ChangeLog.gcc43 revision 252080
12007-05-01 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com> (r124341) 2 3 * doc/invoke.texi: Fix typo, 'AMD Family 10h core' instead of 4 'AMD Family 10 core'. 5 62007-05-01 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com> (r124339) 7 8 * config/i386/i386.c (override_options): Accept k8-sse3, opteron-sse3 9 and athlon64-sse3 as improved versions of k8, opteron and athlon64 10 with SSE3 instruction set support. 11 * doc/invoke.texi: Likewise. 12 132007-05-01 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com> (r124330) 14 15 * config/i386/i386.c (override_options): Tuning 32-byte loop 16 alignment for amdfam10 architecture. Increasing the max loop 17 alignment to 24 bytes. 18 192007-04-12 Richard Guenther <rguenther@suse.de> (r123736) 20 21 PR tree-optimization/24689 22 PR tree-optimization/31307 23 * fold-const.c (operand_equal_p): Compare INTEGER_CST array 24 indices by value. 25 * gimplify.c (canonicalize_addr_expr): To be consistent with 26 gimplify_compound_lval only set operands two and three of 27 ARRAY_REFs if they are not gimple_min_invariant. This makes 28 it never at this place. 29 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Likewise. 30 312007-04-07 H.J. Lu <hongjiu.lu@intel.com> (r123639) 32 33 * config/i386/i386.c (ix86_handle_option): Handle SSSE3. 34 352007-03-28 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com> (r123313) 36 37 * config.gcc: Accept barcelona as a variant of amdfam10. 38 * config/i386/i386.c (override_options): Likewise. 39 * doc/invoke.texi: Likewise. 40 412007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com> (r121763) 42 43 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10. 44 (bit_SSE4a): New. 45 462007-02-08 Harsha Jagasia <harsha.jagasia@amd.com> (r121726) 47 48 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h 49 conditional to __SSE2__. 50 (Entries below should have been added to first ChangeLog 51 entry for amdfam10 dated 2007-02-05) 52 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not 53 defined. 54 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not 55 defined. 56 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not 57 defined. 58 592007-02-07 Jakub Jelinek <jakub@redhat.com> (r121687) 60 61 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2. 62 632007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 64 65 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8, 66 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov, 67 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul, 68 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn, 69 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8, 70 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load, 71 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8, 72 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10. 73 742007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 75 76 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse, 77 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387, 78 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse, 79 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse, 80 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed, 81 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse, 82 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1, 83 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn, 84 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn, 85 umuldi3_highpart_rex64, umulsi3_highpart_insn, 86 umulsi3_highpart_zext, smuldi3_highpart_rex64, 87 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld, 88 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse, 89 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387, 90 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387, 91 sqrtextenddfxf2_i387): Added amdfam10_decode. 92 93 * config/i386/athlon.md (athlon_idirect_amdfam10, 94 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10, 95 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10, 96 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10, 97 athlon_ivector_store_amdfam10): New define_insn_reservation. 98 (athlon_idirect_loadmov, athlon_idirect_movstore): Added 99 amdfam10. 100 1012007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 102 103 * config/i386/athlon.md (athlon_call_amdfam10, 104 athlon_pop_amdfam10, athlon_lea_amdfam10): New 105 define_insn_reservation. 106 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8, 107 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI, 108 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10. 109 1102007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 111 112 * config/i386/athlon.md (athlon_sseld_amdfam10, 113 athlon_mmxld_amdfam10, athlon_ssest_amdfam10, 114 athlon_mmxssest_short_amdfam10): New define_insn_reservation. 115 1162007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 117 118 * config/i386/athlon.md (athlon_sseins_amdfam10): New 119 define_insn_reservation. 120 * config/i386/i386.md (sseins): Added sseins to define_attr type 121 and define_attr unit. 122 * config/i386/sse.md: Set type attribute to sseins for insertq 123 and insertqi. 124 1252007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 126 127 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10, 128 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10, 129 ssecomi_load_amdfam10, ssecomi_amdfam10, 130 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New 131 define_insn_reservation. 132 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10. 133 1342007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 135 136 * config/i386/athlon.md (cvtss2sd_load_amdfam10, 137 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10, 138 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10, 139 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10, 140 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10, 141 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New 142 define_insn_reservation. 143 144 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si, 145 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq, 146 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq, 147 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd, 148 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute. 149 1502007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 151 152 * config/i386/athlon.md (athlon_ssedivvector_amdfam10, 153 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10, 154 athlon_ssemulvector_load_amdfam10): New define_insn_reservation. 155 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul, 156 athlon_ssemul_load_k8): Added amdfam10. 157 1582007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 159 160 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro. 161 (x86_sse_unaligned_move_optimal): New variable. 162 163 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for 164 m_AMDFAM10. 165 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups 166 for unaligned vector SSE double/single precision loads for AMDFAM10. 167 1682007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 169 170 * config/i386/i386.h (TARGET_AMDFAM10): New macro. 171 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10. 172 Define TARGET_CPU_DEFAULT_amdfam10. 173 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10. 174 (processor_type): Add PROCESSOR_AMDFAM10. 175 176 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match 177 processor_type in config/i386/i386.h. 178 Enable imul peepholes for TARGET_AMDFAM10. 179 180 * config.gcc: Add support for --with-cpu option for amdfam10. 181 182 * config/i386/i386.c (amdfam10_cost): New variable. 183 (m_AMDFAM10): New macro. 184 (m_ATHLON_K8_AMDFAM10): New macro. 185 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen, 186 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop, 187 x86_promote_QImode, x86_integer_DFmode_moves, 188 x86_partial_reg_dependency, x86_memory_mismatch_stall, 189 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387, 190 x86_sse_partial_reg_dependency, x86_sse_typeless_stores, 191 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit, 192 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns): 193 Enable/disable for amdfam10. 194 (override_options): Add amdfam10_cost to processor_target_table. 195 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in 196 processor_alias_table. 197 (ix86_issue_rate): Add PROCESSOR_AMDFAM10. 198 (ix86_adjust_cost): Add code for amdfam10. 199 2002007-02-05 Harsha Jagasia <harsha.jagasia@amd.com> (r121625) 201 202 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm) 203 instruction set feature flag. Add new (-mpopcnt) flag for popcnt 204 instruction. Add new SSE4A (-msse4a) instruction set feature flag. 205 * config/i386/i386.h: Add builtin definition for SSE4A. 206 * config/i386/i386.md: Add support for ABM instructions 207 (popcnt and lzcnt). 208 * config/i386/sse.md: Add support for SSE4A instructions 209 (movntss, movntsd, extrq, insertq). 210 * config/i386/i386.c: Add support for ABM and SSE4A builtins. 211 Add -march=amdfam10 flag. 212 * config/i386/ammintrin.h: Add support for SSE4A intrinsics. 213 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt 214 and amdfam10. 215 * doc/extend.texi: Add documentation for SSE4A builtins. 216 2172007-01-24 Jakub Jelinek <jakub@redhat.com> (r121140) 218 219 * config/i386/i386.h (x86_cmpxchg16b): Remove const. 220 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b. 221 * config/i386/i386.c (x86_cmpxchg16b): Remove const. 222 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b 223 for CPUs that have PTA_CX16 set. 224 2252007-01-17 Eric Christopher <echristo@apple.com> (r120846) 226 227 * config.gcc: Support core2 processor. 228 2292006-12-02 H.J. Lu <hongjiu.lu@intel.com> (r119454 - partial) 230 231 PR target/30040 232 * config/i386/driver-i386.c (bit_SSSE3): New. 233 2342006-11-27 Uros Bizjak <ubizjak@gmail.com> (r119260) 235 236 * config/i386/i386.c (x86_ext_80387_constants): Add m_K8, m_CORE2 237 and m_GENERIC64. 238 2392006-11-18 Vladimir Makarov <vmakarov@redhat.com> (r118973) 240 241 * doc/invoke.texi (core2): Add item. 242 243 * config/i386/i386.h (TARGET_CORE2, TARGET_CPU_DEFAULT_core2): New 244 macros. 245 (TARGET_CPU_CPP_BUILTINS): Add code for core2. 246 (TARGET_CPU_DEFAULT_generic): Change value. 247 (TARGET_CPU_DEFAULT_NAMES): Add core2. 248 (processor_type): Add new constant PROCESSOR_CORE2. 249 250 * config/i386/i386.md (cpu): Add core2. 251 252 * config/i386/i386.c (core2_cost): New initialized variable. 253 (m_CORE2): New macro. 254 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen, 255 x86_deep_branch, x86_partial_reg_stall, x86_use_simode_fiop, 256 x86_use_cltd, x86_promote_QImode, x86_sub_esp_4, x86_sub_esp_8, 257 x86_add_esp_4, x86_add_esp_8, x86_integer_DFmode_moves, 258 x86_partial_reg_dependency, x86_memory_mismatch_stall, 259 x86_accumulate_outgoing_args, x86_prologue_using_move, 260 x86_epilogue_using_move, x86_arch_always_fancy_math_387, 261 x86_sse_partial_reg_dependency, x86_rep_movl_optimal, 262 x86_use_incdec, x86_four_jump_limit, x86_schedule, 263 x86_pad_returns): Add m_CORE2. 264 (override_options): Add entries for Core2. 265 (ix86_issue_rate): Add case for Core2. 266 2672006-10-27 Vladimir Makarov <vmakarov@redhat.com> (r118090) 268 269 * config/i386/i386.h (TARGET_GEODE): 270 (TARGET_CPU_CPP_BUILTINS): Add code for geode. 271 (TARGET_CPU_DEFAULT_geode): New macro. 272 (TARGET_CPU_DEFAULT_k6, TARGET_CPU_DEFAULT_k6_2, 273 TARGET_CPU_DEFAULT_k6_3, TARGET_CPU_DEFAULT_athlon, 274 TARGET_CPU_DEFAULT_athlon_sse, TARGET_CPU_DEFAULT_k8, 275 TARGET_CPU_DEFAULT_pentium_m, TARGET_CPU_DEFAULT_prescott, 276 TARGET_CPU_DEFAULT_nocona, TARGET_CPU_DEFAULT_generic): Increase 277 the macro values. 278 (TARGET_CPU_DEFAULT_NAMES): Add geode. 279 (processor_type): Add PROCESSOR_GEODE. 280 281 * config/i386/i386.md: Include geode.md. 282 (cpu): Add geode. 283 284 * config/i386/i386.c (geode_cost): New initialized global 285 variable. 286 (m_GEODE, m_K6_GEODE): New macros. 287 (x86_use_leave, x86_push_memory, x86_deep_branch, x86_use_sahf, 288 x86_use_himode_fiop, x86_promote_QImode, x86_add_esp_4, 289 x86_add_esp_8, x86_rep_movl_optimal, x86_ext_80387_constants, 290 x86_schedule): Use m_K6_GEODE instead of m_K6. 291 (x86_movx, x86_cmove): Set up m_GEODE. 292 (x86_integer_DFmode_moves): Clear m_GEODE. 293 (processor_target_table): Add entry for geode. 294 (processor_alias_table): Ditto. 295 296 * config/i386/geode.md: New file. 297 298 * doc/invoke.texi: Add entry about geode processor. 299 3002006-10-24 Richard Guenther <rguenther@suse.de> (r118001) 301 302 PR middle-end/28796 303 * builtins.c (fold_builtin_classify): Use HONOR_INFINITIES 304 and HONOR_NANS instead of MODE_HAS_INFINITIES and MODE_HAS_NANS 305 for deciding optimizations in consistency with fold-const.c 306 (fold_builtin_unordered_cmp): Likewise. 307 3082006-10-22 H.J. Lu <hongjiu.lu@intel.com> (r117958) 309 310 * config.gcc (i[34567]86-*-*): Add tmmintrin.h to extra_headers. 311 (x86_64-*-*): Likewise. 312 313 * config/i386/i386.c (pta_flags): Add PTA_SSSE3. 314 (override_options): Check SSSE3. 315 (ix86_builtins): Add IX86_BUILTIN_PHADDW, IX86_BUILTIN_PHADDD, 316 IX86_BUILTIN_PHADDSW, IX86_BUILTIN_PHSUBW, IX86_BUILTIN_PHSUBD, 317 IX86_BUILTIN_PHSUBSW, IX86_BUILTIN_PMADDUBSW, 318 IX86_BUILTIN_PMULHRSW, IX86_BUILTIN_PSHUFB, 319 IX86_BUILTIN_PSIGNB, IX86_BUILTIN_PSIGNW, IX86_BUILTIN_PSIGND, 320 IX86_BUILTIN_PALIGNR, IX86_BUILTIN_PABSB, IX86_BUILTIN_PABSW, 321 IX86_BUILTIN_PABSD, IX86_BUILTIN_PHADDW128, 322 IX86_BUILTIN_PHADDD128, IX86_BUILTIN_PHADDSW128, 323 IX86_BUILTIN_PHSUBW128, IX86_BUILTIN_PHSUBD128, 324 IX86_BUILTIN_PHSUBSW128, IX86_BUILTIN_PMADDUBSW128, 325 IX86_BUILTIN_PMULHRSW128, IX86_BUILTIN_PSHUFB128, 326 IX86_BUILTIN_PSIGNB128, IX86_BUILTIN_PSIGNW128, 327 IX86_BUILTIN_PSIGND128, IX86_BUILTIN_PALIGNR128, 328 IX86_BUILTIN_PABSB128, IX86_BUILTIN_PABSW128 and 329 IX86_BUILTIN_PABSD128. 330 (bdesc_2arg): Add SSSE3. 331 (bdesc_1arg): Likewise. 332 (ix86_init_mmx_sse_builtins): Support SSSE3. 333 (ix86_expand_builtin): Likewise. 334 * config/i386/i386.h (TARGET_CPU_CPP_BUILTINS): Likewise. 335 336 * config/i386/i386.md (UNSPEC_PSHUFB): New. 337 (UNSPEC_PSIGN): Likewise. 338 (UNSPEC_PALIGNR): Likewise. 339 Include mmx.md before sse.md. 340 341 * config/i386/i386.opt: Add -mssse3. 342 343 * config/i386/sse.md (ssse3_phaddwv8hi3): New pattern for SSSE3. 344 (ssse3_phaddwv4hi3): Likewise. 345 (ssse3_phadddv4si3): Likewise. 346 (ssse3_phadddv2si3): Likewise. 347 (ssse3_phaddswv8hi3): Likewise. 348 (ssse3_phaddswv4hi3): Likewise. 349 (ssse3_phsubwv8hi3): Likewise. 350 (ssse3_phsubwv4hi3): Likewise. 351 (ssse3_phsubdv4si3): Likewise. 352 (ssse3_phsubdv2si3): Likewise. 353 (ssse3_phsubswv8hi3): Likewise. 354 (ssse3_phsubswv4hi3): Likewise. 355 (ssse3_pmaddubswv8hi3): Likewise. 356 (ssse3_pmaddubswv4hi3): Likewise. 357 (ssse3_pmulhrswv8hi3): Likewise. 358 (ssse3_pmulhrswv4hi3): Likewise. 359 (ssse3_pshufbv16qi3): Likewise. 360 (ssse3_pshufbv8qi3): Likewise. 361 (ssse3_psign<mode>3): Likewise. 362 (ssse3_psign<mode>3): Likewise. 363 (ssse3_palignrti): Likewise. 364 (ssse3_palignrdi): Likewise. 365 (abs<mode>2): Likewise. 366 (abs<mode>2): Likewise. 367 368 * config/i386/tmmintrin.h: New file. 369 370 * doc/extend.texi: Document SSSE3 built-in functions. 371 372 * doc/invoke.texi: Document -mssse3/-mno-ssse3 switches. 373 3742006-10-22 H.J. Lu <hongjiu.lu@intel.com> (r117959) 375 376 * config/i386/tmmintrin.h: Remove the duplicated content. 377 3782006-10-21 Richard Guenther <rguenther@suse.de> (r117932) 379 380 PR tree-optimization/3511 381 * tree-ssa-pre.c (phi_translate): Fold CALL_EXPRs that 382 got new invariant arguments during PHI translation. 383 3842006-10-21 Richard Guenther <rguenther@suse.de> (r117929) 385 386 * builtins.c (fold_builtin_classify): Fix typo. 387 388