1 2020-01-10 Martin Jambor <mjambor@suse.cz>
3 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
5 to get param_ipa_sra_max_replacements.
6 (param_splitting_across_edge): Pass the caller to
7 pull_accesses_from_callee.
9 2020-01-10 Martin Jambor <mjambor@suse.cz>
11 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
12 * ipa-cp.c (max_new_size): Removed.
13 (orig_overall_size): New variable.
14 (get_max_overall_size): New function.
15 (estimate_local_effects): Use it. Adjust dump.
16 (decide_about_value): Likewise.
17 (ipcp_propagate_stage): Do not calculate max_new_size, just store
18 orig_overall_size. Adjust dump.
19 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
21 2020-01-10 Martin Jambor <mjambor@suse.cz>
23 * params.opt (param_ipa_max_agg_items): Mark as Optimization
24 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
25 instead of param_ipa_max_agg_items.
26 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
27 optimization info for the callee.
29 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
31 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
32 markers if debug_inline_points is false.
34 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
36 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
38 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
39 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
40 aarch64-sve-builtins-sve2.h.
41 (aarch64-sve-builtins-sve2.o): New rule.
42 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
43 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
44 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
45 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
46 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
47 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
49 * config/aarch64/aarch64-sve.md: Update comments with SVE2
50 instructions that are handled here.
51 (@cond_asrd<mode>): Generalize to...
52 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
53 (*cond_asrd<mode>_2): Generalize to...
54 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
55 (*cond_asrd<mode>_z): Generalize to...
56 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
57 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
58 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
59 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
60 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
62 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
63 (@aarch64_scatter_stnt<mode>): Likewise.
64 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
65 (@aarch64_mul_lane_<mode>): Likewise.
66 (@aarch64_sve_suqadd<mode>_const): Likewise.
67 (*<sur>h<addsub><mode>): Generalize to...
68 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
70 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
71 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
72 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
73 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
74 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
75 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
76 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
77 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
78 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
79 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
80 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
81 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
82 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
83 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
84 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
85 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
86 (@aarch64_sve2_xar<mode>): Likewise.
87 (@aarch64_sve2_bcax<mode>): Likewise.
88 (*aarch64_sve2_eor3<mode>): Rename to...
89 (@aarch64_sve2_eor3<mode>): ...this.
90 (@aarch64_sve2_bsl<mode>): New expander.
91 (@aarch64_sve2_nbsl<mode>): Likewise.
92 (@aarch64_sve2_bsl1n<mode>): Likewise.
93 (@aarch64_sve2_bsl2n<mode>): Likewise.
94 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
95 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
96 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
97 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
98 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
99 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
100 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
101 (<su>mull<bt><Vwide>): Generalize to...
102 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
104 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
105 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
106 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
107 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
108 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
109 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
110 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
111 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
112 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
113 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
114 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
115 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
116 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
117 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
118 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
119 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
120 (<SHRNB:r>shrnb<mode>): Generalize to...
121 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
123 (<SHRNT:r>shrnt<mode>): Generalize to...
124 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
126 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
127 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
128 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
129 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
130 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
131 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
132 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
133 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
134 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
135 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
136 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
137 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
138 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
139 (@aarch64_sve2_cvtnt<mode>): Likewise.
140 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
141 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
142 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
143 (@aarch64_sve2_cvtxnt<mode>): Likewise.
144 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
145 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
146 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
147 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
148 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
149 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
150 (@aarch64_sve2_pmul<mode>): Likewise.
151 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
152 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
153 (@aarch64_sve2_tbl2<mode>): Likewise.
154 (@aarch64_sve2_tbx<mode>): Likewise.
155 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
156 (@aarch64_sve2_histcnt<mode>): Likewise.
157 (@aarch64_sve2_histseg<mode>): Likewise.
158 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
159 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
160 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
161 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
162 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
163 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
164 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
165 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
166 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
167 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
168 (SVE2_PMULL_PAIR_I): New mode iterators.
169 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
170 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
171 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
172 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
173 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
174 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
175 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
176 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
177 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
178 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
179 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
180 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
181 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
182 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
183 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
184 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
185 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
186 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
187 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
188 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
189 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
190 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
191 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
192 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
193 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
194 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
195 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
196 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
197 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
198 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
199 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
201 (VNARROW, Ventype): New mode attributes.
202 (Vewtype): Handle VNx2DI. Fix typo in comment.
203 (VDOUBLE): New mode attribute.
204 (sve_lane_con): Handle VNx8HI.
205 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
206 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
207 (sve_int_op, sve_int_op_rev): Handle the above codes.
208 (sve_pred_int_rhs2_operand): Likewise.
209 (MULLBT, SHRNB, SHRNT): Delete.
210 (SVE_INT_SHIFT_IMM): New int iterator.
211 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
212 and UNSPEC_WHILEHS for TARGET_SVE2.
213 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
214 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
215 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
216 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
217 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
218 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
219 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
220 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
221 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
222 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
223 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
224 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
225 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
226 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
227 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
228 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
229 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
230 (optab): Handle the new unspecs.
231 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
233 (lr): Handle the new unspecs.
235 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
236 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
237 (sve_int_qsub_op): New int attributes.
238 (sve_fp_op, rot): Handle the new unspecs.
239 * config/aarch64/aarch64-sve-builtins.h
240 (function_resolver::require_matching_pointer_type): Declare.
241 (function_resolver::resolve_unary): Add an optional boolean argument.
242 (function_resolver::finish_opt_n_resolution): Add an optional
243 type_suffix_index argument.
244 (gimple_folder::redirect_call): Declare.
245 (gimple_expander::prepare_gather_address_operands): Add an optional
247 * config/aarch64/aarch64-sve-builtins.cc: Include
248 aarch64-sve-builtins-sve2.h.
249 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
250 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
251 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
252 (TYPES_hsd_integer): Use TYPES_hsd_signed.
253 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
254 (TYPES_s_unsigned): Likewise.
255 (TYPES_s_integer): Use TYPES_s_unsigned.
256 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
257 (TYPES_sd_integer): Use them.
258 (TYPES_d_unsigned): New macro.
259 (TYPES_d_integer): Use it.
260 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
261 (TYPES_cvt_narrow): Likewise.
262 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
263 (preds_mx): New variable.
264 (function_builder::add_overloaded_function): Allow the new feature
265 set to be more restrictive than the original one.
266 (function_resolver::infer_pointer_type): Remove qualifiers from
267 the pointer type before printing it.
268 (function_resolver::require_matching_pointer_type): New function.
269 (function_resolver::resolve_sv_displacement): Handle functions
270 that don't support 32-bit vector indices or svint32_t vector offsets.
271 (function_resolver::finish_opt_n_resolution): Take the inferred type
272 as a separate argument.
273 (function_resolver::resolve_unary): Optionally treat all forms in
274 the same way as normal merging functions.
275 (gimple_folder::redirect_call): New function.
276 (function_expander::prepare_gather_address_operands): Add an argument
277 that says whether scaled forms are available. If they aren't,
278 handle scaling of vector indices and don't add the extension and
280 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
281 fall back to using cond_* instead.
282 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
283 Split out the member variables into...
284 (rtx_code_function_base): ...this new base class.
285 (rtx_code_function_rotated): Inherit rtx_code_function_base.
286 (unspec_based_function): Split out the member variables into...
287 (unspec_based_function_base): ...this new base class.
288 (unspec_based_function_rotated): Inherit unspec_based_function_base.
289 (unspec_based_function_exact_insn): New class.
290 (unspec_based_add_function, unspec_based_add_lane_function)
291 (unspec_based_lane_function, unspec_based_pred_function)
292 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
293 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
294 (unspec_based_sub_function, unspec_based_sub_lane_function): New
296 (unspec_based_fused_function): New class.
297 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
298 (unspec_based_fused_lane_function): New class.
299 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
301 (CODE_FOR_MODE1): New macro.
302 (fixed_insn_function): New class.
303 (while_comparison): Likewise.
304 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
305 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
306 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
307 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
308 (load_gather_sv_restricted, shift_left_imm_long): Declare.
309 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
310 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
311 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
312 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
313 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
314 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
315 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
316 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
317 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
318 Also add an initial argument for unary_convert_narrowt, regardless
319 of the predication type.
320 (build_32_64): Allow loads and stores to specify MODE_none.
321 (build_sv_index64, build_sv_uint_offset): New functions.
322 (long_type_suffix): New function.
323 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
324 (binary_imm_long_base, load_gather_sv_base): Likewise.
325 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
326 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
327 (unary_narrowb_base, unary_narrowt_base): Likewise.
328 (binary_long_lane_def, binary_long_lane): New shape.
329 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
330 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
331 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
332 (binary_to_uint_def, binary_to_uint): Likewise.
333 (binary_wide_def, binary_wide): Likewise.
334 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
335 (compare_def, compare): Likewise.
336 (compare_ptr_def, compare_ptr): Likewise.
337 (load_ext_gather_index_restricted_def,
338 load_ext_gather_index_restricted): Likewise.
339 (load_ext_gather_offset_restricted_def,
340 load_ext_gather_offset_restricted): Likewise.
341 (load_gather_sv_def): Inherit from load_gather_sv_base.
342 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
343 (shift_left_imm_def, shift_left_imm): Likewise.
344 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
345 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
346 (store_scatter_index_restricted_def,
347 store_scatter_index_restricted): Likewise.
348 (store_scatter_offset_restricted_def,
349 store_scatter_offset_restricted): Likewise.
350 (tbl_tuple_def, tbl_tuple): Likewise.
351 (ternary_long_lane_def, ternary_long_lane): Likewise.
352 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
353 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
354 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
355 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
356 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
357 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
358 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
359 (ternary_uint_def, ternary_uint): Likewise.
360 (unary_convert): Fix typo in comment.
361 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
362 (unary_long_def, unary_long): Likewise.
363 (unary_narrowb_def, unary_narrowb): Likewise.
364 (unary_narrowt_def, unary_narrowt): Likewise.
365 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
366 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
367 (unary_to_int_def, unary_to_int): Likewise.
368 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
369 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
370 (svasrd_impl): Delete.
371 (svcadd_impl::expand): Handle integer operations too.
372 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
373 new functions to derive the unspec numbers.
374 (svmla_svmls_lane_impl): Replace with...
375 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
376 integer operations too.
377 (svwhile_impl): Rename to...
378 (svwhilelx_impl): ...this and inherit from while_comparison.
379 (svasrd): Use unspec_based_function.
380 (svmla_lane): Use svmla_lane_impl.
381 (svmls_lane): Use svmls_lane_impl.
382 (svrecpe, svrsqrte): Handle unsigned integer operations too.
383 (svwhilele, svwhilelt): Use svwhilelx_impl.
384 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
385 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
386 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
387 * config/aarch64/aarch64-sve-builtins.def: Include
388 aarch64-sve-builtins-sve2.def.
390 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
392 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
393 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
394 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
395 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
396 immediates as well as vector ones.
397 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
398 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
399 (aarch64_sve_qsub_immediate): Update calls accordingly.
401 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
403 * config/aarch64/aarch64-sve2.md: Add banner comments.
404 (<su>mulh<r>s<mode>3): Move further up file.
405 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
406 (*aarch64_sve2_sra<mode>): Move further down file.
407 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
409 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
411 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
413 (while_optab_cmp): Handle them.
414 * config/aarch64/aarch64-sve.md
415 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
416 and add a "@" marker.
417 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
418 instead of gen_aarch64_sve2_while_ptest.
419 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
421 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
423 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
424 (UNSPEC_WHILELE): ...this.
425 (UNSPEC_WHILE_LO): Rename to...
426 (UNSPEC_WHILELO): ...this.
427 (UNSPEC_WHILE_LS): Rename to...
428 (UNSPEC_WHILELS): ...this.
429 (UNSPEC_WHILE_LT): Rename to...
430 (UNSPEC_WHILELT): ...this.
431 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
432 (cmp_op, while_optab_cmp): Likewise.
433 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
434 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
435 (svwhilelt): Likewise.
437 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
439 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
440 (unary_to_uint): Define.
441 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
442 (unary_count): Rename to...
443 (unary_to_uint_def, unary_to_uint): ...this.
444 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
446 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
448 * config/aarch64/aarch64-sve-builtins-functions.h
449 (code_for_mode_function): New class.
450 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
451 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
452 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
453 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
454 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
456 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
458 * config/aarch64/iterators.md (addsub): New code attribute.
459 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
461 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
462 in the asm string and attributes. Fix indentation.
463 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
465 (@aarch64_sve_<optab><mode>): ...this.
466 * config/aarch64/aarch64-sve-builtins.h
467 (function_expander::expand_signed_unpred_op): Delete.
468 * config/aarch64/aarch64-sve-builtins.cc
469 (function_expander::expand_signed_unpred_op): Likewise.
470 (function_expander::map_to_rtx_codes): If the optab isn't defined,
471 try using code_for_aarch64_sve instead.
472 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
473 (svqsub_impl): Likewise.
474 (svqadd, svqsub): Use rtx_code_function instead.
476 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
478 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
479 (HADDSUB, sur, addsub): Remove them.
481 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
483 * tree-nrv.c (pass_return_slot::execute): Handle all internal
484 functions the same way, rather than singling out those that
485 aren't mapped directly to optabs.
487 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
489 * target.def (compatible_vector_types_p): New target hook.
490 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
491 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
492 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
493 * doc/tm.texi: Regenerate.
494 * gimple-expr.c: Include target.h.
495 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
496 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
498 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
499 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
500 Use the original predicate if it already has a suitable type.
502 2020-01-09 Martin Jambor <mjambor@suse.cz>
504 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
505 resolve_speculation and redirect_call_stmt_to_callee static. Change
506 return type of set_call_stmt to cgraph_edge *.
507 * auto-profile.c (afdo_indirect_call): Adjust call to
508 redirect_call_stmt_to_callee.
509 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
510 make the this pointer explicit, adjust self-recursive calls and the
511 call top make_direct. Return the resulting edge.
512 (cgraph_edge::remove): Make this pointer explicit.
513 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
514 (cgraph_edge::make_direct): Likewise, adjust call to
516 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
517 call to set_call_stmt.
518 (cgraph_update_edges_for_call_stmt_node): Update call to
519 set_call_stmt and remove.
520 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
521 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
522 (cgraph_node::create_edge_including_clones): Moved "first" definition
523 of edge to the block where it was used. Adjusted calls to
525 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
527 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
528 make_direct and redirect_call_stmt_to_callee.
529 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
530 resolve_speculation and make_direct.
531 * ipa-inline-transform.c (inline_transform): Adjust call to
532 redirect_call_stmt_to_callee.
533 (check_speculations_1):: Adjust call to resolve_speculation.
534 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
536 (inline_small_functions): Adjust call to resolve_speculation.
537 (ipa_inline): Likewise.
538 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
540 * ipa-visibility.c (function_and_variable_visibility): Make iteration
541 safe with regards to edge removal, adjust calls to
542 redirect_call_stmt_to_callee.
543 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
544 and redirect_call_stmt_to_callee.
545 * multiple_target.c (create_dispatcher_calls): Adjust call to
546 redirect_call_stmt_to_callee
547 (redirect_to_specific_clone): Likewise.
548 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
549 Adjust calls to cgraph_edge::remove.
550 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
551 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
552 (expand_call_inline): Adjust call to cgraph_edge::remove.
554 2020-01-09 Martin Liska <mliska@suse.cz>
556 * params.opt: Set Optimization for
557 param_max_speculative_devirt_maydefs.
559 2020-01-09 Martin Sebor <msebor@redhat.com>
563 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
565 2020-01-09 Martin Liska <mliska@suse.cz>
567 * auto-profile.c (auto_profile): Use opt_for_fn
569 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
570 (propagate_vals_across_arith_jfunc): Likewise.
571 (hint_time_bonus): Likewise.
572 (incorporate_penalties): Likewise.
573 (good_cloning_opportunity_p): Likewise.
574 (perform_estimation_of_a_value): Likewise.
575 (estimate_local_effects): Likewise.
576 (ipcp_propagate_stage): Likewise.
577 * ipa-fnsummary.c (decompose_param_expr): Likewise.
578 (set_switch_stmt_execution_predicate): Likewise.
579 (analyze_function_body): Likewise.
580 * ipa-inline-analysis.c (offline_size): Likewise.
581 * ipa-inline.c (early_inliner): Likewise.
582 * ipa-prop.c (ipa_analyze_node): Likewise.
583 (ipcp_transform_function): Likewise.
584 * ipa-sra.c (process_scan_results): Likewise.
585 (ipa_sra_summarize_function): Likewise.
586 * params.opt: Rename ipcp-unit-growth to
587 ipa-cp-unit-growth. Add Optimization for various
588 IPA-related parameters.
590 2020-01-09 Richard Biener <rguenther@suse.de>
593 * gimplify.c (gimplify_expr): Deal with NOP definitions.
595 2020-01-09 Richard Biener <rguenther@suse.de>
597 PR tree-optimization/93040
598 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
600 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
602 * common/config/avr/avr-common.c (avr_option_optimization_table)
603 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
605 2020-01-09 Martin Liska <mliska@suse.cz>
607 * cgraphclones.c (symbol_table::materialize_all_clones):
608 Use cgraph_node::dump_name.
610 2020-01-09 Jakub Jelinek <jakub@redhat.com>
613 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
614 output_operand_lossage instead of gcc_unreachable.
615 * doc/md.texi (riscv f constraint): Fix typo.
618 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
619 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
620 CONST_SCALAR_INT_P instead of CONST_INT_P.
621 (*subv<mode>4_1): Rename to ...
622 (subv<mode>4_1): ... this.
623 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
624 define_insn_and_split patterns.
625 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
628 2020-01-08 David Malcolm <dmalcolm@redhat.com>
630 * vec.c (class selftest::count_dtor): New class.
631 (selftest::test_auto_delete_vec): New test.
632 (selftest::vec_c_tests): Call it.
633 * vec.h (class auto_delete_vec): New class template.
634 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
636 2020-01-08 David Malcolm <dmalcolm@redhat.com>
638 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
640 2020-01-08 Jim Wilson <jimw@sifive.com>
642 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
643 use of TLS_MODEL_LOCAL_EXEC when not pic.
645 2020-01-08 David Malcolm <dmalcolm@redhat.com>
647 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
650 2020-01-08 Jakub Jelinek <jakub@redhat.com>
653 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
654 *stack_protect_set_3 peephole2): Also check that the second
655 insns source is general_operand.
658 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
659 predicate for output operand instead of register_operand.
660 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
661 memory destination and non-memory operands[2].
663 2020-01-08 Martin Liska <mliska@suse.cz>
665 * cgraph.c (cgraph_node::dump): Use ::dump_name or
666 ::dump_asm_name instead of (::name or ::asm_name).
667 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
668 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
669 (analyze_functions): Likewise.
670 (expand_all_functions): Likewise.
671 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
672 (propagate_bits_across_jump_function): Likewise.
673 (dump_profile_updates): Likewise.
674 (ipcp_store_bits_results): Likewise.
675 (ipcp_store_vr_results): Likewise.
676 * ipa-devirt.c (dump_targets): Likewise.
677 * ipa-fnsummary.c (analyze_function_body): Likewise.
678 * ipa-hsa.c (check_warn_node_versionable): Likewise.
679 (process_hsa_functions): Likewise.
680 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
681 (set_alias_uids): Likewise.
682 * ipa-inline-transform.c (save_inline_function_body): Likewise.
683 * ipa-inline.c (recursive_inlining): Likewise.
684 (inline_to_all_callers_1): Likewise.
685 (ipa_inline): Likewise.
686 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
687 (ipa_propagate_frequency): Likewise.
688 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
689 (remove_described_reference): Likewise.
690 * ipa-pure-const.c (worse_state): Likewise.
691 (check_retval_uses): Likewise.
692 (analyze_function): Likewise.
693 (propagate_pure_const): Likewise.
694 (propagate_nothrow): Likewise.
695 (dump_malloc_lattice): Likewise.
696 (propagate_malloc): Likewise.
697 (pass_local_pure_const::execute): Likewise.
698 * ipa-visibility.c (optimize_weakref): Likewise.
699 (function_and_variable_visibility): Likewise.
700 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
701 (ipa_discover_variable_flags): Likewise.
702 * lto-streamer-out.c (output_function): Likewise.
703 (output_constructor): Likewise.
704 * tree-inline.c (copy_bb): Likewise.
705 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
706 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
708 2020-01-08 Richard Biener <rguenther@suse.de>
711 * tree-eh.c (sink_clobbers): Update virtual operands for
712 the first and last stmt only. Add a dry-run capability.
713 (pass_lower_eh_dispatch::execute): Perform clobber sinking
714 after CFG manipulations and in RPO order to catch all
715 secondary opportunities reliably.
717 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
720 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
722 2019-01-08 Richard Biener <rguenther@suse.de>
725 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
726 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
727 virtual operand, also updating SSA use.
728 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
729 Update stmt after resetting virtual operand.
730 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
731 * gimple-iterator.c (gsi_remove): When not removing the stmt
732 permanently do not delink immediate uses or mark the stmt modified.
734 2020-01-08 Martin Liska <mliska@suse.cz>
736 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
737 (ipa_call_context::estimate_size_and_time): Likewise.
738 (inline_analyze_function): Likewise.
740 2020-01-08 Martin Liska <mliska@suse.cz>
742 * cgraph.c (cgraph_node::dump): Use systematically
745 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
747 Add -nodevicespecs option for avr.
750 * config/avr/avr.opt (-nodevicespecs): New driver option.
751 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
752 "-specs=device-specs/..." if that option is not set.
753 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
755 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
757 Implement 64-bit double functions for avr.
760 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
761 --with-double-comparison.
762 * doc/install.texi: Document them.
763 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
764 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
765 <WITH_DOUBLE_COMPARISON>: New built-in defines.
766 * doc/invoke.texi (AVR Built-in Macros): Document them.
767 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
768 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
769 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
771 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
774 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
775 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
776 when only building rm-profile multilibs.
778 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
781 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
782 lattice for a value to check.
783 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
784 finite propagation in self-recursive scc.
786 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
788 * ipa-inline.c (caller_growth_limits): Restore the AND.
790 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
792 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
793 (VEC_ALLREG_ALT): New iterator.
794 (VEC_ALLREG_INT_MODE): New iterator.
795 (VCMP_MODE): New iterator.
796 (VCMP_MODE_INT): New iterator.
797 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
798 (vec_cmp<u>v64qidi): New define_expand.
799 (vec_cmp<mode>di_exec): Use VCMP_MODE.
800 (vec_cmpu<mode>di_exec): New define_expand.
801 (vec_cmp<u>v64qidi_exec): New define_expand.
802 (vec_cmp<mode>di_dup): Use VCMP_MODE.
803 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
804 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
805 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
806 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
807 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
808 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
809 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
810 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
811 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
813 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
814 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
816 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
818 * config/gcn/constraints.md (DA): Update description and match.
820 (Db): New constraint.
821 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
823 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
824 Implement 'Db' mixed immediate type.
825 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
826 (addcv64si3_dup<exec_vcc>): Delete.
827 (subcv64si3<exec_vcc>): Rework constraints.
828 (addv64di3): Rework constraints.
829 (addv64di3_exec): Rework constraints.
830 (subv64di3): Rework constraints.
831 (addv64di3_dup): Delete.
832 (addv64di3_dup_exec): Delete.
833 (addv64di3_zext): Rework constraints.
834 (addv64di3_zext_exec): Rework constraints.
835 (addv64di3_zext_dup): Rework constraints.
836 (addv64di3_zext_dup_exec): Rework constraints.
837 (addv64di3_zext_dup2): Rework constraints.
838 (addv64di3_zext_dup2_exec): Rework constraints.
839 (addv64di3_sext_dup2): Rework constraints.
840 (addv64di3_sext_dup2_exec): Rework constraints.
842 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
844 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
845 existing target checks.
847 2020-01-07 Richard Biener <rguenther@suse.de>
849 * doc/install.texi: Bump minimal supported MPC version.
851 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
853 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
854 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
855 * langhooks.c: Include stor-layout.h.
856 (lhd_simulate_enum_decl): New function.
857 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
858 handle_arm_sve_h for the LTO frontend.
859 (register_vector_type): Cope with null returns from pushdecl.
861 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
863 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
864 (aarch64_sve::nvectors_if_data_type): Replace with...
865 (aarch64_sve::builtin_type_p): ...this.
866 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
867 (find_vector_type): Delete.
868 (add_sve_type_attribute): New function.
869 (lookup_sve_type_attribute): Likewise.
870 (register_builtin_types): Add an "SVE type" attribute to each type.
871 (register_tuple_type): Likewise.
872 (svbool_type_p, nvectors_if_data_type): Delete.
873 (mangle_builtin_type): Use lookup_sve_type_attribute.
874 (builtin_type_p): Likewise. Add an overload that returns the
875 number of constituent vector and predicate registers.
876 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
877 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
878 instead of aarch64_sve_argument_p.
879 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
880 (aarch64_pass_by_reference): Likewise.
881 (aarch64_function_value_1): Likewise.
882 (aarch64_return_in_memory): Likewise.
883 (aarch64_layout_arg): Likewise.
885 2020-01-07 Jakub Jelinek <jakub@redhat.com>
887 PR tree-optimization/93156
888 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
889 least significant bit is always clear.
891 PR tree-optimization/93118
892 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
893 simplifier with two intermediate conversions.
895 2020-01-07 Martin Liska <mliska@suse.cz>
897 * params.opt: Add Optimization for various parameters.
899 2020-01-07 Martin Liska <mliska@suse.cz>
902 * doc/extend.texi: Explain cloning for target_clone
905 2020-01-07 Martin Liska <mliska@suse.cz>
907 PR tree-optimization/92860
908 * common.opt: Make in Optimization option
909 as it is affected by -O0, which is an Optimization
911 * tree-inline.c (tree_inlinable_function_p):
912 Use opt_for_fn for warn_inline.
913 (expand_call_inline): Likewise.
915 2020-01-07 Martin Liska <mliska@suse.cz>
917 PR tree-optimization/92860
918 * common.opt: Make flag_ree as optimization
921 2020-01-07 Martin Liska <mliska@suse.cz>
923 PR optimization/92860
924 * params.opt: Mark param_min_crossjump_insns with Optimization
927 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
929 * ipa-inline-analysis.c (estimate_growth): Fix typo.
930 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
932 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
934 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
935 helper function to return the valid addressing formats for a given
936 hard register and mode.
937 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
939 * config/rs6000/constraints.md (Q constraint): Update
941 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
944 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
945 Use 'Q' for doing vector extract from memory.
946 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
948 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
949 doing vector extract from memory.
950 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
953 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
954 for the offset being 34-bits when -mcpu=future is used.
956 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
958 * config/pa/pa.md: Revert change to use ordered_comparison_operator
959 instead of cmpib_comparison_operator in cmpib patterns.
960 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
961 of cmpib_comparison_operator. Revise comment.
963 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
965 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
966 in an IFN_DIV_POW2 node to be equal.
968 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
970 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
971 (vect_check_scalar_mask): ...this.
972 (vectorizable_store, vectorizable_load): Update call accordingly.
973 (vectorizable_call): Use vect_check_scalar_mask to check the mask
974 argument in calls to conditional internal functions.
976 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
978 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
980 (subv64di3_exec): Likewise.
982 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
984 * config/mips/mips.c (vr4130_align_insns): Fix typo.
985 * doc/md.texi (movstr): Likewise.
987 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
989 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
992 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
994 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
996 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
997 to a temporary file and use move-if-change to update the real
998 file where necessary.
1000 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1002 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
1003 rather than Upa for CPY /M.
1005 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1007 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
1010 2020-01-06 Martin Liska <mliska@suse.cz>
1012 PR tree-optimization/92860
1013 * params.opt: Mark param_max_combine_insns with Optimization
1016 2020-01-05 Jakub Jelinek <jakub@redhat.com>
1019 * config/i386/i386.md (SWIDWI): New mode iterator.
1020 (DWI, dwi): Add TImode variants.
1021 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
1022 <general_hilo_operand> instead of <general_operand>. Use
1023 CONST_SCALAR_INT_P instead of CONST_INT_P.
1024 (*addv<mode>4_1): Rename to ...
1025 (addv<mode>4_1): ... this.
1026 (QWI): New mode attribute.
1027 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
1028 define_insn_and_split patterns.
1029 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
1031 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
1032 <general_hilo_operand> instead of <general_operand>.
1033 (*addcarry<mode>_1): New define_insn.
1034 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
1036 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
1038 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
1039 Use "call" instead of "set".
1041 2020-01-03 Martin Jambor <mjambor@suse.cz>
1044 * ipa-cp.c (print_all_lattices): Skip functions without info.
1046 2020-01-03 Jakub Jelinek <jakub@redhat.com>
1049 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
1050 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
1051 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
1052 for 'e' simd clones.
1055 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
1057 (mprefer-vector-width=): Add Save.
1058 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
1059 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
1060 (ix86_debug_options, ix86_function_specific_print): Adjust
1061 ix86_target_string callers.
1062 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
1063 (ix86_valid_target_attribute_tree): Likewise.
1064 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
1065 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
1066 ix86_target_string caller.
1069 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
1070 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
1071 instead of gen_int_shift_amount + convert_modes.
1073 PR rtl-optimization/93088
1074 * loop-iv.c (find_single_def_src): Punt after looking through
1075 128 reg copies for regs with single definitions. Move definitions
1078 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
1080 * config/arm/arm-c.c (arm_cpu_builtins): Define
1081 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
1082 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
1083 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
1084 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
1085 * config/arm/arm-tables.opt: Regenerated.
1086 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
1087 arm_arch_i8mm and arm_arch_bf16 when enabled.
1088 * config/arm/arm.h (TARGET_I8MM): New macro.
1089 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
1090 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
1091 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
1092 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
1093 (v8_6_a_simd_variants): New.
1094 (v8_*_a_simd_variants): Add i8mm and bf16.
1095 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
1097 2020-01-02 Jakub Jelinek <jakub@redhat.com>
1100 * predict.c (compute_function_frequency): Don't call
1101 warn_function_cold on functions that already have cold attribute.
1103 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
1106 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
1107 COMDAT group function labels in .data.rel.ro.local section.
1108 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
1111 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
1112 comparison_operator in B and S integer comparisons. Likewise, use
1113 ordered_comparison_operator instead of cmpib_comparison_operator in
1115 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
1117 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1119 Update copyright years.
1121 * gcc.c (process_command): Update copyright notice dates.
1122 * gcov-dump.c (print_version): Ditto.
1123 * gcov.c (print_version): Ditto.
1124 * gcov-tool.c (print_version): Ditto.
1125 * gengtype.c (create_file): Ditto.
1126 * doc/cpp.texi: Bump @copying's copyright year.
1127 * doc/cppinternals.texi: Ditto.
1128 * doc/gcc.texi: Ditto.
1129 * doc/gccint.texi: Ditto.
1130 * doc/gcov.texi: Ditto.
1131 * doc/install.texi: Ditto.
1132 * doc/invoke.texi: Ditto.
1134 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
1136 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
1139 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1141 PR tree-optimization/93098
1142 * match.pd (popcount): For shift amounts, use integer_onep
1143 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
1144 tests. Make sure that precision is power of two larger than or equal
1145 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
1146 instead of ULL suffixed constants. Formatting fixes.
1148 Copyright (C) 2020 Free Software Foundation, Inc.
1150 Copying and distribution of this file, with or without modification,
1151 are permitted in any medium without royalty provided the copyright
1152 notice and this notice are preserved.