1 2020-01-10 Richard Biener <rguenther@suse.de>
3 * gimple-ssa-store-merging
4 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
6 2020-01-10 Martin Liska <mliska@suse.cz>
9 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
10 encapsulation that was there before r280040.
12 2020-01-10 Richard Biener <rguenther@suse.de>
15 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
16 sequences to avoid walking them again for secondary opportunities.
17 (pass_lower_eh_dispatch::execute): Instead actually insert
20 2020-01-10 Richard Biener <rguenther@suse.de>
23 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
24 (cleanup_all_empty_eh): Walk landing pads in reverse order to
27 2020-01-10 Martin Jambor <mjambor@suse.cz>
29 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
30 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
31 to get param_ipa_sra_max_replacements.
32 (param_splitting_across_edge): Pass the caller to
33 pull_accesses_from_callee.
35 2020-01-10 Martin Jambor <mjambor@suse.cz>
37 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
38 * ipa-cp.c (max_new_size): Removed.
39 (orig_overall_size): New variable.
40 (get_max_overall_size): New function.
41 (estimate_local_effects): Use it. Adjust dump.
42 (decide_about_value): Likewise.
43 (ipcp_propagate_stage): Do not calculate max_new_size, just store
44 orig_overall_size. Adjust dump.
45 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
47 2020-01-10 Martin Jambor <mjambor@suse.cz>
49 * params.opt (param_ipa_max_agg_items): Mark as Optimization
50 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
51 instead of param_ipa_max_agg_items.
52 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
53 optimization info for the callee.
55 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
57 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
58 markers if debug_inline_points is false.
60 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
62 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
64 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
65 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
66 aarch64-sve-builtins-sve2.h.
67 (aarch64-sve-builtins-sve2.o): New rule.
68 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
69 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
70 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
71 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
72 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
73 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
75 * config/aarch64/aarch64-sve.md: Update comments with SVE2
76 instructions that are handled here.
77 (@cond_asrd<mode>): Generalize to...
78 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
79 (*cond_asrd<mode>_2): Generalize to...
80 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
81 (*cond_asrd<mode>_z): Generalize to...
82 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
83 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
84 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
85 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
86 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
88 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
89 (@aarch64_scatter_stnt<mode>): Likewise.
90 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
91 (@aarch64_mul_lane_<mode>): Likewise.
92 (@aarch64_sve_suqadd<mode>_const): Likewise.
93 (*<sur>h<addsub><mode>): Generalize to...
94 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
96 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
97 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
98 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
99 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
100 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
101 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
102 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
103 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
104 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
105 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
106 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
107 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
108 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
109 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
110 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
111 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
112 (@aarch64_sve2_xar<mode>): Likewise.
113 (@aarch64_sve2_bcax<mode>): Likewise.
114 (*aarch64_sve2_eor3<mode>): Rename to...
115 (@aarch64_sve2_eor3<mode>): ...this.
116 (@aarch64_sve2_bsl<mode>): New expander.
117 (@aarch64_sve2_nbsl<mode>): Likewise.
118 (@aarch64_sve2_bsl1n<mode>): Likewise.
119 (@aarch64_sve2_bsl2n<mode>): Likewise.
120 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
121 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
122 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
123 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
124 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
125 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
126 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
127 (<su>mull<bt><Vwide>): Generalize to...
128 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
130 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
131 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
132 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
133 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
134 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
135 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
136 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
137 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
138 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
139 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
140 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
141 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
142 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
143 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
144 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
145 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
146 (<SHRNB:r>shrnb<mode>): Generalize to...
147 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
149 (<SHRNT:r>shrnt<mode>): Generalize to...
150 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
152 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
153 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
154 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
155 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
156 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
157 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
158 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
159 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
160 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
161 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
162 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
163 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
164 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
165 (@aarch64_sve2_cvtnt<mode>): Likewise.
166 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
167 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
168 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
169 (@aarch64_sve2_cvtxnt<mode>): Likewise.
170 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
171 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
172 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
173 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
174 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
175 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
176 (@aarch64_sve2_pmul<mode>): Likewise.
177 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
178 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
179 (@aarch64_sve2_tbl2<mode>): Likewise.
180 (@aarch64_sve2_tbx<mode>): Likewise.
181 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
182 (@aarch64_sve2_histcnt<mode>): Likewise.
183 (@aarch64_sve2_histseg<mode>): Likewise.
184 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
185 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
186 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
187 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
188 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
189 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
190 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
191 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
192 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
193 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
194 (SVE2_PMULL_PAIR_I): New mode iterators.
195 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
196 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
197 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
198 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
199 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
200 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
201 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
202 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
203 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
204 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
205 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
206 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
207 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
208 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
209 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
210 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
211 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
212 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
213 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
214 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
215 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
216 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
217 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
218 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
219 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
220 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
221 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
222 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
223 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
224 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
225 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
227 (VNARROW, Ventype): New mode attributes.
228 (Vewtype): Handle VNx2DI. Fix typo in comment.
229 (VDOUBLE): New mode attribute.
230 (sve_lane_con): Handle VNx8HI.
231 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
232 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
233 (sve_int_op, sve_int_op_rev): Handle the above codes.
234 (sve_pred_int_rhs2_operand): Likewise.
235 (MULLBT, SHRNB, SHRNT): Delete.
236 (SVE_INT_SHIFT_IMM): New int iterator.
237 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
238 and UNSPEC_WHILEHS for TARGET_SVE2.
239 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
240 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
241 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
242 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
243 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
244 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
245 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
246 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
247 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
248 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
249 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
250 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
251 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
252 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
253 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
254 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
255 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
256 (optab): Handle the new unspecs.
257 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
259 (lr): Handle the new unspecs.
261 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
262 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
263 (sve_int_qsub_op): New int attributes.
264 (sve_fp_op, rot): Handle the new unspecs.
265 * config/aarch64/aarch64-sve-builtins.h
266 (function_resolver::require_matching_pointer_type): Declare.
267 (function_resolver::resolve_unary): Add an optional boolean argument.
268 (function_resolver::finish_opt_n_resolution): Add an optional
269 type_suffix_index argument.
270 (gimple_folder::redirect_call): Declare.
271 (gimple_expander::prepare_gather_address_operands): Add an optional
273 * config/aarch64/aarch64-sve-builtins.cc: Include
274 aarch64-sve-builtins-sve2.h.
275 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
276 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
277 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
278 (TYPES_hsd_integer): Use TYPES_hsd_signed.
279 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
280 (TYPES_s_unsigned): Likewise.
281 (TYPES_s_integer): Use TYPES_s_unsigned.
282 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
283 (TYPES_sd_integer): Use them.
284 (TYPES_d_unsigned): New macro.
285 (TYPES_d_integer): Use it.
286 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
287 (TYPES_cvt_narrow): Likewise.
288 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
289 (preds_mx): New variable.
290 (function_builder::add_overloaded_function): Allow the new feature
291 set to be more restrictive than the original one.
292 (function_resolver::infer_pointer_type): Remove qualifiers from
293 the pointer type before printing it.
294 (function_resolver::require_matching_pointer_type): New function.
295 (function_resolver::resolve_sv_displacement): Handle functions
296 that don't support 32-bit vector indices or svint32_t vector offsets.
297 (function_resolver::finish_opt_n_resolution): Take the inferred type
298 as a separate argument.
299 (function_resolver::resolve_unary): Optionally treat all forms in
300 the same way as normal merging functions.
301 (gimple_folder::redirect_call): New function.
302 (function_expander::prepare_gather_address_operands): Add an argument
303 that says whether scaled forms are available. If they aren't,
304 handle scaling of vector indices and don't add the extension and
306 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
307 fall back to using cond_* instead.
308 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
309 Split out the member variables into...
310 (rtx_code_function_base): ...this new base class.
311 (rtx_code_function_rotated): Inherit rtx_code_function_base.
312 (unspec_based_function): Split out the member variables into...
313 (unspec_based_function_base): ...this new base class.
314 (unspec_based_function_rotated): Inherit unspec_based_function_base.
315 (unspec_based_function_exact_insn): New class.
316 (unspec_based_add_function, unspec_based_add_lane_function)
317 (unspec_based_lane_function, unspec_based_pred_function)
318 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
319 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
320 (unspec_based_sub_function, unspec_based_sub_lane_function): New
322 (unspec_based_fused_function): New class.
323 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
324 (unspec_based_fused_lane_function): New class.
325 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
327 (CODE_FOR_MODE1): New macro.
328 (fixed_insn_function): New class.
329 (while_comparison): Likewise.
330 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
331 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
332 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
333 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
334 (load_gather_sv_restricted, shift_left_imm_long): Declare.
335 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
336 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
337 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
338 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
339 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
340 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
341 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
342 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
343 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
344 Also add an initial argument for unary_convert_narrowt, regardless
345 of the predication type.
346 (build_32_64): Allow loads and stores to specify MODE_none.
347 (build_sv_index64, build_sv_uint_offset): New functions.
348 (long_type_suffix): New function.
349 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
350 (binary_imm_long_base, load_gather_sv_base): Likewise.
351 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
352 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
353 (unary_narrowb_base, unary_narrowt_base): Likewise.
354 (binary_long_lane_def, binary_long_lane): New shape.
355 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
356 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
357 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
358 (binary_to_uint_def, binary_to_uint): Likewise.
359 (binary_wide_def, binary_wide): Likewise.
360 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
361 (compare_def, compare): Likewise.
362 (compare_ptr_def, compare_ptr): Likewise.
363 (load_ext_gather_index_restricted_def,
364 load_ext_gather_index_restricted): Likewise.
365 (load_ext_gather_offset_restricted_def,
366 load_ext_gather_offset_restricted): Likewise.
367 (load_gather_sv_def): Inherit from load_gather_sv_base.
368 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
369 (shift_left_imm_def, shift_left_imm): Likewise.
370 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
371 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
372 (store_scatter_index_restricted_def,
373 store_scatter_index_restricted): Likewise.
374 (store_scatter_offset_restricted_def,
375 store_scatter_offset_restricted): Likewise.
376 (tbl_tuple_def, tbl_tuple): Likewise.
377 (ternary_long_lane_def, ternary_long_lane): Likewise.
378 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
379 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
380 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
381 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
382 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
383 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
384 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
385 (ternary_uint_def, ternary_uint): Likewise.
386 (unary_convert): Fix typo in comment.
387 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
388 (unary_long_def, unary_long): Likewise.
389 (unary_narrowb_def, unary_narrowb): Likewise.
390 (unary_narrowt_def, unary_narrowt): Likewise.
391 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
392 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
393 (unary_to_int_def, unary_to_int): Likewise.
394 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
395 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
396 (svasrd_impl): Delete.
397 (svcadd_impl::expand): Handle integer operations too.
398 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
399 new functions to derive the unspec numbers.
400 (svmla_svmls_lane_impl): Replace with...
401 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
402 integer operations too.
403 (svwhile_impl): Rename to...
404 (svwhilelx_impl): ...this and inherit from while_comparison.
405 (svasrd): Use unspec_based_function.
406 (svmla_lane): Use svmla_lane_impl.
407 (svmls_lane): Use svmls_lane_impl.
408 (svrecpe, svrsqrte): Handle unsigned integer operations too.
409 (svwhilele, svwhilelt): Use svwhilelx_impl.
410 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
411 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
412 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
413 * config/aarch64/aarch64-sve-builtins.def: Include
414 aarch64-sve-builtins-sve2.def.
416 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
418 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
419 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
420 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
421 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
422 immediates as well as vector ones.
423 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
424 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
425 (aarch64_sve_qsub_immediate): Update calls accordingly.
427 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
429 * config/aarch64/aarch64-sve2.md: Add banner comments.
430 (<su>mulh<r>s<mode>3): Move further up file.
431 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
432 (*aarch64_sve2_sra<mode>): Move further down file.
433 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
435 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
437 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
439 (while_optab_cmp): Handle them.
440 * config/aarch64/aarch64-sve.md
441 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
442 and add a "@" marker.
443 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
444 instead of gen_aarch64_sve2_while_ptest.
445 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
447 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
449 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
450 (UNSPEC_WHILELE): ...this.
451 (UNSPEC_WHILE_LO): Rename to...
452 (UNSPEC_WHILELO): ...this.
453 (UNSPEC_WHILE_LS): Rename to...
454 (UNSPEC_WHILELS): ...this.
455 (UNSPEC_WHILE_LT): Rename to...
456 (UNSPEC_WHILELT): ...this.
457 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
458 (cmp_op, while_optab_cmp): Likewise.
459 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
460 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
461 (svwhilelt): Likewise.
463 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
465 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
466 (unary_to_uint): Define.
467 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
468 (unary_count): Rename to...
469 (unary_to_uint_def, unary_to_uint): ...this.
470 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
472 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
474 * config/aarch64/aarch64-sve-builtins-functions.h
475 (code_for_mode_function): New class.
476 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
477 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
478 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
479 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
480 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
482 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
484 * config/aarch64/iterators.md (addsub): New code attribute.
485 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
487 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
488 in the asm string and attributes. Fix indentation.
489 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
491 (@aarch64_sve_<optab><mode>): ...this.
492 * config/aarch64/aarch64-sve-builtins.h
493 (function_expander::expand_signed_unpred_op): Delete.
494 * config/aarch64/aarch64-sve-builtins.cc
495 (function_expander::expand_signed_unpred_op): Likewise.
496 (function_expander::map_to_rtx_codes): If the optab isn't defined,
497 try using code_for_aarch64_sve instead.
498 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
499 (svqsub_impl): Likewise.
500 (svqadd, svqsub): Use rtx_code_function instead.
502 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
504 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
505 (HADDSUB, sur, addsub): Remove them.
507 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
509 * tree-nrv.c (pass_return_slot::execute): Handle all internal
510 functions the same way, rather than singling out those that
511 aren't mapped directly to optabs.
513 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
515 * target.def (compatible_vector_types_p): New target hook.
516 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
517 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
518 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
519 * doc/tm.texi: Regenerate.
520 * gimple-expr.c: Include target.h.
521 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
522 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
524 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
525 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
526 Use the original predicate if it already has a suitable type.
528 2020-01-09 Martin Jambor <mjambor@suse.cz>
530 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
531 resolve_speculation and redirect_call_stmt_to_callee static. Change
532 return type of set_call_stmt to cgraph_edge *.
533 * auto-profile.c (afdo_indirect_call): Adjust call to
534 redirect_call_stmt_to_callee.
535 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
536 make the this pointer explicit, adjust self-recursive calls and the
537 call top make_direct. Return the resulting edge.
538 (cgraph_edge::remove): Make this pointer explicit.
539 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
540 (cgraph_edge::make_direct): Likewise, adjust call to
542 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
543 call to set_call_stmt.
544 (cgraph_update_edges_for_call_stmt_node): Update call to
545 set_call_stmt and remove.
546 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
547 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
548 (cgraph_node::create_edge_including_clones): Moved "first" definition
549 of edge to the block where it was used. Adjusted calls to
551 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
553 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
554 make_direct and redirect_call_stmt_to_callee.
555 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
556 resolve_speculation and make_direct.
557 * ipa-inline-transform.c (inline_transform): Adjust call to
558 redirect_call_stmt_to_callee.
559 (check_speculations_1):: Adjust call to resolve_speculation.
560 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
562 (inline_small_functions): Adjust call to resolve_speculation.
563 (ipa_inline): Likewise.
564 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
566 * ipa-visibility.c (function_and_variable_visibility): Make iteration
567 safe with regards to edge removal, adjust calls to
568 redirect_call_stmt_to_callee.
569 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
570 and redirect_call_stmt_to_callee.
571 * multiple_target.c (create_dispatcher_calls): Adjust call to
572 redirect_call_stmt_to_callee
573 (redirect_to_specific_clone): Likewise.
574 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
575 Adjust calls to cgraph_edge::remove.
576 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
577 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
578 (expand_call_inline): Adjust call to cgraph_edge::remove.
580 2020-01-09 Martin Liska <mliska@suse.cz>
582 * params.opt: Set Optimization for
583 param_max_speculative_devirt_maydefs.
585 2020-01-09 Martin Sebor <msebor@redhat.com>
589 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
591 2020-01-09 Martin Liska <mliska@suse.cz>
593 * auto-profile.c (auto_profile): Use opt_for_fn
595 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
596 (propagate_vals_across_arith_jfunc): Likewise.
597 (hint_time_bonus): Likewise.
598 (incorporate_penalties): Likewise.
599 (good_cloning_opportunity_p): Likewise.
600 (perform_estimation_of_a_value): Likewise.
601 (estimate_local_effects): Likewise.
602 (ipcp_propagate_stage): Likewise.
603 * ipa-fnsummary.c (decompose_param_expr): Likewise.
604 (set_switch_stmt_execution_predicate): Likewise.
605 (analyze_function_body): Likewise.
606 * ipa-inline-analysis.c (offline_size): Likewise.
607 * ipa-inline.c (early_inliner): Likewise.
608 * ipa-prop.c (ipa_analyze_node): Likewise.
609 (ipcp_transform_function): Likewise.
610 * ipa-sra.c (process_scan_results): Likewise.
611 (ipa_sra_summarize_function): Likewise.
612 * params.opt: Rename ipcp-unit-growth to
613 ipa-cp-unit-growth. Add Optimization for various
614 IPA-related parameters.
616 2020-01-09 Richard Biener <rguenther@suse.de>
619 * gimplify.c (gimplify_expr): Deal with NOP definitions.
621 2020-01-09 Richard Biener <rguenther@suse.de>
623 PR tree-optimization/93040
624 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
626 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
628 * common/config/avr/avr-common.c (avr_option_optimization_table)
629 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
631 2020-01-09 Martin Liska <mliska@suse.cz>
633 * cgraphclones.c (symbol_table::materialize_all_clones):
634 Use cgraph_node::dump_name.
636 2020-01-09 Jakub Jelinek <jakub@redhat.com>
639 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
640 output_operand_lossage instead of gcc_unreachable.
641 * doc/md.texi (riscv f constraint): Fix typo.
644 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
645 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
646 CONST_SCALAR_INT_P instead of CONST_INT_P.
647 (*subv<mode>4_1): Rename to ...
648 (subv<mode>4_1): ... this.
649 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
650 define_insn_and_split patterns.
651 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
654 2020-01-08 David Malcolm <dmalcolm@redhat.com>
656 * vec.c (class selftest::count_dtor): New class.
657 (selftest::test_auto_delete_vec): New test.
658 (selftest::vec_c_tests): Call it.
659 * vec.h (class auto_delete_vec): New class template.
660 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
662 2020-01-08 David Malcolm <dmalcolm@redhat.com>
664 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
666 2020-01-08 Jim Wilson <jimw@sifive.com>
668 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
669 use of TLS_MODEL_LOCAL_EXEC when not pic.
671 2020-01-08 David Malcolm <dmalcolm@redhat.com>
673 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
676 2020-01-08 Jakub Jelinek <jakub@redhat.com>
679 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
680 *stack_protect_set_3 peephole2): Also check that the second
681 insns source is general_operand.
684 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
685 predicate for output operand instead of register_operand.
686 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
687 memory destination and non-memory operands[2].
689 2020-01-08 Martin Liska <mliska@suse.cz>
691 * cgraph.c (cgraph_node::dump): Use ::dump_name or
692 ::dump_asm_name instead of (::name or ::asm_name).
693 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
694 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
695 (analyze_functions): Likewise.
696 (expand_all_functions): Likewise.
697 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
698 (propagate_bits_across_jump_function): Likewise.
699 (dump_profile_updates): Likewise.
700 (ipcp_store_bits_results): Likewise.
701 (ipcp_store_vr_results): Likewise.
702 * ipa-devirt.c (dump_targets): Likewise.
703 * ipa-fnsummary.c (analyze_function_body): Likewise.
704 * ipa-hsa.c (check_warn_node_versionable): Likewise.
705 (process_hsa_functions): Likewise.
706 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
707 (set_alias_uids): Likewise.
708 * ipa-inline-transform.c (save_inline_function_body): Likewise.
709 * ipa-inline.c (recursive_inlining): Likewise.
710 (inline_to_all_callers_1): Likewise.
711 (ipa_inline): Likewise.
712 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
713 (ipa_propagate_frequency): Likewise.
714 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
715 (remove_described_reference): Likewise.
716 * ipa-pure-const.c (worse_state): Likewise.
717 (check_retval_uses): Likewise.
718 (analyze_function): Likewise.
719 (propagate_pure_const): Likewise.
720 (propagate_nothrow): Likewise.
721 (dump_malloc_lattice): Likewise.
722 (propagate_malloc): Likewise.
723 (pass_local_pure_const::execute): Likewise.
724 * ipa-visibility.c (optimize_weakref): Likewise.
725 (function_and_variable_visibility): Likewise.
726 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
727 (ipa_discover_variable_flags): Likewise.
728 * lto-streamer-out.c (output_function): Likewise.
729 (output_constructor): Likewise.
730 * tree-inline.c (copy_bb): Likewise.
731 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
732 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
734 2020-01-08 Richard Biener <rguenther@suse.de>
737 * tree-eh.c (sink_clobbers): Update virtual operands for
738 the first and last stmt only. Add a dry-run capability.
739 (pass_lower_eh_dispatch::execute): Perform clobber sinking
740 after CFG manipulations and in RPO order to catch all
741 secondary opportunities reliably.
743 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
746 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
748 2019-01-08 Richard Biener <rguenther@suse.de>
751 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
752 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
753 virtual operand, also updating SSA use.
754 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
755 Update stmt after resetting virtual operand.
756 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
757 * gimple-iterator.c (gsi_remove): When not removing the stmt
758 permanently do not delink immediate uses or mark the stmt modified.
760 2020-01-08 Martin Liska <mliska@suse.cz>
762 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
763 (ipa_call_context::estimate_size_and_time): Likewise.
764 (inline_analyze_function): Likewise.
766 2020-01-08 Martin Liska <mliska@suse.cz>
768 * cgraph.c (cgraph_node::dump): Use systematically
771 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
773 Add -nodevicespecs option for avr.
776 * config/avr/avr.opt (-nodevicespecs): New driver option.
777 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
778 "-specs=device-specs/..." if that option is not set.
779 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
781 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
783 Implement 64-bit double functions for avr.
786 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
787 --with-double-comparison.
788 * doc/install.texi: Document them.
789 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
790 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
791 <WITH_DOUBLE_COMPARISON>: New built-in defines.
792 * doc/invoke.texi (AVR Built-in Macros): Document them.
793 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
794 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
795 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
797 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
800 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
801 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
802 when only building rm-profile multilibs.
804 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
807 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
808 lattice for a value to check.
809 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
810 finite propagation in self-recursive scc.
812 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
814 * ipa-inline.c (caller_growth_limits): Restore the AND.
816 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
818 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
819 (VEC_ALLREG_ALT): New iterator.
820 (VEC_ALLREG_INT_MODE): New iterator.
821 (VCMP_MODE): New iterator.
822 (VCMP_MODE_INT): New iterator.
823 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
824 (vec_cmp<u>v64qidi): New define_expand.
825 (vec_cmp<mode>di_exec): Use VCMP_MODE.
826 (vec_cmpu<mode>di_exec): New define_expand.
827 (vec_cmp<u>v64qidi_exec): New define_expand.
828 (vec_cmp<mode>di_dup): Use VCMP_MODE.
829 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
830 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
831 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
832 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
833 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
834 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
835 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
836 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
837 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
839 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
840 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
842 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
844 * config/gcn/constraints.md (DA): Update description and match.
846 (Db): New constraint.
847 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
849 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
850 Implement 'Db' mixed immediate type.
851 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
852 (addcv64si3_dup<exec_vcc>): Delete.
853 (subcv64si3<exec_vcc>): Rework constraints.
854 (addv64di3): Rework constraints.
855 (addv64di3_exec): Rework constraints.
856 (subv64di3): Rework constraints.
857 (addv64di3_dup): Delete.
858 (addv64di3_dup_exec): Delete.
859 (addv64di3_zext): Rework constraints.
860 (addv64di3_zext_exec): Rework constraints.
861 (addv64di3_zext_dup): Rework constraints.
862 (addv64di3_zext_dup_exec): Rework constraints.
863 (addv64di3_zext_dup2): Rework constraints.
864 (addv64di3_zext_dup2_exec): Rework constraints.
865 (addv64di3_sext_dup2): Rework constraints.
866 (addv64di3_sext_dup2_exec): Rework constraints.
868 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
870 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
871 existing target checks.
873 2020-01-07 Richard Biener <rguenther@suse.de>
875 * doc/install.texi: Bump minimal supported MPC version.
877 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
879 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
880 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
881 * langhooks.c: Include stor-layout.h.
882 (lhd_simulate_enum_decl): New function.
883 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
884 handle_arm_sve_h for the LTO frontend.
885 (register_vector_type): Cope with null returns from pushdecl.
887 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
889 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
890 (aarch64_sve::nvectors_if_data_type): Replace with...
891 (aarch64_sve::builtin_type_p): ...this.
892 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
893 (find_vector_type): Delete.
894 (add_sve_type_attribute): New function.
895 (lookup_sve_type_attribute): Likewise.
896 (register_builtin_types): Add an "SVE type" attribute to each type.
897 (register_tuple_type): Likewise.
898 (svbool_type_p, nvectors_if_data_type): Delete.
899 (mangle_builtin_type): Use lookup_sve_type_attribute.
900 (builtin_type_p): Likewise. Add an overload that returns the
901 number of constituent vector and predicate registers.
902 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
903 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
904 instead of aarch64_sve_argument_p.
905 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
906 (aarch64_pass_by_reference): Likewise.
907 (aarch64_function_value_1): Likewise.
908 (aarch64_return_in_memory): Likewise.
909 (aarch64_layout_arg): Likewise.
911 2020-01-07 Jakub Jelinek <jakub@redhat.com>
913 PR tree-optimization/93156
914 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
915 least significant bit is always clear.
917 PR tree-optimization/93118
918 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
919 simplifier with two intermediate conversions.
921 2020-01-07 Martin Liska <mliska@suse.cz>
923 * params.opt: Add Optimization for various parameters.
925 2020-01-07 Martin Liska <mliska@suse.cz>
928 * doc/extend.texi: Explain cloning for target_clone
931 2020-01-07 Martin Liska <mliska@suse.cz>
933 PR tree-optimization/92860
934 * common.opt: Make in Optimization option
935 as it is affected by -O0, which is an Optimization
937 * tree-inline.c (tree_inlinable_function_p):
938 Use opt_for_fn for warn_inline.
939 (expand_call_inline): Likewise.
941 2020-01-07 Martin Liska <mliska@suse.cz>
943 PR tree-optimization/92860
944 * common.opt: Make flag_ree as optimization
947 2020-01-07 Martin Liska <mliska@suse.cz>
949 PR optimization/92860
950 * params.opt: Mark param_min_crossjump_insns with Optimization
953 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
955 * ipa-inline-analysis.c (estimate_growth): Fix typo.
956 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
958 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
960 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
961 helper function to return the valid addressing formats for a given
962 hard register and mode.
963 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
965 * config/rs6000/constraints.md (Q constraint): Update
967 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
970 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
971 Use 'Q' for doing vector extract from memory.
972 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
974 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
975 doing vector extract from memory.
976 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
979 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
980 for the offset being 34-bits when -mcpu=future is used.
982 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
984 * config/pa/pa.md: Revert change to use ordered_comparison_operator
985 instead of cmpib_comparison_operator in cmpib patterns.
986 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
987 of cmpib_comparison_operator. Revise comment.
989 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
991 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
992 in an IFN_DIV_POW2 node to be equal.
994 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
996 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
997 (vect_check_scalar_mask): ...this.
998 (vectorizable_store, vectorizable_load): Update call accordingly.
999 (vectorizable_call): Use vect_check_scalar_mask to check the mask
1000 argument in calls to conditional internal functions.
1002 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1004 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
1005 '0' matching inputs.
1006 (subv64di3_exec): Likewise.
1008 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
1010 * config/mips/mips.c (vr4130_align_insns): Fix typo.
1011 * doc/md.texi (movstr): Likewise.
1013 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1015 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
1018 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1020 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
1022 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
1023 to a temporary file and use move-if-change to update the real
1024 file where necessary.
1026 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1028 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
1029 rather than Upa for CPY /M.
1031 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1033 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
1036 2020-01-06 Martin Liska <mliska@suse.cz>
1038 PR tree-optimization/92860
1039 * params.opt: Mark param_max_combine_insns with Optimization
1042 2020-01-05 Jakub Jelinek <jakub@redhat.com>
1045 * config/i386/i386.md (SWIDWI): New mode iterator.
1046 (DWI, dwi): Add TImode variants.
1047 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
1048 <general_hilo_operand> instead of <general_operand>. Use
1049 CONST_SCALAR_INT_P instead of CONST_INT_P.
1050 (*addv<mode>4_1): Rename to ...
1051 (addv<mode>4_1): ... this.
1052 (QWI): New mode attribute.
1053 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
1054 define_insn_and_split patterns.
1055 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
1057 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
1058 <general_hilo_operand> instead of <general_operand>.
1059 (*addcarry<mode>_1): New define_insn.
1060 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
1062 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
1064 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
1065 Use "call" instead of "set".
1067 2020-01-03 Martin Jambor <mjambor@suse.cz>
1070 * ipa-cp.c (print_all_lattices): Skip functions without info.
1072 2020-01-03 Jakub Jelinek <jakub@redhat.com>
1075 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
1076 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
1077 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
1078 for 'e' simd clones.
1081 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
1083 (mprefer-vector-width=): Add Save.
1084 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
1085 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
1086 (ix86_debug_options, ix86_function_specific_print): Adjust
1087 ix86_target_string callers.
1088 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
1089 (ix86_valid_target_attribute_tree): Likewise.
1090 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
1091 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
1092 ix86_target_string caller.
1095 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
1096 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
1097 instead of gen_int_shift_amount + convert_modes.
1099 PR rtl-optimization/93088
1100 * loop-iv.c (find_single_def_src): Punt after looking through
1101 128 reg copies for regs with single definitions. Move definitions
1104 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
1106 * config/arm/arm-c.c (arm_cpu_builtins): Define
1107 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
1108 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
1109 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
1110 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
1111 * config/arm/arm-tables.opt: Regenerated.
1112 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
1113 arm_arch_i8mm and arm_arch_bf16 when enabled.
1114 * config/arm/arm.h (TARGET_I8MM): New macro.
1115 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
1116 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
1117 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
1118 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
1119 (v8_6_a_simd_variants): New.
1120 (v8_*_a_simd_variants): Add i8mm and bf16.
1121 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
1123 2020-01-02 Jakub Jelinek <jakub@redhat.com>
1126 * predict.c (compute_function_frequency): Don't call
1127 warn_function_cold on functions that already have cold attribute.
1129 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
1132 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
1133 COMDAT group function labels in .data.rel.ro.local section.
1134 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
1137 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
1138 comparison_operator in B and S integer comparisons. Likewise, use
1139 ordered_comparison_operator instead of cmpib_comparison_operator in
1141 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
1143 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1145 Update copyright years.
1147 * gcc.c (process_command): Update copyright notice dates.
1148 * gcov-dump.c (print_version): Ditto.
1149 * gcov.c (print_version): Ditto.
1150 * gcov-tool.c (print_version): Ditto.
1151 * gengtype.c (create_file): Ditto.
1152 * doc/cpp.texi: Bump @copying's copyright year.
1153 * doc/cppinternals.texi: Ditto.
1154 * doc/gcc.texi: Ditto.
1155 * doc/gccint.texi: Ditto.
1156 * doc/gcov.texi: Ditto.
1157 * doc/install.texi: Ditto.
1158 * doc/invoke.texi: Ditto.
1160 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
1162 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
1165 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1167 PR tree-optimization/93098
1168 * match.pd (popcount): For shift amounts, use integer_onep
1169 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
1170 tests. Make sure that precision is power of two larger than or equal
1171 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
1172 instead of ULL suffixed constants. Formatting fixes.
1174 Copyright (C) 2020 Free Software Foundation, Inc.
1176 Copying and distribution of this file, with or without modification,
1177 are permitted in any medium without royalty provided the copyright
1178 notice and this notice are preserved.