1 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
5 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
6 parameter and its use to reset DR_OFFSET's.
7 (vect_transform_loop): Remove orig_drs_init argument.
8 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
9 member of dr_vec_info rather than the offset of the associated
10 data_reference's innermost_loop_behavior.
11 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
12 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
13 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
15 (vectorizable_store): Likewise.
16 (vectorizable_load): Likewise.
18 2020-01-10 Richard Biener <rguenther@suse.de>
20 * gimple-ssa-store-merging
21 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
23 2020-01-10 Martin Liska <mliska@suse.cz>
26 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
27 encapsulation that was there before r280040.
29 2020-01-10 Richard Biener <rguenther@suse.de>
32 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
33 sequences to avoid walking them again for secondary opportunities.
34 (pass_lower_eh_dispatch::execute): Instead actually insert
37 2020-01-10 Richard Biener <rguenther@suse.de>
40 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
41 (cleanup_all_empty_eh): Walk landing pads in reverse order to
44 2020-01-10 Martin Jambor <mjambor@suse.cz>
46 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
47 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
48 to get param_ipa_sra_max_replacements.
49 (param_splitting_across_edge): Pass the caller to
50 pull_accesses_from_callee.
52 2020-01-10 Martin Jambor <mjambor@suse.cz>
54 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
55 * ipa-cp.c (max_new_size): Removed.
56 (orig_overall_size): New variable.
57 (get_max_overall_size): New function.
58 (estimate_local_effects): Use it. Adjust dump.
59 (decide_about_value): Likewise.
60 (ipcp_propagate_stage): Do not calculate max_new_size, just store
61 orig_overall_size. Adjust dump.
62 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
64 2020-01-10 Martin Jambor <mjambor@suse.cz>
66 * params.opt (param_ipa_max_agg_items): Mark as Optimization
67 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
68 instead of param_ipa_max_agg_items.
69 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
70 optimization info for the callee.
72 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
74 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
75 markers if debug_inline_points is false.
77 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
79 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
81 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
82 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
83 aarch64-sve-builtins-sve2.h.
84 (aarch64-sve-builtins-sve2.o): New rule.
85 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
86 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
87 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
88 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
89 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
90 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
92 * config/aarch64/aarch64-sve.md: Update comments with SVE2
93 instructions that are handled here.
94 (@cond_asrd<mode>): Generalize to...
95 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
96 (*cond_asrd<mode>_2): Generalize to...
97 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
98 (*cond_asrd<mode>_z): Generalize to...
99 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
100 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
101 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
102 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
103 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
105 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
106 (@aarch64_scatter_stnt<mode>): Likewise.
107 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
108 (@aarch64_mul_lane_<mode>): Likewise.
109 (@aarch64_sve_suqadd<mode>_const): Likewise.
110 (*<sur>h<addsub><mode>): Generalize to...
111 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
113 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
114 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
115 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
116 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
117 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
118 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
119 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
120 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
121 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
122 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
123 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
124 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
125 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
126 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
127 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
128 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
129 (@aarch64_sve2_xar<mode>): Likewise.
130 (@aarch64_sve2_bcax<mode>): Likewise.
131 (*aarch64_sve2_eor3<mode>): Rename to...
132 (@aarch64_sve2_eor3<mode>): ...this.
133 (@aarch64_sve2_bsl<mode>): New expander.
134 (@aarch64_sve2_nbsl<mode>): Likewise.
135 (@aarch64_sve2_bsl1n<mode>): Likewise.
136 (@aarch64_sve2_bsl2n<mode>): Likewise.
137 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
138 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
139 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
140 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
141 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
142 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
143 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
144 (<su>mull<bt><Vwide>): Generalize to...
145 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
147 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
148 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
149 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
150 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
151 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
152 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
153 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
154 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
155 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
156 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
157 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
158 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
159 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
160 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
161 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
162 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
163 (<SHRNB:r>shrnb<mode>): Generalize to...
164 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
166 (<SHRNT:r>shrnt<mode>): Generalize to...
167 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
169 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
170 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
171 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
172 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
173 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
174 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
175 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
176 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
177 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
178 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
179 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
180 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
181 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
182 (@aarch64_sve2_cvtnt<mode>): Likewise.
183 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
184 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
185 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
186 (@aarch64_sve2_cvtxnt<mode>): Likewise.
187 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
188 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
189 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
190 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
191 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
192 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
193 (@aarch64_sve2_pmul<mode>): Likewise.
194 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
195 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
196 (@aarch64_sve2_tbl2<mode>): Likewise.
197 (@aarch64_sve2_tbx<mode>): Likewise.
198 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
199 (@aarch64_sve2_histcnt<mode>): Likewise.
200 (@aarch64_sve2_histseg<mode>): Likewise.
201 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
202 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
203 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
204 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
205 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
206 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
207 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
208 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
209 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
210 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
211 (SVE2_PMULL_PAIR_I): New mode iterators.
212 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
213 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
214 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
215 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
216 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
217 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
218 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
219 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
220 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
221 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
222 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
223 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
224 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
225 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
226 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
227 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
228 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
229 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
230 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
231 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
232 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
233 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
234 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
235 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
236 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
237 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
238 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
239 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
240 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
241 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
242 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
244 (VNARROW, Ventype): New mode attributes.
245 (Vewtype): Handle VNx2DI. Fix typo in comment.
246 (VDOUBLE): New mode attribute.
247 (sve_lane_con): Handle VNx8HI.
248 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
249 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
250 (sve_int_op, sve_int_op_rev): Handle the above codes.
251 (sve_pred_int_rhs2_operand): Likewise.
252 (MULLBT, SHRNB, SHRNT): Delete.
253 (SVE_INT_SHIFT_IMM): New int iterator.
254 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
255 and UNSPEC_WHILEHS for TARGET_SVE2.
256 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
257 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
258 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
259 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
260 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
261 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
262 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
263 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
264 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
265 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
266 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
267 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
268 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
269 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
270 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
271 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
272 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
273 (optab): Handle the new unspecs.
274 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
276 (lr): Handle the new unspecs.
278 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
279 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
280 (sve_int_qsub_op): New int attributes.
281 (sve_fp_op, rot): Handle the new unspecs.
282 * config/aarch64/aarch64-sve-builtins.h
283 (function_resolver::require_matching_pointer_type): Declare.
284 (function_resolver::resolve_unary): Add an optional boolean argument.
285 (function_resolver::finish_opt_n_resolution): Add an optional
286 type_suffix_index argument.
287 (gimple_folder::redirect_call): Declare.
288 (gimple_expander::prepare_gather_address_operands): Add an optional
290 * config/aarch64/aarch64-sve-builtins.cc: Include
291 aarch64-sve-builtins-sve2.h.
292 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
293 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
294 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
295 (TYPES_hsd_integer): Use TYPES_hsd_signed.
296 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
297 (TYPES_s_unsigned): Likewise.
298 (TYPES_s_integer): Use TYPES_s_unsigned.
299 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
300 (TYPES_sd_integer): Use them.
301 (TYPES_d_unsigned): New macro.
302 (TYPES_d_integer): Use it.
303 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
304 (TYPES_cvt_narrow): Likewise.
305 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
306 (preds_mx): New variable.
307 (function_builder::add_overloaded_function): Allow the new feature
308 set to be more restrictive than the original one.
309 (function_resolver::infer_pointer_type): Remove qualifiers from
310 the pointer type before printing it.
311 (function_resolver::require_matching_pointer_type): New function.
312 (function_resolver::resolve_sv_displacement): Handle functions
313 that don't support 32-bit vector indices or svint32_t vector offsets.
314 (function_resolver::finish_opt_n_resolution): Take the inferred type
315 as a separate argument.
316 (function_resolver::resolve_unary): Optionally treat all forms in
317 the same way as normal merging functions.
318 (gimple_folder::redirect_call): New function.
319 (function_expander::prepare_gather_address_operands): Add an argument
320 that says whether scaled forms are available. If they aren't,
321 handle scaling of vector indices and don't add the extension and
323 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
324 fall back to using cond_* instead.
325 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
326 Split out the member variables into...
327 (rtx_code_function_base): ...this new base class.
328 (rtx_code_function_rotated): Inherit rtx_code_function_base.
329 (unspec_based_function): Split out the member variables into...
330 (unspec_based_function_base): ...this new base class.
331 (unspec_based_function_rotated): Inherit unspec_based_function_base.
332 (unspec_based_function_exact_insn): New class.
333 (unspec_based_add_function, unspec_based_add_lane_function)
334 (unspec_based_lane_function, unspec_based_pred_function)
335 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
336 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
337 (unspec_based_sub_function, unspec_based_sub_lane_function): New
339 (unspec_based_fused_function): New class.
340 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
341 (unspec_based_fused_lane_function): New class.
342 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
344 (CODE_FOR_MODE1): New macro.
345 (fixed_insn_function): New class.
346 (while_comparison): Likewise.
347 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
348 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
349 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
350 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
351 (load_gather_sv_restricted, shift_left_imm_long): Declare.
352 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
353 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
354 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
355 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
356 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
357 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
358 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
359 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
360 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
361 Also add an initial argument for unary_convert_narrowt, regardless
362 of the predication type.
363 (build_32_64): Allow loads and stores to specify MODE_none.
364 (build_sv_index64, build_sv_uint_offset): New functions.
365 (long_type_suffix): New function.
366 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
367 (binary_imm_long_base, load_gather_sv_base): Likewise.
368 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
369 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
370 (unary_narrowb_base, unary_narrowt_base): Likewise.
371 (binary_long_lane_def, binary_long_lane): New shape.
372 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
373 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
374 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
375 (binary_to_uint_def, binary_to_uint): Likewise.
376 (binary_wide_def, binary_wide): Likewise.
377 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
378 (compare_def, compare): Likewise.
379 (compare_ptr_def, compare_ptr): Likewise.
380 (load_ext_gather_index_restricted_def,
381 load_ext_gather_index_restricted): Likewise.
382 (load_ext_gather_offset_restricted_def,
383 load_ext_gather_offset_restricted): Likewise.
384 (load_gather_sv_def): Inherit from load_gather_sv_base.
385 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
386 (shift_left_imm_def, shift_left_imm): Likewise.
387 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
388 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
389 (store_scatter_index_restricted_def,
390 store_scatter_index_restricted): Likewise.
391 (store_scatter_offset_restricted_def,
392 store_scatter_offset_restricted): Likewise.
393 (tbl_tuple_def, tbl_tuple): Likewise.
394 (ternary_long_lane_def, ternary_long_lane): Likewise.
395 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
396 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
397 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
398 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
399 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
400 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
401 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
402 (ternary_uint_def, ternary_uint): Likewise.
403 (unary_convert): Fix typo in comment.
404 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
405 (unary_long_def, unary_long): Likewise.
406 (unary_narrowb_def, unary_narrowb): Likewise.
407 (unary_narrowt_def, unary_narrowt): Likewise.
408 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
409 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
410 (unary_to_int_def, unary_to_int): Likewise.
411 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
412 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
413 (svasrd_impl): Delete.
414 (svcadd_impl::expand): Handle integer operations too.
415 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
416 new functions to derive the unspec numbers.
417 (svmla_svmls_lane_impl): Replace with...
418 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
419 integer operations too.
420 (svwhile_impl): Rename to...
421 (svwhilelx_impl): ...this and inherit from while_comparison.
422 (svasrd): Use unspec_based_function.
423 (svmla_lane): Use svmla_lane_impl.
424 (svmls_lane): Use svmls_lane_impl.
425 (svrecpe, svrsqrte): Handle unsigned integer operations too.
426 (svwhilele, svwhilelt): Use svwhilelx_impl.
427 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
428 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
429 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
430 * config/aarch64/aarch64-sve-builtins.def: Include
431 aarch64-sve-builtins-sve2.def.
433 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
435 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
436 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
437 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
438 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
439 immediates as well as vector ones.
440 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
441 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
442 (aarch64_sve_qsub_immediate): Update calls accordingly.
444 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
446 * config/aarch64/aarch64-sve2.md: Add banner comments.
447 (<su>mulh<r>s<mode>3): Move further up file.
448 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
449 (*aarch64_sve2_sra<mode>): Move further down file.
450 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
452 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
454 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
456 (while_optab_cmp): Handle them.
457 * config/aarch64/aarch64-sve.md
458 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
459 and add a "@" marker.
460 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
461 instead of gen_aarch64_sve2_while_ptest.
462 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
464 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
466 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
467 (UNSPEC_WHILELE): ...this.
468 (UNSPEC_WHILE_LO): Rename to...
469 (UNSPEC_WHILELO): ...this.
470 (UNSPEC_WHILE_LS): Rename to...
471 (UNSPEC_WHILELS): ...this.
472 (UNSPEC_WHILE_LT): Rename to...
473 (UNSPEC_WHILELT): ...this.
474 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
475 (cmp_op, while_optab_cmp): Likewise.
476 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
477 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
478 (svwhilelt): Likewise.
480 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
482 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
483 (unary_to_uint): Define.
484 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
485 (unary_count): Rename to...
486 (unary_to_uint_def, unary_to_uint): ...this.
487 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
489 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
491 * config/aarch64/aarch64-sve-builtins-functions.h
492 (code_for_mode_function): New class.
493 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
494 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
495 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
496 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
497 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
499 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
501 * config/aarch64/iterators.md (addsub): New code attribute.
502 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
504 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
505 in the asm string and attributes. Fix indentation.
506 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
508 (@aarch64_sve_<optab><mode>): ...this.
509 * config/aarch64/aarch64-sve-builtins.h
510 (function_expander::expand_signed_unpred_op): Delete.
511 * config/aarch64/aarch64-sve-builtins.cc
512 (function_expander::expand_signed_unpred_op): Likewise.
513 (function_expander::map_to_rtx_codes): If the optab isn't defined,
514 try using code_for_aarch64_sve instead.
515 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
516 (svqsub_impl): Likewise.
517 (svqadd, svqsub): Use rtx_code_function instead.
519 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
521 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
522 (HADDSUB, sur, addsub): Remove them.
524 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
526 * tree-nrv.c (pass_return_slot::execute): Handle all internal
527 functions the same way, rather than singling out those that
528 aren't mapped directly to optabs.
530 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
532 * target.def (compatible_vector_types_p): New target hook.
533 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
534 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
535 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
536 * doc/tm.texi: Regenerate.
537 * gimple-expr.c: Include target.h.
538 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
539 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
541 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
542 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
543 Use the original predicate if it already has a suitable type.
545 2020-01-09 Martin Jambor <mjambor@suse.cz>
547 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
548 resolve_speculation and redirect_call_stmt_to_callee static. Change
549 return type of set_call_stmt to cgraph_edge *.
550 * auto-profile.c (afdo_indirect_call): Adjust call to
551 redirect_call_stmt_to_callee.
552 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
553 make the this pointer explicit, adjust self-recursive calls and the
554 call top make_direct. Return the resulting edge.
555 (cgraph_edge::remove): Make this pointer explicit.
556 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
557 (cgraph_edge::make_direct): Likewise, adjust call to
559 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
560 call to set_call_stmt.
561 (cgraph_update_edges_for_call_stmt_node): Update call to
562 set_call_stmt and remove.
563 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
564 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
565 (cgraph_node::create_edge_including_clones): Moved "first" definition
566 of edge to the block where it was used. Adjusted calls to
568 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
570 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
571 make_direct and redirect_call_stmt_to_callee.
572 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
573 resolve_speculation and make_direct.
574 * ipa-inline-transform.c (inline_transform): Adjust call to
575 redirect_call_stmt_to_callee.
576 (check_speculations_1):: Adjust call to resolve_speculation.
577 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
579 (inline_small_functions): Adjust call to resolve_speculation.
580 (ipa_inline): Likewise.
581 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
583 * ipa-visibility.c (function_and_variable_visibility): Make iteration
584 safe with regards to edge removal, adjust calls to
585 redirect_call_stmt_to_callee.
586 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
587 and redirect_call_stmt_to_callee.
588 * multiple_target.c (create_dispatcher_calls): Adjust call to
589 redirect_call_stmt_to_callee
590 (redirect_to_specific_clone): Likewise.
591 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
592 Adjust calls to cgraph_edge::remove.
593 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
594 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
595 (expand_call_inline): Adjust call to cgraph_edge::remove.
597 2020-01-09 Martin Liska <mliska@suse.cz>
599 * params.opt: Set Optimization for
600 param_max_speculative_devirt_maydefs.
602 2020-01-09 Martin Sebor <msebor@redhat.com>
606 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
608 2020-01-09 Martin Liska <mliska@suse.cz>
610 * auto-profile.c (auto_profile): Use opt_for_fn
612 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
613 (propagate_vals_across_arith_jfunc): Likewise.
614 (hint_time_bonus): Likewise.
615 (incorporate_penalties): Likewise.
616 (good_cloning_opportunity_p): Likewise.
617 (perform_estimation_of_a_value): Likewise.
618 (estimate_local_effects): Likewise.
619 (ipcp_propagate_stage): Likewise.
620 * ipa-fnsummary.c (decompose_param_expr): Likewise.
621 (set_switch_stmt_execution_predicate): Likewise.
622 (analyze_function_body): Likewise.
623 * ipa-inline-analysis.c (offline_size): Likewise.
624 * ipa-inline.c (early_inliner): Likewise.
625 * ipa-prop.c (ipa_analyze_node): Likewise.
626 (ipcp_transform_function): Likewise.
627 * ipa-sra.c (process_scan_results): Likewise.
628 (ipa_sra_summarize_function): Likewise.
629 * params.opt: Rename ipcp-unit-growth to
630 ipa-cp-unit-growth. Add Optimization for various
631 IPA-related parameters.
633 2020-01-09 Richard Biener <rguenther@suse.de>
636 * gimplify.c (gimplify_expr): Deal with NOP definitions.
638 2020-01-09 Richard Biener <rguenther@suse.de>
640 PR tree-optimization/93040
641 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
643 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
645 * common/config/avr/avr-common.c (avr_option_optimization_table)
646 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
648 2020-01-09 Martin Liska <mliska@suse.cz>
650 * cgraphclones.c (symbol_table::materialize_all_clones):
651 Use cgraph_node::dump_name.
653 2020-01-09 Jakub Jelinek <jakub@redhat.com>
656 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
657 output_operand_lossage instead of gcc_unreachable.
658 * doc/md.texi (riscv f constraint): Fix typo.
661 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
662 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
663 CONST_SCALAR_INT_P instead of CONST_INT_P.
664 (*subv<mode>4_1): Rename to ...
665 (subv<mode>4_1): ... this.
666 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
667 define_insn_and_split patterns.
668 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
671 2020-01-08 David Malcolm <dmalcolm@redhat.com>
673 * vec.c (class selftest::count_dtor): New class.
674 (selftest::test_auto_delete_vec): New test.
675 (selftest::vec_c_tests): Call it.
676 * vec.h (class auto_delete_vec): New class template.
677 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
679 2020-01-08 David Malcolm <dmalcolm@redhat.com>
681 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
683 2020-01-08 Jim Wilson <jimw@sifive.com>
685 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
686 use of TLS_MODEL_LOCAL_EXEC when not pic.
688 2020-01-08 David Malcolm <dmalcolm@redhat.com>
690 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
693 2020-01-08 Jakub Jelinek <jakub@redhat.com>
696 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
697 *stack_protect_set_3 peephole2): Also check that the second
698 insns source is general_operand.
701 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
702 predicate for output operand instead of register_operand.
703 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
704 memory destination and non-memory operands[2].
706 2020-01-08 Martin Liska <mliska@suse.cz>
708 * cgraph.c (cgraph_node::dump): Use ::dump_name or
709 ::dump_asm_name instead of (::name or ::asm_name).
710 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
711 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
712 (analyze_functions): Likewise.
713 (expand_all_functions): Likewise.
714 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
715 (propagate_bits_across_jump_function): Likewise.
716 (dump_profile_updates): Likewise.
717 (ipcp_store_bits_results): Likewise.
718 (ipcp_store_vr_results): Likewise.
719 * ipa-devirt.c (dump_targets): Likewise.
720 * ipa-fnsummary.c (analyze_function_body): Likewise.
721 * ipa-hsa.c (check_warn_node_versionable): Likewise.
722 (process_hsa_functions): Likewise.
723 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
724 (set_alias_uids): Likewise.
725 * ipa-inline-transform.c (save_inline_function_body): Likewise.
726 * ipa-inline.c (recursive_inlining): Likewise.
727 (inline_to_all_callers_1): Likewise.
728 (ipa_inline): Likewise.
729 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
730 (ipa_propagate_frequency): Likewise.
731 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
732 (remove_described_reference): Likewise.
733 * ipa-pure-const.c (worse_state): Likewise.
734 (check_retval_uses): Likewise.
735 (analyze_function): Likewise.
736 (propagate_pure_const): Likewise.
737 (propagate_nothrow): Likewise.
738 (dump_malloc_lattice): Likewise.
739 (propagate_malloc): Likewise.
740 (pass_local_pure_const::execute): Likewise.
741 * ipa-visibility.c (optimize_weakref): Likewise.
742 (function_and_variable_visibility): Likewise.
743 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
744 (ipa_discover_variable_flags): Likewise.
745 * lto-streamer-out.c (output_function): Likewise.
746 (output_constructor): Likewise.
747 * tree-inline.c (copy_bb): Likewise.
748 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
749 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
751 2020-01-08 Richard Biener <rguenther@suse.de>
754 * tree-eh.c (sink_clobbers): Update virtual operands for
755 the first and last stmt only. Add a dry-run capability.
756 (pass_lower_eh_dispatch::execute): Perform clobber sinking
757 after CFG manipulations and in RPO order to catch all
758 secondary opportunities reliably.
760 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
763 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
765 2019-01-08 Richard Biener <rguenther@suse.de>
768 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
769 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
770 virtual operand, also updating SSA use.
771 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
772 Update stmt after resetting virtual operand.
773 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
774 * gimple-iterator.c (gsi_remove): When not removing the stmt
775 permanently do not delink immediate uses or mark the stmt modified.
777 2020-01-08 Martin Liska <mliska@suse.cz>
779 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
780 (ipa_call_context::estimate_size_and_time): Likewise.
781 (inline_analyze_function): Likewise.
783 2020-01-08 Martin Liska <mliska@suse.cz>
785 * cgraph.c (cgraph_node::dump): Use systematically
788 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
790 Add -nodevicespecs option for avr.
793 * config/avr/avr.opt (-nodevicespecs): New driver option.
794 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
795 "-specs=device-specs/..." if that option is not set.
796 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
798 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
800 Implement 64-bit double functions for avr.
803 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
804 --with-double-comparison.
805 * doc/install.texi: Document them.
806 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
807 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
808 <WITH_DOUBLE_COMPARISON>: New built-in defines.
809 * doc/invoke.texi (AVR Built-in Macros): Document them.
810 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
811 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
812 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
814 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
817 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
818 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
819 when only building rm-profile multilibs.
821 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
824 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
825 lattice for a value to check.
826 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
827 finite propagation in self-recursive scc.
829 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
831 * ipa-inline.c (caller_growth_limits): Restore the AND.
833 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
835 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
836 (VEC_ALLREG_ALT): New iterator.
837 (VEC_ALLREG_INT_MODE): New iterator.
838 (VCMP_MODE): New iterator.
839 (VCMP_MODE_INT): New iterator.
840 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
841 (vec_cmp<u>v64qidi): New define_expand.
842 (vec_cmp<mode>di_exec): Use VCMP_MODE.
843 (vec_cmpu<mode>di_exec): New define_expand.
844 (vec_cmp<u>v64qidi_exec): New define_expand.
845 (vec_cmp<mode>di_dup): Use VCMP_MODE.
846 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
847 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
848 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
849 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
850 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
851 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
852 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
853 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
854 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
856 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
857 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
859 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
861 * config/gcn/constraints.md (DA): Update description and match.
863 (Db): New constraint.
864 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
866 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
867 Implement 'Db' mixed immediate type.
868 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
869 (addcv64si3_dup<exec_vcc>): Delete.
870 (subcv64si3<exec_vcc>): Rework constraints.
871 (addv64di3): Rework constraints.
872 (addv64di3_exec): Rework constraints.
873 (subv64di3): Rework constraints.
874 (addv64di3_dup): Delete.
875 (addv64di3_dup_exec): Delete.
876 (addv64di3_zext): Rework constraints.
877 (addv64di3_zext_exec): Rework constraints.
878 (addv64di3_zext_dup): Rework constraints.
879 (addv64di3_zext_dup_exec): Rework constraints.
880 (addv64di3_zext_dup2): Rework constraints.
881 (addv64di3_zext_dup2_exec): Rework constraints.
882 (addv64di3_sext_dup2): Rework constraints.
883 (addv64di3_sext_dup2_exec): Rework constraints.
885 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
887 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
888 existing target checks.
890 2020-01-07 Richard Biener <rguenther@suse.de>
892 * doc/install.texi: Bump minimal supported MPC version.
894 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
896 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
897 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
898 * langhooks.c: Include stor-layout.h.
899 (lhd_simulate_enum_decl): New function.
900 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
901 handle_arm_sve_h for the LTO frontend.
902 (register_vector_type): Cope with null returns from pushdecl.
904 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
906 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
907 (aarch64_sve::nvectors_if_data_type): Replace with...
908 (aarch64_sve::builtin_type_p): ...this.
909 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
910 (find_vector_type): Delete.
911 (add_sve_type_attribute): New function.
912 (lookup_sve_type_attribute): Likewise.
913 (register_builtin_types): Add an "SVE type" attribute to each type.
914 (register_tuple_type): Likewise.
915 (svbool_type_p, nvectors_if_data_type): Delete.
916 (mangle_builtin_type): Use lookup_sve_type_attribute.
917 (builtin_type_p): Likewise. Add an overload that returns the
918 number of constituent vector and predicate registers.
919 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
920 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
921 instead of aarch64_sve_argument_p.
922 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
923 (aarch64_pass_by_reference): Likewise.
924 (aarch64_function_value_1): Likewise.
925 (aarch64_return_in_memory): Likewise.
926 (aarch64_layout_arg): Likewise.
928 2020-01-07 Jakub Jelinek <jakub@redhat.com>
930 PR tree-optimization/93156
931 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
932 least significant bit is always clear.
934 PR tree-optimization/93118
935 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
936 simplifier with two intermediate conversions.
938 2020-01-07 Martin Liska <mliska@suse.cz>
940 * params.opt: Add Optimization for various parameters.
942 2020-01-07 Martin Liska <mliska@suse.cz>
945 * doc/extend.texi: Explain cloning for target_clone
948 2020-01-07 Martin Liska <mliska@suse.cz>
950 PR tree-optimization/92860
951 * common.opt: Make in Optimization option
952 as it is affected by -O0, which is an Optimization
954 * tree-inline.c (tree_inlinable_function_p):
955 Use opt_for_fn for warn_inline.
956 (expand_call_inline): Likewise.
958 2020-01-07 Martin Liska <mliska@suse.cz>
960 PR tree-optimization/92860
961 * common.opt: Make flag_ree as optimization
964 2020-01-07 Martin Liska <mliska@suse.cz>
966 PR optimization/92860
967 * params.opt: Mark param_min_crossjump_insns with Optimization
970 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
972 * ipa-inline-analysis.c (estimate_growth): Fix typo.
973 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
975 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
977 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
978 helper function to return the valid addressing formats for a given
979 hard register and mode.
980 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
982 * config/rs6000/constraints.md (Q constraint): Update
984 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
987 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
988 Use 'Q' for doing vector extract from memory.
989 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
991 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
992 doing vector extract from memory.
993 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
996 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
997 for the offset being 34-bits when -mcpu=future is used.
999 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
1001 * config/pa/pa.md: Revert change to use ordered_comparison_operator
1002 instead of cmpib_comparison_operator in cmpib patterns.
1003 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
1004 of cmpib_comparison_operator. Revise comment.
1006 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1008 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
1009 in an IFN_DIV_POW2 node to be equal.
1011 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1013 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
1014 (vect_check_scalar_mask): ...this.
1015 (vectorizable_store, vectorizable_load): Update call accordingly.
1016 (vectorizable_call): Use vect_check_scalar_mask to check the mask
1017 argument in calls to conditional internal functions.
1019 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1021 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
1022 '0' matching inputs.
1023 (subv64di3_exec): Likewise.
1025 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
1027 * config/mips/mips.c (vr4130_align_insns): Fix typo.
1028 * doc/md.texi (movstr): Likewise.
1030 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1032 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
1035 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1037 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
1039 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
1040 to a temporary file and use move-if-change to update the real
1041 file where necessary.
1043 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1045 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
1046 rather than Upa for CPY /M.
1048 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1050 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
1053 2020-01-06 Martin Liska <mliska@suse.cz>
1055 PR tree-optimization/92860
1056 * params.opt: Mark param_max_combine_insns with Optimization
1059 2020-01-05 Jakub Jelinek <jakub@redhat.com>
1062 * config/i386/i386.md (SWIDWI): New mode iterator.
1063 (DWI, dwi): Add TImode variants.
1064 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
1065 <general_hilo_operand> instead of <general_operand>. Use
1066 CONST_SCALAR_INT_P instead of CONST_INT_P.
1067 (*addv<mode>4_1): Rename to ...
1068 (addv<mode>4_1): ... this.
1069 (QWI): New mode attribute.
1070 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
1071 define_insn_and_split patterns.
1072 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
1074 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
1075 <general_hilo_operand> instead of <general_operand>.
1076 (*addcarry<mode>_1): New define_insn.
1077 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
1079 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
1081 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
1082 Use "call" instead of "set".
1084 2020-01-03 Martin Jambor <mjambor@suse.cz>
1087 * ipa-cp.c (print_all_lattices): Skip functions without info.
1089 2020-01-03 Jakub Jelinek <jakub@redhat.com>
1092 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
1093 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
1094 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
1095 for 'e' simd clones.
1098 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
1100 (mprefer-vector-width=): Add Save.
1101 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
1102 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
1103 (ix86_debug_options, ix86_function_specific_print): Adjust
1104 ix86_target_string callers.
1105 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
1106 (ix86_valid_target_attribute_tree): Likewise.
1107 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
1108 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
1109 ix86_target_string caller.
1112 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
1113 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
1114 instead of gen_int_shift_amount + convert_modes.
1116 PR rtl-optimization/93088
1117 * loop-iv.c (find_single_def_src): Punt after looking through
1118 128 reg copies for regs with single definitions. Move definitions
1121 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
1123 * config/arm/arm-c.c (arm_cpu_builtins): Define
1124 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
1125 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
1126 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
1127 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
1128 * config/arm/arm-tables.opt: Regenerated.
1129 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
1130 arm_arch_i8mm and arm_arch_bf16 when enabled.
1131 * config/arm/arm.h (TARGET_I8MM): New macro.
1132 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
1133 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
1134 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
1135 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
1136 (v8_6_a_simd_variants): New.
1137 (v8_*_a_simd_variants): Add i8mm and bf16.
1138 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
1140 2020-01-02 Jakub Jelinek <jakub@redhat.com>
1143 * predict.c (compute_function_frequency): Don't call
1144 warn_function_cold on functions that already have cold attribute.
1146 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
1149 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
1150 COMDAT group function labels in .data.rel.ro.local section.
1151 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
1154 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
1155 comparison_operator in B and S integer comparisons. Likewise, use
1156 ordered_comparison_operator instead of cmpib_comparison_operator in
1158 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
1160 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1162 Update copyright years.
1164 * gcc.c (process_command): Update copyright notice dates.
1165 * gcov-dump.c (print_version): Ditto.
1166 * gcov.c (print_version): Ditto.
1167 * gcov-tool.c (print_version): Ditto.
1168 * gengtype.c (create_file): Ditto.
1169 * doc/cpp.texi: Bump @copying's copyright year.
1170 * doc/cppinternals.texi: Ditto.
1171 * doc/gcc.texi: Ditto.
1172 * doc/gccint.texi: Ditto.
1173 * doc/gcov.texi: Ditto.
1174 * doc/install.texi: Ditto.
1175 * doc/invoke.texi: Ditto.
1177 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
1179 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
1182 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1184 PR tree-optimization/93098
1185 * match.pd (popcount): For shift amounts, use integer_onep
1186 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
1187 tests. Make sure that precision is power of two larger than or equal
1188 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
1189 instead of ULL suffixed constants. Formatting fixes.
1191 Copyright (C) 2020 Free Software Foundation, Inc.
1193 Copying and distribution of this file, with or without modification,
1194 are permitted in any medium without royalty provided the copyright
1195 notice and this notice are preserved.