv64di3_dup_exec): Delete.
881 (addv64di3_zext): Rework constraints.
882 (addv64di3_zext_exec): Rework constraints.
883 (addv64di3_zext_dup): Rework constraints.
884 (addv64di3_zext_dup_exec): Rework constraints.
885 (addv64di3_zext_dup2): Rework constraints.
886 (addv64di3_zext_dup2_exec): Rework constraints.
887 (addv64di3_sext_dup2): Rework constraints.
888 (addv64di3_sext_dup2_exec): Rework constraints.
890 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
892 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
893 existing target checks.
895 2020-01-07 Richard Biener <rguenther@suse.de>
897 * doc/install.texi: Bump minimal supported MPC version.
899 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
901 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
902 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
903 * langhooks.c: Include stor-layout.h.
904 (lhd_simulate_enum_decl): New function.
905 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
906 handle_arm_sve_h for the LTO frontend.
907 (register_vector_type): Cope with null returns from pushdecl.
909 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
911 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
912 (aarch64_sve::nvectors_if_data_type): Replace with...
913 (aarch64_sve::builtin_type_p): ...this.
914 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
915 (find_vector_type): Delete.
916 (add_sve_type_attribute): New function.
917 (lookup_sve_type_attribute): Likewise.
918 (register_builtin_types): Add an "SVE type" attribute to each type.
919 (register_tuple_type): Likewise.
920 (svbool_type_p, nvectors_if_data_type): Delete.
921 (mangle_builtin_type): Use lookup_sve_type_attribute.
922 (builtin_type_p): Likewise. Add an overload that returns the
923 number of constituent vector and predicate registers.
924 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
925 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
926 instead of aarch64_sve_argument_p.
927 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
928 (aarch64_pass_by_reference): Likewise.
929 (aarch64_function_value_1): Likewise.
930 (aarch64_return_in_memory): Likewise.
931 (aarch64_layout_arg): Likewise.
933 2020-01-07 Jakub Jelinek <jakub@redhat.com>
935 PR tree-optimization/93156
936 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
937 least significant bit is always clear.
939 PR tree-optimization/93118
940 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
941 simplifier with two intermediate conversions.
943 2020-01-07 Martin Liska <mliska@suse.cz>
945 * params.opt: Add Optimization for various parameters.
947 2020-01-07 Martin Liska <mliska@suse.cz>
950 * doc/extend.texi: Explain cloning for target_clone
953 2020-01-07 Martin Liska <mliska@suse.cz>
955 PR tree-optimization/92860
956 * common.opt: Make in Optimization option
957 as it is affected by -O0, which is an Optimization
959 * tree-inline.c (tree_inlinable_function_p):
960 Use opt_for_fn for warn_inline.
961 (expand_call_inline): Likewise.
963 2020-01-07 Martin Liska <mliska@suse.cz>
965 PR tree-optimization/92860
966 * common.opt: Make flag_ree as optimization
969 2020-01-07 Martin Liska <mliska@suse.cz>
971 PR optimization/92860
972 * params.opt: Mark param_min_crossjump_insns with Optimization
975 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
977 * ipa-inline-analysis.c (estimate_growth): Fix typo.
978 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
980 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
982 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
983 helper function to return the valid addressing formats for a given
984 hard register and mode.
985 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
987 * config/rs6000/constraints.md (Q constraint): Update
989 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
992 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
993 Use 'Q' for doing vector extract from memory.
994 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
996 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
997 doing vector extract from memory.
998 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
1001 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
1002 for the offset being 34-bits when -mcpu=future is used.
1004 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
1006 * config/pa/pa.md: Revert change to use ordered_comparison_operator
1007 instead of cmpib_comparison_operator in cmpib patterns.
1008 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
1009 of cmpib_comparison_operator. Revise comment.
1011 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1013 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
1014 in an IFN_DIV_POW2 node to be equal.
1016 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1018 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
1019 (vect_check_scalar_mask): ...this.
1020 (vectorizable_store, vectorizable_load): Update call accordingly.
1021 (vectorizable_call): Use vect_check_scalar_mask to check the mask
1022 argument in calls to conditional internal functions.
1024 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1026 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
1027 '0' matching inputs.
1028 (subv64di3_exec): Likewise.
1030 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
1032 * config/mips/mips.c (vr4130_align_insns): Fix typo.
1033 * doc/md.texi (movstr): Likewise.
1035 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1037 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
1040 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1042 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
1044 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
1045 to a temporary file and use move-if-change to update the real
1046 file where necessary.
1048 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1050 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
1051 rather than Upa for CPY /M.
1053 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1055 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
1058 2020-01-06 Martin Liska <mliska@suse.cz>
1060 PR tree-optimization/92860
1061 * params.opt: Mark param_max_combine_insns with Optimization
1064 2020-01-05 Jakub Jelinek <jakub@redhat.com>
1067 * config/i386/i386.md (SWIDWI): New mode iterator.
1068 (DWI, dwi): Add TImode variants.
1069 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
1070 <general_hilo_operand> instead of <general_operand>. Use
1071 CONST_SCALAR_INT_P instead of CONST_INT_P.
1072 (*addv<mode>4_1): Rename to ...
1073 (addv<mode>4_1): ... this.
1074 (QWI): New mode attribute.
1075 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
1076 define_insn_and_split patterns.
1077 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
1079 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
1080 <general_hilo_operand> instead of <general_operand>.
1081 (*addcarry<mode>_1): New define_insn.
1082 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
1084 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
1086 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
1087 Use "call" instead of "set".
1089 2020-01-03 Martin Jambor <mjambor@suse.cz>
1092 * ipa-cp.c (print_all_lattices): Skip functions without info.
1094 2020-01-03 Jakub Jelinek <jakub@redhat.com>
1097 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
1098 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
1099 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
1100 for 'e' simd clones.
1103 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
1105 (mprefer-vector-width=): Add Save.
1106 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
1107 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
1108 (ix86_debug_options, ix86_function_specific_print): Adjust
1109 ix86_target_string callers.
1110 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
1111 (ix86_valid_target_attribute_tree): Likewise.
1112 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
1113 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
1114 ix86_target_string caller.
1117 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
1118 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
1119 instead of gen_int_shift_amount + convert_modes.
1121 PR rtl-optimization/93088
1122 * loop-iv.c (find_single_def_src): Punt after looking through
1123 128 reg copies for regs with single definitions. Move definitions
1126 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
1128 * config/arm/arm-c.c (arm_cpu_builtins): Define
1129 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
1130 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
1131 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
1132 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
1133 * config/arm/arm-tables.opt: Regenerated.
1134 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
1135 arm_arch_i8mm and arm_arch_bf16 when enabled.
1136 * config/arm/arm.h (TARGET_I8MM): New macro.
1137 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
1138 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
1139 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
1140 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
1141 (v8_6_a_simd_variants): New.
1142 (v8_*_a_simd_variants): Add i8mm and bf16.
1143 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
1145 2020-01-02 Jakub Jelinek <jakub@redhat.com>
1148 * predict.c (compute_function_frequency): Don't call
1149 warn_function_cold on functions that already have cold attribute.
1151 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
1154 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
1155 COMDAT group function labels in .data.rel.ro.local section.
1156 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
1159 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
1160 comparison_operator in B and S integer comparisons. Likewise, use
1161 ordered_comparison_operator instead of cmpib_comparison_operator in
1163 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
1165 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1167 Update copyright years.
1169 * gcc.c (process_command): Update copyright notice dates.
1170 * gcov-dump.c (print_version): Ditto.
1171 * gcov.c (print_version): Ditto.
1172 * gcov-tool.c (print_version): Ditto.
1173 * gengtype.c (create_file): Ditto.
1174 * doc/cpp.texi: Bump @copying's copyright year.
1175 * doc/cppinternals.texi: Ditto.
1176 * doc/gcc.texi: Ditto.
1177 * doc/gccint.texi: Ditto.
1178 * doc/gcov.texi: Ditto.
1179 * doc/install.texi: Ditto.
1180 * doc/invoke.texi: Ditto.
1182 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
1184 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
1187 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1189 PR tree-optimization/93098
1190 * match.pd (popcount): For shift amounts, use integer_onep
1191 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
1192 tests. Make sure that precision is power of two larger than or equal
1193 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
1194 instead of ULL suffixed constants. Formatting fixes.
1196 Copyright (C) 2020 Free Software Foundation, Inc.
1198 Copying and distribution of this file, with or without modification,
1199 are permitted in any medium without royalty provided the copyright
1200 notice and this notice are preserved.