| author | |
| committer | |
| log | 7cd8321b82bb142d6bc03b918e883ad8c4bcdffb |
| tree | e8fa748a606d7150fe4571c1b4777905b85cd74d |
| parent | 107e01eb8e3260fe76b2a65f80c8fe707d6d67c0 |
| signature |
18 files changed, 2754 insertions(+), 442 deletions(-)
lib/compiler/aro/aro/Compilation.zig-1| ... | ... | @@ -560,7 +560,6 @@ fn generateSystemDefines(comp: *Compilation, w: *Io.Writer) !void { |
| 560 | 560 | .{ .amx_fp8, "__AMX_FP8__" }, |
| 561 | 561 | .{ .amx_movrs, "__AMX_MOVRS__" }, |
| 562 | 562 | .{ .amx_avx512, "__AMX_AVX512__" }, |
| 563 | .{ .amx_tf32, "__AMX_TF32__" }, | |
| 564 | 563 | .{ .cmpccxadd, "__CMPCCXADD__" }, |
| 565 | 564 | .{ .raoint, "__RAOINT__" }, |
| 566 | 565 | .{ .avxifma, "__AVXIFMA__" }, |
lib/std/Target.zig+1-1| ... | ... | @@ -1238,7 +1238,7 @@ pub const Cpu = struct { |
| 1238 | 1238 | pub const Set = struct { |
| 1239 | 1239 | ints: [usize_count]usize, |
| 1240 | 1240 | |
| 1241 | pub const needed_bit_count = 347; | |
| 1241 | pub const needed_bit_count = 378; | |
| 1242 | 1242 | pub const byte_count = @divCeil(needed_bit_count, 8); |
| 1243 | 1243 | pub const usize_count = (byte_count + (@sizeOf(usize) - 1)) / @sizeOf(usize); |
| 1244 | 1244 | pub const Index = std.math.Log2Int(@Int(.unsigned, usize_count * @bitSizeOf(usize))); |
lib/std/Target/aarch64.zig+191-33| ... | ... | @@ -10,6 +10,7 @@ pub const Feature = enum { |
| 10 | 10 | aes, |
| 11 | 11 | aggressive_fma, |
| 12 | 12 | aggressive_interleaving, |
| 13 | align_cmp_csel_pairs, | |
| 13 | 14 | alternate_sextload_cvt_f32_pattern, |
| 14 | 15 | altnzcv, |
| 15 | 16 | alu_lsl_fast, |
| ... | ... | @@ -38,7 +39,6 @@ pub const Feature = enum { |
| 38 | 39 | ccpp, |
| 39 | 40 | chk, |
| 40 | 41 | clrbhb, |
| 41 | cmh, | |
| 42 | 42 | cmp_bcc_fusion, |
| 43 | 43 | cmpbr, |
| 44 | 44 | complxnum, |
| ... | ... | @@ -48,6 +48,7 @@ pub const Feature = enum { |
| 48 | 48 | crypto, |
| 49 | 49 | cssc, |
| 50 | 50 | d128, |
| 51 | disable_distinct_predicate_dst_reg, | |
| 51 | 52 | disable_fast_inc_vl, |
| 52 | 53 | disable_latency_sched_heuristic, |
| 53 | 54 | disable_ldp, |
| ... | ... | @@ -71,8 +72,10 @@ pub const Feature = enum { |
| 71 | 72 | f8f16mm, |
| 72 | 73 | f8f32mm, |
| 73 | 74 | faminmax, |
| 75 | fast_ld1_single, | |
| 74 | 76 | fgt, |
| 75 | 77 | fix_cortex_a53_835769, |
| 78 | fixed_load_latency_4, | |
| 76 | 79 | flagm, |
| 77 | 80 | fmv, |
| 78 | 81 | force_32bit_jump_tables, |
| ... | ... | @@ -95,14 +98,16 @@ pub const Feature = enum { |
| 95 | 98 | fuse_crypto_eor, |
| 96 | 99 | fuse_csel, |
| 97 | 100 | fuse_cset, |
| 101 | fuse_fcsel, | |
| 98 | 102 | fuse_literals, |
| 99 | 103 | gcie, |
| 100 | 104 | gcs, |
| 101 | 105 | harden_sls_blr, |
| 102 | 106 | harden_sls_nocomdat, |
| 103 | 107 | harden_sls_retbr, |
| 108 | has_limited_64bit_vector_mul_bandwidth, | |
| 104 | 109 | hbc, |
| 105 | hcx, | |
| 110 | hinte, | |
| 106 | 111 | i8mm, |
| 107 | 112 | ite, |
| 108 | 113 | jsconv, |
| ... | ... | @@ -121,12 +126,13 @@ pub const Feature = enum { |
| 121 | 126 | mops, |
| 122 | 127 | mops_go, |
| 123 | 128 | mpam, |
| 124 | mpamv2, | |
| 125 | 129 | mte, |
| 126 | 130 | mtetc, |
| 127 | 131 | neon, |
| 128 | 132 | nmi, |
| 129 | 133 | no_bti_at_return_twice, |
| 134 | no_lfi_loads, | |
| 135 | no_lfi_stores, | |
| 130 | 136 | no_neg_immediates, |
| 131 | 137 | no_sve_fp_ld1r, |
| 132 | 138 | no_zcz_fpr64, |
| ... | ... | @@ -138,7 +144,6 @@ pub const Feature = enum { |
| 138 | 144 | pan_rwv, |
| 139 | 145 | pauth, |
| 140 | 146 | pauth_lr, |
| 141 | pcdphint, | |
| 142 | 147 | perfmon, |
| 143 | 148 | poe2, |
| 144 | 149 | pops, |
| ... | ... | @@ -328,6 +333,11 @@ pub const all_features = blk: { |
| 328 | 333 | .description = "Make use of aggressive interleaving during vectorization", |
| 329 | 334 | .dependencies = featureSet(&[_]Feature{}), |
| 330 | 335 | }; |
| 336 | result[@backingInt(Feature.align_cmp_csel_pairs)] = .{ | |
| 337 | .llvm_name = "align-cmp-csel-pairs", | |
| 338 | .description = "Align certain CMP/FCMP and CSEL/FCSEL instruction pairs", | |
| 339 | .dependencies = featureSet(&[_]Feature{}), | |
| 340 | }; | |
| 331 | 341 | result[@backingInt(Feature.alternate_sextload_cvt_f32_pattern)] = .{ |
| 332 | 342 | .llvm_name = "alternate-sextload-cvt-f32-pattern", |
| 333 | 343 | .description = "Use alternative pattern for sextload convert to f32", |
| ... | ... | @@ -476,11 +486,6 @@ pub const all_features = blk: { |
| 476 | 486 | .description = "Enable Clear BHB instruction", |
| 477 | 487 | .dependencies = featureSet(&[_]Feature{}), |
| 478 | 488 | }; |
| 479 | result[@backingInt(Feature.cmh)] = .{ | |
| 480 | .llvm_name = "cmh", | |
| 481 | .description = "Enable Armv9.7-A Contention Management Hints", | |
| 482 | .dependencies = featureSet(&[_]Feature{}), | |
| 483 | }; | |
| 484 | 489 | result[@backingInt(Feature.cmp_bcc_fusion)] = .{ |
| 485 | 490 | .llvm_name = "cmp-bcc-fusion", |
| 486 | 491 | .description = "CPU fuses cmp+bcc operations", |
| ... | ... | @@ -533,6 +538,11 @@ pub const all_features = blk: { |
| 533 | 538 | .lse128, |
| 534 | 539 | }), |
| 535 | 540 | }; |
| 541 | result[@backingInt(Feature.disable_distinct_predicate_dst_reg)] = .{ | |
| 542 | .llvm_name = "disable-distinct-predicate-dst-reg", | |
| 543 | .description = "Disabling selecting a distinct predicate register for the destination operand of instructions that take a governing predicate", | |
| 544 | .dependencies = featureSet(&[_]Feature{}), | |
| 545 | }; | |
| 536 | 546 | result[@backingInt(Feature.disable_fast_inc_vl)] = .{ |
| 537 | 547 | .llvm_name = "disable-fast-inc-vl", |
| 538 | 548 | .description = "Do not prefer INC/DEC, ALL, { 1, 2, 4 } over ADDVL", |
| ... | ... | @@ -545,7 +555,7 @@ pub const all_features = blk: { |
| 545 | 555 | }; |
| 546 | 556 | result[@backingInt(Feature.disable_ldp)] = .{ |
| 547 | 557 | .llvm_name = "disable-ldp", |
| 548 | .description = "Do not emit ldp", | |
| 558 | .description = "Prefer not to emit ldp", | |
| 549 | 559 | .dependencies = featureSet(&[_]Feature{}), |
| 550 | 560 | }; |
| 551 | 561 | result[@backingInt(Feature.disable_maximize_scalable_bandwidth)] = .{ |
| ... | ... | @@ -555,7 +565,7 @@ pub const all_features = blk: { |
| 555 | 565 | }; |
| 556 | 566 | result[@backingInt(Feature.disable_stp)] = .{ |
| 557 | 567 | .llvm_name = "disable-stp", |
| 558 | .description = "Do not emit stp", | |
| 568 | .description = "Prefer not to emit stp", | |
| 559 | 569 | .dependencies = featureSet(&[_]Feature{}), |
| 560 | 570 | }; |
| 561 | 571 | result[@backingInt(Feature.disable_unpredicated_ld_st_lower)] = .{ |
| ... | ... | @@ -671,6 +681,11 @@ pub const all_features = blk: { |
| 671 | 681 | .neon, |
| 672 | 682 | }), |
| 673 | 683 | }; |
| 684 | result[@backingInt(Feature.fast_ld1_single)] = .{ | |
| 685 | .llvm_name = "fast-ld1-single", | |
| 686 | .description = "Single-element LD1 to vector lane has same performance as regular load", | |
| 687 | .dependencies = featureSet(&[_]Feature{}), | |
| 688 | }; | |
| 674 | 689 | result[@backingInt(Feature.fgt)] = .{ |
| 675 | 690 | .llvm_name = "fgt", |
| 676 | 691 | .description = "Enable fine grained virtualization traps extension", |
| ... | ... | @@ -681,6 +696,11 @@ pub const all_features = blk: { |
| 681 | 696 | .description = "Mitigate Cortex-A53 Erratum 835769", |
| 682 | 697 | .dependencies = featureSet(&[_]Feature{}), |
| 683 | 698 | }; |
| 699 | result[@backingInt(Feature.fixed_load_latency_4)] = .{ | |
| 700 | .llvm_name = "fixed-load-latency-4", | |
| 701 | .description = "Use a fixed latency of 4 for loads instead of querying the scheduling model", | |
| 702 | .dependencies = featureSet(&[_]Feature{}), | |
| 703 | }; | |
| 684 | 704 | result[@backingInt(Feature.flagm)] = .{ |
| 685 | 705 | .llvm_name = "flagm", |
| 686 | 706 | .description = "Enable Armv8.4-A Flag Manipulation instructions", |
| ... | ... | @@ -813,6 +833,11 @@ pub const all_features = blk: { |
| 813 | 833 | .description = "CPU can fuse CMP and CSET operations", |
| 814 | 834 | .dependencies = featureSet(&[_]Feature{}), |
| 815 | 835 | }; |
| 836 | result[@backingInt(Feature.fuse_fcsel)] = .{ | |
| 837 | .llvm_name = "fuse-fcsel", | |
| 838 | .description = "CPU can fuse FCMP and FCSEL operations", | |
| 839 | .dependencies = featureSet(&[_]Feature{}), | |
| 840 | }; | |
| 816 | 841 | result[@backingInt(Feature.fuse_literals)] = .{ |
| 817 | 842 | .llvm_name = "fuse-literals", |
| 818 | 843 | .description = "CPU fuses literal generation operations", |
| ... | ... | @@ -845,14 +870,19 @@ pub const all_features = blk: { |
| 845 | 870 | .description = "Harden against straight line speculation across RET and BR instructions", |
| 846 | 871 | .dependencies = featureSet(&[_]Feature{}), |
| 847 | 872 | }; |
| 873 | result[@backingInt(Feature.has_limited_64bit_vector_mul_bandwidth)] = .{ | |
| 874 | .llvm_name = "has-limited-64bit-vector-mul-bandwidth", | |
| 875 | .description = "Has limited 64bit vector multiply bandwidth compared to scalar multiply", | |
| 876 | .dependencies = featureSet(&[_]Feature{}), | |
| 877 | }; | |
| 848 | 878 | result[@backingInt(Feature.hbc)] = .{ |
| 849 | 879 | .llvm_name = "hbc", |
| 850 | 880 | .description = "Enable Armv8.8-A Hinted Conditional Branches Extension", |
| 851 | 881 | .dependencies = featureSet(&[_]Feature{}), |
| 852 | 882 | }; |
| 853 | result[@backingInt(Feature.hcx)] = .{ | |
| 854 | .llvm_name = "hcx", | |
| 855 | .description = "Enable Armv8.7-A HCRX_EL2 system register", | |
| 883 | result[@backingInt(Feature.hinte)] = .{ | |
| 884 | .llvm_name = "hinte", | |
| 885 | .description = "Enable extended A64 hint instruction space", | |
| 856 | 886 | .dependencies = featureSet(&[_]Feature{}), |
| 857 | 887 | }; |
| 858 | 888 | result[@backingInt(Feature.i8mm)] = .{ |
| ... | ... | @@ -962,11 +992,6 @@ pub const all_features = blk: { |
| 962 | 992 | .description = "Enable Armv8.4-A Memory system Partitioning and Monitoring extension", |
| 963 | 993 | .dependencies = featureSet(&[_]Feature{}), |
| 964 | 994 | }; |
| 965 | result[@backingInt(Feature.mpamv2)] = .{ | |
| 966 | .llvm_name = "mpamv2", | |
| 967 | .description = "Enable Armv9.7-A MPAMv2 Lookaside Buffer Invalidate instructions", | |
| 968 | .dependencies = featureSet(&[_]Feature{}), | |
| 969 | }; | |
| 970 | 995 | result[@backingInt(Feature.mte)] = .{ |
| 971 | 996 | .llvm_name = "mte", |
| 972 | 997 | .description = "Enable Memory Tagging Extension", |
| ... | ... | @@ -996,6 +1021,16 @@ pub const all_features = blk: { |
| 996 | 1021 | .description = "Don't place a BTI instruction after a return-twice", |
| 997 | 1022 | .dependencies = featureSet(&[_]Feature{}), |
| 998 | 1023 | }; |
| 1024 | result[@backingInt(Feature.no_lfi_loads)] = .{ | |
| 1025 | .llvm_name = "no-lfi-loads", | |
| 1026 | .description = "Disable LFI sandboxing for load instructions", | |
| 1027 | .dependencies = featureSet(&[_]Feature{}), | |
| 1028 | }; | |
| 1029 | result[@backingInt(Feature.no_lfi_stores)] = .{ | |
| 1030 | .llvm_name = "no-lfi-stores", | |
| 1031 | .description = "Disable LFI sandboxing for store instructions", | |
| 1032 | .dependencies = featureSet(&[_]Feature{}), | |
| 1033 | }; | |
| 999 | 1034 | result[@backingInt(Feature.no_neg_immediates)] = .{ |
| 1000 | 1035 | .llvm_name = "no-neg-immediates", |
| 1001 | 1036 | .description = "Convert immediates and instructions to their negated or complemented equivalent when the immediate does not fit in the encoding.", |
| ... | ... | @@ -1063,11 +1098,6 @@ pub const all_features = blk: { |
| 1063 | 1098 | .description = "Enable Armv9.5-A PAC enhancements", |
| 1064 | 1099 | .dependencies = featureSet(&[_]Feature{}), |
| 1065 | 1100 | }; |
| 1066 | result[@backingInt(Feature.pcdphint)] = .{ | |
| 1067 | .llvm_name = "pcdphint", | |
| 1068 | .description = "Enable Armv9.6-A Producer Consumer Data Placement hints", | |
| 1069 | .dependencies = featureSet(&[_]Feature{}), | |
| 1070 | }; | |
| 1071 | 1101 | result[@backingInt(Feature.perfmon)] = .{ |
| 1072 | 1102 | .llvm_name = "perfmon", |
| 1073 | 1103 | .description = "Enable Armv8.0-A PMUv3 Performance Monitors extension", |
| ... | ... | @@ -1813,7 +1843,6 @@ pub const all_features = blk: { |
| 1813 | 1843 | .llvm_name = "v8.7a", |
| 1814 | 1844 | .description = "Support ARM v8.7a architecture", |
| 1815 | 1845 | .dependencies = featureSet(&[_]Feature{ |
| 1816 | .hcx, | |
| 1817 | 1846 | .spe_eef, |
| 1818 | 1847 | .v8_6a, |
| 1819 | 1848 | .wfxt, |
| ... | ... | @@ -1931,6 +1960,7 @@ pub const all_features = blk: { |
| 1931 | 1960 | .description = "Support ARM v9.6a architecture", |
| 1932 | 1961 | .dependencies = featureSet(&[_]Feature{ |
| 1933 | 1962 | .cmpbr, |
| 1963 | .hinte, | |
| 1934 | 1964 | .lsui, |
| 1935 | 1965 | .occmo, |
| 1936 | 1966 | .v9_5a, |
| ... | ... | @@ -2265,12 +2295,14 @@ pub const cpu = struct { |
| 2265 | 2295 | .features = featureSet(&[_]Feature{ |
| 2266 | 2296 | .aes, |
| 2267 | 2297 | .aggressive_fma, |
| 2298 | .align_cmp_csel_pairs, | |
| 2268 | 2299 | .alternate_sextload_cvt_f32_pattern, |
| 2269 | 2300 | .altnzcv, |
| 2270 | 2301 | .arith_bcc_fusion, |
| 2271 | 2302 | .arith_cbz_fusion, |
| 2272 | 2303 | .ccdp, |
| 2273 | 2304 | .disable_latency_sched_heuristic, |
| 2305 | .fast_ld1_single, | |
| 2274 | 2306 | .fp16fml, |
| 2275 | 2307 | .fptoint, |
| 2276 | 2308 | .fuse_address, |
| ... | ... | @@ -2301,10 +2333,12 @@ pub const cpu = struct { |
| 2301 | 2333 | .llvm_name = "apple-a15", |
| 2302 | 2334 | .features = featureSet(&[_]Feature{ |
| 2303 | 2335 | .aes, |
| 2336 | .align_cmp_csel_pairs, | |
| 2304 | 2337 | .alternate_sextload_cvt_f32_pattern, |
| 2305 | 2338 | .arith_bcc_fusion, |
| 2306 | 2339 | .arith_cbz_fusion, |
| 2307 | 2340 | .disable_latency_sched_heuristic, |
| 2341 | .fast_ld1_single, | |
| 2308 | 2342 | .fp16fml, |
| 2309 | 2343 | .fpac, |
| 2310 | 2344 | .fuse_address, |
| ... | ... | @@ -2332,10 +2366,12 @@ pub const cpu = struct { |
| 2332 | 2366 | .llvm_name = "apple-a16", |
| 2333 | 2367 | .features = featureSet(&[_]Feature{ |
| 2334 | 2368 | .aes, |
| 2369 | .align_cmp_csel_pairs, | |
| 2335 | 2370 | .alternate_sextload_cvt_f32_pattern, |
| 2336 | 2371 | .arith_bcc_fusion, |
| 2337 | 2372 | .arith_cbz_fusion, |
| 2338 | 2373 | .disable_latency_sched_heuristic, |
| 2374 | .fast_ld1_single, | |
| 2339 | 2375 | .fp16fml, |
| 2340 | 2376 | .fpac, |
| 2341 | 2377 | .fuse_address, |
| ... | ... | @@ -2344,8 +2380,8 @@ pub const cpu = struct { |
| 2344 | 2380 | .fuse_arith_logic, |
| 2345 | 2381 | .fuse_crypto_eor, |
| 2346 | 2382 | .fuse_csel, |
| 2383 | .fuse_fcsel, | |
| 2347 | 2384 | .fuse_literals, |
| 2348 | .hcx, | |
| 2349 | 2385 | .max_interleave_factor_4, |
| 2350 | 2386 | .no_zcz_fpr64, |
| 2351 | 2387 | .perfmon, |
| ... | ... | @@ -2364,10 +2400,12 @@ pub const cpu = struct { |
| 2364 | 2400 | .llvm_name = "apple-a17", |
| 2365 | 2401 | .features = featureSet(&[_]Feature{ |
| 2366 | 2402 | .aes, |
| 2403 | .align_cmp_csel_pairs, | |
| 2367 | 2404 | .alternate_sextload_cvt_f32_pattern, |
| 2368 | 2405 | .arith_bcc_fusion, |
| 2369 | 2406 | .arith_cbz_fusion, |
| 2370 | 2407 | .disable_latency_sched_heuristic, |
| 2408 | .fast_ld1_single, | |
| 2371 | 2409 | .fp16fml, |
| 2372 | 2410 | .fpac, |
| 2373 | 2411 | .fuse_address, |
| ... | ... | @@ -2376,8 +2414,8 @@ pub const cpu = struct { |
| 2376 | 2414 | .fuse_arith_logic, |
| 2377 | 2415 | .fuse_crypto_eor, |
| 2378 | 2416 | .fuse_csel, |
| 2417 | .fuse_fcsel, | |
| 2379 | 2418 | .fuse_literals, |
| 2380 | .hcx, | |
| 2381 | 2419 | .max_interleave_factor_4, |
| 2382 | 2420 | .no_zcz_fpr64, |
| 2383 | 2421 | .perfmon, |
| ... | ... | @@ -2396,10 +2434,12 @@ pub const cpu = struct { |
| 2396 | 2434 | .llvm_name = "apple-a18", |
| 2397 | 2435 | .features = featureSet(&[_]Feature{ |
| 2398 | 2436 | .aes, |
| 2437 | .align_cmp_csel_pairs, | |
| 2399 | 2438 | .alternate_sextload_cvt_f32_pattern, |
| 2400 | 2439 | .arith_bcc_fusion, |
| 2401 | 2440 | .arith_cbz_fusion, |
| 2402 | 2441 | .disable_latency_sched_heuristic, |
| 2442 | .fast_ld1_single, | |
| 2403 | 2443 | .fp16fml, |
| 2404 | 2444 | .fpac, |
| 2405 | 2445 | .fuse_address, |
| ... | ... | @@ -2408,6 +2448,7 @@ pub const cpu = struct { |
| 2408 | 2448 | .fuse_arith_logic, |
| 2409 | 2449 | .fuse_crypto_eor, |
| 2410 | 2450 | .fuse_csel, |
| 2451 | .fuse_fcsel, | |
| 2411 | 2452 | .fuse_literals, |
| 2412 | 2453 | .max_interleave_factor_4, |
| 2413 | 2454 | .no_zcz_fpr64, |
| ... | ... | @@ -2429,11 +2470,13 @@ pub const cpu = struct { |
| 2429 | 2470 | .llvm_name = "apple-a19", |
| 2430 | 2471 | .features = featureSet(&[_]Feature{ |
| 2431 | 2472 | .aes, |
| 2473 | .align_cmp_csel_pairs, | |
| 2432 | 2474 | .alternate_sextload_cvt_f32_pattern, |
| 2433 | 2475 | .arith_bcc_fusion, |
| 2434 | 2476 | .arith_cbz_fusion, |
| 2435 | 2477 | .cssc, |
| 2436 | 2478 | .disable_latency_sched_heuristic, |
| 2479 | .fast_ld1_single, | |
| 2437 | 2480 | .fp16fml, |
| 2438 | 2481 | .fpac, |
| 2439 | 2482 | .fuse_address, |
| ... | ... | @@ -2442,6 +2485,7 @@ pub const cpu = struct { |
| 2442 | 2485 | .fuse_arith_logic, |
| 2443 | 2486 | .fuse_crypto_eor, |
| 2444 | 2487 | .fuse_csel, |
| 2488 | .fuse_fcsel, | |
| 2445 | 2489 | .fuse_literals, |
| 2446 | 2490 | .hbc, |
| 2447 | 2491 | .max_interleave_factor_4, |
| ... | ... | @@ -2541,12 +2585,14 @@ pub const cpu = struct { |
| 2541 | 2585 | .features = featureSet(&[_]Feature{ |
| 2542 | 2586 | .aes, |
| 2543 | 2587 | .aggressive_fma, |
| 2588 | .align_cmp_csel_pairs, | |
| 2544 | 2589 | .alternate_sextload_cvt_f32_pattern, |
| 2545 | 2590 | .altnzcv, |
| 2546 | 2591 | .arith_bcc_fusion, |
| 2547 | 2592 | .arith_cbz_fusion, |
| 2548 | 2593 | .ccdp, |
| 2549 | 2594 | .disable_latency_sched_heuristic, |
| 2595 | .fast_ld1_single, | |
| 2550 | 2596 | .fp16fml, |
| 2551 | 2597 | .fptoint, |
| 2552 | 2598 | .fuse_address, |
| ... | ... | @@ -2577,10 +2623,12 @@ pub const cpu = struct { |
| 2577 | 2623 | .llvm_name = "apple-m2", |
| 2578 | 2624 | .features = featureSet(&[_]Feature{ |
| 2579 | 2625 | .aes, |
| 2626 | .align_cmp_csel_pairs, | |
| 2580 | 2627 | .alternate_sextload_cvt_f32_pattern, |
| 2581 | 2628 | .arith_bcc_fusion, |
| 2582 | 2629 | .arith_cbz_fusion, |
| 2583 | 2630 | .disable_latency_sched_heuristic, |
| 2631 | .fast_ld1_single, | |
| 2584 | 2632 | .fp16fml, |
| 2585 | 2633 | .fpac, |
| 2586 | 2634 | .fuse_address, |
| ... | ... | @@ -2608,10 +2656,12 @@ pub const cpu = struct { |
| 2608 | 2656 | .llvm_name = "apple-m3", |
| 2609 | 2657 | .features = featureSet(&[_]Feature{ |
| 2610 | 2658 | .aes, |
| 2659 | .align_cmp_csel_pairs, | |
| 2611 | 2660 | .alternate_sextload_cvt_f32_pattern, |
| 2612 | 2661 | .arith_bcc_fusion, |
| 2613 | 2662 | .arith_cbz_fusion, |
| 2614 | 2663 | .disable_latency_sched_heuristic, |
| 2664 | .fast_ld1_single, | |
| 2615 | 2665 | .fp16fml, |
| 2616 | 2666 | .fpac, |
| 2617 | 2667 | .fuse_address, |
| ... | ... | @@ -2620,8 +2670,8 @@ pub const cpu = struct { |
| 2620 | 2670 | .fuse_arith_logic, |
| 2621 | 2671 | .fuse_crypto_eor, |
| 2622 | 2672 | .fuse_csel, |
| 2673 | .fuse_fcsel, | |
| 2623 | 2674 | .fuse_literals, |
| 2624 | .hcx, | |
| 2625 | 2675 | .max_interleave_factor_4, |
| 2626 | 2676 | .no_zcz_fpr64, |
| 2627 | 2677 | .perfmon, |
| ... | ... | @@ -2640,10 +2690,12 @@ pub const cpu = struct { |
| 2640 | 2690 | .llvm_name = "apple-m4", |
| 2641 | 2691 | .features = featureSet(&[_]Feature{ |
| 2642 | 2692 | .aes, |
| 2693 | .align_cmp_csel_pairs, | |
| 2643 | 2694 | .alternate_sextload_cvt_f32_pattern, |
| 2644 | 2695 | .arith_bcc_fusion, |
| 2645 | 2696 | .arith_cbz_fusion, |
| 2646 | 2697 | .disable_latency_sched_heuristic, |
| 2698 | .fast_ld1_single, | |
| 2647 | 2699 | .fp16fml, |
| 2648 | 2700 | .fpac, |
| 2649 | 2701 | .fuse_address, |
| ... | ... | @@ -2652,6 +2704,7 @@ pub const cpu = struct { |
| 2652 | 2704 | .fuse_arith_logic, |
| 2653 | 2705 | .fuse_crypto_eor, |
| 2654 | 2706 | .fuse_csel, |
| 2707 | .fuse_fcsel, | |
| 2655 | 2708 | .fuse_literals, |
| 2656 | 2709 | .max_interleave_factor_4, |
| 2657 | 2710 | .no_zcz_fpr64, |
| ... | ... | @@ -2673,11 +2726,13 @@ pub const cpu = struct { |
| 2673 | 2726 | .llvm_name = "apple-m5", |
| 2674 | 2727 | .features = featureSet(&[_]Feature{ |
| 2675 | 2728 | .aes, |
| 2729 | .align_cmp_csel_pairs, | |
| 2676 | 2730 | .alternate_sextload_cvt_f32_pattern, |
| 2677 | 2731 | .arith_bcc_fusion, |
| 2678 | 2732 | .arith_cbz_fusion, |
| 2679 | 2733 | .cssc, |
| 2680 | 2734 | .disable_latency_sched_heuristic, |
| 2735 | .fast_ld1_single, | |
| 2681 | 2736 | .fp16fml, |
| 2682 | 2737 | .fpac, |
| 2683 | 2738 | .fuse_address, |
| ... | ... | @@ -2686,6 +2741,7 @@ pub const cpu = struct { |
| 2686 | 2741 | .fuse_arith_logic, |
| 2687 | 2742 | .fuse_crypto_eor, |
| 2688 | 2743 | .fuse_csel, |
| 2744 | .fuse_fcsel, | |
| 2689 | 2745 | .fuse_literals, |
| 2690 | 2746 | .hbc, |
| 2691 | 2747 | .max_interleave_factor_4, |
| ... | ... | @@ -2712,10 +2768,12 @@ pub const cpu = struct { |
| 2712 | 2768 | .llvm_name = "apple-s10", |
| 2713 | 2769 | .features = featureSet(&[_]Feature{ |
| 2714 | 2770 | .aes, |
| 2771 | .align_cmp_csel_pairs, | |
| 2715 | 2772 | .alternate_sextload_cvt_f32_pattern, |
| 2716 | 2773 | .arith_bcc_fusion, |
| 2717 | 2774 | .arith_cbz_fusion, |
| 2718 | 2775 | .disable_latency_sched_heuristic, |
| 2776 | .fast_ld1_single, | |
| 2719 | 2777 | .fp16fml, |
| 2720 | 2778 | .fpac, |
| 2721 | 2779 | .fuse_address, |
| ... | ... | @@ -2724,8 +2782,8 @@ pub const cpu = struct { |
| 2724 | 2782 | .fuse_arith_logic, |
| 2725 | 2783 | .fuse_crypto_eor, |
| 2726 | 2784 | .fuse_csel, |
| 2785 | .fuse_fcsel, | |
| 2727 | 2786 | .fuse_literals, |
| 2728 | .hcx, | |
| 2729 | 2787 | .max_interleave_factor_4, |
| 2730 | 2788 | .no_zcz_fpr64, |
| 2731 | 2789 | .perfmon, |
| ... | ... | @@ -2864,10 +2922,12 @@ pub const cpu = struct { |
| 2864 | 2922 | .llvm_name = "apple-s9", |
| 2865 | 2923 | .features = featureSet(&[_]Feature{ |
| 2866 | 2924 | .aes, |
| 2925 | .align_cmp_csel_pairs, | |
| 2867 | 2926 | .alternate_sextload_cvt_f32_pattern, |
| 2868 | 2927 | .arith_bcc_fusion, |
| 2869 | 2928 | .arith_cbz_fusion, |
| 2870 | 2929 | .disable_latency_sched_heuristic, |
| 2930 | .fast_ld1_single, | |
| 2871 | 2931 | .fp16fml, |
| 2872 | 2932 | .fpac, |
| 2873 | 2933 | .fuse_address, |
| ... | ... | @@ -2876,8 +2936,8 @@ pub const cpu = struct { |
| 2876 | 2936 | .fuse_arith_logic, |
| 2877 | 2937 | .fuse_crypto_eor, |
| 2878 | 2938 | .fuse_csel, |
| 2939 | .fuse_fcsel, | |
| 2879 | 2940 | .fuse_literals, |
| 2880 | .hcx, | |
| 2881 | 2941 | .max_interleave_factor_4, |
| 2882 | 2942 | .no_zcz_fpr64, |
| 2883 | 2943 | .perfmon, |
| ... | ... | @@ -2891,6 +2951,34 @@ pub const cpu = struct { |
| 2891 | 2951 | .zcz_gpr64, |
| 2892 | 2952 | }), |
| 2893 | 2953 | }; |
| 2954 | pub const armagicpu: CpuModel = .{ | |
| 2955 | .name = "armagicpu", | |
| 2956 | .llvm_name = "armagicpu", | |
| 2957 | .features = featureSet(&[_]Feature{ | |
| 2958 | .alu_lsl_fast, | |
| 2959 | .avoid_ldapur, | |
| 2960 | .brbe, | |
| 2961 | .enable_select_opt, | |
| 2962 | .ete, | |
| 2963 | .fp16fml, | |
| 2964 | .fpac, | |
| 2965 | .fuse_adrp_add, | |
| 2966 | .fuse_aes, | |
| 2967 | .fuse_csel, | |
| 2968 | .fuse_cset, | |
| 2969 | .has_limited_64bit_vector_mul_bandwidth, | |
| 2970 | .ls64, | |
| 2971 | .mte, | |
| 2972 | .perfmon, | |
| 2973 | .predictable_select_expensive, | |
| 2974 | .rand, | |
| 2975 | .spe, | |
| 2976 | .sve_bitperm, | |
| 2977 | .use_fixed_over_scalable_if_equal_cost, | |
| 2978 | .use_postra_scheduler, | |
| 2979 | .v9_2a, | |
| 2980 | }), | |
| 2981 | }; | |
| 2894 | 2982 | pub const c1_nano: CpuModel = .{ |
| 2895 | 2983 | .name = "c1_nano", |
| 2896 | 2984 | .llvm_name = "c1-nano", |
| ... | ... | @@ -2987,6 +3075,7 @@ pub const cpu = struct { |
| 2987 | 3075 | .fuse_aes, |
| 2988 | 3076 | .fuse_csel, |
| 2989 | 3077 | .fuse_cset, |
| 3078 | .max_interleave_factor_4, | |
| 2990 | 3079 | .mte, |
| 2991 | 3080 | .perfmon, |
| 2992 | 3081 | .predictable_select_expensive, |
| ... | ... | @@ -3016,7 +3105,6 @@ pub const cpu = struct { |
| 3016 | 3105 | .features = featureSet(&[_]Feature{ |
| 3017 | 3106 | .alu_lsl_fast, |
| 3018 | 3107 | .bf16, |
| 3019 | .disable_maximize_scalable_bandwidth, | |
| 3020 | 3108 | .enable_select_opt, |
| 3021 | 3109 | .ete, |
| 3022 | 3110 | .fp16fml, |
| ... | ... | @@ -3025,11 +3113,13 @@ pub const cpu = struct { |
| 3025 | 3113 | .fuse_aes, |
| 3026 | 3114 | .fuse_csel, |
| 3027 | 3115 | .fuse_cset, |
| 3116 | .has_limited_64bit_vector_mul_bandwidth, | |
| 3028 | 3117 | .i8mm, |
| 3029 | 3118 | .mte, |
| 3030 | 3119 | .perfmon, |
| 3031 | 3120 | .predictable_select_expensive, |
| 3032 | 3121 | .sve_bitperm, |
| 3122 | .use_fixed_over_scalable_if_equal_cost, | |
| 3033 | 3123 | .use_postra_scheduler, |
| 3034 | 3124 | .v9a, |
| 3035 | 3125 | }), |
| ... | ... | @@ -3562,6 +3652,7 @@ pub const cpu = struct { |
| 3562 | 3652 | .fullfp16, |
| 3563 | 3653 | .fuse_adrp_add, |
| 3564 | 3654 | .fuse_aes, |
| 3655 | .max_interleave_factor_4, | |
| 3565 | 3656 | .perfmon, |
| 3566 | 3657 | .predictable_select_expensive, |
| 3567 | 3658 | .rcpc, |
| ... | ... | @@ -3587,6 +3678,7 @@ pub const cpu = struct { |
| 3587 | 3678 | .fuse_adrp_add, |
| 3588 | 3679 | .fuse_aes, |
| 3589 | 3680 | .lse2, |
| 3681 | .max_interleave_factor_4, | |
| 3590 | 3682 | .pauth, |
| 3591 | 3683 | .perfmon, |
| 3592 | 3684 | .predictable_select_expensive, |
| ... | ... | @@ -3612,6 +3704,7 @@ pub const cpu = struct { |
| 3612 | 3704 | .fuse_adrp_add, |
| 3613 | 3705 | .fuse_aes, |
| 3614 | 3706 | .i8mm, |
| 3707 | .max_interleave_factor_4, | |
| 3615 | 3708 | .mte, |
| 3616 | 3709 | .perfmon, |
| 3617 | 3710 | .predictable_select_expensive, |
| ... | ... | @@ -3635,6 +3728,7 @@ pub const cpu = struct { |
| 3635 | 3728 | .fuse_adrp_add, |
| 3636 | 3729 | .fuse_aes, |
| 3637 | 3730 | .i8mm, |
| 3731 | .max_interleave_factor_4, | |
| 3638 | 3732 | .mte, |
| 3639 | 3733 | .perfmon, |
| 3640 | 3734 | .predictable_select_expensive, |
| ... | ... | @@ -3659,6 +3753,7 @@ pub const cpu = struct { |
| 3659 | 3753 | .fuse_aes, |
| 3660 | 3754 | .fuse_csel, |
| 3661 | 3755 | .fuse_cset, |
| 3756 | .max_interleave_factor_4, | |
| 3662 | 3757 | .mte, |
| 3663 | 3758 | .perfmon, |
| 3664 | 3759 | .predictable_select_expensive, |
| ... | ... | @@ -3683,6 +3778,7 @@ pub const cpu = struct { |
| 3683 | 3778 | .fuse_aes, |
| 3684 | 3779 | .fuse_csel, |
| 3685 | 3780 | .fuse_cset, |
| 3781 | .max_interleave_factor_4, | |
| 3686 | 3782 | .mte, |
| 3687 | 3783 | .perfmon, |
| 3688 | 3784 | .predictable_select_expensive, |
| ... | ... | @@ -3898,6 +3994,7 @@ pub const cpu = struct { |
| 3898 | 3994 | .fuse_aes, |
| 3899 | 3995 | .fuse_csel, |
| 3900 | 3996 | .fuse_cset, |
| 3997 | .max_interleave_factor_4, | |
| 3901 | 3998 | .mte, |
| 3902 | 3999 | .perfmon, |
| 3903 | 4000 | .predictable_select_expensive, |
| ... | ... | @@ -3917,6 +4014,7 @@ pub const cpu = struct { |
| 3917 | 4014 | .features = featureSet(&[_]Feature{ |
| 3918 | 4015 | .enable_select_opt, |
| 3919 | 4016 | .ete, |
| 4017 | .fixed_load_latency_4, | |
| 3920 | 4018 | .fuse_adrp_add, |
| 3921 | 4019 | .fuse_aes, |
| 3922 | 4020 | .neon, |
| ... | ... | @@ -3955,6 +4053,36 @@ pub const cpu = struct { |
| 3955 | 4053 | .v9a, |
| 3956 | 4054 | }), |
| 3957 | 4055 | }; |
| 4056 | pub const hip12: CpuModel = .{ | |
| 4057 | .name = "hip12", | |
| 4058 | .llvm_name = "hip12", | |
| 4059 | .features = featureSet(&[_]Feature{ | |
| 4060 | .addr_lsl_slow_14, | |
| 4061 | .alu_lsl_fast, | |
| 4062 | .brbe, | |
| 4063 | .cmp_bcc_fusion, | |
| 4064 | .ete, | |
| 4065 | .fp16fml, | |
| 4066 | .fpac, | |
| 4067 | .fuse_aes, | |
| 4068 | .hbc, | |
| 4069 | .ls64, | |
| 4070 | .nmi, | |
| 4071 | .perfmon, | |
| 4072 | .rand, | |
| 4073 | .rcpc3, | |
| 4074 | .rme, | |
| 4075 | .spe, | |
| 4076 | .store_pair_suppress, | |
| 4077 | .sve2, | |
| 4078 | .sve_aes, | |
| 4079 | .sve_bitperm, | |
| 4080 | .sve_sha3, | |
| 4081 | .sve_sm4, | |
| 4082 | .use_postra_scheduler, | |
| 4083 | .v8_7a, | |
| 4084 | }), | |
| 4085 | }; | |
| 3958 | 4086 | pub const kryo: CpuModel = .{ |
| 3959 | 4087 | .name = "kryo", |
| 3960 | 4088 | .llvm_name = "kryo", |
| ... | ... | @@ -4045,7 +4173,6 @@ pub const cpu = struct { |
| 4045 | 4173 | .features = featureSet(&[_]Feature{ |
| 4046 | 4174 | .alu_lsl_fast, |
| 4047 | 4175 | .bf16, |
| 4048 | .disable_maximize_scalable_bandwidth, | |
| 4049 | 4176 | .enable_select_opt, |
| 4050 | 4177 | .ete, |
| 4051 | 4178 | .fp16fml, |
| ... | ... | @@ -4054,11 +4181,13 @@ pub const cpu = struct { |
| 4054 | 4181 | .fuse_aes, |
| 4055 | 4182 | .fuse_csel, |
| 4056 | 4183 | .fuse_cset, |
| 4184 | .has_limited_64bit_vector_mul_bandwidth, | |
| 4057 | 4185 | .i8mm, |
| 4058 | 4186 | .mte, |
| 4059 | 4187 | .perfmon, |
| 4060 | 4188 | .predictable_select_expensive, |
| 4061 | 4189 | .sve_bitperm, |
| 4190 | .use_fixed_over_scalable_if_equal_cost, | |
| 4062 | 4191 | .use_postra_scheduler, |
| 4063 | 4192 | .v9a, |
| 4064 | 4193 | }), |
| ... | ... | @@ -4169,6 +4298,7 @@ pub const cpu = struct { |
| 4169 | 4298 | .rand, |
| 4170 | 4299 | .spe, |
| 4171 | 4300 | .sve_bitperm, |
| 4301 | .use_fixed_over_scalable_if_equal_cost, | |
| 4172 | 4302 | .use_postra_scheduler, |
| 4173 | 4303 | .v9_2a, |
| 4174 | 4304 | }), |
| ... | ... | @@ -4188,14 +4318,15 @@ pub const cpu = struct { |
| 4188 | 4318 | .fuse_aes, |
| 4189 | 4319 | .fuse_csel, |
| 4190 | 4320 | .fuse_cset, |
| 4321 | .has_limited_64bit_vector_mul_bandwidth, | |
| 4191 | 4322 | .ls64, |
| 4192 | .max_interleave_factor_4, | |
| 4193 | 4323 | .mte, |
| 4194 | 4324 | .perfmon, |
| 4195 | 4325 | .predictable_select_expensive, |
| 4196 | 4326 | .rand, |
| 4197 | 4327 | .spe, |
| 4198 | 4328 | .sve_bitperm, |
| 4329 | .use_fixed_over_scalable_if_equal_cost, | |
| 4199 | 4330 | .use_postra_scheduler, |
| 4200 | 4331 | .v9_2a, |
| 4201 | 4332 | }), |
| ... | ... | @@ -4248,6 +4379,33 @@ pub const cpu = struct { |
| 4248 | 4379 | .v8_6a, |
| 4249 | 4380 | }), |
| 4250 | 4381 | }; |
| 4382 | pub const rigel: CpuModel = .{ | |
| 4383 | .name = "rigel", | |
| 4384 | .llvm_name = "rigel", | |
| 4385 | .features = featureSet(&[_]Feature{ | |
| 4386 | .brbe, | |
| 4387 | .chk, | |
| 4388 | .ete, | |
| 4389 | .faminmax, | |
| 4390 | .fp16fml, | |
| 4391 | .fp8dot2, | |
| 4392 | .fp8dot4, | |
| 4393 | .fp8fma, | |
| 4394 | .fpac, | |
| 4395 | .ls64, | |
| 4396 | .lut, | |
| 4397 | .mte, | |
| 4398 | .olympus, | |
| 4399 | .perfmon, | |
| 4400 | .rand, | |
| 4401 | .spe, | |
| 4402 | .sve_aes, | |
| 4403 | .sve_bitperm, | |
| 4404 | .sve_sha3, | |
| 4405 | .sve_sm4, | |
| 4406 | .v9_2a, | |
| 4407 | }), | |
| 4408 | }; | |
| 4251 | 4409 | pub const saphira: CpuModel = .{ |
| 4252 | 4410 | .name = "saphira", |
| 4253 | 4411 | .llvm_name = "saphira", |
lib/std/Target/amdgcn.zig+998-133| ... | ... | @@ -6,6 +6,7 @@ const CpuModel = std.Target.Cpu.Model; |
| 6 | 6 | |
| 7 | 7 | pub const Feature = enum { |
| 8 | 8 | @"1024_addressable_vgprs", |
| 9 | @"1536_physical_vgprs", | |
| 9 | 10 | @"16_bit_insts", |
| 10 | 11 | @"45_bit_num_records_buffer_resource", |
| 11 | 12 | @"64_bit_literals", |
| ... | ... | @@ -18,12 +19,12 @@ pub const Feature = enum { |
| 18 | 19 | addressablelocalmemorysize327680, |
| 19 | 20 | addressablelocalmemorysize65536, |
| 20 | 21 | agent_scope_fine_grained_remote_memory_atomics, |
| 21 | allocate1_5xvgprs, | |
| 22 | 22 | aperture_regs, |
| 23 | 23 | architected_flat_scratch, |
| 24 | 24 | architected_sgprs, |
| 25 | 25 | ashr_pk_insts, |
| 26 | 26 | assembler_permissive_wavesize, |
| 27 | asynccnt, | |
| 27 | 28 | atomic_buffer_global_pk_add_f16_insts, |
| 28 | 29 | atomic_buffer_global_pk_add_f16_no_rtn_insts, |
| 29 | 30 | atomic_buffer_pk_add_bf16_inst, |
| ... | ... | @@ -75,9 +76,12 @@ pub const Feature = enum { |
| 75 | 76 | dpp, |
| 76 | 77 | dpp8, |
| 77 | 78 | dpp_64bit, |
| 79 | dpp_row_bcast, | |
| 78 | 80 | dpp_src1_sgpr, |
| 81 | dpp_wavefront_shifts, | |
| 79 | 82 | ds128, |
| 80 | 83 | ds_src2_insts, |
| 84 | dx10_clamp_and_ieee_mode, | |
| 81 | 85 | emulated_system_scope_atomics, |
| 82 | 86 | extended_image_insts, |
| 83 | 87 | f16bf16_to_fp6bf6_cvt_scale_insts, |
| ... | ... | @@ -91,13 +95,17 @@ pub const Feature = enum { |
| 91 | 95 | flat_global_insts, |
| 92 | 96 | flat_gvs_mode, |
| 93 | 97 | flat_inst_offsets, |
| 98 | flat_offset_bits_12, | |
| 99 | flat_offset_bits_24, | |
| 94 | 100 | flat_scratch, |
| 95 | 101 | flat_scratch_insts, |
| 96 | 102 | flat_segment_offset_bug, |
| 103 | flat_signed_offset, | |
| 97 | 104 | fma_mix_bf16_insts, |
| 98 | 105 | fma_mix_insts, |
| 99 | 106 | fmacf64_inst, |
| 100 | 107 | fmaf, |
| 108 | formatted_mubuf_insts, | |
| 101 | 109 | fp4_cvt_scale_insts, |
| 102 | 110 | fp64, |
| 103 | 111 | fp6bf6_cvt_scale_insts, |
| ... | ... | @@ -116,10 +124,15 @@ pub const Feature = enum { |
| 116 | 124 | gfx10_b_encoding, |
| 117 | 125 | gfx10_insts, |
| 118 | 126 | gfx11, |
| 127 | gfx11_7_insts, | |
| 119 | 128 | gfx11_insts, |
| 120 | 129 | gfx12, |
| 121 | 130 | gfx1250_insts, |
| 131 | gfx1251_gemm_insts, | |
| 132 | gfx125x_lowest_rate_wmma, | |
| 122 | 133 | gfx12_insts, |
| 134 | gfx13, | |
| 135 | gfx13_insts, | |
| 123 | 136 | gfx7_gfx8_gfx9_insts, |
| 124 | 137 | gfx8_insts, |
| 125 | 138 | gfx9, |
| ... | ... | @@ -135,6 +148,8 @@ pub const Feature = enum { |
| 135 | 148 | image_insts, |
| 136 | 149 | image_store_d16_bug, |
| 137 | 150 | inst_fwd_prefetch_bug, |
| 151 | instcachelinesize128, | |
| 152 | instcachelinesize64, | |
| 138 | 153 | int_clamp_insts, |
| 139 | 154 | inv_2pi_inline_imm, |
| 140 | 155 | kernarg_preload, |
| ... | ... | @@ -149,6 +164,7 @@ pub const Feature = enum { |
| 149 | 164 | mad_intra_fwd_bug, |
| 150 | 165 | mad_mac_f32_insts, |
| 151 | 166 | mad_mix_insts, |
| 167 | mad_nc_64_32_insts, | |
| 152 | 168 | mad_u32_inst, |
| 153 | 169 | mai_insts, |
| 154 | 170 | max_hard_clause_length_32, |
| ... | ... | @@ -161,22 +177,31 @@ pub const Feature = enum { |
| 161 | 177 | mfma_inline_literal_bug, |
| 162 | 178 | mimg_r128, |
| 163 | 179 | min3_max3_pkf16, |
| 180 | min_max_i64_insts, | |
| 164 | 181 | minimum3_maximum3_f16, |
| 165 | 182 | minimum3_maximum3_f32, |
| 166 | 183 | minimum3_maximum3_pkf16, |
| 167 | 184 | movrel, |
| 185 | mqsad_insts, | |
| 186 | mqsad_pk_insts, | |
| 168 | 187 | msaa_load_dst_sel_bug, |
| 188 | msad_insts, | |
| 189 | mtbuf_insts, | |
| 169 | 190 | negative_scratch_offset_bug, |
| 170 | 191 | negative_unaligned_scratch_offset_bug, |
| 171 | 192 | no_data_dep_hazard, |
| 193 | no_f16_pseudo_scalar_trans_inline_constants, | |
| 172 | 194 | no_sdst_cmpx, |
| 173 | 195 | nsa_clause_bug, |
| 174 | 196 | nsa_encoding, |
| 175 | 197 | nsa_to_vmem_bug, |
| 176 | 198 | offset_3f_bug, |
| 177 | 199 | packed_fp32_ops, |
| 200 | packed_fp64_ops, | |
| 178 | 201 | packed_tid, |
| 202 | packed_u64_ops, | |
| 179 | 203 | partial_nsa_encoding, |
| 204 | permlane16_insts, | |
| 180 | 205 | permlane16_swap, |
| 181 | 206 | permlane32_swap, |
| 182 | 207 | pk_add_min_max_insts, |
| ... | ... | @@ -185,23 +210,25 @@ pub const Feature = enum { |
| 185 | 210 | precise_memory, |
| 186 | 211 | priv_enabled_trap2_nop_bug, |
| 187 | 212 | prng_inst, |
| 188 | promote_alloca, | |
| 189 | 213 | prt_strict_null, |
| 190 | 214 | pseudo_scalar_trans, |
| 191 | 215 | qsad_insts, |
| 192 | 216 | r128_a16, |
| 193 | 217 | real_true16, |
| 194 | relaxed_buffer_oob_mode, | |
| 195 | 218 | required_export_priority, |
| 196 | 219 | requires_cov6, |
| 220 | requires_initial_unclause_vmem, | |
| 197 | 221 | restricted_soffset, |
| 222 | s_barrier_leave_imm, | |
| 198 | 223 | s_memrealtime, |
| 199 | 224 | s_memtime_inst, |
| 200 | 225 | s_wakeup_barrier_inst, |
| 226 | s_wakeup_imm, | |
| 201 | 227 | sad_insts, |
| 202 | 228 | safe_cu_prefetch, |
| 203 | 229 | safe_smem_prefetch, |
| 204 | 230 | salu_float, |
| 231 | salu_minimum_maximum_insts, | |
| 205 | 232 | scalar_atomics, |
| 206 | 233 | scalar_dwordx3_loads, |
| 207 | 234 | scalar_flat_scratch_insts, |
| ... | ... | @@ -219,13 +246,17 @@ pub const Feature = enum { |
| 219 | 246 | shader_cycles_hi_lo_registers, |
| 220 | 247 | shader_cycles_register, |
| 221 | 248 | si_scheduler, |
| 249 | smem_prefetch_insts, | |
| 222 | 250 | smem_to_vector_write_hazard, |
| 223 | 251 | southern_islands, |
| 224 | 252 | sramecc, |
| 225 | 253 | sramecc_support, |
| 254 | swmmac_gfx1200_insts, | |
| 255 | swmmac_gfx1250_insts, | |
| 226 | 256 | tanh_insts, |
| 227 | 257 | tensor_cvt_lut_insts, |
| 228 | tgsplit, | |
| 258 | tgsplit_support, | |
| 259 | trans_coexecution_hazard, | |
| 229 | 260 | transpose_load_f4f6_insts, |
| 230 | 261 | trap_handler, |
| 231 | 262 | trig_reduced_range, |
| ... | ... | @@ -236,7 +267,10 @@ pub const Feature = enum { |
| 236 | 267 | unaligned_scratch_access, |
| 237 | 268 | unpacked_d16_vmem, |
| 238 | 269 | unsafe_ds_offset_folding, |
| 270 | use_add_pc64_inst, | |
| 239 | 271 | user_sgpr_init16_bug, |
| 272 | v_mov_b64_inst, | |
| 273 | v_mul_u64_inst, | |
| 240 | 274 | valu_trans_use_hazard, |
| 241 | 275 | vcmpx_exec_war_hazard, |
| 242 | 276 | vcmpx_permlane_hazard, |
| ... | ... | @@ -249,6 +283,7 @@ pub const Feature = enum { |
| 249 | 283 | volcanic_islands, |
| 250 | 284 | vop3_literal, |
| 251 | 285 | vop3p, |
| 286 | vop3px2_increments_va_vdst_twice, | |
| 252 | 287 | vopd, |
| 253 | 288 | vscnt, |
| 254 | 289 | wait_xcnt, |
| ... | ... | @@ -256,8 +291,12 @@ pub const Feature = enum { |
| 256 | 291 | wavefrontsize16, |
| 257 | 292 | wavefrontsize32, |
| 258 | 293 | wavefrontsize64, |
| 294 | wmma_128b_insts, | |
| 295 | wmma_256b_insts, | |
| 296 | wmma_coexecution_hazards, | |
| 259 | 297 | xf32_insts, |
| 260 | 298 | xnack, |
| 299 | xnack_on_off_modes, | |
| 261 | 300 | xnack_support, |
| 262 | 301 | }; |
| 263 | 302 | |
| ... | ... | @@ -276,6 +315,11 @@ pub const all_features = blk: { |
| 276 | 315 | .description = "Has 1024 addressable VGPRs", |
| 277 | 316 | .dependencies = featureSet(&[_]Feature{}), |
| 278 | 317 | }; |
| 318 | result[@backingInt(Feature.@"1536_physical_vgprs")] = .{ | |
| 319 | .llvm_name = "1536-physical-vgprs", | |
| 320 | .description = "Has 1536 physical VGPRs per SIMD", | |
| 321 | .dependencies = featureSet(&[_]Feature{}), | |
| 322 | }; | |
| 279 | 323 | result[@backingInt(Feature.@"16_bit_insts")] = .{ |
| 280 | 324 | .llvm_name = "16-bit-insts", |
| 281 | 325 | .description = "Has i16/f16 instructions", |
| ... | ... | @@ -336,11 +380,6 @@ pub const all_features = blk: { |
| 336 | 380 | .description = "Agent (device) scoped atomic operations, excluding those directly supported by PCIe (i.e. integer atomic add, exchange, and compare-and-swap), are functional for allocations in host or peer device memory.", |
| 337 | 381 | .dependencies = featureSet(&[_]Feature{}), |
| 338 | 382 | }; |
| 339 | result[@backingInt(Feature.allocate1_5xvgprs)] = .{ | |
| 340 | .llvm_name = "allocate1_5xvgprs", | |
| 341 | .description = "Has 50% more physical VGPRs and 50% larger allocation granule", | |
| 342 | .dependencies = featureSet(&[_]Feature{}), | |
| 343 | }; | |
| 344 | 383 | result[@backingInt(Feature.aperture_regs)] = .{ |
| 345 | 384 | .llvm_name = "aperture-regs", |
| 346 | 385 | .description = "Has Memory Aperture Base and Size Registers", |
| ... | ... | @@ -363,7 +402,12 @@ pub const all_features = blk: { |
| 363 | 402 | }; |
| 364 | 403 | result[@backingInt(Feature.assembler_permissive_wavesize)] = .{ |
| 365 | 404 | .llvm_name = "assembler-permissive-wavesize", |
| 366 | .description = "allow parsing wave32 and wave64 variants of instructions", | |
| 405 | .description = "Allow parsing wave32 and wave64 variants of instructions", | |
| 406 | .dependencies = featureSet(&[_]Feature{}), | |
| 407 | }; | |
| 408 | result[@backingInt(Feature.asynccnt)] = .{ | |
| 409 | .llvm_name = "asynccnt", | |
| 410 | .description = "Has separate asynccnt counter", | |
| 367 | 411 | .dependencies = featureSet(&[_]Feature{}), |
| 368 | 412 | }; |
| 369 | 413 | result[@backingInt(Feature.atomic_buffer_global_pk_add_f16_insts)] = .{ |
| ... | ... | @@ -637,11 +681,21 @@ pub const all_features = blk: { |
| 637 | 681 | .description = "Support DPP (Data Parallel Primitives) extension in DP ALU", |
| 638 | 682 | .dependencies = featureSet(&[_]Feature{}), |
| 639 | 683 | }; |
| 684 | result[@backingInt(Feature.dpp_row_bcast)] = .{ | |
| 685 | .llvm_name = "dpp-row-bcast", | |
| 686 | .description = "Support DPP row_bcast15/row_bcast31", | |
| 687 | .dependencies = featureSet(&[_]Feature{}), | |
| 688 | }; | |
| 640 | 689 | result[@backingInt(Feature.dpp_src1_sgpr)] = .{ |
| 641 | 690 | .llvm_name = "dpp-src1-sgpr", |
| 642 | 691 | .description = "Support SGPR for Src1 of DPP instructions", |
| 643 | 692 | .dependencies = featureSet(&[_]Feature{}), |
| 644 | 693 | }; |
| 694 | result[@backingInt(Feature.dpp_wavefront_shifts)] = .{ | |
| 695 | .llvm_name = "dpp-wavefront-shifts", | |
| 696 | .description = "Support DPP wave_shl/wave_rol/wave_shr/wave_ror", | |
| 697 | .dependencies = featureSet(&[_]Feature{}), | |
| 698 | }; | |
| 645 | 699 | result[@backingInt(Feature.ds128)] = .{ |
| 646 | 700 | .llvm_name = "enable-ds128", |
| 647 | 701 | .description = "Use ds_{read|write}_b128", |
| ... | ... | @@ -652,6 +706,11 @@ pub const all_features = blk: { |
| 652 | 706 | .description = "Has ds_*_src2 instructions", |
| 653 | 707 | .dependencies = featureSet(&[_]Feature{}), |
| 654 | 708 | }; |
| 709 | result[@backingInt(Feature.dx10_clamp_and_ieee_mode)] = .{ | |
| 710 | .llvm_name = "dx10-clamp-and-ieee-mode", | |
| 711 | .description = "Target has DX10_CLAMP and IEEE_MODE kernel descriptor bits", | |
| 712 | .dependencies = featureSet(&[_]Feature{}), | |
| 713 | }; | |
| 655 | 714 | result[@backingInt(Feature.emulated_system_scope_atomics)] = .{ |
| 656 | 715 | .llvm_name = "emulated-system-scope-atomics", |
| 657 | 716 | .description = "System scope atomics unsupported by the PCI-e are emulated in HW via CAS loop and functional.", |
| ... | ... | @@ -723,6 +782,16 @@ pub const all_features = blk: { |
| 723 | 782 | .description = "Flat instructions have immediate offset addressing mode", |
| 724 | 783 | .dependencies = featureSet(&[_]Feature{}), |
| 725 | 784 | }; |
| 785 | result[@backingInt(Feature.flat_offset_bits_12)] = .{ | |
| 786 | .llvm_name = "flat-offset-bits-12", | |
| 787 | .description = "Number of bits for flat offset encoding", | |
| 788 | .dependencies = featureSet(&[_]Feature{}), | |
| 789 | }; | |
| 790 | result[@backingInt(Feature.flat_offset_bits_24)] = .{ | |
| 791 | .llvm_name = "flat-offset-bits-24", | |
| 792 | .description = "Number of bits for flat offset encoding", | |
| 793 | .dependencies = featureSet(&[_]Feature{}), | |
| 794 | }; | |
| 726 | 795 | result[@backingInt(Feature.flat_scratch)] = .{ |
| 727 | 796 | .llvm_name = "enable-flat-scratch", |
| 728 | 797 | .description = "Use scratch_* flat memory instructions to access scratch", |
| ... | ... | @@ -740,6 +809,11 @@ pub const all_features = blk: { |
| 740 | 809 | .description = "GFX10 bug where inst_offset is ignored when flat instructions access global memory", |
| 741 | 810 | .dependencies = featureSet(&[_]Feature{}), |
| 742 | 811 | }; |
| 812 | result[@backingInt(Feature.flat_signed_offset)] = .{ | |
| 813 | .llvm_name = "flat-signed-offset", | |
| 814 | .description = "Immediate offset of FLAT instructions are always signed", | |
| 815 | .dependencies = featureSet(&[_]Feature{}), | |
| 816 | }; | |
| 743 | 817 | result[@backingInt(Feature.fma_mix_bf16_insts)] = .{ |
| 744 | 818 | .llvm_name = "fma-mix-bf16-insts", |
| 745 | 819 | .description = "Has v_fma_mix_f32_bf16, v_fma_mixlo_bf16, v_fma_mixhi_bf16 instructions", |
| ... | ... | @@ -760,6 +834,11 @@ pub const all_features = blk: { |
| 760 | 834 | .description = "Enable single precision FMA (not as fast as mul+add, but fused)", |
| 761 | 835 | .dependencies = featureSet(&[_]Feature{}), |
| 762 | 836 | }; |
| 837 | result[@backingInt(Feature.formatted_mubuf_insts)] = .{ | |
| 838 | .llvm_name = "formatted-mubuf-insts", | |
| 839 | .description = "Has formatted memory untyped buffer instructions.", | |
| 840 | .dependencies = featureSet(&[_]Feature{}), | |
| 841 | }; | |
| 763 | 842 | result[@backingInt(Feature.fp4_cvt_scale_insts)] = .{ |
| 764 | 843 | .llvm_name = "fp4-cvt-scale-insts", |
| 765 | 844 | .description = "Has fp4 conversion scale instructions", |
| ... | ... | @@ -841,13 +920,16 @@ pub const all_features = blk: { |
| 841 | 920 | .default_component_zero, |
| 842 | 921 | .dpp, |
| 843 | 922 | .dpp8, |
| 923 | .dx10_clamp_and_ieee_mode, | |
| 844 | 924 | .extended_image_insts, |
| 845 | 925 | .fast_denormal_f32, |
| 846 | 926 | .fast_fmaf, |
| 847 | 927 | .flat_global_insts, |
| 848 | 928 | .flat_inst_offsets, |
| 929 | .flat_offset_bits_12, | |
| 849 | 930 | .flat_scratch_insts, |
| 850 | 931 | .fma_mix_insts, |
| 932 | .formatted_mubuf_insts, | |
| 851 | 933 | .fp64, |
| 852 | 934 | .g16, |
| 853 | 935 | .gds, |
| ... | ... | @@ -856,14 +938,20 @@ pub const all_features = blk: { |
| 856 | 938 | .gfx9_insts, |
| 857 | 939 | .gws, |
| 858 | 940 | .image_insts, |
| 941 | .instcachelinesize64, | |
| 859 | 942 | .int_clamp_insts, |
| 860 | 943 | .inv_2pi_inline_imm, |
| 861 | 944 | .lerp_inst, |
| 862 | 945 | .max_hard_clause_length_63, |
| 863 | 946 | .mimg_r128, |
| 864 | 947 | .movrel, |
| 948 | .mqsad_insts, | |
| 949 | .mqsad_pk_insts, | |
| 950 | .msad_insts, | |
| 951 | .mtbuf_insts, | |
| 865 | 952 | .no_data_dep_hazard, |
| 866 | 953 | .no_sdst_cmpx, |
| 954 | .permlane16_insts, | |
| 867 | 955 | .pk_fmac_f16_inst, |
| 868 | 956 | .qsad_insts, |
| 869 | 957 | .s_memrealtime, |
| ... | ... | @@ -929,6 +1017,7 @@ pub const all_features = blk: { |
| 929 | 1017 | .flat_inst_offsets, |
| 930 | 1018 | .flat_scratch_insts, |
| 931 | 1019 | .fma_mix_insts, |
| 1020 | .formatted_mubuf_insts, | |
| 932 | 1021 | .fp64, |
| 933 | 1022 | .g16, |
| 934 | 1023 | .gds, |
| ... | ... | @@ -940,14 +1029,20 @@ pub const all_features = blk: { |
| 940 | 1029 | .gfx8_insts, |
| 941 | 1030 | .gfx9_insts, |
| 942 | 1031 | .gws, |
| 1032 | .instcachelinesize128, | |
| 943 | 1033 | .int_clamp_insts, |
| 944 | 1034 | .inv_2pi_inline_imm, |
| 945 | 1035 | .lerp_inst, |
| 946 | 1036 | .max_hard_clause_length_32, |
| 947 | 1037 | .mimg_r128, |
| 948 | 1038 | .movrel, |
| 1039 | .mqsad_insts, | |
| 1040 | .mqsad_pk_insts, | |
| 1041 | .msad_insts, | |
| 1042 | .mtbuf_insts, | |
| 949 | 1043 | .no_data_dep_hazard, |
| 950 | 1044 | .no_sdst_cmpx, |
| 1045 | .permlane16_insts, | |
| 951 | 1046 | .pk_fmac_f16_inst, |
| 952 | 1047 | .qsad_insts, |
| 953 | 1048 | .sad_insts, |
| ... | ... | @@ -962,6 +1057,11 @@ pub const all_features = blk: { |
| 962 | 1057 | .vscnt, |
| 963 | 1058 | }), |
| 964 | 1059 | }; |
| 1060 | result[@backingInt(Feature.gfx11_7_insts)] = .{ | |
| 1061 | .llvm_name = "gfx11-7-insts", | |
| 1062 | .description = "Additional instructions for GFX11.7", | |
| 1063 | .dependencies = featureSet(&[_]Feature{}), | |
| 1064 | }; | |
| 965 | 1065 | result[@backingInt(Feature.gfx11_insts)] = .{ |
| 966 | 1066 | .llvm_name = "gfx11-insts", |
| 967 | 1067 | .description = "Additional instructions for GFX11+", |
| ... | ... | @@ -986,7 +1086,9 @@ pub const all_features = blk: { |
| 986 | 1086 | .fast_fmaf, |
| 987 | 1087 | .flat_global_insts, |
| 988 | 1088 | .flat_inst_offsets, |
| 1089 | .flat_offset_bits_24, | |
| 989 | 1090 | .flat_scratch_insts, |
| 1091 | .flat_signed_offset, | |
| 990 | 1092 | .fma_mix_insts, |
| 991 | 1093 | .fp64, |
| 992 | 1094 | .g16, |
| ... | ... | @@ -999,6 +1101,7 @@ pub const all_features = blk: { |
| 999 | 1101 | .gfx8_insts, |
| 1000 | 1102 | .gfx9_insts, |
| 1001 | 1103 | .ieee_minimum_maximum_insts, |
| 1104 | .instcachelinesize128, | |
| 1002 | 1105 | .int_clamp_insts, |
| 1003 | 1106 | .inv_2pi_inline_imm, |
| 1004 | 1107 | .max_hard_clause_length_32, |
| ... | ... | @@ -1008,7 +1111,9 @@ pub const all_features = blk: { |
| 1008 | 1111 | .movrel, |
| 1009 | 1112 | .no_data_dep_hazard, |
| 1010 | 1113 | .no_sdst_cmpx, |
| 1114 | .permlane16_insts, | |
| 1011 | 1115 | .pk_fmac_f16_inst, |
| 1116 | .salu_minimum_maximum_insts, | |
| 1012 | 1117 | .true16, |
| 1013 | 1118 | .unaligned_buffer_access, |
| 1014 | 1119 | .unaligned_ds_access, |
| ... | ... | @@ -1024,11 +1129,91 @@ pub const all_features = blk: { |
| 1024 | 1129 | .description = "Additional instructions for GFX1250+", |
| 1025 | 1130 | .dependencies = featureSet(&[_]Feature{}), |
| 1026 | 1131 | }; |
| 1132 | result[@backingInt(Feature.gfx1251_gemm_insts)] = .{ | |
| 1133 | .llvm_name = "gfx1251-gemm-insts", | |
| 1134 | .description = "Has additional gfx1251 DGEMM instructions", | |
| 1135 | .dependencies = featureSet(&[_]Feature{}), | |
| 1136 | }; | |
| 1137 | result[@backingInt(Feature.gfx125x_lowest_rate_wmma)] = .{ | |
| 1138 | .llvm_name = "gfx125x-lowest-rate-wmma", | |
| 1139 | .description = "Has the lowest rate wmma in the gfx125x family, mainly for the gfx12-5-generic", | |
| 1140 | .dependencies = featureSet(&[_]Feature{}), | |
| 1141 | }; | |
| 1027 | 1142 | result[@backingInt(Feature.gfx12_insts)] = .{ |
| 1028 | 1143 | .llvm_name = "gfx12-insts", |
| 1029 | 1144 | .description = "Additional instructions for GFX12+", |
| 1030 | 1145 | .dependencies = featureSet(&[_]Feature{}), |
| 1031 | 1146 | }; |
| 1147 | result[@backingInt(Feature.gfx13)] = .{ | |
| 1148 | .llvm_name = "gfx13", | |
| 1149 | .description = "GFX13 GPU generation", | |
| 1150 | .dependencies = featureSet(&[_]Feature{ | |
| 1151 | .@"16_bit_insts", | |
| 1152 | .a16, | |
| 1153 | .add_no_carry_insts, | |
| 1154 | .agent_scope_fine_grained_remote_memory_atomics, | |
| 1155 | .aperture_regs, | |
| 1156 | .atomic_fmin_fmax_flat_f32, | |
| 1157 | .atomic_fmin_fmax_global_f32, | |
| 1158 | .ci_insts, | |
| 1159 | .default_component_broadcast, | |
| 1160 | .dpp, | |
| 1161 | .dpp8, | |
| 1162 | .extended_image_insts, | |
| 1163 | .fast_denormal_f32, | |
| 1164 | .fast_fmaf, | |
| 1165 | .flat_global_insts, | |
| 1166 | .flat_inst_offsets, | |
| 1167 | .flat_offset_bits_24, | |
| 1168 | .flat_scratch_insts, | |
| 1169 | .flat_signed_offset, | |
| 1170 | .fma_mix_insts, | |
| 1171 | .formatted_mubuf_insts, | |
| 1172 | .fp64, | |
| 1173 | .g16, | |
| 1174 | .gfx10_3_insts, | |
| 1175 | .gfx10_a_encoding, | |
| 1176 | .gfx10_b_encoding, | |
| 1177 | .gfx10_insts, | |
| 1178 | .gfx11_insts, | |
| 1179 | .gfx12_insts, | |
| 1180 | .gfx13_insts, | |
| 1181 | .gfx8_insts, | |
| 1182 | .gfx9_insts, | |
| 1183 | .ieee_minimum_maximum_insts, | |
| 1184 | .image_insts, | |
| 1185 | .instcachelinesize128, | |
| 1186 | .int_clamp_insts, | |
| 1187 | .inv_2pi_inline_imm, | |
| 1188 | .max_hard_clause_length_32, | |
| 1189 | .mimg_r128, | |
| 1190 | .minimum3_maximum3_f16, | |
| 1191 | .minimum3_maximum3_f32, | |
| 1192 | .movrel, | |
| 1193 | .mtbuf_insts, | |
| 1194 | .no_data_dep_hazard, | |
| 1195 | .no_sdst_cmpx, | |
| 1196 | .permlane16_insts, | |
| 1197 | .pk_fmac_f16_inst, | |
| 1198 | .salu_minimum_maximum_insts, | |
| 1199 | .true16, | |
| 1200 | .unaligned_buffer_access, | |
| 1201 | .unaligned_ds_access, | |
| 1202 | .unaligned_scratch_access, | |
| 1203 | .vop3_literal, | |
| 1204 | .vop3p, | |
| 1205 | .vopd, | |
| 1206 | .vscnt, | |
| 1207 | }), | |
| 1208 | }; | |
| 1209 | result[@backingInt(Feature.gfx13_insts)] = .{ | |
| 1210 | .llvm_name = "gfx13-insts", | |
| 1211 | .description = "Additional instructions for GFX13+", | |
| 1212 | .dependencies = featureSet(&[_]Feature{ | |
| 1213 | .s_barrier_leave_imm, | |
| 1214 | .s_wakeup_imm, | |
| 1215 | }), | |
| 1216 | }; | |
| 1032 | 1217 | result[@backingInt(Feature.gfx7_gfx8_gfx9_insts)] = .{ |
| 1033 | 1218 | .llvm_name = "gfx7-gfx8-gfx9-insts", |
| 1034 | 1219 | .description = "Instructions shared in GFX7, GFX8, GFX9", |
| ... | ... | @@ -1054,20 +1239,29 @@ pub const all_features = blk: { |
| 1054 | 1239 | .cvt_pknorm_vop3_insts, |
| 1055 | 1240 | .default_component_zero, |
| 1056 | 1241 | .dpp, |
| 1242 | .dpp_row_bcast, | |
| 1243 | .dpp_wavefront_shifts, | |
| 1244 | .dx10_clamp_and_ieee_mode, | |
| 1057 | 1245 | .fast_denormal_f32, |
| 1058 | 1246 | .fast_fmaf, |
| 1059 | 1247 | .flat_global_insts, |
| 1060 | 1248 | .flat_inst_offsets, |
| 1061 | 1249 | .flat_scratch_insts, |
| 1250 | .formatted_mubuf_insts, | |
| 1062 | 1251 | .fp64, |
| 1063 | 1252 | .gcn3_encoding, |
| 1064 | 1253 | .gfx7_gfx8_gfx9_insts, |
| 1065 | 1254 | .gfx8_insts, |
| 1066 | 1255 | .gfx9_insts, |
| 1067 | 1256 | .gws, |
| 1257 | .instcachelinesize64, | |
| 1068 | 1258 | .int_clamp_insts, |
| 1069 | 1259 | .inv_2pi_inline_imm, |
| 1070 | 1260 | .lerp_inst, |
| 1261 | .mqsad_insts, | |
| 1262 | .mqsad_pk_insts, | |
| 1263 | .msad_insts, | |
| 1264 | .mtbuf_insts, | |
| 1071 | 1265 | .negative_scratch_offset_bug, |
| 1072 | 1266 | .qsad_insts, |
| 1073 | 1267 | .r128_a16, |
| ... | ... | @@ -1089,7 +1283,7 @@ pub const all_features = blk: { |
| 1089 | 1283 | .vmem_write_vgpr_in_order, |
| 1090 | 1284 | .vop3p, |
| 1091 | 1285 | .wavefrontsize64, |
| 1092 | .xnack_support, | |
| 1286 | .xnack_on_off_modes, | |
| 1093 | 1287 | }), |
| 1094 | 1288 | }; |
| 1095 | 1289 | result[@backingInt(Feature.gfx90a_insts)] = .{ |
| ... | ... | @@ -1142,7 +1336,7 @@ pub const all_features = blk: { |
| 1142 | 1336 | }; |
| 1143 | 1337 | result[@backingInt(Feature.ieee_minimum_maximum_insts)] = .{ |
| 1144 | 1338 | .llvm_name = "ieee-minimum-maximum-insts", |
| 1145 | .description = "Has v_minimum/maximum_f16/f32/f64, v_minimummaximum/maximumminimum_f16/f32 and v_pk_minimum/maximum_f16 instructions", | |
| 1339 | .description = "Has v_minimum/maximum_f16/f32/f64, v_minimummaximum/maximumminimum_f16/f32 andv_pk_minimum/maximum_f16 instructions", | |
| 1146 | 1340 | .dependencies = featureSet(&[_]Feature{}), |
| 1147 | 1341 | }; |
| 1148 | 1342 | result[@backingInt(Feature.image_gather4_d16_bug)] = .{ |
| ... | ... | @@ -1165,6 +1359,16 @@ pub const all_features = blk: { |
| 1165 | 1359 | .description = "S_INST_PREFETCH instruction causes shader to hang", |
| 1166 | 1360 | .dependencies = featureSet(&[_]Feature{}), |
| 1167 | 1361 | }; |
| 1362 | result[@backingInt(Feature.instcachelinesize128)] = .{ | |
| 1363 | .llvm_name = "instcachelinesize128", | |
| 1364 | .description = "Instruction cache line size in bytes.", | |
| 1365 | .dependencies = featureSet(&[_]Feature{}), | |
| 1366 | }; | |
| 1367 | result[@backingInt(Feature.instcachelinesize64)] = .{ | |
| 1368 | .llvm_name = "instcachelinesize64", | |
| 1369 | .description = "Instruction cache line size in bytes.", | |
| 1370 | .dependencies = featureSet(&[_]Feature{}), | |
| 1371 | }; | |
| 1168 | 1372 | result[@backingInt(Feature.int_clamp_insts)] = .{ |
| 1169 | 1373 | .llvm_name = "int-clamp-insts", |
| 1170 | 1374 | .description = "Support clamp for integer destination", |
| ... | ... | @@ -1235,6 +1439,11 @@ pub const all_features = blk: { |
| 1235 | 1439 | .description = "Has v_mad_mix_f32, v_mad_mixlo_f16, v_mad_mixhi_f16 instructions", |
| 1236 | 1440 | .dependencies = featureSet(&[_]Feature{}), |
| 1237 | 1441 | }; |
| 1442 | result[@backingInt(Feature.mad_nc_64_32_insts)] = .{ | |
| 1443 | .llvm_name = "mad-nc-64-32-insts", | |
| 1444 | .description = "Has v_mad_nc_{u64_u32|i64_i32} instructions", | |
| 1445 | .dependencies = featureSet(&[_]Feature{}), | |
| 1446 | }; | |
| 1238 | 1447 | result[@backingInt(Feature.mad_u32_inst)] = .{ |
| 1239 | 1448 | .llvm_name = "mad-u32-inst", |
| 1240 | 1449 | .description = "Has v_mad_u32 instruction", |
| ... | ... | @@ -1295,6 +1504,11 @@ pub const all_features = blk: { |
| 1295 | 1504 | .description = "Has v_pk_min3_num_f16 and v_pk_max3_num_f16 instructions", |
| 1296 | 1505 | .dependencies = featureSet(&[_]Feature{}), |
| 1297 | 1506 | }; |
| 1507 | result[@backingInt(Feature.min_max_i64_insts)] = .{ | |
| 1508 | .llvm_name = "min-max-i64-insts", | |
| 1509 | .description = "Has v_{min|max}_{i|u}64 instructions", | |
| 1510 | .dependencies = featureSet(&[_]Feature{}), | |
| 1511 | }; | |
| 1298 | 1512 | result[@backingInt(Feature.minimum3_maximum3_f16)] = .{ |
| 1299 | 1513 | .llvm_name = "minimum3-maximum3-f16", |
| 1300 | 1514 | .description = "Has v_minimum3_f16 and v_maximum3_f16 instructions", |
| ... | ... | @@ -1315,11 +1529,31 @@ pub const all_features = blk: { |
| 1315 | 1529 | .description = "Has v_movrel*_b32 instructions", |
| 1316 | 1530 | .dependencies = featureSet(&[_]Feature{}), |
| 1317 | 1531 | }; |
| 1532 | result[@backingInt(Feature.mqsad_insts)] = .{ | |
| 1533 | .llvm_name = "mqsad-insts", | |
| 1534 | .description = "Has v_mqsad_u32_u8 instruction", | |
| 1535 | .dependencies = featureSet(&[_]Feature{}), | |
| 1536 | }; | |
| 1537 | result[@backingInt(Feature.mqsad_pk_insts)] = .{ | |
| 1538 | .llvm_name = "mqsad-pk-insts", | |
| 1539 | .description = "Has v_mqsad_pk_u16_u8 instruction", | |
| 1540 | .dependencies = featureSet(&[_]Feature{}), | |
| 1541 | }; | |
| 1318 | 1542 | result[@backingInt(Feature.msaa_load_dst_sel_bug)] = .{ |
| 1319 | 1543 | .llvm_name = "msaa-load-dst-sel-bug", |
| 1320 | 1544 | .description = "MSAA loads not honoring dst_sel bug", |
| 1321 | 1545 | .dependencies = featureSet(&[_]Feature{}), |
| 1322 | 1546 | }; |
| 1547 | result[@backingInt(Feature.msad_insts)] = .{ | |
| 1548 | .llvm_name = "msad-insts", | |
| 1549 | .description = "Has v_msad_u8 instruction", | |
| 1550 | .dependencies = featureSet(&[_]Feature{}), | |
| 1551 | }; | |
| 1552 | result[@backingInt(Feature.mtbuf_insts)] = .{ | |
| 1553 | .llvm_name = "mtbuf-insts", | |
| 1554 | .description = "Has memory typed buffer instructions.", | |
| 1555 | .dependencies = featureSet(&[_]Feature{}), | |
| 1556 | }; | |
| 1323 | 1557 | result[@backingInt(Feature.negative_scratch_offset_bug)] = .{ |
| 1324 | 1558 | .llvm_name = "negative-scratch-offset-bug", |
| 1325 | 1559 | .description = "Negative immediate offsets in scratch instructions with an SGPR offset page fault on GFX9", |
| ... | ... | @@ -1327,7 +1561,7 @@ pub const all_features = blk: { |
| 1327 | 1561 | }; |
| 1328 | 1562 | result[@backingInt(Feature.negative_unaligned_scratch_offset_bug)] = .{ |
| 1329 | 1563 | .llvm_name = "negative-unaligned-scratch-offset-bug", |
| 1330 | .description = "Scratch instructions with a VGPR offset and a negative immediate offset that is not a multiple of 4 read wrong memory on GFX10", | |
| 1564 | .description = "Scratch instructions with a VGPR offset and a negative immediate offset thatis not a multiple of 4 read wrong memory on GFX10", | |
| 1331 | 1565 | .dependencies = featureSet(&[_]Feature{}), |
| 1332 | 1566 | }; |
| 1333 | 1567 | result[@backingInt(Feature.no_data_dep_hazard)] = .{ |
| ... | ... | @@ -1335,6 +1569,11 @@ pub const all_features = blk: { |
| 1335 | 1569 | .description = "Does not need SW waitstates", |
| 1336 | 1570 | .dependencies = featureSet(&[_]Feature{}), |
| 1337 | 1571 | }; |
| 1572 | result[@backingInt(Feature.no_f16_pseudo_scalar_trans_inline_constants)] = .{ | |
| 1573 | .llvm_name = "no-f16-pseudo-scalar-trans-inline-constants", | |
| 1574 | .description = "Inline constants are not supported for F16 pseudo scalar transcendentals", | |
| 1575 | .dependencies = featureSet(&[_]Feature{}), | |
| 1576 | }; | |
| 1338 | 1577 | result[@backingInt(Feature.no_sdst_cmpx)] = .{ |
| 1339 | 1578 | .llvm_name = "no-sdst-cmpx", |
| 1340 | 1579 | .description = "V_CMPX does not write VCC/SGPR in addition to EXEC", |
| ... | ... | @@ -1365,16 +1604,31 @@ pub const all_features = blk: { |
| 1365 | 1604 | .description = "Support packed fp32 instructions", |
| 1366 | 1605 | .dependencies = featureSet(&[_]Feature{}), |
| 1367 | 1606 | }; |
| 1607 | result[@backingInt(Feature.packed_fp64_ops)] = .{ | |
| 1608 | .llvm_name = "packed-fp64-ops", | |
| 1609 | .description = "Support packed fp64 instructions", | |
| 1610 | .dependencies = featureSet(&[_]Feature{}), | |
| 1611 | }; | |
| 1368 | 1612 | result[@backingInt(Feature.packed_tid)] = .{ |
| 1369 | 1613 | .llvm_name = "packed-tid", |
| 1370 | 1614 | .description = "Workitem IDs are packed into v0 at kernel launch", |
| 1371 | 1615 | .dependencies = featureSet(&[_]Feature{}), |
| 1372 | 1616 | }; |
| 1617 | result[@backingInt(Feature.packed_u64_ops)] = .{ | |
| 1618 | .llvm_name = "packed-u64-ops", | |
| 1619 | .description = "Support packed uint64 instructions", | |
| 1620 | .dependencies = featureSet(&[_]Feature{}), | |
| 1621 | }; | |
| 1373 | 1622 | result[@backingInt(Feature.partial_nsa_encoding)] = .{ |
| 1374 | 1623 | .llvm_name = "partial-nsa-encoding", |
| 1375 | 1624 | .description = "Support partial NSA encoding for image instructions", |
| 1376 | 1625 | .dependencies = featureSet(&[_]Feature{}), |
| 1377 | 1626 | }; |
| 1627 | result[@backingInt(Feature.permlane16_insts)] = .{ | |
| 1628 | .llvm_name = "permlane16-insts", | |
| 1629 | .description = "Has v_permlane16_b32/v_permlanex16_b32 instructions", | |
| 1630 | .dependencies = featureSet(&[_]Feature{}), | |
| 1631 | }; | |
| 1378 | 1632 | result[@backingInt(Feature.permlane16_swap)] = .{ |
| 1379 | 1633 | .llvm_name = "permlane16-swap", |
| 1380 | 1634 | .description = "Has v_permlane16_swap_b32 instructions", |
| ... | ... | @@ -1415,11 +1669,6 @@ pub const all_features = blk: { |
| 1415 | 1669 | .description = "Has v_prng_b32 instruction", |
| 1416 | 1670 | .dependencies = featureSet(&[_]Feature{}), |
| 1417 | 1671 | }; |
| 1418 | result[@backingInt(Feature.promote_alloca)] = .{ | |
| 1419 | .llvm_name = "promote-alloca", | |
| 1420 | .description = "Enable promote alloca pass", | |
| 1421 | .dependencies = featureSet(&[_]Feature{}), | |
| 1422 | }; | |
| 1423 | 1672 | result[@backingInt(Feature.prt_strict_null)] = .{ |
| 1424 | 1673 | .llvm_name = "enable-prt-strict-null", |
| 1425 | 1674 | .description = "Enable zeroing of result registers for sparse texture fetches", |
| ... | ... | @@ -1445,11 +1694,6 @@ pub const all_features = blk: { |
| 1445 | 1694 | .description = "Use true 16-bit registers", |
| 1446 | 1695 | .dependencies = featureSet(&[_]Feature{}), |
| 1447 | 1696 | }; |
| 1448 | result[@backingInt(Feature.relaxed_buffer_oob_mode)] = .{ | |
| 1449 | .llvm_name = "relaxed-buffer-oob-mode", | |
| 1450 | .description = "Disable strict out-of-bounds buffer guarantees. An OOB access may potentially cause an adjacent access to be treated as if it were also OOB", | |
| 1451 | .dependencies = featureSet(&[_]Feature{}), | |
| 1452 | }; | |
| 1453 | 1697 | result[@backingInt(Feature.required_export_priority)] = .{ |
| 1454 | 1698 | .llvm_name = "required-export-priority", |
| 1455 | 1699 | .description = "Export priority must be explicitly manipulated on GFX11.5", |
| ... | ... | @@ -1460,11 +1704,21 @@ pub const all_features = blk: { |
| 1460 | 1704 | .description = "Target Requires Code Object V6", |
| 1461 | 1705 | .dependencies = featureSet(&[_]Feature{}), |
| 1462 | 1706 | }; |
| 1707 | result[@backingInt(Feature.requires_initial_unclause_vmem)] = .{ | |
| 1708 | .llvm_name = "requires-initial-unclause-vmem", | |
| 1709 | .description = "Hardware entrypoints require first VMEM instruction to be unclaused", | |
| 1710 | .dependencies = featureSet(&[_]Feature{}), | |
| 1711 | }; | |
| 1463 | 1712 | result[@backingInt(Feature.restricted_soffset)] = .{ |
| 1464 | 1713 | .llvm_name = "restricted-soffset", |
| 1465 | 1714 | .description = "Has restricted SOffset (immediate not supported).", |
| 1466 | 1715 | .dependencies = featureSet(&[_]Feature{}), |
| 1467 | 1716 | }; |
| 1717 | result[@backingInt(Feature.s_barrier_leave_imm)] = .{ | |
| 1718 | .llvm_name = "s-barrier-leave-imm", | |
| 1719 | .description = "s_barrier_leave takes an immediate operand", | |
| 1720 | .dependencies = featureSet(&[_]Feature{}), | |
| 1721 | }; | |
| 1468 | 1722 | result[@backingInt(Feature.s_memrealtime)] = .{ |
| 1469 | 1723 | .llvm_name = "s-memrealtime", |
| 1470 | 1724 | .description = "Has s_memrealtime instruction", |
| ... | ... | @@ -1480,6 +1734,11 @@ pub const all_features = blk: { |
| 1480 | 1734 | .description = "Has s_wakeup_barrier instruction.", |
| 1481 | 1735 | .dependencies = featureSet(&[_]Feature{}), |
| 1482 | 1736 | }; |
| 1737 | result[@backingInt(Feature.s_wakeup_imm)] = .{ | |
| 1738 | .llvm_name = "s-wakeup-imm", | |
| 1739 | .description = "s_wakeup takes an immediate operand", | |
| 1740 | .dependencies = featureSet(&[_]Feature{}), | |
| 1741 | }; | |
| 1483 | 1742 | result[@backingInt(Feature.sad_insts)] = .{ |
| 1484 | 1743 | .llvm_name = "sad-insts", |
| 1485 | 1744 | .description = "Has v_sad* instructions", |
| ... | ... | @@ -1500,6 +1759,11 @@ pub const all_features = blk: { |
| 1500 | 1759 | .description = "Has SALU floating point instructions", |
| 1501 | 1760 | .dependencies = featureSet(&[_]Feature{}), |
| 1502 | 1761 | }; |
| 1762 | result[@backingInt(Feature.salu_minimum_maximum_insts)] = .{ | |
| 1763 | .llvm_name = "salu-minimum-maximum-insts", | |
| 1764 | .description = "Has s_minimum/maximum_f16/f32 instructions", | |
| 1765 | .dependencies = featureSet(&[_]Feature{}), | |
| 1766 | }; | |
| 1503 | 1767 | result[@backingInt(Feature.scalar_atomics)] = .{ |
| 1504 | 1768 | .llvm_name = "scalar-atomics", |
| 1505 | 1769 | .description = "Has atomic scalar memory instructions", |
| ... | ... | @@ -1564,16 +1828,23 @@ pub const all_features = blk: { |
| 1564 | 1828 | .cvt_pknorm_vop2_insts, |
| 1565 | 1829 | .default_component_zero, |
| 1566 | 1830 | .ds_src2_insts, |
| 1831 | .dx10_clamp_and_ieee_mode, | |
| 1567 | 1832 | .extended_image_insts, |
| 1833 | .formatted_mubuf_insts, | |
| 1568 | 1834 | .fp64, |
| 1569 | 1835 | .gds, |
| 1570 | 1836 | .gfx7_gfx8_gfx9_insts, |
| 1571 | 1837 | .gws, |
| 1572 | 1838 | .image_insts, |
| 1839 | .instcachelinesize64, | |
| 1573 | 1840 | .lerp_inst, |
| 1574 | 1841 | .mad_mac_f32_insts, |
| 1575 | 1842 | .mimg_r128, |
| 1576 | 1843 | .movrel, |
| 1844 | .mqsad_insts, | |
| 1845 | .mqsad_pk_insts, | |
| 1846 | .msad_insts, | |
| 1847 | .mtbuf_insts, | |
| 1577 | 1848 | .qsad_insts, |
| 1578 | 1849 | .s_memtime_inst, |
| 1579 | 1850 | .sad_insts, |
| ... | ... | @@ -1613,6 +1884,11 @@ pub const all_features = blk: { |
| 1613 | 1884 | .description = "Enable SI Machine Scheduler", |
| 1614 | 1885 | .dependencies = featureSet(&[_]Feature{}), |
| 1615 | 1886 | }; |
| 1887 | result[@backingInt(Feature.smem_prefetch_insts)] = .{ | |
| 1888 | .llvm_name = "smem-prefetch-insts", | |
| 1889 | .description = "Has s_prefetch_inst, s_prefetch_inst_pc_rel, s_prefetch_data and s_prefetch_data_pc_rel instructions", | |
| 1890 | .dependencies = featureSet(&[_]Feature{}), | |
| 1891 | }; | |
| 1616 | 1892 | result[@backingInt(Feature.smem_to_vector_write_hazard)] = .{ |
| 1617 | 1893 | .llvm_name = "smem-to-vector-write-hazard", |
| 1618 | 1894 | .description = "s_load_dword followed by v_cmp page faults", |
| ... | ... | @@ -1629,16 +1905,22 @@ pub const all_features = blk: { |
| 1629 | 1905 | .cvt_pknorm_vop2_insts, |
| 1630 | 1906 | .default_component_zero, |
| 1631 | 1907 | .ds_src2_insts, |
| 1908 | .dx10_clamp_and_ieee_mode, | |
| 1632 | 1909 | .extended_image_insts, |
| 1910 | .formatted_mubuf_insts, | |
| 1633 | 1911 | .fp64, |
| 1634 | 1912 | .gds, |
| 1635 | 1913 | .gws, |
| 1636 | 1914 | .image_insts, |
| 1915 | .instcachelinesize64, | |
| 1637 | 1916 | .ldsbankcount32, |
| 1638 | 1917 | .lerp_inst, |
| 1639 | 1918 | .mad_mac_f32_insts, |
| 1640 | 1919 | .mimg_r128, |
| 1641 | 1920 | .movrel, |
| 1921 | .mqsad_pk_insts, | |
| 1922 | .msad_insts, | |
| 1923 | .mtbuf_insts, | |
| 1642 | 1924 | .s_memtime_inst, |
| 1643 | 1925 | .sad_insts, |
| 1644 | 1926 | .trig_reduced_range, |
| ... | ... | @@ -1656,6 +1938,16 @@ pub const all_features = blk: { |
| 1656 | 1938 | .description = "Hardware supports SRAMECC", |
| 1657 | 1939 | .dependencies = featureSet(&[_]Feature{}), |
| 1658 | 1940 | }; |
| 1941 | result[@backingInt(Feature.swmmac_gfx1200_insts)] = .{ | |
| 1942 | .llvm_name = "swmmac-gfx1200-insts", | |
| 1943 | .description = "Has GFX1200 SWMMAC instructions", | |
| 1944 | .dependencies = featureSet(&[_]Feature{}), | |
| 1945 | }; | |
| 1946 | result[@backingInt(Feature.swmmac_gfx1250_insts)] = .{ | |
| 1947 | .llvm_name = "swmmac-gfx1250-insts", | |
| 1948 | .description = "Has GFX1250 SWMMAC instructions", | |
| 1949 | .dependencies = featureSet(&[_]Feature{}), | |
| 1950 | }; | |
| 1659 | 1951 | result[@backingInt(Feature.tanh_insts)] = .{ |
| 1660 | 1952 | .llvm_name = "tanh-insts", |
| 1661 | 1953 | .description = "Has v_tanh_f32/f16 instructions", |
| ... | ... | @@ -1666,9 +1958,14 @@ pub const all_features = blk: { |
| 1666 | 1958 | .description = "Has v_perm_pk16* instructions", |
| 1667 | 1959 | .dependencies = featureSet(&[_]Feature{}), |
| 1668 | 1960 | }; |
| 1669 | result[@backingInt(Feature.tgsplit)] = .{ | |
| 1670 | .llvm_name = "tgsplit", | |
| 1671 | .description = "Enable threadgroup split execution", | |
| 1961 | result[@backingInt(Feature.tgsplit_support)] = .{ | |
| 1962 | .llvm_name = "tgsplit-support", | |
| 1963 | .description = "Hardware support for threadgroup split execution mode", | |
| 1964 | .dependencies = featureSet(&[_]Feature{}), | |
| 1965 | }; | |
| 1966 | result[@backingInt(Feature.trans_coexecution_hazard)] = .{ | |
| 1967 | .llvm_name = "trans-coexecution-hazard", | |
| 1968 | .description = "Hazard when TRANS co-executes with other VALU instruction", | |
| 1672 | 1969 | .dependencies = featureSet(&[_]Feature{}), |
| 1673 | 1970 | }; |
| 1674 | 1971 | result[@backingInt(Feature.transpose_load_f4f6_insts)] = .{ |
| ... | ... | @@ -1721,11 +2018,26 @@ pub const all_features = blk: { |
| 1721 | 2018 | .description = "Force using DS instruction immediate offsets on SI", |
| 1722 | 2019 | .dependencies = featureSet(&[_]Feature{}), |
| 1723 | 2020 | }; |
| 2021 | result[@backingInt(Feature.use_add_pc64_inst)] = .{ | |
| 2022 | .llvm_name = "use-add-pc64-inst", | |
| 2023 | .description = "Use s_add_pc_i64 instruction.", | |
| 2024 | .dependencies = featureSet(&[_]Feature{}), | |
| 2025 | }; | |
| 1724 | 2026 | result[@backingInt(Feature.user_sgpr_init16_bug)] = .{ |
| 1725 | 2027 | .llvm_name = "user-sgpr-init16-bug", |
| 1726 | 2028 | .description = "Bug requiring at least 16 user+system SGPRs to be enabled", |
| 1727 | 2029 | .dependencies = featureSet(&[_]Feature{}), |
| 1728 | 2030 | }; |
| 2031 | result[@backingInt(Feature.v_mov_b64_inst)] = .{ | |
| 2032 | .llvm_name = "v-mov-b64-inst", | |
| 2033 | .description = "Has v_mov_b64 instruction", | |
| 2034 | .dependencies = featureSet(&[_]Feature{}), | |
| 2035 | }; | |
| 2036 | result[@backingInt(Feature.v_mul_u64_inst)] = .{ | |
| 2037 | .llvm_name = "v-mul-u64-inst", | |
| 2038 | .description = "Has v_mul_u64 instruction.", | |
| 2039 | .dependencies = featureSet(&[_]Feature{}), | |
| 2040 | }; | |
| 1729 | 2041 | result[@backingInt(Feature.valu_trans_use_hazard)] = .{ |
| 1730 | 2042 | .llvm_name = "valu-trans-use-hazard", |
| 1731 | 2043 | .description = "Hazard when TRANS instructions are closely followed by a use of the result", |
| ... | ... | @@ -1758,7 +2070,7 @@ pub const all_features = blk: { |
| 1758 | 2070 | }; |
| 1759 | 2071 | result[@backingInt(Feature.vmem_to_lds_load_insts)] = .{ |
| 1760 | 2072 | .llvm_name = "vmem-to-lds-load-insts", |
| 1761 | .description = "The platform has memory to lds instructions (global_load w/lds bit set, buffer_load w/lds bit set or global_load_lds. This does not include scratch_load_lds.", | |
| 2073 | .description = "The platform has memory to lds instructions (global_load w/lds bit set, buffer_loadw/lds bit set or global_load_lds. This does not include scratch_load_lds.", | |
| 1762 | 2074 | .dependencies = featureSet(&[_]Feature{}), |
| 1763 | 2075 | }; |
| 1764 | 2076 | result[@backingInt(Feature.vmem_to_scalar_write_hazard)] = .{ |
| ... | ... | @@ -1782,10 +2094,14 @@ pub const all_features = blk: { |
| 1782 | 2094 | .cvt_pknorm_vop2_insts, |
| 1783 | 2095 | .default_component_zero, |
| 1784 | 2096 | .dpp, |
| 2097 | .dpp_row_bcast, | |
| 2098 | .dpp_wavefront_shifts, | |
| 1785 | 2099 | .ds_src2_insts, |
| 2100 | .dx10_clamp_and_ieee_mode, | |
| 1786 | 2101 | .extended_image_insts, |
| 1787 | 2102 | .fast_denormal_f32, |
| 1788 | 2103 | .flat_address_space, |
| 2104 | .formatted_mubuf_insts, | |
| 1789 | 2105 | .fp64, |
| 1790 | 2106 | .gcn3_encoding, |
| 1791 | 2107 | .gds, |
| ... | ... | @@ -1793,12 +2109,17 @@ pub const all_features = blk: { |
| 1793 | 2109 | .gfx8_insts, |
| 1794 | 2110 | .gws, |
| 1795 | 2111 | .image_insts, |
| 2112 | .instcachelinesize64, | |
| 1796 | 2113 | .int_clamp_insts, |
| 1797 | 2114 | .inv_2pi_inline_imm, |
| 1798 | 2115 | .lerp_inst, |
| 1799 | 2116 | .mad_mac_f32_insts, |
| 1800 | 2117 | .mimg_r128, |
| 1801 | 2118 | .movrel, |
| 2119 | .mqsad_insts, | |
| 2120 | .mqsad_pk_insts, | |
| 2121 | .msad_insts, | |
| 2122 | .mtbuf_insts, | |
| 1802 | 2123 | .qsad_insts, |
| 1803 | 2124 | .s_memrealtime, |
| 1804 | 2125 | .s_memtime_inst, |
| ... | ... | @@ -1824,6 +2145,11 @@ pub const all_features = blk: { |
| 1824 | 2145 | .description = "Has VOP3P packed instructions", |
| 1825 | 2146 | .dependencies = featureSet(&[_]Feature{}), |
| 1826 | 2147 | }; |
| 2148 | result[@backingInt(Feature.vop3px2_increments_va_vdst_twice)] = .{ | |
| 2149 | .llvm_name = "vop3px2-increments-va-vdst-twice", | |
| 2150 | .description = "VOP3PX2 encoded instructions increment the VaVdst counter twice", | |
| 2151 | .dependencies = featureSet(&[_]Feature{}), | |
| 2152 | }; | |
| 1827 | 2153 | result[@backingInt(Feature.vopd)] = .{ |
| 1828 | 2154 | .llvm_name = "vopd", |
| 1829 | 2155 | .description = "Has VOPD dual issue wave32 instructions", |
| ... | ... | @@ -1859,6 +2185,21 @@ pub const all_features = blk: { |
| 1859 | 2185 | .description = "The number of threads per wavefront", |
| 1860 | 2186 | .dependencies = featureSet(&[_]Feature{}), |
| 1861 | 2187 | }; |
| 2188 | result[@backingInt(Feature.wmma_128b_insts)] = .{ | |
| 2189 | .llvm_name = "wmma-128b-insts", | |
| 2190 | .description = "Has WMMA instructions where A and B matrices do not have duplicated data", | |
| 2191 | .dependencies = featureSet(&[_]Feature{}), | |
| 2192 | }; | |
| 2193 | result[@backingInt(Feature.wmma_256b_insts)] = .{ | |
| 2194 | .llvm_name = "wmma-256b-insts", | |
| 2195 | .description = "Has WMMA instructions where A and B matrices have duplicated data", | |
| 2196 | .dependencies = featureSet(&[_]Feature{}), | |
| 2197 | }; | |
| 2198 | result[@backingInt(Feature.wmma_coexecution_hazards)] = .{ | |
| 2199 | .llvm_name = "wmma-coexecution-hazards", | |
| 2200 | .description = "Hazards when WMMA co-executes with other WMMA/VALU instruction", | |
| 2201 | .dependencies = featureSet(&[_]Feature{}), | |
| 2202 | }; | |
| 1862 | 2203 | result[@backingInt(Feature.xf32_insts)] = .{ |
| 1863 | 2204 | .llvm_name = "xf32-insts", |
| 1864 | 2205 | .description = "Has instructions that support xf32 format, such as v_mfma_f32_16x16x8_xf32 and v_mfma_f32_32x32x4_xf32", |
| ... | ... | @@ -1867,7 +2208,16 @@ pub const all_features = blk: { |
| 1867 | 2208 | result[@backingInt(Feature.xnack)] = .{ |
| 1868 | 2209 | .llvm_name = "xnack", |
| 1869 | 2210 | .description = "Enable XNACK support", |
| 1870 | .dependencies = featureSet(&[_]Feature{}), | |
| 2211 | .dependencies = featureSet(&[_]Feature{ | |
| 2212 | .xnack_support, | |
| 2213 | }), | |
| 2214 | }; | |
| 2215 | result[@backingInt(Feature.xnack_on_off_modes)] = .{ | |
| 2216 | .llvm_name = "xnack-on-off-modes", | |
| 2217 | .description = "Target supports XNACK on/off modes", | |
| 2218 | .dependencies = featureSet(&[_]Feature{ | |
| 2219 | .xnack_support, | |
| 2220 | }), | |
| 1871 | 2221 | }; |
| 1872 | 2222 | result[@backingInt(Feature.xnack_support)] = .{ |
| 1873 | 2223 | .llvm_name = "xnack-support", |
| ... | ... | @@ -1900,7 +2250,7 @@ pub const cpu = struct { |
| 1900 | 2250 | .ldsbankcount32, |
| 1901 | 2251 | .unpacked_d16_vmem, |
| 1902 | 2252 | .volcanic_islands, |
| 1903 | .xnack_support, | |
| 2253 | .xnack_on_off_modes, | |
| 1904 | 2254 | }), |
| 1905 | 2255 | }; |
| 1906 | 2256 | pub const fiji: CpuModel = .{ |
| ... | ... | @@ -1951,7 +2301,7 @@ pub const cpu = struct { |
| 1951 | 2301 | .vcmpx_exec_war_hazard, |
| 1952 | 2302 | .vcmpx_permlane_hazard, |
| 1953 | 2303 | .vmem_to_scalar_write_hazard, |
| 1954 | .xnack_support, | |
| 2304 | .xnack_on_off_modes, | |
| 1955 | 2305 | }), |
| 1956 | 2306 | }; |
| 1957 | 2307 | pub const gfx1011: CpuModel = .{ |
| ... | ... | @@ -1987,7 +2337,7 @@ pub const cpu = struct { |
| 1987 | 2337 | .vcmpx_exec_war_hazard, |
| 1988 | 2338 | .vcmpx_permlane_hazard, |
| 1989 | 2339 | .vmem_to_scalar_write_hazard, |
| 1990 | .xnack_support, | |
| 2340 | .xnack_on_off_modes, | |
| 1991 | 2341 | }), |
| 1992 | 2342 | }; |
| 1993 | 2343 | pub const gfx1012: CpuModel = .{ |
| ... | ... | @@ -2023,7 +2373,7 @@ pub const cpu = struct { |
| 2023 | 2373 | .vcmpx_exec_war_hazard, |
| 2024 | 2374 | .vcmpx_permlane_hazard, |
| 2025 | 2375 | .vmem_to_scalar_write_hazard, |
| 2026 | .xnack_support, | |
| 2376 | .xnack_on_off_modes, | |
| 2027 | 2377 | }), |
| 2028 | 2378 | }; |
| 2029 | 2379 | pub const gfx1013: CpuModel = .{ |
| ... | ... | @@ -2054,7 +2404,7 @@ pub const cpu = struct { |
| 2054 | 2404 | .vcmpx_exec_war_hazard, |
| 2055 | 2405 | .vcmpx_permlane_hazard, |
| 2056 | 2406 | .vmem_to_scalar_write_hazard, |
| 2057 | .xnack_support, | |
| 2407 | .xnack_on_off_modes, | |
| 2058 | 2408 | }), |
| 2059 | 2409 | }; |
| 2060 | 2410 | pub const gfx1030: CpuModel = .{ |
| ... | ... | @@ -2232,7 +2582,7 @@ pub const cpu = struct { |
| 2232 | 2582 | .vcmpx_exec_war_hazard, |
| 2233 | 2583 | .vcmpx_permlane_hazard, |
| 2234 | 2584 | .vmem_to_scalar_write_hazard, |
| 2235 | .xnack_support, | |
| 2585 | .xnack_on_off_modes, | |
| 2236 | 2586 | }), |
| 2237 | 2587 | }; |
| 2238 | 2588 | pub const gfx10_3_generic: CpuModel = .{ |
| ... | ... | @@ -2261,7 +2611,7 @@ pub const cpu = struct { |
| 2261 | 2611 | .name = "gfx1100", |
| 2262 | 2612 | .llvm_name = "gfx1100", |
| 2263 | 2613 | .features = featureSet(&[_]Feature{ |
| 2264 | .allocate1_5xvgprs, | |
| 2614 | .@"1536_physical_vgprs", | |
| 2265 | 2615 | .architected_flat_scratch, |
| 2266 | 2616 | .atomic_fadd_no_rtn_insts, |
| 2267 | 2617 | .atomic_fadd_rtn_insts, |
| ... | ... | @@ -2274,6 +2624,7 @@ pub const cpu = struct { |
| 2274 | 2624 | .dot7_insts, |
| 2275 | 2625 | .dot8_insts, |
| 2276 | 2626 | .dot9_insts, |
| 2627 | .dx10_clamp_and_ieee_mode, | |
| 2277 | 2628 | .flat_atomic_fadd_f32_inst, |
| 2278 | 2629 | .gfx11, |
| 2279 | 2630 | .image_insts, |
| ... | ... | @@ -2290,13 +2641,14 @@ pub const cpu = struct { |
| 2290 | 2641 | .user_sgpr_init16_bug, |
| 2291 | 2642 | .valu_trans_use_hazard, |
| 2292 | 2643 | .vcmpx_permlane_hazard, |
| 2644 | .wmma_256b_insts, | |
| 2293 | 2645 | }), |
| 2294 | 2646 | }; |
| 2295 | 2647 | pub const gfx1101: CpuModel = .{ |
| 2296 | 2648 | .name = "gfx1101", |
| 2297 | 2649 | .llvm_name = "gfx1101", |
| 2298 | 2650 | .features = featureSet(&[_]Feature{ |
| 2299 | .allocate1_5xvgprs, | |
| 2651 | .@"1536_physical_vgprs", | |
| 2300 | 2652 | .architected_flat_scratch, |
| 2301 | 2653 | .atomic_fadd_no_rtn_insts, |
| 2302 | 2654 | .atomic_fadd_rtn_insts, |
| ... | ... | @@ -2309,6 +2661,7 @@ pub const cpu = struct { |
| 2309 | 2661 | .dot7_insts, |
| 2310 | 2662 | .dot8_insts, |
| 2311 | 2663 | .dot9_insts, |
| 2664 | .dx10_clamp_and_ieee_mode, | |
| 2312 | 2665 | .flat_atomic_fadd_f32_inst, |
| 2313 | 2666 | .gfx11, |
| 2314 | 2667 | .image_insts, |
| ... | ... | @@ -2324,6 +2677,7 @@ pub const cpu = struct { |
| 2324 | 2677 | .shader_cycles_register, |
| 2325 | 2678 | .valu_trans_use_hazard, |
| 2326 | 2679 | .vcmpx_permlane_hazard, |
| 2680 | .wmma_256b_insts, | |
| 2327 | 2681 | }), |
| 2328 | 2682 | }; |
| 2329 | 2683 | pub const gfx1102: CpuModel = .{ |
| ... | ... | @@ -2342,6 +2696,7 @@ pub const cpu = struct { |
| 2342 | 2696 | .dot7_insts, |
| 2343 | 2697 | .dot8_insts, |
| 2344 | 2698 | .dot9_insts, |
| 2699 | .dx10_clamp_and_ieee_mode, | |
| 2345 | 2700 | .flat_atomic_fadd_f32_inst, |
| 2346 | 2701 | .gfx11, |
| 2347 | 2702 | .image_insts, |
| ... | ... | @@ -2358,6 +2713,7 @@ pub const cpu = struct { |
| 2358 | 2713 | .user_sgpr_init16_bug, |
| 2359 | 2714 | .valu_trans_use_hazard, |
| 2360 | 2715 | .vcmpx_permlane_hazard, |
| 2716 | .wmma_256b_insts, | |
| 2361 | 2717 | }), |
| 2362 | 2718 | }; |
| 2363 | 2719 | pub const gfx1103: CpuModel = .{ |
| ... | ... | @@ -2376,6 +2732,7 @@ pub const cpu = struct { |
| 2376 | 2732 | .dot7_insts, |
| 2377 | 2733 | .dot8_insts, |
| 2378 | 2734 | .dot9_insts, |
| 2735 | .dx10_clamp_and_ieee_mode, | |
| 2379 | 2736 | .flat_atomic_fadd_f32_inst, |
| 2380 | 2737 | .gfx11, |
| 2381 | 2738 | .image_insts, |
| ... | ... | @@ -2391,6 +2748,7 @@ pub const cpu = struct { |
| 2391 | 2748 | .shader_cycles_register, |
| 2392 | 2749 | .valu_trans_use_hazard, |
| 2393 | 2750 | .vcmpx_permlane_hazard, |
| 2751 | .wmma_256b_insts, | |
| 2394 | 2752 | }), |
| 2395 | 2753 | }; |
| 2396 | 2754 | pub const gfx1150: CpuModel = .{ |
| ... | ... | @@ -2410,6 +2768,7 @@ pub const cpu = struct { |
| 2410 | 2768 | .dot8_insts, |
| 2411 | 2769 | .dot9_insts, |
| 2412 | 2770 | .dpp_src1_sgpr, |
| 2771 | .dx10_clamp_and_ieee_mode, | |
| 2413 | 2772 | .flat_atomic_fadd_f32_inst, |
| 2414 | 2773 | .gfx11, |
| 2415 | 2774 | .image_insts, |
| ... | ... | @@ -2424,13 +2783,14 @@ pub const cpu = struct { |
| 2424 | 2783 | .salu_float, |
| 2425 | 2784 | .shader_cycles_register, |
| 2426 | 2785 | .vcmpx_permlane_hazard, |
| 2786 | .wmma_256b_insts, | |
| 2427 | 2787 | }), |
| 2428 | 2788 | }; |
| 2429 | 2789 | pub const gfx1151: CpuModel = .{ |
| 2430 | 2790 | .name = "gfx1151", |
| 2431 | 2791 | .llvm_name = "gfx1151", |
| 2432 | 2792 | .features = featureSet(&[_]Feature{ |
| 2433 | .allocate1_5xvgprs, | |
| 2793 | .@"1536_physical_vgprs", | |
| 2434 | 2794 | .architected_flat_scratch, |
| 2435 | 2795 | .atomic_fadd_no_rtn_insts, |
| 2436 | 2796 | .atomic_fadd_rtn_insts, |
| ... | ... | @@ -2444,6 +2804,7 @@ pub const cpu = struct { |
| 2444 | 2804 | .dot8_insts, |
| 2445 | 2805 | .dot9_insts, |
| 2446 | 2806 | .dpp_src1_sgpr, |
| 2807 | .dx10_clamp_and_ieee_mode, | |
| 2447 | 2808 | .flat_atomic_fadd_f32_inst, |
| 2448 | 2809 | .gfx11, |
| 2449 | 2810 | .image_insts, |
| ... | ... | @@ -2458,6 +2819,7 @@ pub const cpu = struct { |
| 2458 | 2819 | .salu_float, |
| 2459 | 2820 | .shader_cycles_register, |
| 2460 | 2821 | .vcmpx_permlane_hazard, |
| 2822 | .wmma_256b_insts, | |
| 2461 | 2823 | }), |
| 2462 | 2824 | }; |
| 2463 | 2825 | pub const gfx1152: CpuModel = .{ |
| ... | ... | @@ -2477,6 +2839,7 @@ pub const cpu = struct { |
| 2477 | 2839 | .dot8_insts, |
| 2478 | 2840 | .dot9_insts, |
| 2479 | 2841 | .dpp_src1_sgpr, |
| 2842 | .dx10_clamp_and_ieee_mode, | |
| 2480 | 2843 | .flat_atomic_fadd_f32_inst, |
| 2481 | 2844 | .gfx11, |
| 2482 | 2845 | .image_insts, |
| ... | ... | @@ -2491,6 +2854,7 @@ pub const cpu = struct { |
| 2491 | 2854 | .salu_float, |
| 2492 | 2855 | .shader_cycles_register, |
| 2493 | 2856 | .vcmpx_permlane_hazard, |
| 2857 | .wmma_256b_insts, | |
| 2494 | 2858 | }), |
| 2495 | 2859 | }; |
| 2496 | 2860 | pub const gfx1153: CpuModel = .{ |
| ... | ... | @@ -2510,6 +2874,7 @@ pub const cpu = struct { |
| 2510 | 2874 | .dot8_insts, |
| 2511 | 2875 | .dot9_insts, |
| 2512 | 2876 | .dpp_src1_sgpr, |
| 2877 | .dx10_clamp_and_ieee_mode, | |
| 2513 | 2878 | .flat_atomic_fadd_f32_inst, |
| 2514 | 2879 | .gfx11, |
| 2515 | 2880 | .image_insts, |
| ... | ... | @@ -2523,11 +2888,12 @@ pub const cpu = struct { |
| 2523 | 2888 | .salu_float, |
| 2524 | 2889 | .shader_cycles_register, |
| 2525 | 2890 | .vcmpx_permlane_hazard, |
| 2891 | .wmma_256b_insts, | |
| 2526 | 2892 | }), |
| 2527 | 2893 | }; |
| 2528 | pub const gfx11_generic: CpuModel = .{ | |
| 2529 | .name = "gfx11_generic", | |
| 2530 | .llvm_name = "gfx11-generic", | |
| 2894 | pub const gfx1154: CpuModel = .{ | |
| 2895 | .name = "gfx1154", | |
| 2896 | .llvm_name = "gfx1154", | |
| 2531 | 2897 | .features = featureSet(&[_]Feature{ |
| 2532 | 2898 | .architected_flat_scratch, |
| 2533 | 2899 | .atomic_fadd_no_rtn_insts, |
| ... | ... | @@ -2541,47 +2907,32 @@ pub const cpu = struct { |
| 2541 | 2907 | .dot7_insts, |
| 2542 | 2908 | .dot8_insts, |
| 2543 | 2909 | .dot9_insts, |
| 2910 | .dpp_src1_sgpr, | |
| 2911 | .dx10_clamp_and_ieee_mode, | |
| 2544 | 2912 | .flat_atomic_fadd_f32_inst, |
| 2545 | 2913 | .gfx11, |
| 2546 | 2914 | .image_insts, |
| 2547 | 2915 | .ldsbankcount32, |
| 2548 | .mad_intra_fwd_bug, | |
| 2549 | 2916 | .memory_atomic_fadd_f32_denormal_support, |
| 2550 | .msaa_load_dst_sel_bug, | |
| 2551 | 2917 | .nsa_encoding, |
| 2552 | 2918 | .packed_tid, |
| 2553 | 2919 | .partial_nsa_encoding, |
| 2554 | .priv_enabled_trap2_nop_bug, | |
| 2555 | 2920 | .real_true16, |
| 2556 | 2921 | .required_export_priority, |
| 2557 | .requires_cov6, | |
| 2922 | .salu_float, | |
| 2558 | 2923 | .shader_cycles_register, |
| 2559 | .user_sgpr_init16_bug, | |
| 2560 | .valu_trans_use_hazard, | |
| 2561 | 2924 | .vcmpx_permlane_hazard, |
| 2925 | .wmma_256b_insts, | |
| 2562 | 2926 | }), |
| 2563 | 2927 | }; |
| 2564 | pub const gfx1200: CpuModel = .{ | |
| 2565 | .name = "gfx1200", | |
| 2566 | .llvm_name = "gfx1200", | |
| 2928 | pub const gfx1170: CpuModel = .{ | |
| 2929 | .name = "gfx1170", | |
| 2930 | .llvm_name = "gfx1170", | |
| 2567 | 2931 | .features = featureSet(&[_]Feature{ |
| 2568 | .addressablelocalmemorysize65536, | |
| 2569 | .allocate1_5xvgprs, | |
| 2570 | 2932 | .architected_flat_scratch, |
| 2571 | .architected_sgprs, | |
| 2572 | .atomic_buffer_global_pk_add_f16_insts, | |
| 2573 | .atomic_buffer_pk_add_bf16_inst, | |
| 2574 | .atomic_ds_pk_add_16_insts, | |
| 2575 | 2933 | .atomic_fadd_no_rtn_insts, |
| 2576 | 2934 | .atomic_fadd_rtn_insts, |
| 2577 | .atomic_flat_pk_add_16_insts, | |
| 2578 | .atomic_global_pk_add_bf16_inst, | |
| 2579 | 2935 | .back_off_barrier, |
| 2580 | .bvh_dual_bvh_8_insts, | |
| 2581 | .cube_insts, | |
| 2582 | .cvt_norm_insts, | |
| 2583 | .cvt_pknorm_vop2_insts, | |
| 2584 | .cvt_pknorm_vop3_insts, | |
| 2585 | 2936 | .d16_write_vgpr32, |
| 2586 | 2937 | .dl_insts, |
| 2587 | 2938 | .dot10_insts, |
| ... | ... | @@ -2591,49 +2942,36 @@ pub const cpu = struct { |
| 2591 | 2942 | .dot8_insts, |
| 2592 | 2943 | .dot9_insts, |
| 2593 | 2944 | .dpp_src1_sgpr, |
| 2594 | .extended_image_insts, | |
| 2595 | 2945 | .flat_atomic_fadd_f32_inst, |
| 2596 | 2946 | .fp8_conversion_insts, |
| 2597 | .gfx12, | |
| 2947 | .gfx11, | |
| 2948 | .gfx11_7_insts, | |
| 2949 | .ieee_minimum_maximum_insts, | |
| 2598 | 2950 | .image_insts, |
| 2599 | 2951 | .ldsbankcount32, |
| 2600 | .lerp_inst, | |
| 2601 | 2952 | .memory_atomic_fadd_f32_denormal_support, |
| 2953 | .minimum3_maximum3_f16, | |
| 2954 | .minimum3_maximum3_f32, | |
| 2955 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 2602 | 2956 | .nsa_encoding, |
| 2603 | 2957 | .packed_tid, |
| 2604 | 2958 | .partial_nsa_encoding, |
| 2605 | .pseudo_scalar_trans, | |
| 2606 | .qsad_insts, | |
| 2607 | .restricted_soffset, | |
| 2608 | .sad_insts, | |
| 2959 | .real_true16, | |
| 2609 | 2960 | .salu_float, |
| 2610 | .scalar_dwordx3_loads, | |
| 2611 | .shader_cycles_hi_lo_registers, | |
| 2961 | .shader_cycles_register, | |
| 2962 | .swmmac_gfx1200_insts, | |
| 2612 | 2963 | .vcmpx_permlane_hazard, |
| 2613 | .waits_before_system_scope_stores, | |
| 2964 | .wmma_128b_insts, | |
| 2614 | 2965 | }), |
| 2615 | 2966 | }; |
| 2616 | pub const gfx1201: CpuModel = .{ | |
| 2617 | .name = "gfx1201", | |
| 2618 | .llvm_name = "gfx1201", | |
| 2967 | pub const gfx1171: CpuModel = .{ | |
| 2968 | .name = "gfx1171", | |
| 2969 | .llvm_name = "gfx1171", | |
| 2619 | 2970 | .features = featureSet(&[_]Feature{ |
| 2620 | .addressablelocalmemorysize65536, | |
| 2621 | .allocate1_5xvgprs, | |
| 2622 | 2971 | .architected_flat_scratch, |
| 2623 | .architected_sgprs, | |
| 2624 | .atomic_buffer_global_pk_add_f16_insts, | |
| 2625 | .atomic_buffer_pk_add_bf16_inst, | |
| 2626 | .atomic_ds_pk_add_16_insts, | |
| 2627 | 2972 | .atomic_fadd_no_rtn_insts, |
| 2628 | 2973 | .atomic_fadd_rtn_insts, |
| 2629 | .atomic_flat_pk_add_16_insts, | |
| 2630 | .atomic_global_pk_add_bf16_inst, | |
| 2631 | 2974 | .back_off_barrier, |
| 2632 | .bvh_dual_bvh_8_insts, | |
| 2633 | .cube_insts, | |
| 2634 | .cvt_norm_insts, | |
| 2635 | .cvt_pknorm_vop2_insts, | |
| 2636 | .cvt_pknorm_vop3_insts, | |
| 2637 | 2975 | .d16_write_vgpr32, |
| 2638 | 2976 | .dl_insts, |
| 2639 | 2977 | .dot10_insts, |
| ... | ... | @@ -2643,61 +2981,304 @@ pub const cpu = struct { |
| 2643 | 2981 | .dot8_insts, |
| 2644 | 2982 | .dot9_insts, |
| 2645 | 2983 | .dpp_src1_sgpr, |
| 2646 | .extended_image_insts, | |
| 2647 | 2984 | .flat_atomic_fadd_f32_inst, |
| 2648 | 2985 | .fp8_conversion_insts, |
| 2649 | .gfx12, | |
| 2986 | .gfx11, | |
| 2987 | .gfx11_7_insts, | |
| 2988 | .ieee_minimum_maximum_insts, | |
| 2650 | 2989 | .image_insts, |
| 2651 | 2990 | .ldsbankcount32, |
| 2652 | .lerp_inst, | |
| 2653 | 2991 | .memory_atomic_fadd_f32_denormal_support, |
| 2992 | .minimum3_maximum3_f16, | |
| 2993 | .minimum3_maximum3_f32, | |
| 2994 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 2654 | 2995 | .nsa_encoding, |
| 2655 | 2996 | .packed_tid, |
| 2656 | 2997 | .partial_nsa_encoding, |
| 2657 | .pseudo_scalar_trans, | |
| 2658 | .qsad_insts, | |
| 2659 | .restricted_soffset, | |
| 2660 | .sad_insts, | |
| 2998 | .real_true16, | |
| 2661 | 2999 | .salu_float, |
| 2662 | .scalar_dwordx3_loads, | |
| 2663 | .shader_cycles_hi_lo_registers, | |
| 3000 | .shader_cycles_register, | |
| 3001 | .swmmac_gfx1200_insts, | |
| 2664 | 3002 | .vcmpx_permlane_hazard, |
| 2665 | .waits_before_system_scope_stores, | |
| 3003 | .wmma_128b_insts, | |
| 2666 | 3004 | }), |
| 2667 | 3005 | }; |
| 2668 | pub const gfx1250: CpuModel = .{ | |
| 2669 | .name = "gfx1250", | |
| 2670 | .llvm_name = "gfx1250", | |
| 3006 | pub const gfx1172: CpuModel = .{ | |
| 3007 | .name = "gfx1172", | |
| 3008 | .llvm_name = "gfx1172", | |
| 2671 | 3009 | .features = featureSet(&[_]Feature{ |
| 2672 | .@"1024_addressable_vgprs", | |
| 2673 | .@"45_bit_num_records_buffer_resource", | |
| 2674 | .@"64_bit_literals", | |
| 2675 | .add_min_max_insts, | |
| 2676 | .add_sub_u64_insts, | |
| 2677 | .addressablelocalmemorysize327680, | |
| 2678 | 3010 | .architected_flat_scratch, |
| 2679 | .architected_sgprs, | |
| 2680 | .ashr_pk_insts, | |
| 2681 | .atomic_buffer_global_pk_add_f16_insts, | |
| 2682 | .atomic_buffer_pk_add_bf16_inst, | |
| 2683 | .atomic_ds_pk_add_16_insts, | |
| 2684 | 3011 | .atomic_fadd_no_rtn_insts, |
| 2685 | 3012 | .atomic_fadd_rtn_insts, |
| 2686 | .atomic_flat_pk_add_16_insts, | |
| 2687 | .atomic_fmin_fmax_flat_f64, | |
| 2688 | .atomic_fmin_fmax_global_f64, | |
| 2689 | .atomic_global_pk_add_bf16_inst, | |
| 2690 | .bf16_cvt_insts, | |
| 2691 | .bf16_pk_insts, | |
| 2692 | .bf16_trans_insts, | |
| 2693 | .bitop3_insts, | |
| 2694 | .clusters, | |
| 2695 | .cube_insts, | |
| 2696 | .cumode, | |
| 2697 | .cvt_norm_insts, | |
| 2698 | .cvt_pk_f16_f32_inst, | |
| 2699 | .cvt_pknorm_vop2_insts, | |
| 2700 | .cvt_pknorm_vop3_insts, | |
| 3013 | .back_off_barrier, | |
| 3014 | .d16_write_vgpr32, | |
| 3015 | .dl_insts, | |
| 3016 | .dot10_insts, | |
| 3017 | .dot11_insts, | |
| 3018 | .dot12_insts, | |
| 3019 | .dot7_insts, | |
| 3020 | .dot8_insts, | |
| 3021 | .dot9_insts, | |
| 3022 | .dpp_src1_sgpr, | |
| 3023 | .flat_atomic_fadd_f32_inst, | |
| 3024 | .fp8_conversion_insts, | |
| 3025 | .gfx11, | |
| 3026 | .gfx11_7_insts, | |
| 3027 | .ieee_minimum_maximum_insts, | |
| 3028 | .image_insts, | |
| 3029 | .ldsbankcount32, | |
| 3030 | .memory_atomic_fadd_f32_denormal_support, | |
| 3031 | .minimum3_maximum3_f16, | |
| 3032 | .minimum3_maximum3_f32, | |
| 3033 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 3034 | .nsa_encoding, | |
| 3035 | .packed_tid, | |
| 3036 | .partial_nsa_encoding, | |
| 3037 | .real_true16, | |
| 3038 | .salu_float, | |
| 3039 | .shader_cycles_register, | |
| 3040 | .swmmac_gfx1200_insts, | |
| 3041 | .vcmpx_permlane_hazard, | |
| 3042 | .wmma_128b_insts, | |
| 3043 | }), | |
| 3044 | }; | |
| 3045 | pub const gfx11_7_generic: CpuModel = .{ | |
| 3046 | .name = "gfx11_7_generic", | |
| 3047 | .llvm_name = "gfx11-7-generic", | |
| 3048 | .features = featureSet(&[_]Feature{ | |
| 3049 | .architected_flat_scratch, | |
| 3050 | .atomic_fadd_no_rtn_insts, | |
| 3051 | .atomic_fadd_rtn_insts, | |
| 3052 | .back_off_barrier, | |
| 3053 | .d16_write_vgpr32, | |
| 3054 | .dl_insts, | |
| 3055 | .dot10_insts, | |
| 3056 | .dot11_insts, | |
| 3057 | .dot12_insts, | |
| 3058 | .dot7_insts, | |
| 3059 | .dot8_insts, | |
| 3060 | .dot9_insts, | |
| 3061 | .dpp_src1_sgpr, | |
| 3062 | .flat_atomic_fadd_f32_inst, | |
| 3063 | .fp8_conversion_insts, | |
| 3064 | .gfx11, | |
| 3065 | .gfx11_7_insts, | |
| 3066 | .ieee_minimum_maximum_insts, | |
| 3067 | .image_insts, | |
| 3068 | .ldsbankcount32, | |
| 3069 | .memory_atomic_fadd_f32_denormal_support, | |
| 3070 | .minimum3_maximum3_f16, | |
| 3071 | .minimum3_maximum3_f32, | |
| 3072 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 3073 | .nsa_encoding, | |
| 3074 | .packed_tid, | |
| 3075 | .partial_nsa_encoding, | |
| 3076 | .real_true16, | |
| 3077 | .requires_cov6, | |
| 3078 | .salu_float, | |
| 3079 | .shader_cycles_register, | |
| 3080 | .swmmac_gfx1200_insts, | |
| 3081 | .vcmpx_permlane_hazard, | |
| 3082 | .wmma_128b_insts, | |
| 3083 | }), | |
| 3084 | }; | |
| 3085 | pub const gfx11_generic: CpuModel = .{ | |
| 3086 | .name = "gfx11_generic", | |
| 3087 | .llvm_name = "gfx11-generic", | |
| 3088 | .features = featureSet(&[_]Feature{ | |
| 3089 | .architected_flat_scratch, | |
| 3090 | .atomic_fadd_no_rtn_insts, | |
| 3091 | .atomic_fadd_rtn_insts, | |
| 3092 | .back_off_barrier, | |
| 3093 | .d16_write_vgpr32, | |
| 3094 | .dl_insts, | |
| 3095 | .dot10_insts, | |
| 3096 | .dot12_insts, | |
| 3097 | .dot5_insts, | |
| 3098 | .dot7_insts, | |
| 3099 | .dot8_insts, | |
| 3100 | .dot9_insts, | |
| 3101 | .dx10_clamp_and_ieee_mode, | |
| 3102 | .flat_atomic_fadd_f32_inst, | |
| 3103 | .gfx11, | |
| 3104 | .image_insts, | |
| 3105 | .ldsbankcount32, | |
| 3106 | .mad_intra_fwd_bug, | |
| 3107 | .memory_atomic_fadd_f32_denormal_support, | |
| 3108 | .msaa_load_dst_sel_bug, | |
| 3109 | .nsa_encoding, | |
| 3110 | .packed_tid, | |
| 3111 | .partial_nsa_encoding, | |
| 3112 | .priv_enabled_trap2_nop_bug, | |
| 3113 | .real_true16, | |
| 3114 | .required_export_priority, | |
| 3115 | .requires_cov6, | |
| 3116 | .shader_cycles_register, | |
| 3117 | .user_sgpr_init16_bug, | |
| 3118 | .valu_trans_use_hazard, | |
| 3119 | .vcmpx_permlane_hazard, | |
| 3120 | .wmma_256b_insts, | |
| 3121 | }), | |
| 3122 | }; | |
| 3123 | pub const gfx1200: CpuModel = .{ | |
| 3124 | .name = "gfx1200", | |
| 3125 | .llvm_name = "gfx1200", | |
| 3126 | .features = featureSet(&[_]Feature{ | |
| 3127 | .@"1536_physical_vgprs", | |
| 3128 | .addressablelocalmemorysize65536, | |
| 3129 | .architected_flat_scratch, | |
| 3130 | .architected_sgprs, | |
| 3131 | .atomic_buffer_global_pk_add_f16_insts, | |
| 3132 | .atomic_buffer_pk_add_bf16_inst, | |
| 3133 | .atomic_ds_pk_add_16_insts, | |
| 3134 | .atomic_fadd_no_rtn_insts, | |
| 3135 | .atomic_fadd_rtn_insts, | |
| 3136 | .atomic_flat_pk_add_16_insts, | |
| 3137 | .atomic_global_pk_add_bf16_inst, | |
| 3138 | .back_off_barrier, | |
| 3139 | .bvh_dual_bvh_8_insts, | |
| 3140 | .cube_insts, | |
| 3141 | .cvt_norm_insts, | |
| 3142 | .cvt_pknorm_vop2_insts, | |
| 3143 | .cvt_pknorm_vop3_insts, | |
| 3144 | .d16_write_vgpr32, | |
| 3145 | .dl_insts, | |
| 3146 | .dot10_insts, | |
| 3147 | .dot11_insts, | |
| 3148 | .dot12_insts, | |
| 3149 | .dot7_insts, | |
| 3150 | .dot8_insts, | |
| 3151 | .dot9_insts, | |
| 3152 | .dpp_src1_sgpr, | |
| 3153 | .extended_image_insts, | |
| 3154 | .flat_atomic_fadd_f32_inst, | |
| 3155 | .formatted_mubuf_insts, | |
| 3156 | .fp8_conversion_insts, | |
| 3157 | .gfx12, | |
| 3158 | .image_insts, | |
| 3159 | .ldsbankcount32, | |
| 3160 | .lerp_inst, | |
| 3161 | .memory_atomic_fadd_f32_denormal_support, | |
| 3162 | .mqsad_insts, | |
| 3163 | .mqsad_pk_insts, | |
| 3164 | .msad_insts, | |
| 3165 | .mtbuf_insts, | |
| 3166 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 3167 | .nsa_encoding, | |
| 3168 | .packed_tid, | |
| 3169 | .partial_nsa_encoding, | |
| 3170 | .pseudo_scalar_trans, | |
| 3171 | .qsad_insts, | |
| 3172 | .real_true16, | |
| 3173 | .restricted_soffset, | |
| 3174 | .sad_insts, | |
| 3175 | .salu_float, | |
| 3176 | .scalar_dwordx3_loads, | |
| 3177 | .shader_cycles_hi_lo_registers, | |
| 3178 | .smem_prefetch_insts, | |
| 3179 | .swmmac_gfx1200_insts, | |
| 3180 | .vcmpx_permlane_hazard, | |
| 3181 | .waits_before_system_scope_stores, | |
| 3182 | .wmma_128b_insts, | |
| 3183 | }), | |
| 3184 | }; | |
| 3185 | pub const gfx1201: CpuModel = .{ | |
| 3186 | .name = "gfx1201", | |
| 3187 | .llvm_name = "gfx1201", | |
| 3188 | .features = featureSet(&[_]Feature{ | |
| 3189 | .@"1536_physical_vgprs", | |
| 3190 | .addressablelocalmemorysize65536, | |
| 3191 | .architected_flat_scratch, | |
| 3192 | .architected_sgprs, | |
| 3193 | .atomic_buffer_global_pk_add_f16_insts, | |
| 3194 | .atomic_buffer_pk_add_bf16_inst, | |
| 3195 | .atomic_ds_pk_add_16_insts, | |
| 3196 | .atomic_fadd_no_rtn_insts, | |
| 3197 | .atomic_fadd_rtn_insts, | |
| 3198 | .atomic_flat_pk_add_16_insts, | |
| 3199 | .atomic_global_pk_add_bf16_inst, | |
| 3200 | .back_off_barrier, | |
| 3201 | .bvh_dual_bvh_8_insts, | |
| 3202 | .cube_insts, | |
| 3203 | .cvt_norm_insts, | |
| 3204 | .cvt_pknorm_vop2_insts, | |
| 3205 | .cvt_pknorm_vop3_insts, | |
| 3206 | .d16_write_vgpr32, | |
| 3207 | .dl_insts, | |
| 3208 | .dot10_insts, | |
| 3209 | .dot11_insts, | |
| 3210 | .dot12_insts, | |
| 3211 | .dot7_insts, | |
| 3212 | .dot8_insts, | |
| 3213 | .dot9_insts, | |
| 3214 | .dpp_src1_sgpr, | |
| 3215 | .extended_image_insts, | |
| 3216 | .flat_atomic_fadd_f32_inst, | |
| 3217 | .formatted_mubuf_insts, | |
| 3218 | .fp8_conversion_insts, | |
| 3219 | .gfx12, | |
| 3220 | .image_insts, | |
| 3221 | .ldsbankcount32, | |
| 3222 | .lerp_inst, | |
| 3223 | .memory_atomic_fadd_f32_denormal_support, | |
| 3224 | .mqsad_insts, | |
| 3225 | .mqsad_pk_insts, | |
| 3226 | .msad_insts, | |
| 3227 | .mtbuf_insts, | |
| 3228 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 3229 | .nsa_encoding, | |
| 3230 | .packed_tid, | |
| 3231 | .partial_nsa_encoding, | |
| 3232 | .pseudo_scalar_trans, | |
| 3233 | .qsad_insts, | |
| 3234 | .real_true16, | |
| 3235 | .restricted_soffset, | |
| 3236 | .sad_insts, | |
| 3237 | .salu_float, | |
| 3238 | .scalar_dwordx3_loads, | |
| 3239 | .shader_cycles_hi_lo_registers, | |
| 3240 | .smem_prefetch_insts, | |
| 3241 | .swmmac_gfx1200_insts, | |
| 3242 | .vcmpx_permlane_hazard, | |
| 3243 | .waits_before_system_scope_stores, | |
| 3244 | .wmma_128b_insts, | |
| 3245 | }), | |
| 3246 | }; | |
| 3247 | pub const gfx1250: CpuModel = .{ | |
| 3248 | .name = "gfx1250", | |
| 3249 | .llvm_name = "gfx1250", | |
| 3250 | .features = featureSet(&[_]Feature{ | |
| 3251 | .@"1024_addressable_vgprs", | |
| 3252 | .@"45_bit_num_records_buffer_resource", | |
| 3253 | .@"64_bit_literals", | |
| 3254 | .add_min_max_insts, | |
| 3255 | .add_sub_u64_insts, | |
| 3256 | .addressablelocalmemorysize327680, | |
| 3257 | .architected_flat_scratch, | |
| 3258 | .architected_sgprs, | |
| 3259 | .ashr_pk_insts, | |
| 3260 | .asynccnt, | |
| 3261 | .atomic_buffer_global_pk_add_f16_insts, | |
| 3262 | .atomic_buffer_pk_add_bf16_inst, | |
| 3263 | .atomic_ds_pk_add_16_insts, | |
| 3264 | .atomic_fadd_no_rtn_insts, | |
| 3265 | .atomic_fadd_rtn_insts, | |
| 3266 | .atomic_flat_pk_add_16_insts, | |
| 3267 | .atomic_fmin_fmax_flat_f64, | |
| 3268 | .atomic_fmin_fmax_global_f64, | |
| 3269 | .atomic_global_pk_add_bf16_inst, | |
| 3270 | .back_off_barrier, | |
| 3271 | .bf16_cvt_insts, | |
| 3272 | .bf16_pk_insts, | |
| 3273 | .bf16_trans_insts, | |
| 3274 | .bitop3_insts, | |
| 3275 | .clusters, | |
| 3276 | .cube_insts, | |
| 3277 | .cumode, | |
| 3278 | .cvt_norm_insts, | |
| 3279 | .cvt_pk_f16_f32_inst, | |
| 3280 | .cvt_pknorm_vop2_insts, | |
| 3281 | .cvt_pknorm_vop3_insts, | |
| 2701 | 3282 | .d16_write_vgpr32, |
| 2702 | 3283 | .dl_insts, |
| 2703 | 3284 | .dot7_insts, |
| ... | ... | @@ -2719,12 +3300,18 @@ pub const cpu = struct { |
| 2719 | 3300 | .ldsbankcount32, |
| 2720 | 3301 | .lerp_inst, |
| 2721 | 3302 | .lshl_add_u64_inst, |
| 3303 | .mad_nc_64_32_insts, | |
| 2722 | 3304 | .mad_u32_inst, |
| 2723 | 3305 | .max_hard_clause_length_63, |
| 2724 | 3306 | .mcast_load_insts, |
| 2725 | 3307 | .memory_atomic_fadd_f32_denormal_support, |
| 2726 | 3308 | .min3_max3_pkf16, |
| 3309 | .min_max_i64_insts, | |
| 2727 | 3310 | .minimum3_maximum3_pkf16, |
| 3311 | .mqsad_insts, | |
| 3312 | .mqsad_pk_insts, | |
| 3313 | .msad_insts, | |
| 3314 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 2728 | 3315 | .packed_fp32_ops, |
| 2729 | 3316 | .packed_tid, |
| 2730 | 3317 | .permlane16_swap, |
| ... | ... | @@ -2732,6 +3319,8 @@ pub const cpu = struct { |
| 2732 | 3319 | .prng_inst, |
| 2733 | 3320 | .pseudo_scalar_trans, |
| 2734 | 3321 | .qsad_insts, |
| 3322 | .real_true16, | |
| 3323 | .requires_initial_unclause_vmem, | |
| 2735 | 3324 | .restricted_soffset, |
| 2736 | 3325 | .s_wakeup_barrier_inst, |
| 2737 | 3326 | .sad_insts, |
| ... | ... | @@ -2740,17 +3329,24 @@ pub const cpu = struct { |
| 2740 | 3329 | .setprio_inc_wg_inst, |
| 2741 | 3330 | .setreg_vgpr_msb_fixup, |
| 2742 | 3331 | .shader_cycles_hi_lo_registers, |
| 3332 | .smem_prefetch_insts, | |
| 2743 | 3333 | .sramecc_support, |
| 3334 | .swmmac_gfx1200_insts, | |
| 3335 | .swmmac_gfx1250_insts, | |
| 2744 | 3336 | .tanh_insts, |
| 2745 | 3337 | .tensor_cvt_lut_insts, |
| 3338 | .trans_coexecution_hazard, | |
| 2746 | 3339 | .transpose_load_f4f6_insts, |
| 3340 | .v_mov_b64_inst, | |
| 3341 | .v_mul_u64_inst, | |
| 2747 | 3342 | .vcmpx_permlane_hazard, |
| 2748 | 3343 | .vgpr_align2, |
| 2749 | 3344 | .vmem_pref_insts, |
| 3345 | .vop3px2_increments_va_vdst_twice, | |
| 2750 | 3346 | .wait_xcnt, |
| 2751 | 3347 | .wavefrontsize32, |
| 3348 | .wmma_coexecution_hazards, | |
| 2752 | 3349 | .xnack, |
| 2753 | .xnack_support, | |
| 2754 | 3350 | }), |
| 2755 | 3351 | }; |
| 2756 | 3352 | pub const gfx1251: CpuModel = .{ |
| ... | ... | @@ -2766,6 +3362,7 @@ pub const cpu = struct { |
| 2766 | 3362 | .architected_flat_scratch, |
| 2767 | 3363 | .architected_sgprs, |
| 2768 | 3364 | .ashr_pk_insts, |
| 3365 | .asynccnt, | |
| 2769 | 3366 | .atomic_buffer_global_pk_add_f16_insts, |
| 2770 | 3367 | .atomic_buffer_pk_add_bf16_inst, |
| 2771 | 3368 | .atomic_ds_pk_add_16_insts, |
| ... | ... | @@ -2775,6 +3372,7 @@ pub const cpu = struct { |
| 2775 | 3372 | .atomic_fmin_fmax_flat_f64, |
| 2776 | 3373 | .atomic_fmin_fmax_global_f64, |
| 2777 | 3374 | .atomic_global_pk_add_bf16_inst, |
| 3375 | .back_off_barrier, | |
| 2778 | 3376 | .bf16_cvt_insts, |
| 2779 | 3377 | .bf16_pk_insts, |
| 2780 | 3378 | .bf16_trans_insts, |
| ... | ... | @@ -2800,27 +3398,40 @@ pub const cpu = struct { |
| 2800 | 3398 | .fmacf64_inst, |
| 2801 | 3399 | .fp8_conversion_insts, |
| 2802 | 3400 | .fp8e5m3_insts, |
| 3401 | .full_rate_64_ops, | |
| 2803 | 3402 | .gfx12, |
| 2804 | 3403 | .gfx1250_insts, |
| 3404 | .gfx1251_gemm_insts, | |
| 3405 | .gfx125x_lowest_rate_wmma, | |
| 2805 | 3406 | .globally_addressable_scratch, |
| 2806 | 3407 | .kernarg_preload, |
| 2807 | 3408 | .lds_barrier_arrive_atomic, |
| 2808 | 3409 | .ldsbankcount32, |
| 2809 | 3410 | .lerp_inst, |
| 2810 | 3411 | .lshl_add_u64_inst, |
| 3412 | .mad_nc_64_32_insts, | |
| 2811 | 3413 | .mad_u32_inst, |
| 2812 | 3414 | .max_hard_clause_length_63, |
| 2813 | 3415 | .mcast_load_insts, |
| 2814 | 3416 | .memory_atomic_fadd_f32_denormal_support, |
| 2815 | 3417 | .min3_max3_pkf16, |
| 3418 | .min_max_i64_insts, | |
| 2816 | 3419 | .minimum3_maximum3_pkf16, |
| 3420 | .mqsad_insts, | |
| 3421 | .mqsad_pk_insts, | |
| 3422 | .msad_insts, | |
| 3423 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 2817 | 3424 | .packed_fp32_ops, |
| 3425 | .packed_fp64_ops, | |
| 2818 | 3426 | .packed_tid, |
| 3427 | .packed_u64_ops, | |
| 2819 | 3428 | .permlane16_swap, |
| 2820 | 3429 | .pk_add_min_max_insts, |
| 2821 | 3430 | .prng_inst, |
| 2822 | 3431 | .pseudo_scalar_trans, |
| 2823 | 3432 | .qsad_insts, |
| 3433 | .real_true16, | |
| 3434 | .requires_initial_unclause_vmem, | |
| 2824 | 3435 | .restricted_soffset, |
| 2825 | 3436 | .s_wakeup_barrier_inst, |
| 2826 | 3437 | .sad_insts, |
| ... | ... | @@ -2828,25 +3439,127 @@ pub const cpu = struct { |
| 2828 | 3439 | .scalar_dwordx3_loads, |
| 2829 | 3440 | .setprio_inc_wg_inst, |
| 2830 | 3441 | .shader_cycles_hi_lo_registers, |
| 3442 | .smem_prefetch_insts, | |
| 2831 | 3443 | .sramecc_support, |
| 3444 | .swmmac_gfx1200_insts, | |
| 3445 | .swmmac_gfx1250_insts, | |
| 2832 | 3446 | .tanh_insts, |
| 2833 | 3447 | .tensor_cvt_lut_insts, |
| 3448 | .trans_coexecution_hazard, | |
| 2834 | 3449 | .transpose_load_f4f6_insts, |
| 3450 | .use_add_pc64_inst, | |
| 3451 | .v_mov_b64_inst, | |
| 3452 | .v_mul_u64_inst, | |
| 2835 | 3453 | .vcmpx_permlane_hazard, |
| 2836 | 3454 | .vgpr_align2, |
| 2837 | 3455 | .vmem_pref_insts, |
| 3456 | .vop3px2_increments_va_vdst_twice, | |
| 2838 | 3457 | .wait_xcnt, |
| 2839 | 3458 | .wavefrontsize32, |
| 3459 | .wmma_coexecution_hazards, | |
| 3460 | .xnack, | |
| 3461 | }), | |
| 3462 | }; | |
| 3463 | pub const gfx12_5_generic: CpuModel = .{ | |
| 3464 | .name = "gfx12_5_generic", | |
| 3465 | .llvm_name = "gfx12-5-generic", | |
| 3466 | .features = featureSet(&[_]Feature{ | |
| 3467 | .@"1024_addressable_vgprs", | |
| 3468 | .@"45_bit_num_records_buffer_resource", | |
| 3469 | .@"64_bit_literals", | |
| 3470 | .add_min_max_insts, | |
| 3471 | .add_sub_u64_insts, | |
| 3472 | .addressablelocalmemorysize327680, | |
| 3473 | .architected_flat_scratch, | |
| 3474 | .architected_sgprs, | |
| 3475 | .ashr_pk_insts, | |
| 3476 | .asynccnt, | |
| 3477 | .atomic_buffer_global_pk_add_f16_insts, | |
| 3478 | .atomic_buffer_pk_add_bf16_inst, | |
| 3479 | .atomic_ds_pk_add_16_insts, | |
| 3480 | .atomic_fadd_no_rtn_insts, | |
| 3481 | .atomic_fadd_rtn_insts, | |
| 3482 | .atomic_flat_pk_add_16_insts, | |
| 3483 | .atomic_fmin_fmax_flat_f64, | |
| 3484 | .atomic_fmin_fmax_global_f64, | |
| 3485 | .atomic_global_pk_add_bf16_inst, | |
| 3486 | .back_off_barrier, | |
| 3487 | .bf16_cvt_insts, | |
| 3488 | .bf16_pk_insts, | |
| 3489 | .bf16_trans_insts, | |
| 3490 | .bitop3_insts, | |
| 3491 | .clusters, | |
| 3492 | .cumode, | |
| 3493 | .cvt_pk_f16_f32_inst, | |
| 3494 | .d16_write_vgpr32, | |
| 3495 | .dl_insts, | |
| 3496 | .dot7_insts, | |
| 3497 | .dot8_insts, | |
| 3498 | .dpp_src1_sgpr, | |
| 3499 | .emulated_system_scope_atomics, | |
| 3500 | .flat_atomic_fadd_f32_inst, | |
| 3501 | .flat_buffer_global_fadd_f64_inst, | |
| 3502 | .flat_gvs_mode, | |
| 3503 | .fma_mix_bf16_insts, | |
| 3504 | .fmacf64_inst, | |
| 3505 | .fp8_conversion_insts, | |
| 3506 | .fp8e5m3_insts, | |
| 3507 | .gfx12, | |
| 3508 | .gfx1250_insts, | |
| 3509 | .gfx125x_lowest_rate_wmma, | |
| 3510 | .globally_addressable_scratch, | |
| 3511 | .kernarg_preload, | |
| 3512 | .lds_barrier_arrive_atomic, | |
| 3513 | .ldsbankcount32, | |
| 3514 | .lshl_add_u64_inst, | |
| 3515 | .mad_nc_64_32_insts, | |
| 3516 | .mad_u32_inst, | |
| 3517 | .max_hard_clause_length_63, | |
| 3518 | .mcast_load_insts, | |
| 3519 | .memory_atomic_fadd_f32_denormal_support, | |
| 3520 | .min3_max3_pkf16, | |
| 3521 | .min_max_i64_insts, | |
| 3522 | .minimum3_maximum3_pkf16, | |
| 3523 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 3524 | .packed_fp32_ops, | |
| 3525 | .packed_tid, | |
| 3526 | .permlane16_swap, | |
| 3527 | .pk_add_min_max_insts, | |
| 3528 | .prng_inst, | |
| 3529 | .pseudo_scalar_trans, | |
| 3530 | .real_true16, | |
| 3531 | .requires_cov6, | |
| 3532 | .requires_initial_unclause_vmem, | |
| 3533 | .restricted_soffset, | |
| 3534 | .s_wakeup_barrier_inst, | |
| 3535 | .salu_float, | |
| 3536 | .scalar_dwordx3_loads, | |
| 3537 | .setprio_inc_wg_inst, | |
| 3538 | .setreg_vgpr_msb_fixup, | |
| 3539 | .shader_cycles_hi_lo_registers, | |
| 3540 | .smem_prefetch_insts, | |
| 3541 | .sramecc_support, | |
| 3542 | .tanh_insts, | |
| 3543 | .tensor_cvt_lut_insts, | |
| 3544 | .trans_coexecution_hazard, | |
| 3545 | .transpose_load_f4f6_insts, | |
| 3546 | .v_mov_b64_inst, | |
| 3547 | .v_mul_u64_inst, | |
| 3548 | .vcmpx_permlane_hazard, | |
| 3549 | .vgpr_align2, | |
| 3550 | .vmem_pref_insts, | |
| 3551 | .wait_xcnt, | |
| 3552 | .wavefrontsize32, | |
| 3553 | .wmma_coexecution_hazards, | |
| 2840 | 3554 | .xnack, |
| 2841 | .xnack_support, | |
| 2842 | 3555 | }), |
| 2843 | 3556 | }; |
| 2844 | 3557 | pub const gfx12_generic: CpuModel = .{ |
| 2845 | 3558 | .name = "gfx12_generic", |
| 2846 | 3559 | .llvm_name = "gfx12-generic", |
| 2847 | 3560 | .features = featureSet(&[_]Feature{ |
| 3561 | .@"1536_physical_vgprs", | |
| 2848 | 3562 | .addressablelocalmemorysize65536, |
| 2849 | .allocate1_5xvgprs, | |
| 2850 | 3563 | .architected_flat_scratch, |
| 2851 | 3564 | .architected_sgprs, |
| 2852 | 3565 | .atomic_buffer_global_pk_add_f16_insts, |
| ... | ... | @@ -2873,25 +3586,170 @@ pub const cpu = struct { |
| 2873 | 3586 | .dpp_src1_sgpr, |
| 2874 | 3587 | .extended_image_insts, |
| 2875 | 3588 | .flat_atomic_fadd_f32_inst, |
| 3589 | .formatted_mubuf_insts, | |
| 2876 | 3590 | .fp8_conversion_insts, |
| 2877 | 3591 | .gfx12, |
| 2878 | 3592 | .image_insts, |
| 2879 | 3593 | .ldsbankcount32, |
| 2880 | 3594 | .lerp_inst, |
| 2881 | 3595 | .memory_atomic_fadd_f32_denormal_support, |
| 3596 | .mqsad_insts, | |
| 3597 | .mqsad_pk_insts, | |
| 3598 | .msad_insts, | |
| 3599 | .mtbuf_insts, | |
| 3600 | .no_f16_pseudo_scalar_trans_inline_constants, | |
| 2882 | 3601 | .nsa_encoding, |
| 2883 | 3602 | .packed_tid, |
| 2884 | 3603 | .partial_nsa_encoding, |
| 2885 | 3604 | .pseudo_scalar_trans, |
| 2886 | 3605 | .qsad_insts, |
| 3606 | .real_true16, | |
| 2887 | 3607 | .requires_cov6, |
| 2888 | 3608 | .restricted_soffset, |
| 2889 | 3609 | .sad_insts, |
| 2890 | 3610 | .salu_float, |
| 2891 | 3611 | .scalar_dwordx3_loads, |
| 2892 | 3612 | .shader_cycles_hi_lo_registers, |
| 3613 | .smem_prefetch_insts, | |
| 3614 | .swmmac_gfx1200_insts, | |
| 2893 | 3615 | .vcmpx_permlane_hazard, |
| 2894 | 3616 | .waits_before_system_scope_stores, |
| 3617 | .wmma_128b_insts, | |
| 3618 | }), | |
| 3619 | }; | |
| 3620 | pub const gfx1310: CpuModel = .{ | |
| 3621 | .name = "gfx1310", | |
| 3622 | .llvm_name = "gfx1310", | |
| 3623 | .features = featureSet(&[_]Feature{ | |
| 3624 | .@"1536_physical_vgprs", | |
| 3625 | .@"64_bit_literals", | |
| 3626 | .addressablelocalmemorysize65536, | |
| 3627 | .architected_flat_scratch, | |
| 3628 | .architected_sgprs, | |
| 3629 | .ashr_pk_insts, | |
| 3630 | .atomic_buffer_global_pk_add_f16_insts, | |
| 3631 | .atomic_buffer_pk_add_bf16_inst, | |
| 3632 | .atomic_ds_pk_add_16_insts, | |
| 3633 | .atomic_fadd_no_rtn_insts, | |
| 3634 | .atomic_fadd_rtn_insts, | |
| 3635 | .atomic_flat_pk_add_16_insts, | |
| 3636 | .atomic_fmin_fmax_flat_f64, | |
| 3637 | .atomic_fmin_fmax_global_f64, | |
| 3638 | .atomic_global_pk_add_bf16_inst, | |
| 3639 | .bf16_cvt_insts, | |
| 3640 | .bf16_pk_insts, | |
| 3641 | .bf16_trans_insts, | |
| 3642 | .bitop3_insts, | |
| 3643 | .clusters, | |
| 3644 | .cube_insts, | |
| 3645 | .cvt_norm_insts, | |
| 3646 | .cvt_pk_f16_f32_inst, | |
| 3647 | .cvt_pknorm_vop2_insts, | |
| 3648 | .cvt_pknorm_vop3_insts, | |
| 3649 | .dl_insts, | |
| 3650 | .dot7_insts, | |
| 3651 | .dot8_insts, | |
| 3652 | .dpp_src1_sgpr, | |
| 3653 | .f16bf16_to_fp6bf6_cvt_scale_insts, | |
| 3654 | .flat_atomic_fadd_f32_inst, | |
| 3655 | .flat_gvs_mode, | |
| 3656 | .fma_mix_bf16_insts, | |
| 3657 | .fmacf64_inst, | |
| 3658 | .fp8_conversion_insts, | |
| 3659 | .gfx1250_insts, | |
| 3660 | .gfx13, | |
| 3661 | .globally_addressable_scratch, | |
| 3662 | .ldsbankcount32, | |
| 3663 | .lerp_inst, | |
| 3664 | .mqsad_insts, | |
| 3665 | .mqsad_pk_insts, | |
| 3666 | .msad_insts, | |
| 3667 | .nsa_encoding, | |
| 3668 | .packed_tid, | |
| 3669 | .partial_nsa_encoding, | |
| 3670 | .permlane16_swap, | |
| 3671 | .prng_inst, | |
| 3672 | .pseudo_scalar_trans, | |
| 3673 | .qsad_insts, | |
| 3674 | .restricted_soffset, | |
| 3675 | .s_wakeup_barrier_inst, | |
| 3676 | .sad_insts, | |
| 3677 | .salu_float, | |
| 3678 | .scalar_dwordx3_loads, | |
| 3679 | .shader_cycles_register, | |
| 3680 | .smem_prefetch_insts, | |
| 3681 | .tanh_insts, | |
| 3682 | .tensor_cvt_lut_insts, | |
| 3683 | .transpose_load_f4f6_insts, | |
| 3684 | .vcmpx_permlane_hazard, | |
| 3685 | }), | |
| 3686 | }; | |
| 3687 | pub const gfx13_generic: CpuModel = .{ | |
| 3688 | .name = "gfx13_generic", | |
| 3689 | .llvm_name = "gfx13-generic", | |
| 3690 | .features = featureSet(&[_]Feature{ | |
| 3691 | .@"1536_physical_vgprs", | |
| 3692 | .@"64_bit_literals", | |
| 3693 | .addressablelocalmemorysize65536, | |
| 3694 | .architected_flat_scratch, | |
| 3695 | .architected_sgprs, | |
| 3696 | .ashr_pk_insts, | |
| 3697 | .atomic_buffer_global_pk_add_f16_insts, | |
| 3698 | .atomic_buffer_pk_add_bf16_inst, | |
| 3699 | .atomic_ds_pk_add_16_insts, | |
| 3700 | .atomic_fadd_no_rtn_insts, | |
| 3701 | .atomic_fadd_rtn_insts, | |
| 3702 | .atomic_flat_pk_add_16_insts, | |
| 3703 | .atomic_fmin_fmax_flat_f64, | |
| 3704 | .atomic_fmin_fmax_global_f64, | |
| 3705 | .atomic_global_pk_add_bf16_inst, | |
| 3706 | .bf16_cvt_insts, | |
| 3707 | .bf16_pk_insts, | |
| 3708 | .bf16_trans_insts, | |
| 3709 | .bitop3_insts, | |
| 3710 | .clusters, | |
| 3711 | .cube_insts, | |
| 3712 | .cvt_norm_insts, | |
| 3713 | .cvt_pk_f16_f32_inst, | |
| 3714 | .cvt_pknorm_vop2_insts, | |
| 3715 | .cvt_pknorm_vop3_insts, | |
| 3716 | .dl_insts, | |
| 3717 | .dot7_insts, | |
| 3718 | .dot8_insts, | |
| 3719 | .dpp_src1_sgpr, | |
| 3720 | .f16bf16_to_fp6bf6_cvt_scale_insts, | |
| 3721 | .flat_atomic_fadd_f32_inst, | |
| 3722 | .flat_gvs_mode, | |
| 3723 | .fma_mix_bf16_insts, | |
| 3724 | .fmacf64_inst, | |
| 3725 | .fp8_conversion_insts, | |
| 3726 | .gfx1250_insts, | |
| 3727 | .gfx13, | |
| 3728 | .globally_addressable_scratch, | |
| 3729 | .ldsbankcount32, | |
| 3730 | .lerp_inst, | |
| 3731 | .mqsad_insts, | |
| 3732 | .mqsad_pk_insts, | |
| 3733 | .msad_insts, | |
| 3734 | .nsa_encoding, | |
| 3735 | .packed_tid, | |
| 3736 | .partial_nsa_encoding, | |
| 3737 | .permlane16_swap, | |
| 3738 | .prng_inst, | |
| 3739 | .pseudo_scalar_trans, | |
| 3740 | .qsad_insts, | |
| 3741 | .requires_cov6, | |
| 3742 | .restricted_soffset, | |
| 3743 | .s_wakeup_barrier_inst, | |
| 3744 | .sad_insts, | |
| 3745 | .salu_float, | |
| 3746 | .scalar_dwordx3_loads, | |
| 3747 | .shader_cycles_register, | |
| 3748 | .smem_prefetch_insts, | |
| 3749 | .tanh_insts, | |
| 3750 | .tensor_cvt_lut_insts, | |
| 3751 | .transpose_load_f4f6_insts, | |
| 3752 | .vcmpx_permlane_hazard, | |
| 2895 | 3753 | }), |
| 2896 | 3754 | }; |
| 2897 | 3755 | pub const gfx600: CpuModel = .{ |
| ... | ... | @@ -2977,7 +3835,7 @@ pub const cpu = struct { |
| 2977 | 3835 | .ldsbankcount32, |
| 2978 | 3836 | .unpacked_d16_vmem, |
| 2979 | 3837 | .volcanic_islands, |
| 2980 | .xnack_support, | |
| 3838 | .xnack_on_off_modes, | |
| 2981 | 3839 | }), |
| 2982 | 3840 | }; |
| 2983 | 3841 | pub const gfx802: CpuModel = .{ |
| ... | ... | @@ -3017,7 +3875,7 @@ pub const cpu = struct { |
| 3017 | 3875 | .image_store_d16_bug, |
| 3018 | 3876 | .ldsbankcount16, |
| 3019 | 3877 | .volcanic_islands, |
| 3020 | .xnack_support, | |
| 3878 | .xnack_on_off_modes, | |
| 3021 | 3879 | }), |
| 3022 | 3880 | }; |
| 3023 | 3881 | pub const gfx900: CpuModel = .{ |
| ... | ... | @@ -3175,6 +4033,7 @@ pub const cpu = struct { |
| 3175 | 4033 | .packed_tid, |
| 3176 | 4034 | .pk_fmac_f16_inst, |
| 3177 | 4035 | .sramecc_support, |
| 4036 | .tgsplit_support, | |
| 3178 | 4037 | .vgpr_align2, |
| 3179 | 4038 | }), |
| 3180 | 4039 | }; |
| ... | ... | @@ -3239,6 +4098,8 @@ pub const cpu = struct { |
| 3239 | 4098 | .packed_tid, |
| 3240 | 4099 | .pk_fmac_f16_inst, |
| 3241 | 4100 | .sramecc_support, |
| 4101 | .tgsplit_support, | |
| 4102 | .v_mov_b64_inst, | |
| 3242 | 4103 | .vgpr_align2, |
| 3243 | 4104 | .xf32_insts, |
| 3244 | 4105 | }), |
| ... | ... | @@ -3295,6 +4156,8 @@ pub const cpu = struct { |
| 3295 | 4156 | .pk_fmac_f16_inst, |
| 3296 | 4157 | .prng_inst, |
| 3297 | 4158 | .sramecc_support, |
| 4159 | .tgsplit_support, | |
| 4160 | .v_mov_b64_inst, | |
| 3298 | 4161 | .vgpr_align2, |
| 3299 | 4162 | }), |
| 3300 | 4163 | }; |
| ... | ... | @@ -3342,6 +4205,8 @@ pub const cpu = struct { |
| 3342 | 4205 | .pk_fmac_f16_inst, |
| 3343 | 4206 | .requires_cov6, |
| 3344 | 4207 | .sramecc_support, |
| 4208 | .tgsplit_support, | |
| 4209 | .v_mov_b64_inst, | |
| 3345 | 4210 | .vgpr_align2, |
| 3346 | 4211 | }), |
| 3347 | 4212 | }; |
| ... | ... | @@ -3452,7 +4317,7 @@ pub const cpu = struct { |
| 3452 | 4317 | .image_store_d16_bug, |
| 3453 | 4318 | .ldsbankcount16, |
| 3454 | 4319 | .volcanic_islands, |
| 3455 | .xnack_support, | |
| 4320 | .xnack_on_off_modes, | |
| 3456 | 4321 | }), |
| 3457 | 4322 | }; |
| 3458 | 4323 | pub const tahiti: CpuModel = .{ |
lib/std/Target/hexagon.zig+204| ... | ... | @@ -34,7 +34,41 @@ pub const Feature = enum { |
| 34 | 34 | nvs, |
| 35 | 35 | packets, |
| 36 | 36 | prev65, |
| 37 | reserved_r10, | |
| 38 | reserved_r11, | |
| 39 | reserved_r12, | |
| 40 | reserved_r13, | |
| 41 | reserved_r14, | |
| 42 | reserved_r15, | |
| 43 | reserved_r16, | |
| 44 | reserved_r17, | |
| 45 | reserved_r18, | |
| 37 | 46 | reserved_r19, |
| 47 | reserved_r20, | |
| 48 | reserved_r21, | |
| 49 | reserved_r22, | |
| 50 | reserved_r23, | |
| 51 | reserved_r24, | |
| 52 | reserved_r25, | |
| 53 | reserved_r26, | |
| 54 | reserved_r27, | |
| 55 | reserved_r28, | |
| 56 | reserved_r6, | |
| 57 | reserved_r7, | |
| 58 | reserved_r8, | |
| 59 | reserved_r9, | |
| 60 | scs_reg_r16, | |
| 61 | scs_reg_r17, | |
| 62 | scs_reg_r18, | |
| 63 | scs_reg_r19, | |
| 64 | scs_reg_r20, | |
| 65 | scs_reg_r21, | |
| 66 | scs_reg_r22, | |
| 67 | scs_reg_r23, | |
| 68 | scs_reg_r24, | |
| 69 | scs_reg_r25, | |
| 70 | scs_reg_r26, | |
| 71 | scs_reg_r27, | |
| 38 | 72 | small_data, |
| 39 | 73 | tinycore, |
| 40 | 74 | v5, |
| ... | ... | @@ -241,11 +275,181 @@ pub const all_features = blk: { |
| 241 | 275 | .description = "Support features deprecated in v65", |
| 242 | 276 | .dependencies = featureSet(&[_]Feature{}), |
| 243 | 277 | }; |
| 278 | result[@backingInt(Feature.reserved_r10)] = .{ | |
| 279 | .llvm_name = "reserved-r10", | |
| 280 | .description = "Reserve register R10", | |
| 281 | .dependencies = featureSet(&[_]Feature{}), | |
| 282 | }; | |
| 283 | result[@backingInt(Feature.reserved_r11)] = .{ | |
| 284 | .llvm_name = "reserved-r11", | |
| 285 | .description = "Reserve register R11", | |
| 286 | .dependencies = featureSet(&[_]Feature{}), | |
| 287 | }; | |
| 288 | result[@backingInt(Feature.reserved_r12)] = .{ | |
| 289 | .llvm_name = "reserved-r12", | |
| 290 | .description = "Reserve register R12", | |
| 291 | .dependencies = featureSet(&[_]Feature{}), | |
| 292 | }; | |
| 293 | result[@backingInt(Feature.reserved_r13)] = .{ | |
| 294 | .llvm_name = "reserved-r13", | |
| 295 | .description = "Reserve register R13", | |
| 296 | .dependencies = featureSet(&[_]Feature{}), | |
| 297 | }; | |
| 298 | result[@backingInt(Feature.reserved_r14)] = .{ | |
| 299 | .llvm_name = "reserved-r14", | |
| 300 | .description = "Reserve register R14", | |
| 301 | .dependencies = featureSet(&[_]Feature{}), | |
| 302 | }; | |
| 303 | result[@backingInt(Feature.reserved_r15)] = .{ | |
| 304 | .llvm_name = "reserved-r15", | |
| 305 | .description = "Reserve register R15", | |
| 306 | .dependencies = featureSet(&[_]Feature{}), | |
| 307 | }; | |
| 308 | result[@backingInt(Feature.reserved_r16)] = .{ | |
| 309 | .llvm_name = "reserved-r16", | |
| 310 | .description = "Reserve register R16", | |
| 311 | .dependencies = featureSet(&[_]Feature{}), | |
| 312 | }; | |
| 313 | result[@backingInt(Feature.reserved_r17)] = .{ | |
| 314 | .llvm_name = "reserved-r17", | |
| 315 | .description = "Reserve register R17", | |
| 316 | .dependencies = featureSet(&[_]Feature{}), | |
| 317 | }; | |
| 318 | result[@backingInt(Feature.reserved_r18)] = .{ | |
| 319 | .llvm_name = "reserved-r18", | |
| 320 | .description = "Reserve register R18", | |
| 321 | .dependencies = featureSet(&[_]Feature{}), | |
| 322 | }; | |
| 244 | 323 | result[@backingInt(Feature.reserved_r19)] = .{ |
| 245 | 324 | .llvm_name = "reserved-r19", |
| 246 | 325 | .description = "Reserve register R19", |
| 247 | 326 | .dependencies = featureSet(&[_]Feature{}), |
| 248 | 327 | }; |
| 328 | result[@backingInt(Feature.reserved_r20)] = .{ | |
| 329 | .llvm_name = "reserved-r20", | |
| 330 | .description = "Reserve register R20", | |
| 331 | .dependencies = featureSet(&[_]Feature{}), | |
| 332 | }; | |
| 333 | result[@backingInt(Feature.reserved_r21)] = .{ | |
| 334 | .llvm_name = "reserved-r21", | |
| 335 | .description = "Reserve register R21", | |
| 336 | .dependencies = featureSet(&[_]Feature{}), | |
| 337 | }; | |
| 338 | result[@backingInt(Feature.reserved_r22)] = .{ | |
| 339 | .llvm_name = "reserved-r22", | |
| 340 | .description = "Reserve register R22", | |
| 341 | .dependencies = featureSet(&[_]Feature{}), | |
| 342 | }; | |
| 343 | result[@backingInt(Feature.reserved_r23)] = .{ | |
| 344 | .llvm_name = "reserved-r23", | |
| 345 | .description = "Reserve register R23", | |
| 346 | .dependencies = featureSet(&[_]Feature{}), | |
| 347 | }; | |
| 348 | result[@backingInt(Feature.reserved_r24)] = .{ | |
| 349 | .llvm_name = "reserved-r24", | |
| 350 | .description = "Reserve register R24", | |
| 351 | .dependencies = featureSet(&[_]Feature{}), | |
| 352 | }; | |
| 353 | result[@backingInt(Feature.reserved_r25)] = .{ | |
| 354 | .llvm_name = "reserved-r25", | |
| 355 | .description = "Reserve register R25", | |
| 356 | .dependencies = featureSet(&[_]Feature{}), | |
| 357 | }; | |
| 358 | result[@backingInt(Feature.reserved_r26)] = .{ | |
| 359 | .llvm_name = "reserved-r26", | |
| 360 | .description = "Reserve register R26", | |
| 361 | .dependencies = featureSet(&[_]Feature{}), | |
| 362 | }; | |
| 363 | result[@backingInt(Feature.reserved_r27)] = .{ | |
| 364 | .llvm_name = "reserved-r27", | |
| 365 | .description = "Reserve register R27", | |
| 366 | .dependencies = featureSet(&[_]Feature{}), | |
| 367 | }; | |
| 368 | result[@backingInt(Feature.reserved_r28)] = .{ | |
| 369 | .llvm_name = "reserved-r28", | |
| 370 | .description = "Reserve register R28", | |
| 371 | .dependencies = featureSet(&[_]Feature{}), | |
| 372 | }; | |
| 373 | result[@backingInt(Feature.reserved_r6)] = .{ | |
| 374 | .llvm_name = "reserved-r6", | |
| 375 | .description = "Reserve register R6", | |
| 376 | .dependencies = featureSet(&[_]Feature{}), | |
| 377 | }; | |
| 378 | result[@backingInt(Feature.reserved_r7)] = .{ | |
| 379 | .llvm_name = "reserved-r7", | |
| 380 | .description = "Reserve register R7", | |
| 381 | .dependencies = featureSet(&[_]Feature{}), | |
| 382 | }; | |
| 383 | result[@backingInt(Feature.reserved_r8)] = .{ | |
| 384 | .llvm_name = "reserved-r8", | |
| 385 | .description = "Reserve register R8", | |
| 386 | .dependencies = featureSet(&[_]Feature{}), | |
| 387 | }; | |
| 388 | result[@backingInt(Feature.reserved_r9)] = .{ | |
| 389 | .llvm_name = "reserved-r9", | |
| 390 | .description = "Reserve register R9", | |
| 391 | .dependencies = featureSet(&[_]Feature{}), | |
| 392 | }; | |
| 393 | result[@backingInt(Feature.scs_reg_r16)] = .{ | |
| 394 | .llvm_name = "scs-reg-r16", | |
| 395 | .description = "Use register R16 as the shadow call stack pointer", | |
| 396 | .dependencies = featureSet(&[_]Feature{}), | |
| 397 | }; | |
| 398 | result[@backingInt(Feature.scs_reg_r17)] = .{ | |
| 399 | .llvm_name = "scs-reg-r17", | |
| 400 | .description = "Use register R17 as the shadow call stack pointer", | |
| 401 | .dependencies = featureSet(&[_]Feature{}), | |
| 402 | }; | |
| 403 | result[@backingInt(Feature.scs_reg_r18)] = .{ | |
| 404 | .llvm_name = "scs-reg-r18", | |
| 405 | .description = "Use register R18 as the shadow call stack pointer", | |
| 406 | .dependencies = featureSet(&[_]Feature{}), | |
| 407 | }; | |
| 408 | result[@backingInt(Feature.scs_reg_r19)] = .{ | |
| 409 | .llvm_name = "scs-reg-r19", | |
| 410 | .description = "Use register R19 as the shadow call stack pointer", | |
| 411 | .dependencies = featureSet(&[_]Feature{}), | |
| 412 | }; | |
| 413 | result[@backingInt(Feature.scs_reg_r20)] = .{ | |
| 414 | .llvm_name = "scs-reg-r20", | |
| 415 | .description = "Use register R20 as the shadow call stack pointer", | |
| 416 | .dependencies = featureSet(&[_]Feature{}), | |
| 417 | }; | |
| 418 | result[@backingInt(Feature.scs_reg_r21)] = .{ | |
| 419 | .llvm_name = "scs-reg-r21", | |
| 420 | .description = "Use register R21 as the shadow call stack pointer", | |
| 421 | .dependencies = featureSet(&[_]Feature{}), | |
| 422 | }; | |
| 423 | result[@backingInt(Feature.scs_reg_r22)] = .{ | |
| 424 | .llvm_name = "scs-reg-r22", | |
| 425 | .description = "Use register R22 as the shadow call stack pointer", | |
| 426 | .dependencies = featureSet(&[_]Feature{}), | |
| 427 | }; | |
| 428 | result[@backingInt(Feature.scs_reg_r23)] = .{ | |
| 429 | .llvm_name = "scs-reg-r23", | |
| 430 | .description = "Use register R23 as the shadow call stack pointer", | |
| 431 | .dependencies = featureSet(&[_]Feature{}), | |
| 432 | }; | |
| 433 | result[@backingInt(Feature.scs_reg_r24)] = .{ | |
| 434 | .llvm_name = "scs-reg-r24", | |
| 435 | .description = "Use register R24 as the shadow call stack pointer", | |
| 436 | .dependencies = featureSet(&[_]Feature{}), | |
| 437 | }; | |
| 438 | result[@backingInt(Feature.scs_reg_r25)] = .{ | |
| 439 | .llvm_name = "scs-reg-r25", | |
| 440 | .description = "Use register R25 as the shadow call stack pointer", | |
| 441 | .dependencies = featureSet(&[_]Feature{}), | |
| 442 | }; | |
| 443 | result[@backingInt(Feature.scs_reg_r26)] = .{ | |
| 444 | .llvm_name = "scs-reg-r26", | |
| 445 | .description = "Use register R26 as the shadow call stack pointer", | |
| 446 | .dependencies = featureSet(&[_]Feature{}), | |
| 447 | }; | |
| 448 | result[@backingInt(Feature.scs_reg_r27)] = .{ | |
| 449 | .llvm_name = "scs-reg-r27", | |
| 450 | .description = "Use register R27 as the shadow call stack pointer", | |
| 451 | .dependencies = featureSet(&[_]Feature{}), | |
| 452 | }; | |
| 249 | 453 | result[@backingInt(Feature.small_data)] = .{ |
| 250 | 454 | .llvm_name = "small-data", |
| 251 | 455 | .description = "Allow GP-relative addressing of global variables", |
lib/std/Target/mips.zig+23| ... | ... | @@ -13,6 +13,7 @@ pub const Feature = enum { |
| 13 | 13 | dspr2, |
| 14 | 14 | dspr3, |
| 15 | 15 | eva, |
| 16 | fix_r5900, | |
| 16 | 17 | fp64, |
| 17 | 18 | fpxx, |
| 18 | 19 | ginv, |
| ... | ... | @@ -52,6 +53,7 @@ pub const Feature = enum { |
| 52 | 53 | notraps, |
| 53 | 54 | p5600, |
| 54 | 55 | ptr64, |
| 56 | r5900, | |
| 55 | 57 | single_float, |
| 56 | 58 | soft_float, |
| 57 | 59 | strict_align, |
| ... | ... | @@ -121,6 +123,11 @@ pub const all_features = blk: { |
| 121 | 123 | .description = "Mips EVA ASE", |
| 122 | 124 | .dependencies = featureSet(&[_]Feature{}), |
| 123 | 125 | }; |
| 126 | result[@backingInt(Feature.fix_r5900)] = .{ | |
| 127 | .llvm_name = "fix-r5900", | |
| 128 | .description = "Enable R5900 short loop erratum fix", | |
| 129 | .dependencies = featureSet(&[_]Feature{}), | |
| 130 | }; | |
| 124 | 131 | result[@backingInt(Feature.fp64)] = .{ |
| 125 | 132 | .llvm_name = "fp64", |
| 126 | 133 | .description = "Support 64-bit FP registers", |
| ... | ... | @@ -372,6 +379,15 @@ pub const all_features = blk: { |
| 372 | 379 | .description = "Pointers are 64-bit wide", |
| 373 | 380 | .dependencies = featureSet(&[_]Feature{}), |
| 374 | 381 | }; |
| 382 | result[@backingInt(Feature.r5900)] = .{ | |
| 383 | .llvm_name = "r5900", | |
| 384 | .description = "R5900 (PS2 Emotion Engine) Support", | |
| 385 | .dependencies = featureSet(&[_]Feature{ | |
| 386 | .fix_r5900, | |
| 387 | .mips3, | |
| 388 | .single_float, | |
| 389 | }), | |
| 390 | }; | |
| 375 | 391 | result[@backingInt(Feature.single_float)] = .{ |
| 376 | 392 | .llvm_name = "single-float", |
| 377 | 393 | .description = "Only supports single precision float", |
| ... | ... | @@ -596,4 +612,11 @@ pub const cpu = struct { |
| 596 | 612 | .soft_float, |
| 597 | 613 | }), |
| 598 | 614 | }; |
| 615 | pub const r5900: CpuModel = .{ | |
| 616 | .name = "r5900", | |
| 617 | .llvm_name = "r5900", | |
| 618 | .features = featureSet(&[_]Feature{ | |
| 619 | .r5900, | |
| 620 | }), | |
| 621 | }; | |
| 599 | 622 | }; |
lib/std/Target/nvptx.zig+18| ... | ... | @@ -36,6 +36,9 @@ pub const Feature = enum { |
| 36 | 36 | ptx87, |
| 37 | 37 | ptx88, |
| 38 | 38 | ptx90, |
| 39 | ptx91, | |
| 40 | ptx92, | |
| 41 | ptx93, | |
| 39 | 42 | sm_100, |
| 40 | 43 | sm_100a, |
| 41 | 44 | sm_100f, |
| ... | ... | @@ -242,6 +245,21 @@ pub const all_features = blk: { |
| 242 | 245 | .description = "Use PTX version 90", |
| 243 | 246 | .dependencies = featureSet(&[_]Feature{}), |
| 244 | 247 | }; |
| 248 | result[@backingInt(Feature.ptx91)] = .{ | |
| 249 | .llvm_name = "ptx91", | |
| 250 | .description = "Use PTX version 91", | |
| 251 | .dependencies = featureSet(&[_]Feature{}), | |
| 252 | }; | |
| 253 | result[@backingInt(Feature.ptx92)] = .{ | |
| 254 | .llvm_name = "ptx92", | |
| 255 | .description = "Use PTX version 92", | |
| 256 | .dependencies = featureSet(&[_]Feature{}), | |
| 257 | }; | |
| 258 | result[@backingInt(Feature.ptx93)] = .{ | |
| 259 | .llvm_name = "ptx93", | |
| 260 | .description = "Use PTX version 93", | |
| 261 | .dependencies = featureSet(&[_]Feature{}), | |
| 262 | }; | |
| 245 | 263 | result[@backingInt(Feature.sm_100)] = .{ |
| 246 | 264 | .llvm_name = "sm_100", |
| 247 | 265 | .description = "Target SM 100", |
lib/std/Target/powerpc.zig+17-2| ... | ... | @@ -43,6 +43,7 @@ pub const Feature = enum { |
| 43 | 43 | fuse_wideimm, |
| 44 | 44 | fuse_zeromove, |
| 45 | 45 | fusion, |
| 46 | future_vector, | |
| 46 | 47 | hard_float, |
| 47 | 48 | htm, |
| 48 | 49 | icbt, |
| ... | ... | @@ -83,6 +84,7 @@ pub const Feature = enum { |
| 83 | 84 | spe, |
| 84 | 85 | stfiwx, |
| 85 | 86 | two_const_nr, |
| 87 | use_ptrgl_helper, | |
| 86 | 88 | vectors_use_two_units, |
| 87 | 89 | vsx, |
| 88 | 90 | }; |
| ... | ... | @@ -342,6 +344,14 @@ pub const all_features = blk: { |
| 342 | 344 | .description = "Target supports instruction fusion", |
| 343 | 345 | .dependencies = featureSet(&[_]Feature{}), |
| 344 | 346 | }; |
| 347 | result[@backingInt(Feature.future_vector)] = .{ | |
| 348 | .llvm_name = "future-vector", | |
| 349 | .description = "Enable Future vector instructions", | |
| 350 | .dependencies = featureSet(&[_]Feature{ | |
| 351 | .isa_future_instructions, | |
| 352 | .power10_vector, | |
| 353 | }), | |
| 354 | }; | |
| 345 | 355 | result[@backingInt(Feature.hard_float)] = .{ |
| 346 | 356 | .llvm_name = "hard-float", |
| 347 | 357 | .description = "Enable floating-point instructions", |
| ... | ... | @@ -441,6 +451,7 @@ pub const all_features = blk: { |
| 441 | 451 | .description = "32Byte load and store instructions", |
| 442 | 452 | .dependencies = featureSet(&[_]Feature{ |
| 443 | 453 | .isa_v30_instructions, |
| 454 | .vsx, | |
| 444 | 455 | }), |
| 445 | 456 | }; |
| 446 | 457 | result[@backingInt(Feature.partword_atomics)] = .{ |
| ... | ... | @@ -580,6 +591,11 @@ pub const all_features = blk: { |
| 580 | 591 | .description = "Requires two constant Newton-Raphson computation", |
| 581 | 592 | .dependencies = featureSet(&[_]Feature{}), |
| 582 | 593 | }; |
| 594 | result[@backingInt(Feature.use_ptrgl_helper)] = .{ | |
| 595 | .llvm_name = "use-ptrgl-helper", | |
| 596 | .description = "Use ._ptrgl for indirect calls", | |
| 597 | .dependencies = featureSet(&[_]Feature{}), | |
| 598 | }; | |
| 583 | 599 | result[@backingInt(Feature.vectors_use_two_units)] = .{ |
| 584 | 600 | .llvm_name = "vectors-use-two-units", |
| 585 | 601 | .description = "Vectors use two units", |
| ... | ... | @@ -803,8 +819,8 @@ pub const cpu = struct { |
| 803 | 819 | .fuse_logical_add, |
| 804 | 820 | .fuse_sha3, |
| 805 | 821 | .fuse_store, |
| 822 | .future_vector, | |
| 806 | 823 | .icbt, |
| 807 | .isa_future_instructions, | |
| 808 | 824 | .isa_v206_instructions, |
| 809 | 825 | .isel, |
| 810 | 826 | .ldbrx, |
| ... | ... | @@ -814,7 +830,6 @@ pub const cpu = struct { |
| 814 | 830 | .partword_atomics, |
| 815 | 831 | .pcrelative_memops, |
| 816 | 832 | .popcntd, |
| 817 | .power10_vector, | |
| 818 | 833 | .ppc_postra_sched, |
| 819 | 834 | .ppc_prera_sched, |
| 820 | 835 | .predictable_select_expensive, |
lib/std/Target/riscv.zig+673-134| ... | ... | @@ -8,13 +8,8 @@ pub const Feature = enum { |
| 8 | 8 | @"32bit", |
| 9 | 9 | @"64bit", |
| 10 | 10 | a, |
| 11 | add_load_fusion, | |
| 12 | addi_load_fusion, | |
| 13 | 11 | andes45, |
| 14 | auipc_addi_fusion, | |
| 15 | auipc_load_fusion, | |
| 16 | 12 | b, |
| 17 | bfext_fusion, | |
| 18 | 13 | c, |
| 19 | 14 | conditional_cmv_fusion, |
| 20 | 15 | d, |
| ... | ... | @@ -25,6 +20,7 @@ pub const Feature = enum { |
| 25 | 20 | disable_postmisched_store_clustering, |
| 26 | 21 | dlen_factor_2, |
| 27 | 22 | e, |
| 23 | enable_select_opt, | |
| 28 | 24 | enable_vsetvli_sched_heuristic, |
| 29 | 25 | exact_asm, |
| 30 | 26 | experimental, |
| ... | ... | @@ -32,26 +28,57 @@ pub const Feature = enum { |
| 32 | 28 | experimental_rvm23u32, |
| 33 | 29 | experimental_smpmpmt, |
| 34 | 30 | experimental_svukte, |
| 35 | experimental_xrivosvisni, | |
| 36 | experimental_xrivosvizip, | |
| 31 | experimental_xqccmt, | |
| 37 | 32 | experimental_xsfmclic, |
| 38 | 33 | experimental_xsfsclic, |
| 34 | experimental_y, | |
| 39 | 35 | experimental_zibi, |
| 40 | 36 | experimental_zicfilp, |
| 41 | 37 | experimental_zicfiss, |
| 38 | experimental_zvabd, | |
| 42 | 39 | experimental_zvbc32e, |
| 40 | experimental_zvdot4a8i, | |
| 41 | experimental_zvfbdota32f, | |
| 43 | 42 | experimental_zvfbfa, |
| 44 | 43 | experimental_zvfofp8min, |
| 44 | experimental_zvfqwbdota8f, | |
| 45 | experimental_zvfqwdota8f, | |
| 46 | experimental_zvfwbdota16bf, | |
| 47 | experimental_zvfwdota16bf, | |
| 45 | 48 | experimental_zvkgs, |
| 46 | experimental_zvqdotq, | |
| 49 | experimental_zvqwbdota16i, | |
| 50 | experimental_zvqwbdota8i, | |
| 51 | experimental_zvqwdota16i, | |
| 52 | experimental_zvqwdota8i, | |
| 53 | experimental_zvvfmm, | |
| 54 | experimental_zvvmm, | |
| 55 | experimental_zvvmtls, | |
| 56 | experimental_zvvmttls, | |
| 57 | experimental_zvzip, | |
| 47 | 58 | f, |
| 48 | 59 | forced_atomics, |
| 60 | fusion_add_load, | |
| 61 | fusion_add_mem, | |
| 62 | fusion_addi_load, | |
| 63 | fusion_auipc_addi, | |
| 64 | fusion_auipc_load, | |
| 65 | fusion_bfext, | |
| 66 | fusion_ld_add, | |
| 67 | fusion_logic_imm_reg, | |
| 68 | fusion_logic_reg_imm, | |
| 69 | fusion_logic_reg_reg, | |
| 70 | fusion_lui_addi, | |
| 71 | fusion_lui_load, | |
| 72 | fusion_mul_add, | |
| 73 | fusion_qc_e_li_load_store, | |
| 74 | fusion_shift_bit_extract, | |
| 75 | fusion_shifted_zextw, | |
| 76 | fusion_shxadd_load, | |
| 77 | fusion_zexth, | |
| 78 | fusion_zextw, | |
| 49 | 79 | h, |
| 50 | 80 | i, |
| 51 | ld_add_fusion, | |
| 52 | 81 | log_vrgather, |
| 53 | lui_addi_fusion, | |
| 54 | lui_load_fusion, | |
| 55 | 82 | m, |
| 56 | 83 | mips_p8700, |
| 57 | 84 | no_default_unroll, |
| ... | ... | @@ -67,6 +94,7 @@ pub const Feature = enum { |
| 67 | 94 | optimized_zero_stride_load, |
| 68 | 95 | permissive_zalrsc, |
| 69 | 96 | predictable_select_expensive, |
| 97 | prefer_ascending_load_store, | |
| 70 | 98 | prefer_vsetvli_over_read_vlenb, |
| 71 | 99 | prefer_w_inst, |
| 72 | 100 | q, |
| ... | ... | @@ -118,17 +146,16 @@ pub const Feature = enum { |
| 118 | 146 | sha, |
| 119 | 147 | shcounterenw, |
| 120 | 148 | shgatpa, |
| 121 | shifted_zextw_fusion, | |
| 122 | 149 | shlcofideleg, |
| 123 | 150 | short_forward_branch_ialu, |
| 124 | 151 | short_forward_branch_iload, |
| 125 | 152 | short_forward_branch_iminmax, |
| 153 | short_forward_branch_imm, | |
| 126 | 154 | short_forward_branch_imul, |
| 127 | 155 | shtvala, |
| 128 | 156 | shvsatpa, |
| 129 | 157 | shvstvala, |
| 130 | 158 | shvstvecd, |
| 131 | shxadd_load_fusion, | |
| 132 | 159 | single_element_vec_fp64, |
| 133 | 160 | smaia, |
| 134 | 161 | smcdeleg, |
| ... | ... | @@ -165,6 +192,7 @@ pub const Feature = enum { |
| 165 | 192 | svinval, |
| 166 | 193 | svnapot, |
| 167 | 194 | svpbmt, |
| 195 | svrsw60t59b, | |
| 168 | 196 | svvptc, |
| 169 | 197 | tagged_globals, |
| 170 | 198 | unaligned_scalar_mem, |
| ... | ... | @@ -174,6 +202,7 @@ pub const Feature = enum { |
| 174 | 202 | ventana_veyron, |
| 175 | 203 | vl_dependent_latency, |
| 176 | 204 | vxrm_pipeline_flush, |
| 205 | xaifet, | |
| 177 | 206 | xandesbfhcvt, |
| 178 | 207 | xandesperf, |
| 179 | 208 | xandesvbfhcvt, |
| ... | ... | @@ -181,6 +210,7 @@ pub const Feature = enum { |
| 181 | 210 | xandesvpackfph, |
| 182 | 211 | xandesvsinth, |
| 183 | 212 | xandesvsintload, |
| 213 | xcheriot, | |
| 184 | 214 | xcvalu, |
| 185 | 215 | xcvbi, |
| 186 | 216 | xcvbitmanip, |
| ... | ... | @@ -215,6 +245,7 @@ pub const Feature = enum { |
| 215 | 245 | xsfcease, |
| 216 | 246 | xsfmm128t, |
| 217 | 247 | xsfmm16t, |
| 248 | xsfmm32a, | |
| 218 | 249 | xsfmm32a16f, |
| 219 | 250 | xsfmm32a32f, |
| 220 | 251 | xsfmm32a8f, |
| ... | ... | @@ -236,6 +267,7 @@ pub const Feature = enum { |
| 236 | 267 | xsifivecdiscarddlone, |
| 237 | 268 | xsifivecflushdlone, |
| 238 | 269 | xsmtvdot, |
| 270 | xsmtvdotii, | |
| 239 | 271 | xtheadba, |
| 240 | 272 | xtheadbb, |
| 241 | 273 | xtheadbs, |
| ... | ... | @@ -275,8 +307,6 @@ pub const Feature = enum { |
| 275 | 307 | zcmp, |
| 276 | 308 | zcmt, |
| 277 | 309 | zdinx, |
| 278 | zexth_fusion, | |
| 279 | zextw_fusion, | |
| 280 | 310 | zfa, |
| 281 | 311 | zfbfmin, |
| 282 | 312 | zfh, |
| ... | ... | @@ -290,6 +320,7 @@ pub const Feature = enum { |
| 290 | 320 | zicboz, |
| 291 | 321 | ziccamoa, |
| 292 | 322 | ziccamoc, |
| 323 | ziccid, | |
| 293 | 324 | ziccif, |
| 294 | 325 | zicclsm, |
| 295 | 326 | ziccrse, |
| ... | ... | @@ -301,7 +332,7 @@ pub const Feature = enum { |
| 301 | 332 | zihintpause, |
| 302 | 333 | zihpm, |
| 303 | 334 | zilsd, |
| 304 | zilsd_4byte_align, | |
| 335 | zilsd_word_align, | |
| 305 | 336 | zimop, |
| 306 | 337 | zk, |
| 307 | 338 | zkn, |
| ... | ... | @@ -382,31 +413,11 @@ pub const all_features = blk: { |
| 382 | 413 | .zalrsc, |
| 383 | 414 | }), |
| 384 | 415 | }; |
| 385 | result[@backingInt(Feature.add_load_fusion)] = .{ | |
| 386 | .llvm_name = "add-load-fusion", | |
| 387 | .description = "Enable ADD(.UW) + load macrofusion", | |
| 388 | .dependencies = featureSet(&[_]Feature{}), | |
| 389 | }; | |
| 390 | result[@backingInt(Feature.addi_load_fusion)] = .{ | |
| 391 | .llvm_name = "addi-load-fusion", | |
| 392 | .description = "Enable ADDI + load macrofusion", | |
| 393 | .dependencies = featureSet(&[_]Feature{}), | |
| 394 | }; | |
| 395 | 416 | result[@backingInt(Feature.andes45)] = .{ |
| 396 | 417 | .llvm_name = "andes45", |
| 397 | 418 | .description = "Andes 45-Series processors", |
| 398 | 419 | .dependencies = featureSet(&[_]Feature{}), |
| 399 | 420 | }; |
| 400 | result[@backingInt(Feature.auipc_addi_fusion)] = .{ | |
| 401 | .llvm_name = "auipc-addi-fusion", | |
| 402 | .description = "Enable AUIPC+ADDI macrofusion", | |
| 403 | .dependencies = featureSet(&[_]Feature{}), | |
| 404 | }; | |
| 405 | result[@backingInt(Feature.auipc_load_fusion)] = .{ | |
| 406 | .llvm_name = "auipc-load-fusion", | |
| 407 | .description = "Enable AUIPC + load macrofusion", | |
| 408 | .dependencies = featureSet(&[_]Feature{}), | |
| 409 | }; | |
| 410 | 421 | result[@backingInt(Feature.b)] = .{ |
| 411 | 422 | .llvm_name = "b", |
| 412 | 423 | .description = "'B' (the collection of the Zba, Zbb, Zbs extensions)", |
| ... | ... | @@ -416,11 +427,6 @@ pub const all_features = blk: { |
| 416 | 427 | .zbs, |
| 417 | 428 | }), |
| 418 | 429 | }; |
| 419 | result[@backingInt(Feature.bfext_fusion)] = .{ | |
| 420 | .llvm_name = "bfext-fusion", | |
| 421 | .description = "Enable SLLI+SRLI (bitfield extract) macrofusion", | |
| 422 | .dependencies = featureSet(&[_]Feature{}), | |
| 423 | }; | |
| 424 | 430 | result[@backingInt(Feature.c)] = .{ |
| 425 | 431 | .llvm_name = "c", |
| 426 | 432 | .description = "'C' (Compressed Instructions)", |
| ... | ... | @@ -475,6 +481,11 @@ pub const all_features = blk: { |
| 475 | 481 | .description = "'E' (Embedded Instruction Set with 16 GPRs)", |
| 476 | 482 | .dependencies = featureSet(&[_]Feature{}), |
| 477 | 483 | }; |
| 484 | result[@backingInt(Feature.enable_select_opt)] = .{ | |
| 485 | .llvm_name = "enable-select-opt", | |
| 486 | .description = "Enable the select optimize pass for select loop heuristics", | |
| 487 | .dependencies = featureSet(&[_]Feature{}), | |
| 488 | }; | |
| 478 | 489 | result[@backingInt(Feature.enable_vsetvli_sched_heuristic)] = .{ |
| 479 | 490 | .llvm_name = "enable-vsetvli-sched-heuristic", |
| 480 | 491 | .description = "Enable vsetvli-based scheduling heuristic", |
| ... | ... | @@ -493,7 +504,12 @@ pub const all_features = blk: { |
| 493 | 504 | result[@backingInt(Feature.experimental_p)] = .{ |
| 494 | 505 | .llvm_name = "experimental-p", |
| 495 | 506 | .description = "'P' ('Base P' (Packed SIMD))", |
| 496 | .dependencies = featureSet(&[_]Feature{}), | |
| 507 | .dependencies = featureSet(&[_]Feature{ | |
| 508 | .zba, | |
| 509 | .zbb, | |
| 510 | .zicsr, | |
| 511 | .zmmul, | |
| 512 | }), | |
| 497 | 513 | }; |
| 498 | 514 | result[@backingInt(Feature.experimental_rvm23u32)] = .{ |
| 499 | 515 | .llvm_name = "experimental-rvm23u32", |
| ... | ... | @@ -522,15 +538,13 @@ pub const all_features = blk: { |
| 522 | 538 | .description = "'Svukte' (Address-Independent Latency of User-Mode Faults to Supervisor Addresses)", |
| 523 | 539 | .dependencies = featureSet(&[_]Feature{}), |
| 524 | 540 | }; |
| 525 | result[@backingInt(Feature.experimental_xrivosvisni)] = .{ | |
| 526 | .llvm_name = "experimental-xrivosvisni", | |
| 527 | .description = "'XRivosVisni' (Rivos Vector Integer Small New)", | |
| 528 | .dependencies = featureSet(&[_]Feature{}), | |
| 529 | }; | |
| 530 | result[@backingInt(Feature.experimental_xrivosvizip)] = .{ | |
| 531 | .llvm_name = "experimental-xrivosvizip", | |
| 532 | .description = "'XRivosVizip' (Rivos Vector Register Zips)", | |
| 533 | .dependencies = featureSet(&[_]Feature{}), | |
| 541 | result[@backingInt(Feature.experimental_xqccmt)] = .{ | |
| 542 | .llvm_name = "experimental-xqccmt", | |
| 543 | .description = "'Xqccmt' (Qualcomm 16-bit Table Jump)", | |
| 544 | .dependencies = featureSet(&[_]Feature{ | |
| 545 | .zca, | |
| 546 | .zicsr, | |
| 547 | }), | |
| 534 | 548 | }; |
| 535 | 549 | result[@backingInt(Feature.experimental_xsfmclic)] = .{ |
| 536 | 550 | .llvm_name = "experimental-xsfmclic", |
| ... | ... | @@ -542,6 +556,11 @@ pub const all_features = blk: { |
| 542 | 556 | .description = "'XSfsclic' (SiFive CLIC Supervisor-mode CSRs)", |
| 543 | 557 | .dependencies = featureSet(&[_]Feature{}), |
| 544 | 558 | }; |
| 559 | result[@backingInt(Feature.experimental_y)] = .{ | |
| 560 | .llvm_name = "experimental-y", | |
| 561 | .description = "'Y' ('Base Y' (CHERI))", | |
| 562 | .dependencies = featureSet(&[_]Feature{}), | |
| 563 | }; | |
| 545 | 564 | result[@backingInt(Feature.experimental_zibi)] = .{ |
| 546 | 565 | .llvm_name = "experimental-zibi", |
| 547 | 566 | .description = "'Zibi' (Branch with Immediate)", |
| ... | ... | @@ -562,6 +581,11 @@ pub const all_features = blk: { |
| 562 | 581 | .zimop, |
| 563 | 582 | }), |
| 564 | 583 | }; |
| 584 | result[@backingInt(Feature.experimental_zvabd)] = .{ | |
| 585 | .llvm_name = "experimental-zvabd", | |
| 586 | .description = "'Zvabd' (Vector Absolute Difference)", | |
| 587 | .dependencies = featureSet(&[_]Feature{}), | |
| 588 | }; | |
| 565 | 589 | result[@backingInt(Feature.experimental_zvbc32e)] = .{ |
| 566 | 590 | .llvm_name = "experimental-zvbc32e", |
| 567 | 591 | .description = "'Zvbc32e' (Vector Carryless Multiplication with 32-bits elements)", |
| ... | ... | @@ -569,6 +593,20 @@ pub const all_features = blk: { |
| 569 | 593 | .zve32x, |
| 570 | 594 | }), |
| 571 | 595 | }; |
| 596 | result[@backingInt(Feature.experimental_zvdot4a8i)] = .{ | |
| 597 | .llvm_name = "experimental-zvdot4a8i", | |
| 598 | .description = "'Zvdot4a8i' (Vector 4-element Dot Product of packed 8-bit Integers)", | |
| 599 | .dependencies = featureSet(&[_]Feature{ | |
| 600 | .zve32x, | |
| 601 | }), | |
| 602 | }; | |
| 603 | result[@backingInt(Feature.experimental_zvfbdota32f)] = .{ | |
| 604 | .llvm_name = "experimental-zvfbdota32f", | |
| 605 | .description = "'Zvfbdota32f' (FP32 batched dot-product extension)", | |
| 606 | .dependencies = featureSet(&[_]Feature{ | |
| 607 | .zve32f, | |
| 608 | }), | |
| 609 | }; | |
| 572 | 610 | result[@backingInt(Feature.experimental_zvfbfa)] = .{ |
| 573 | 611 | .llvm_name = "experimental-zvfbfa", |
| 574 | 612 | .description = "'Zvfbfa' (Additional BF16 vector compute support)", |
| ... | ... | @@ -584,6 +622,34 @@ pub const all_features = blk: { |
| 584 | 622 | .zve32f, |
| 585 | 623 | }), |
| 586 | 624 | }; |
| 625 | result[@backingInt(Feature.experimental_zvfqwbdota8f)] = .{ | |
| 626 | .llvm_name = "experimental-zvfqwbdota8f", | |
| 627 | .description = "'Zvfqwbdota8f' (OCP FP8 batched dot-product extension)", | |
| 628 | .dependencies = featureSet(&[_]Feature{ | |
| 629 | .zve32f, | |
| 630 | }), | |
| 631 | }; | |
| 632 | result[@backingInt(Feature.experimental_zvfqwdota8f)] = .{ | |
| 633 | .llvm_name = "experimental-zvfqwdota8f", | |
| 634 | .description = "'Zvfqwdota8f' (OCP FP8 Dot-Product)", | |
| 635 | .dependencies = featureSet(&[_]Feature{ | |
| 636 | .zve32f, | |
| 637 | }), | |
| 638 | }; | |
| 639 | result[@backingInt(Feature.experimental_zvfwbdota16bf)] = .{ | |
| 640 | .llvm_name = "experimental-zvfwbdota16bf", | |
| 641 | .description = "'Zvfwbdota16bf' (BF16 batched dot-product extension)", | |
| 642 | .dependencies = featureSet(&[_]Feature{ | |
| 643 | .zve32f, | |
| 644 | }), | |
| 645 | }; | |
| 646 | result[@backingInt(Feature.experimental_zvfwdota16bf)] = .{ | |
| 647 | .llvm_name = "experimental-zvfwdota16bf", | |
| 648 | .description = "'Zvfwdota16bf' (BF16 Dot-Product)", | |
| 649 | .dependencies = featureSet(&[_]Feature{ | |
| 650 | .zve32f, | |
| 651 | }), | |
| 652 | }; | |
| 587 | 653 | result[@backingInt(Feature.experimental_zvkgs)] = .{ |
| 588 | 654 | .llvm_name = "experimental-zvkgs", |
| 589 | 655 | .description = "'Zvkgs' (Vector-Scalar GCM instructions for Cryptography)", |
| ... | ... | @@ -591,9 +657,65 @@ pub const all_features = blk: { |
| 591 | 657 | .zvkg, |
| 592 | 658 | }), |
| 593 | 659 | }; |
| 594 | result[@backingInt(Feature.experimental_zvqdotq)] = .{ | |
| 595 | .llvm_name = "experimental-zvqdotq", | |
| 596 | .description = "'Zvqdotq' (Vector quad widening 4D Dot Product)", | |
| 660 | result[@backingInt(Feature.experimental_zvqwbdota16i)] = .{ | |
| 661 | .llvm_name = "experimental-zvqwbdota16i", | |
| 662 | .description = "'Zvqwbdota16i' (16-bit integer batched dot-product extension)", | |
| 663 | .dependencies = featureSet(&[_]Feature{ | |
| 664 | .zve64x, | |
| 665 | }), | |
| 666 | }; | |
| 667 | result[@backingInt(Feature.experimental_zvqwbdota8i)] = .{ | |
| 668 | .llvm_name = "experimental-zvqwbdota8i", | |
| 669 | .description = "'Zvqwbdota8i' (8-bit integer batched dot-product extension)", | |
| 670 | .dependencies = featureSet(&[_]Feature{ | |
| 671 | .zve32x, | |
| 672 | }), | |
| 673 | }; | |
| 674 | result[@backingInt(Feature.experimental_zvqwdota16i)] = .{ | |
| 675 | .llvm_name = "experimental-zvqwdota16i", | |
| 676 | .description = "'Zvqwdota16i' (16-bit Integer Dot-Product)", | |
| 677 | .dependencies = featureSet(&[_]Feature{ | |
| 678 | .zve64x, | |
| 679 | }), | |
| 680 | }; | |
| 681 | result[@backingInt(Feature.experimental_zvqwdota8i)] = .{ | |
| 682 | .llvm_name = "experimental-zvqwdota8i", | |
| 683 | .description = "'Zvqwdota8i' (8-bit Integer Dot-Product)", | |
| 684 | .dependencies = featureSet(&[_]Feature{ | |
| 685 | .zve32x, | |
| 686 | }), | |
| 687 | }; | |
| 688 | result[@backingInt(Feature.experimental_zvvfmm)] = .{ | |
| 689 | .llvm_name = "experimental-zvvfmm", | |
| 690 | .description = "'Zvvfmm' (Floating-Point Matrix Multiply-Accumulate)", | |
| 691 | .dependencies = featureSet(&[_]Feature{ | |
| 692 | .zve32f, | |
| 693 | }), | |
| 694 | }; | |
| 695 | result[@backingInt(Feature.experimental_zvvmm)] = .{ | |
| 696 | .llvm_name = "experimental-zvvmm", | |
| 697 | .description = "'Zvvmm' (Integer Matrix Multiply-Accumulate)", | |
| 698 | .dependencies = featureSet(&[_]Feature{ | |
| 699 | .zve32x, | |
| 700 | }), | |
| 701 | }; | |
| 702 | result[@backingInt(Feature.experimental_zvvmtls)] = .{ | |
| 703 | .llvm_name = "experimental-zvvmtls", | |
| 704 | .description = "'Zvvmtls' (Matrix Tile Load/Store)", | |
| 705 | .dependencies = featureSet(&[_]Feature{ | |
| 706 | .zve32x, | |
| 707 | }), | |
| 708 | }; | |
| 709 | result[@backingInt(Feature.experimental_zvvmttls)] = .{ | |
| 710 | .llvm_name = "experimental-zvvmttls", | |
| 711 | .description = "'Zvvmttls' (Transposing Matrix Tile Load/Store)", | |
| 712 | .dependencies = featureSet(&[_]Feature{ | |
| 713 | .zve32x, | |
| 714 | }), | |
| 715 | }; | |
| 716 | result[@backingInt(Feature.experimental_zvzip)] = .{ | |
| 717 | .llvm_name = "experimental-zvzip", | |
| 718 | .description = "'Zvzip' (Vector Reordering Structured Data)", | |
| 597 | 719 | .dependencies = featureSet(&[_]Feature{ |
| 598 | 720 | .zve32x, |
| 599 | 721 | }), |
| ... | ... | @@ -610,36 +732,116 @@ pub const all_features = blk: { |
| 610 | 732 | .description = "Assume that lock-free native-width atomics are available", |
| 611 | 733 | .dependencies = featureSet(&[_]Feature{}), |
| 612 | 734 | }; |
| 613 | result[@backingInt(Feature.h)] = .{ | |
| 614 | .llvm_name = "h", | |
| 615 | .description = "'H' (Hypervisor)", | |
| 735 | result[@backingInt(Feature.fusion_add_load)] = .{ | |
| 736 | .llvm_name = "fusion-add-load", | |
| 737 | .description = "Enable ADD(.UW) + load macrofusion", | |
| 616 | 738 | .dependencies = featureSet(&[_]Feature{}), |
| 617 | 739 | }; |
| 618 | result[@backingInt(Feature.i)] = .{ | |
| 619 | .llvm_name = "i", | |
| 620 | .description = "'I' (Base Integer Instruction Set)", | |
| 740 | result[@backingInt(Feature.fusion_add_mem)] = .{ | |
| 741 | .llvm_name = "fusion-add-mem", | |
| 742 | .description = "Enable ADD+LOAD/STORE macrofusion", | |
| 743 | .dependencies = featureSet(&[_]Feature{}), | |
| 744 | }; | |
| 745 | result[@backingInt(Feature.fusion_addi_load)] = .{ | |
| 746 | .llvm_name = "fusion-addi-load", | |
| 747 | .description = "Enable ADDI + load macrofusion", | |
| 748 | .dependencies = featureSet(&[_]Feature{}), | |
| 749 | }; | |
| 750 | result[@backingInt(Feature.fusion_auipc_addi)] = .{ | |
| 751 | .llvm_name = "fusion-auipc-addi", | |
| 752 | .description = "Enable AUIPC+ADDI macrofusion", | |
| 753 | .dependencies = featureSet(&[_]Feature{}), | |
| 754 | }; | |
| 755 | result[@backingInt(Feature.fusion_auipc_load)] = .{ | |
| 756 | .llvm_name = "fusion-auipc-load", | |
| 757 | .description = "Enable AUIPC + load macrofusion", | |
| 758 | .dependencies = featureSet(&[_]Feature{}), | |
| 759 | }; | |
| 760 | result[@backingInt(Feature.fusion_bfext)] = .{ | |
| 761 | .llvm_name = "fusion-bfext", | |
| 762 | .description = "Enable SLLI+SRLI (bitfield extract) macrofusion", | |
| 621 | 763 | .dependencies = featureSet(&[_]Feature{}), |
| 622 | 764 | }; |
| 623 | result[@backingInt(Feature.ld_add_fusion)] = .{ | |
| 624 | .llvm_name = "ld-add-fusion", | |
| 765 | result[@backingInt(Feature.fusion_ld_add)] = .{ | |
| 766 | .llvm_name = "fusion-ld-add", | |
| 625 | 767 | .description = "Enable LD+ADD macrofusion", |
| 626 | 768 | .dependencies = featureSet(&[_]Feature{}), |
| 627 | 769 | }; |
| 628 | result[@backingInt(Feature.log_vrgather)] = .{ | |
| 629 | .llvm_name = "log-vrgather", | |
| 630 | .description = "Has vrgather.vv with LMUL*log2(LMUL) latency", | |
| 770 | result[@backingInt(Feature.fusion_logic_imm_reg)] = .{ | |
| 771 | .llvm_name = "fusion-logic-imm-reg", | |
| 772 | .description = "Enable ANDI/ORI/XORI+AND/OR/XOR macrofusion", | |
| 773 | .dependencies = featureSet(&[_]Feature{}), | |
| 774 | }; | |
| 775 | result[@backingInt(Feature.fusion_logic_reg_imm)] = .{ | |
| 776 | .llvm_name = "fusion-logic-reg-imm", | |
| 777 | .description = "Enable AND/OR/XOR+ANDI/ORI/XORI macrofusion", | |
| 778 | .dependencies = featureSet(&[_]Feature{}), | |
| 779 | }; | |
| 780 | result[@backingInt(Feature.fusion_logic_reg_reg)] = .{ | |
| 781 | .llvm_name = "fusion-logic-reg-reg", | |
| 782 | .description = "Enable AND/OR/XOR+AND/OR/XOR macrofusion", | |
| 631 | 783 | .dependencies = featureSet(&[_]Feature{}), |
| 632 | 784 | }; |
| 633 | result[@backingInt(Feature.lui_addi_fusion)] = .{ | |
| 634 | .llvm_name = "lui-addi-fusion", | |
| 785 | result[@backingInt(Feature.fusion_lui_addi)] = .{ | |
| 786 | .llvm_name = "fusion-lui-addi", | |
| 635 | 787 | .description = "Enable LUI+ADDI macro fusion", |
| 636 | 788 | .dependencies = featureSet(&[_]Feature{}), |
| 637 | 789 | }; |
| 638 | result[@backingInt(Feature.lui_load_fusion)] = .{ | |
| 639 | .llvm_name = "lui-load-fusion", | |
| 790 | result[@backingInt(Feature.fusion_lui_load)] = .{ | |
| 791 | .llvm_name = "fusion-lui-load", | |
| 640 | 792 | .description = "Enable LUI + load macrofusion", |
| 641 | 793 | .dependencies = featureSet(&[_]Feature{}), |
| 642 | 794 | }; |
| 795 | result[@backingInt(Feature.fusion_mul_add)] = .{ | |
| 796 | .llvm_name = "fusion-mul-add", | |
| 797 | .description = "Enable MUL+ADD macrofusion", | |
| 798 | .dependencies = featureSet(&[_]Feature{}), | |
| 799 | }; | |
| 800 | result[@backingInt(Feature.fusion_qc_e_li_load_store)] = .{ | |
| 801 | .llvm_name = "fusion-qc-e-li-load-store", | |
| 802 | .description = "Enable QC.E.LI + Load/Store macrofusion", | |
| 803 | .dependencies = featureSet(&[_]Feature{}), | |
| 804 | }; | |
| 805 | result[@backingInt(Feature.fusion_shift_bit_extract)] = .{ | |
| 806 | .llvm_name = "fusion-shift-bit-extract", | |
| 807 | .description = "Enable SLLI+SRLI/SRAI macrofusion", | |
| 808 | .dependencies = featureSet(&[_]Feature{}), | |
| 809 | }; | |
| 810 | result[@backingInt(Feature.fusion_shifted_zextw)] = .{ | |
| 811 | .llvm_name = "fusion-shifted-zextw", | |
| 812 | .description = "Enable SLLI+SRLI to be fused when computing (shifted) word zero extension", | |
| 813 | .dependencies = featureSet(&[_]Feature{}), | |
| 814 | }; | |
| 815 | result[@backingInt(Feature.fusion_shxadd_load)] = .{ | |
| 816 | .llvm_name = "fusion-shxadd-load", | |
| 817 | .description = "Enable SH(1|2|3)ADD(.UW) + load macrofusion", | |
| 818 | .dependencies = featureSet(&[_]Feature{}), | |
| 819 | }; | |
| 820 | result[@backingInt(Feature.fusion_zexth)] = .{ | |
| 821 | .llvm_name = "fusion-zexth", | |
| 822 | .description = "Enable SLLI+SRLI to be fused to zero extension of halfword", | |
| 823 | .dependencies = featureSet(&[_]Feature{}), | |
| 824 | }; | |
| 825 | result[@backingInt(Feature.fusion_zextw)] = .{ | |
| 826 | .llvm_name = "fusion-zextw", | |
| 827 | .description = "Enable SLLI+SRLI to be fused to zero extension of word", | |
| 828 | .dependencies = featureSet(&[_]Feature{}), | |
| 829 | }; | |
| 830 | result[@backingInt(Feature.h)] = .{ | |
| 831 | .llvm_name = "h", | |
| 832 | .description = "'H' (Hypervisor)", | |
| 833 | .dependencies = featureSet(&[_]Feature{}), | |
| 834 | }; | |
| 835 | result[@backingInt(Feature.i)] = .{ | |
| 836 | .llvm_name = "i", | |
| 837 | .description = "'I' (Base Integer Instruction Set)", | |
| 838 | .dependencies = featureSet(&[_]Feature{}), | |
| 839 | }; | |
| 840 | result[@backingInt(Feature.log_vrgather)] = .{ | |
| 841 | .llvm_name = "log-vrgather", | |
| 842 | .description = "Has vrgather.vv with LMUL*log2(LMUL) latency", | |
| 843 | .dependencies = featureSet(&[_]Feature{}), | |
| 844 | }; | |
| 643 | 845 | result[@backingInt(Feature.m)] = .{ |
| 644 | 846 | .llvm_name = "m", |
| 645 | 847 | .description = "'M' (Integer Multiplication and Division)", |
| ... | ... | @@ -654,7 +856,7 @@ pub const all_features = blk: { |
| 654 | 856 | }; |
| 655 | 857 | result[@backingInt(Feature.no_default_unroll)] = .{ |
| 656 | 858 | .llvm_name = "no-default-unroll", |
| 657 | .description = "Disable default unroll preference.", | |
| 859 | .description = "Disable default unroll preference", | |
| 658 | 860 | .dependencies = featureSet(&[_]Feature{}), |
| 659 | 861 | }; |
| 660 | 862 | result[@backingInt(Feature.no_sink_splat_operands)] = .{ |
| ... | ... | @@ -664,7 +866,7 @@ pub const all_features = blk: { |
| 664 | 866 | }; |
| 665 | 867 | result[@backingInt(Feature.no_trailing_seq_cst_fence)] = .{ |
| 666 | 868 | .llvm_name = "no-trailing-seq-cst-fence", |
| 667 | .description = "Disable trailing fence for seq-cst store.", | |
| 869 | .description = "Disable trailing fence for seq-cst store", | |
| 668 | 870 | .dependencies = featureSet(&[_]Feature{}), |
| 669 | 871 | }; |
| 670 | 872 | result[@backingInt(Feature.optimized_nf2_segment_load_store)] = .{ |
| ... | ... | @@ -717,6 +919,11 @@ pub const all_features = blk: { |
| 717 | 919 | .description = "Prefer likely predicted branches over selects", |
| 718 | 920 | .dependencies = featureSet(&[_]Feature{}), |
| 719 | 921 | }; |
| 922 | result[@backingInt(Feature.prefer_ascending_load_store)] = .{ | |
| 923 | .llvm_name = "prefer-ascending-load-store", | |
| 924 | .description = "Prefer ascending load/store order for better clustering and performance", | |
| 925 | .dependencies = featureSet(&[_]Feature{}), | |
| 926 | }; | |
| 720 | 927 | result[@backingInt(Feature.prefer_vsetvli_over_read_vlenb)] = .{ |
| 721 | 928 | .llvm_name = "prefer-vsetvli-over-read-vlenb", |
| 722 | 929 | .description = "Prefer vsetvli over read vlenb CSR to calculate VLEN", |
| ... | ... | @@ -736,7 +943,7 @@ pub const all_features = blk: { |
| 736 | 943 | }; |
| 737 | 944 | result[@backingInt(Feature.relax)] = .{ |
| 738 | 945 | .llvm_name = "relax", |
| 739 | .description = "Enable Linker relaxation.", | |
| 946 | .description = "Enable Linker relaxation", | |
| 740 | 947 | .dependencies = featureSet(&[_]Feature{}), |
| 741 | 948 | }; |
| 742 | 949 | result[@backingInt(Feature.reserve_x1)] = .{ |
| ... | ... | @@ -1187,7 +1394,7 @@ pub const all_features = blk: { |
| 1187 | 1394 | }; |
| 1188 | 1395 | result[@backingInt(Feature.save_restore)] = .{ |
| 1189 | 1396 | .llvm_name = "save-restore", |
| 1190 | .description = "Enable save/restore.", | |
| 1397 | .description = "Enable save/restore", | |
| 1191 | 1398 | .dependencies = featureSet(&[_]Feature{}), |
| 1192 | 1399 | }; |
| 1193 | 1400 | result[@backingInt(Feature.sdext)] = .{ |
| ... | ... | @@ -1224,11 +1431,6 @@ pub const all_features = blk: { |
| 1224 | 1431 | .description = "'Shgatpa' (SvNNx4 mode supported for all modes supported by satp, as well as Bare)", |
| 1225 | 1432 | .dependencies = featureSet(&[_]Feature{}), |
| 1226 | 1433 | }; |
| 1227 | result[@backingInt(Feature.shifted_zextw_fusion)] = .{ | |
| 1228 | .llvm_name = "shifted-zextw-fusion", | |
| 1229 | .description = "Enable SLLI+SRLI to be fused when computing (shifted) word zero extension", | |
| 1230 | .dependencies = featureSet(&[_]Feature{}), | |
| 1231 | }; | |
| 1232 | 1434 | result[@backingInt(Feature.shlcofideleg)] = .{ |
| 1233 | 1435 | .llvm_name = "shlcofideleg", |
| 1234 | 1436 | .description = "'Shlcofideleg' (Delegating LCOFI Interrupts to VS-mode)", |
| ... | ... | @@ -1253,6 +1455,13 @@ pub const all_features = blk: { |
| 1253 | 1455 | .short_forward_branch_ialu, |
| 1254 | 1456 | }), |
| 1255 | 1457 | }; |
| 1458 | result[@backingInt(Feature.short_forward_branch_imm)] = .{ | |
| 1459 | .llvm_name = "short-forward-branch-imm", | |
| 1460 | .description = "Enable short forward branch optimization for branches with immediates", | |
| 1461 | .dependencies = featureSet(&[_]Feature{ | |
| 1462 | .short_forward_branch_ialu, | |
| 1463 | }), | |
| 1464 | }; | |
| 1256 | 1465 | result[@backingInt(Feature.short_forward_branch_imul)] = .{ |
| 1257 | 1466 | .llvm_name = "short-forward-branch-imul", |
| 1258 | 1467 | .description = "Enable short forward branch optimization for MUL instruction", |
| ... | ... | @@ -1280,11 +1489,6 @@ pub const all_features = blk: { |
| 1280 | 1489 | .description = "'Shvstvecd' (vstvec supports Direct mode)", |
| 1281 | 1490 | .dependencies = featureSet(&[_]Feature{}), |
| 1282 | 1491 | }; |
| 1283 | result[@backingInt(Feature.shxadd_load_fusion)] = .{ | |
| 1284 | .llvm_name = "shxadd-load-fusion", | |
| 1285 | .description = "Enable SH(1|2|3)ADD(.UW) + load macrofusion", | |
| 1286 | .dependencies = featureSet(&[_]Feature{}), | |
| 1287 | }; | |
| 1288 | 1492 | result[@backingInt(Feature.single_element_vec_fp64)] = .{ |
| 1289 | 1493 | .llvm_name = "single-element-vec-fp64", |
| 1290 | 1494 | .description = "Certain vector FP64 operations produce a single result element per cycle", |
| ... | ... | @@ -1473,6 +1677,11 @@ pub const all_features = blk: { |
| 1473 | 1677 | .description = "'Svpbmt' (Page-Based Memory Types)", |
| 1474 | 1678 | .dependencies = featureSet(&[_]Feature{}), |
| 1475 | 1679 | }; |
| 1680 | result[@backingInt(Feature.svrsw60t59b)] = .{ | |
| 1681 | .llvm_name = "svrsw60t59b", | |
| 1682 | .description = "'Svrsw60t59b' (PTE Reserved-for-Software Bits 60-59)", | |
| 1683 | .dependencies = featureSet(&[_]Feature{}), | |
| 1684 | }; | |
| 1476 | 1685 | result[@backingInt(Feature.svvptc)] = .{ |
| 1477 | 1686 | .llvm_name = "svvptc", |
| 1478 | 1687 | .description = "'Svvptc' (Obviating Memory-Management Instructions after Marking PTEs Valid)", |
| ... | ... | @@ -1521,6 +1730,13 @@ pub const all_features = blk: { |
| 1521 | 1730 | .description = "VXRM writes causes pipeline flush", |
| 1522 | 1731 | .dependencies = featureSet(&[_]Feature{}), |
| 1523 | 1732 | }; |
| 1733 | result[@backingInt(Feature.xaifet)] = .{ | |
| 1734 | .llvm_name = "xaifet", | |
| 1735 | .description = "'XAIFET' (AI Foundry ET Extension)", | |
| 1736 | .dependencies = featureSet(&[_]Feature{ | |
| 1737 | .f, | |
| 1738 | }), | |
| 1739 | }; | |
| 1524 | 1740 | result[@backingInt(Feature.xandesbfhcvt)] = .{ |
| 1525 | 1741 | .llvm_name = "xandesbfhcvt", |
| 1526 | 1742 | .description = "'XAndesBFHCvt' (Andes Scalar BFLOAT16 Conversion Extension)", |
| ... | ... | @@ -1568,6 +1784,15 @@ pub const all_features = blk: { |
| 1568 | 1784 | .zve32x, |
| 1569 | 1785 | }), |
| 1570 | 1786 | }; |
| 1787 | result[@backingInt(Feature.xcheriot)] = .{ | |
| 1788 | .llvm_name = "xcheriot", | |
| 1789 | .description = "'XCheriot' (CHERIoT extension)", | |
| 1790 | .dependencies = featureSet(&[_]Feature{ | |
| 1791 | .e, | |
| 1792 | .m, | |
| 1793 | .zca, | |
| 1794 | }), | |
| 1795 | }; | |
| 1571 | 1796 | result[@backingInt(Feature.xcvalu)] = .{ |
| 1572 | 1797 | .llvm_name = "xcvalu", |
| 1573 | 1798 | .description = "'XCValu' (CORE-V ALU Operations)", |
| ... | ... | @@ -1787,6 +2012,15 @@ pub const all_features = blk: { |
| 1787 | 2012 | .zvl64b, |
| 1788 | 2013 | }), |
| 1789 | 2014 | }; |
| 2015 | result[@backingInt(Feature.xsfmm32a)] = .{ | |
| 2016 | .llvm_name = "xsfmm32a", | |
| 2017 | .description = "'XSfmm32a' (TEW=32-bit accumulation operands - int: 8b; float: fp16, bf16, fp32)", | |
| 2018 | .dependencies = featureSet(&[_]Feature{ | |
| 2019 | .xsfmm32a16f, | |
| 2020 | .xsfmm32a32f, | |
| 2021 | .xsfmm32a8i, | |
| 2022 | }), | |
| 2023 | }; | |
| 1790 | 2024 | result[@backingInt(Feature.xsfmm32a16f)] = .{ |
| 1791 | 2025 | .llvm_name = "xsfmm32a16f", |
| 1792 | 2026 | .description = "'XSfmm32a16f' (TEW=32-bit accumulation, operands - float: 16b, widen=2 (IEEE, BF))", |
| ... | ... | @@ -1938,6 +2172,13 @@ pub const all_features = blk: { |
| 1938 | 2172 | .zve32f, |
| 1939 | 2173 | }), |
| 1940 | 2174 | }; |
| 2175 | result[@backingInt(Feature.xsmtvdotii)] = .{ | |
| 2176 | .llvm_name = "xsmtvdotii", | |
| 2177 | .description = "'XSMTVDotII' (SpacemiT Vector Extension for Matrix 2.0)", | |
| 2178 | .dependencies = featureSet(&[_]Feature{ | |
| 2179 | .zve32f, | |
| 2180 | }), | |
| 2181 | }; | |
| 1941 | 2182 | result[@backingInt(Feature.xtheadba)] = .{ |
| 1942 | 2183 | .llvm_name = "xtheadba", |
| 1943 | 2184 | .description = "'XTHeadBa' (T-Head address calculation instructions)", |
| ... | ... | @@ -2069,7 +2310,9 @@ pub const all_features = blk: { |
| 2069 | 2310 | result[@backingInt(Feature.zbc)] = .{ |
| 2070 | 2311 | .llvm_name = "zbc", |
| 2071 | 2312 | .description = "'Zbc' (Carry-Less Multiplication)", |
| 2072 | .dependencies = featureSet(&[_]Feature{}), | |
| 2313 | .dependencies = featureSet(&[_]Feature{ | |
| 2314 | .zbkc, | |
| 2315 | }), | |
| 2073 | 2316 | }; |
| 2074 | 2317 | result[@backingInt(Feature.zbkb)] = .{ |
| 2075 | 2318 | .llvm_name = "zbkb", |
| ... | ... | @@ -2165,16 +2408,6 @@ pub const all_features = blk: { |
| 2165 | 2408 | .zfinx, |
| 2166 | 2409 | }), |
| 2167 | 2410 | }; |
| 2168 | result[@backingInt(Feature.zexth_fusion)] = .{ | |
| 2169 | .llvm_name = "zexth-fusion", | |
| 2170 | .description = "Enable SLLI+SRLI to be fused to zero extension of halfword", | |
| 2171 | .dependencies = featureSet(&[_]Feature{}), | |
| 2172 | }; | |
| 2173 | result[@backingInt(Feature.zextw_fusion)] = .{ | |
| 2174 | .llvm_name = "zextw-fusion", | |
| 2175 | .description = "Enable SLLI+SRLI to be fused to zero extension of word", | |
| 2176 | .dependencies = featureSet(&[_]Feature{}), | |
| 2177 | }; | |
| 2178 | 2411 | result[@backingInt(Feature.zfa)] = .{ |
| 2179 | 2412 | .llvm_name = "zfa", |
| 2180 | 2413 | .description = "'Zfa' (Additional Floating-Point)", |
| ... | ... | @@ -2254,6 +2487,13 @@ pub const all_features = blk: { |
| 2254 | 2487 | .description = "'Ziccamoc' (Main Memory Supports Atomics in Zacas)", |
| 2255 | 2488 | .dependencies = featureSet(&[_]Feature{}), |
| 2256 | 2489 | }; |
| 2490 | result[@backingInt(Feature.ziccid)] = .{ | |
| 2491 | .llvm_name = "ziccid", | |
| 2492 | .description = "'Ziccid' (Instruction/Data Coherence and Consistency)", | |
| 2493 | .dependencies = featureSet(&[_]Feature{ | |
| 2494 | .ziccif, | |
| 2495 | }), | |
| 2496 | }; | |
| 2257 | 2497 | result[@backingInt(Feature.ziccif)] = .{ |
| 2258 | 2498 | .llvm_name = "ziccif", |
| 2259 | 2499 | .description = "'Ziccif' (Main Memory Supports Instruction Fetch with Atomicity Requirement)", |
| ... | ... | @@ -2313,8 +2553,8 @@ pub const all_features = blk: { |
| 2313 | 2553 | .description = "'Zilsd' (Load/Store Pair Instructions)", |
| 2314 | 2554 | .dependencies = featureSet(&[_]Feature{}), |
| 2315 | 2555 | }; |
| 2316 | result[@backingInt(Feature.zilsd_4byte_align)] = .{ | |
| 2317 | .llvm_name = "zilsd-4byte-align", | |
| 2556 | result[@backingInt(Feature.zilsd_word_align)] = .{ | |
| 2557 | .llvm_name = "zilsd-word-align", | |
| 2318 | 2558 | .description = "Allow 4-byte alignment for Zilsd LD/SD instructions", |
| 2319 | 2559 | .dependencies = featureSet(&[_]Feature{}), |
| 2320 | 2560 | }; |
| ... | ... | @@ -2543,6 +2783,7 @@ pub const all_features = blk: { |
| 2543 | 2783 | .description = "'Zvknhb' (Vector SHA-2 (SHA-256 and SHA-512))", |
| 2544 | 2784 | .dependencies = featureSet(&[_]Feature{ |
| 2545 | 2785 | .zve64x, |
| 2786 | .zvknha, | |
| 2546 | 2787 | }), |
| 2547 | 2788 | }; |
| 2548 | 2789 | result[@backingInt(Feature.zvks)] = .{ |
| ... | ... | @@ -2681,6 +2922,16 @@ pub const all_features = blk: { |
| 2681 | 2922 | }; |
| 2682 | 2923 | |
| 2683 | 2924 | pub const cpu = struct { |
| 2925 | pub const an_erbium: CpuModel = .{ | |
| 2926 | .name = "an_erbium", | |
| 2927 | .llvm_name = "an-erbium", | |
| 2928 | .features = featureSet(&[_]Feature{ | |
| 2929 | .@"64bit", | |
| 2930 | .c, | |
| 2931 | .m, | |
| 2932 | .xaifet, | |
| 2933 | }), | |
| 2934 | }; | |
| 2684 | 2935 | pub const andes_45_series: CpuModel = .{ |
| 2685 | 2936 | .name = "andes_45_series", |
| 2686 | 2937 | .llvm_name = "andes-45-series", |
| ... | ... | @@ -2833,6 +3084,16 @@ pub const cpu = struct { |
| 2833 | 3084 | .m, |
| 2834 | 3085 | }), |
| 2835 | 3086 | }; |
| 3087 | pub const et_soc1: CpuModel = .{ | |
| 3088 | .name = "et_soc1", | |
| 3089 | .llvm_name = "et-soc1", | |
| 3090 | .features = featureSet(&[_]Feature{ | |
| 3091 | .@"64bit", | |
| 3092 | .c, | |
| 3093 | .m, | |
| 3094 | .xaifet, | |
| 3095 | }), | |
| 3096 | }; | |
| 2836 | 3097 | pub const generic: CpuModel = .{ |
| 2837 | 3098 | .name = "generic", |
| 2838 | 3099 | .llvm_name = "generic", |
| ... | ... | @@ -2848,6 +3109,7 @@ pub const cpu = struct { |
| 2848 | 3109 | .llvm_name = "generic-rv32", |
| 2849 | 3110 | .features = featureSet(&[_]Feature{ |
| 2850 | 3111 | .@"32bit", |
| 3112 | .enable_select_opt, | |
| 2851 | 3113 | .i, |
| 2852 | 3114 | .optimized_nf2_segment_load_store, |
| 2853 | 3115 | }), |
| ... | ... | @@ -2857,6 +3119,7 @@ pub const cpu = struct { |
| 2857 | 3119 | .llvm_name = "generic-rv64", |
| 2858 | 3120 | .features = featureSet(&[_]Feature{ |
| 2859 | 3121 | .@"64bit", |
| 3122 | .enable_select_opt, | |
| 2860 | 3123 | .i, |
| 2861 | 3124 | .optimized_nf2_segment_load_store, |
| 2862 | 3125 | }), |
| ... | ... | @@ -3020,13 +3283,13 @@ pub const cpu = struct { |
| 3020 | 3283 | .features = featureSet(&[_]Feature{ |
| 3021 | 3284 | .@"64bit", |
| 3022 | 3285 | .a, |
| 3023 | .auipc_addi_fusion, | |
| 3024 | 3286 | .b, |
| 3025 | 3287 | .c, |
| 3026 | 3288 | .conditional_cmv_fusion, |
| 3027 | 3289 | .d, |
| 3290 | .fusion_auipc_addi, | |
| 3291 | .fusion_lui_addi, | |
| 3028 | 3292 | .i, |
| 3029 | .lui_addi_fusion, | |
| 3030 | 3293 | .m, |
| 3031 | 3294 | .no_default_unroll, |
| 3032 | 3295 | .unaligned_scalar_mem, |
| ... | ... | @@ -3055,12 +3318,12 @@ pub const cpu = struct { |
| 3055 | 3318 | .features = featureSet(&[_]Feature{ |
| 3056 | 3319 | .@"64bit", |
| 3057 | 3320 | .a, |
| 3058 | .auipc_addi_fusion, | |
| 3059 | 3321 | .b, |
| 3060 | 3322 | .c, |
| 3061 | 3323 | .conditional_cmv_fusion, |
| 3324 | .fusion_auipc_addi, | |
| 3325 | .fusion_lui_addi, | |
| 3062 | 3326 | .i, |
| 3063 | .lui_addi_fusion, | |
| 3064 | 3327 | .m, |
| 3065 | 3328 | .no_default_unroll, |
| 3066 | 3329 | .no_sink_splat_operands, |
| ... | ... | @@ -3099,12 +3362,13 @@ pub const cpu = struct { |
| 3099 | 3362 | .features = featureSet(&[_]Feature{ |
| 3100 | 3363 | .@"64bit", |
| 3101 | 3364 | .a, |
| 3102 | .auipc_addi_fusion, | |
| 3103 | 3365 | .c, |
| 3104 | 3366 | .conditional_cmv_fusion, |
| 3105 | 3367 | .d, |
| 3368 | .enable_select_opt, | |
| 3369 | .fusion_auipc_addi, | |
| 3370 | .fusion_lui_addi, | |
| 3106 | 3371 | .i, |
| 3107 | .lui_addi_fusion, | |
| 3108 | 3372 | .m, |
| 3109 | 3373 | .no_default_unroll, |
| 3110 | 3374 | .use_postra_scheduler, |
| ... | ... | @@ -3119,12 +3383,12 @@ pub const cpu = struct { |
| 3119 | 3383 | .features = featureSet(&[_]Feature{ |
| 3120 | 3384 | .@"64bit", |
| 3121 | 3385 | .a, |
| 3122 | .auipc_addi_fusion, | |
| 3123 | 3386 | .b, |
| 3124 | 3387 | .c, |
| 3125 | 3388 | .conditional_cmv_fusion, |
| 3389 | .fusion_auipc_addi, | |
| 3390 | .fusion_lui_addi, | |
| 3126 | 3391 | .i, |
| 3127 | .lui_addi_fusion, | |
| 3128 | 3392 | .m, |
| 3129 | 3393 | .no_default_unroll, |
| 3130 | 3394 | .no_sink_splat_operands, |
| ... | ... | @@ -3155,18 +3419,18 @@ pub const cpu = struct { |
| 3155 | 3419 | .zvksg, |
| 3156 | 3420 | }), |
| 3157 | 3421 | }; |
| 3158 | pub const sifive_p870: CpuModel = .{ | |
| 3159 | .name = "sifive_p870", | |
| 3160 | .llvm_name = "sifive-p870", | |
| 3422 | pub const sifive_p870_d: CpuModel = .{ | |
| 3423 | .name = "sifive_p870_d", | |
| 3424 | .llvm_name = "sifive-p870-d", | |
| 3161 | 3425 | .features = featureSet(&[_]Feature{ |
| 3162 | 3426 | .@"64bit", |
| 3163 | 3427 | .a, |
| 3164 | .auipc_addi_fusion, | |
| 3165 | 3428 | .b, |
| 3166 | 3429 | .c, |
| 3167 | 3430 | .conditional_cmv_fusion, |
| 3431 | .fusion_auipc_addi, | |
| 3432 | .fusion_lui_addi, | |
| 3168 | 3433 | .i, |
| 3169 | .lui_addi_fusion, | |
| 3170 | 3434 | .m, |
| 3171 | 3435 | .no_default_unroll, |
| 3172 | 3436 | .no_sink_splat_operands, |
| ... | ... | @@ -3294,6 +3558,103 @@ pub const cpu = struct { |
| 3294 | 3558 | .zifencei, |
| 3295 | 3559 | }), |
| 3296 | 3560 | }; |
| 3561 | pub const sifive_x160: CpuModel = .{ | |
| 3562 | .name = "sifive_x160", | |
| 3563 | .llvm_name = "sifive-x160", | |
| 3564 | .features = featureSet(&[_]Feature{ | |
| 3565 | .@"32bit", | |
| 3566 | .a, | |
| 3567 | .b, | |
| 3568 | .c, | |
| 3569 | .dlen_factor_2, | |
| 3570 | .experimental_zicfilp, | |
| 3571 | .experimental_zvdot4a8i, | |
| 3572 | .experimental_zvfbfa, | |
| 3573 | .i, | |
| 3574 | .m, | |
| 3575 | .no_default_unroll, | |
| 3576 | .optimized_nf2_segment_load_store, | |
| 3577 | .optimized_zero_stride_load, | |
| 3578 | .short_forward_branch_ialu, | |
| 3579 | .use_postra_scheduler, | |
| 3580 | .vl_dependent_latency, | |
| 3581 | .xsfcease, | |
| 3582 | .zawrs, | |
| 3583 | .zce, | |
| 3584 | .zcf, | |
| 3585 | .zcmop, | |
| 3586 | .zfa, | |
| 3587 | .zfh, | |
| 3588 | .zicbom, | |
| 3589 | .zicbop, | |
| 3590 | .zicboz, | |
| 3591 | .zicond, | |
| 3592 | .zifencei, | |
| 3593 | .zihintntl, | |
| 3594 | .zihintpause, | |
| 3595 | .zihpm, | |
| 3596 | .zimop, | |
| 3597 | .zkt, | |
| 3598 | .zvbb, | |
| 3599 | .zvfbfwma, | |
| 3600 | .zvfh, | |
| 3601 | .zvkt, | |
| 3602 | .zvl128b, | |
| 3603 | }), | |
| 3604 | }; | |
| 3605 | pub const sifive_x180: CpuModel = .{ | |
| 3606 | .name = "sifive_x180", | |
| 3607 | .llvm_name = "sifive-x180", | |
| 3608 | .features = featureSet(&[_]Feature{ | |
| 3609 | .@"64bit", | |
| 3610 | .a, | |
| 3611 | .b, | |
| 3612 | .c, | |
| 3613 | .dlen_factor_2, | |
| 3614 | .experimental_zicfilp, | |
| 3615 | .experimental_zvdot4a8i, | |
| 3616 | .experimental_zvfbfa, | |
| 3617 | .i, | |
| 3618 | .m, | |
| 3619 | .no_default_unroll, | |
| 3620 | .optimized_nf2_segment_load_store, | |
| 3621 | .optimized_zero_stride_load, | |
| 3622 | .short_forward_branch_ialu, | |
| 3623 | .use_postra_scheduler, | |
| 3624 | .v, | |
| 3625 | .vl_dependent_latency, | |
| 3626 | .xsfcease, | |
| 3627 | .xsfvfbfexp16e, | |
| 3628 | .xsfvfexp16e, | |
| 3629 | .xsfvfexp32e, | |
| 3630 | .xsfvfexpa, | |
| 3631 | .xsifivecflushdlone, | |
| 3632 | .za64rs, | |
| 3633 | .zawrs, | |
| 3634 | .zcb, | |
| 3635 | .zcd, | |
| 3636 | .zcmop, | |
| 3637 | .zfa, | |
| 3638 | .zfh, | |
| 3639 | .zic64b, | |
| 3640 | .zicbom, | |
| 3641 | .zicbop, | |
| 3642 | .zicboz, | |
| 3643 | .ziccamoa, | |
| 3644 | .ziccif, | |
| 3645 | .ziccrse, | |
| 3646 | .zicond, | |
| 3647 | .zifencei, | |
| 3648 | .zihintntl, | |
| 3649 | .zihintpause, | |
| 3650 | .zihpm, | |
| 3651 | .zimop, | |
| 3652 | .zkt, | |
| 3653 | .zvbb, | |
| 3654 | .zvfbfwma, | |
| 3655 | .zvkt, | |
| 3656 | }), | |
| 3657 | }; | |
| 3297 | 3658 | pub const sifive_x280: CpuModel = .{ |
| 3298 | 3659 | .name = "sifive_x280", |
| 3299 | 3660 | .llvm_name = "sifive-x280", |
| ... | ... | @@ -3372,7 +3733,7 @@ pub const cpu = struct { |
| 3372 | 3733 | }; |
| 3373 | 3734 | pub const spacemit_a100: CpuModel = .{ |
| 3374 | 3735 | .name = "spacemit_a100", |
| 3375 | .llvm_name = null, | |
| 3736 | .llvm_name = "spacemit-a100", | |
| 3376 | 3737 | .features = featureSet(&[_]Feature{ |
| 3377 | 3738 | .@"64bit", |
| 3378 | 3739 | .a, |
| ... | ... | @@ -3405,10 +3766,10 @@ pub const cpu = struct { |
| 3405 | 3766 | .unaligned_scalar_mem, |
| 3406 | 3767 | .v, |
| 3407 | 3768 | .vxrm_pipeline_flush, |
| 3769 | .xsmtvdotii, | |
| 3408 | 3770 | .za64rs, |
| 3409 | 3771 | .zawrs, |
| 3410 | 3772 | .zbc, |
| 3411 | .zbkc, | |
| 3412 | 3773 | .zcb, |
| 3413 | 3774 | .zcmop, |
| 3414 | 3775 | .zfa, |
| ... | ... | @@ -3433,7 +3794,6 @@ pub const cpu = struct { |
| 3433 | 3794 | .zvfbfwma, |
| 3434 | 3795 | .zvfh, |
| 3435 | 3796 | .zvkng, |
| 3436 | .zvknha, | |
| 3437 | 3797 | .zvksc, |
| 3438 | 3798 | .zvksg, |
| 3439 | 3799 | .zvl1024b, |
| ... | ... | @@ -3441,27 +3801,75 @@ pub const cpu = struct { |
| 3441 | 3801 | }; |
| 3442 | 3802 | pub const spacemit_x100: CpuModel = .{ |
| 3443 | 3803 | .name = "spacemit_x100", |
| 3444 | .llvm_name = null, | |
| 3804 | .llvm_name = "spacemit-x100", | |
| 3445 | 3805 | .features = featureSet(&[_]Feature{ |
| 3806 | .@"64bit", | |
| 3807 | .a, | |
| 3808 | .b, | |
| 3809 | .c, | |
| 3446 | 3810 | .dlen_factor_2, |
| 3811 | .fusion_add_mem, | |
| 3812 | .fusion_logic_imm_reg, | |
| 3813 | .fusion_logic_reg_imm, | |
| 3814 | .fusion_logic_reg_reg, | |
| 3815 | .fusion_mul_add, | |
| 3816 | .fusion_shift_bit_extract, | |
| 3817 | .i, | |
| 3818 | .m, | |
| 3819 | .no_default_unroll, | |
| 3447 | 3820 | .optimized_nf2_segment_load_store, |
| 3448 | 3821 | .optimized_nf3_segment_load_store, |
| 3449 | 3822 | .optimized_nf4_segment_load_store, |
| 3450 | .rva23s64, | |
| 3823 | .sha, | |
| 3451 | 3824 | .smepmp, |
| 3452 | 3825 | .smnpm, |
| 3453 | 3826 | .smstateen, |
| 3827 | .ssccptr, | |
| 3828 | .sscofpmf, | |
| 3829 | .sscounterenw, | |
| 3830 | .ssnpm, | |
| 3454 | 3831 | .sspm, |
| 3832 | .sstc, | |
| 3833 | .sstvala, | |
| 3834 | .sstvecd, | |
| 3835 | .ssu64xl, | |
| 3836 | .supm, | |
| 3837 | .svade, | |
| 3838 | .svbare, | |
| 3839 | .svinval, | |
| 3840 | .svnapot, | |
| 3841 | .svpbmt, | |
| 3455 | 3842 | .unaligned_scalar_mem, |
| 3843 | .v, | |
| 3456 | 3844 | .vxrm_pipeline_flush, |
| 3457 | 3845 | .xsmtvdot, |
| 3846 | .za64rs, | |
| 3847 | .zawrs, | |
| 3458 | 3848 | .zbc, |
| 3459 | .zbkc, | |
| 3849 | .zcb, | |
| 3850 | .zcmop, | |
| 3851 | .zfa, | |
| 3460 | 3852 | .zfh, |
| 3853 | .zic64b, | |
| 3854 | .zicbom, | |
| 3855 | .zicbop, | |
| 3856 | .zicboz, | |
| 3857 | .ziccamoa, | |
| 3858 | .ziccif, | |
| 3859 | .zicclsm, | |
| 3860 | .ziccrse, | |
| 3861 | .zicntr, | |
| 3862 | .zicond, | |
| 3863 | .zifencei, | |
| 3864 | .zihintntl, | |
| 3865 | .zihintpause, | |
| 3866 | .zihpm, | |
| 3867 | .zimop, | |
| 3868 | .zkt, | |
| 3869 | .zvbb, | |
| 3461 | 3870 | .zvfbfwma, |
| 3462 | 3871 | .zvfh, |
| 3463 | 3872 | .zvkng, |
| 3464 | .zvknha, | |
| 3465 | 3873 | .zvksc, |
| 3466 | 3874 | .zvksg, |
| 3467 | 3875 | .zvl256b, |
| ... | ... | @@ -3476,11 +3884,13 @@ pub const cpu = struct { |
| 3476 | 3884 | .b, |
| 3477 | 3885 | .c, |
| 3478 | 3886 | .dlen_factor_2, |
| 3887 | .enable_select_opt, | |
| 3479 | 3888 | .i, |
| 3480 | 3889 | .m, |
| 3481 | 3890 | .optimized_nf2_segment_load_store, |
| 3482 | 3891 | .optimized_nf3_segment_load_store, |
| 3483 | 3892 | .optimized_nf4_segment_load_store, |
| 3893 | .prefer_ascending_load_store, | |
| 3484 | 3894 | .ssccptr, |
| 3485 | 3895 | .sscofpmf, |
| 3486 | 3896 | .sscounterenw, |
| ... | ... | @@ -3498,7 +3908,6 @@ pub const cpu = struct { |
| 3498 | 3908 | .xsmtvdot, |
| 3499 | 3909 | .za64rs, |
| 3500 | 3910 | .zbc, |
| 3501 | .zbkc, | |
| 3502 | 3911 | .zfh, |
| 3503 | 3912 | .zic64b, |
| 3504 | 3913 | .zicbom, |
| ... | ... | @@ -3652,9 +4061,9 @@ pub const cpu = struct { |
| 3652 | 4061 | .zkn, |
| 3653 | 4062 | }), |
| 3654 | 4063 | }; |
| 3655 | pub const tt_ascalon_d8: CpuModel = .{ | |
| 3656 | .name = "tt_ascalon_d8", | |
| 3657 | .llvm_name = "tt-ascalon-d8", | |
| 4064 | pub const tt_ascalon_x: CpuModel = .{ | |
| 4065 | .name = "tt_ascalon_x", | |
| 4066 | .llvm_name = "tt-ascalon-x", | |
| 3658 | 4067 | .features = featureSet(&[_]Feature{ |
| 3659 | 4068 | .@"64bit", |
| 3660 | 4069 | .a, |
| ... | ... | @@ -3674,6 +4083,10 @@ pub const cpu = struct { |
| 3674 | 4083 | .optimized_zero_stride_load, |
| 3675 | 4084 | .sha, |
| 3676 | 4085 | .smaia, |
| 4086 | .smmpm, | |
| 4087 | .smnpm, | |
| 4088 | .smrnmi, | |
| 4089 | .smstateen, | |
| 3677 | 4090 | .ssaia, |
| 3678 | 4091 | .ssccptr, |
| 3679 | 4092 | .sscofpmf, |
| ... | ... | @@ -3715,6 +4128,7 @@ pub const cpu = struct { |
| 3715 | 4128 | .zihintpause, |
| 3716 | 4129 | .zihpm, |
| 3717 | 4130 | .zimop, |
| 4131 | .zkr, | |
| 3718 | 4132 | .zkt, |
| 3719 | 4133 | .zvbb, |
| 3720 | 4134 | .zvbc, |
| ... | ... | @@ -3730,27 +4144,27 @@ pub const cpu = struct { |
| 3730 | 4144 | .features = featureSet(&[_]Feature{ |
| 3731 | 4145 | .@"64bit", |
| 3732 | 4146 | .a, |
| 3733 | .add_load_fusion, | |
| 3734 | .auipc_addi_fusion, | |
| 3735 | .auipc_load_fusion, | |
| 3736 | 4147 | .c, |
| 3737 | 4148 | .d, |
| 3738 | 4149 | .disable_misched_load_clustering, |
| 3739 | 4150 | .disable_postmisched_load_clustering, |
| 3740 | 4151 | .disable_postmisched_store_clustering, |
| 4152 | .fusion_add_load, | |
| 4153 | .fusion_auipc_addi, | |
| 4154 | .fusion_auipc_load, | |
| 4155 | .fusion_lui_addi, | |
| 4156 | .fusion_lui_load, | |
| 4157 | .fusion_shifted_zextw, | |
| 4158 | .fusion_zexth, | |
| 4159 | .fusion_zextw, | |
| 3741 | 4160 | .i, |
| 3742 | .lui_addi_fusion, | |
| 3743 | .lui_load_fusion, | |
| 3744 | 4161 | .m, |
| 3745 | .shifted_zextw_fusion, | |
| 3746 | 4162 | .ventana_veyron, |
| 3747 | 4163 | .xventanacondops, |
| 3748 | 4164 | .zba, |
| 3749 | 4165 | .zbb, |
| 3750 | 4166 | .zbc, |
| 3751 | 4167 | .zbs, |
| 3752 | .zexth_fusion, | |
| 3753 | .zextw_fusion, | |
| 3754 | 4168 | .zicbom, |
| 3755 | 4169 | .zicbop, |
| 3756 | 4170 | .zicboz, |
| ... | ... | @@ -3768,11 +4182,13 @@ pub const cpu = struct { |
| 3768 | 4182 | .a, |
| 3769 | 4183 | .b, |
| 3770 | 4184 | .c, |
| 4185 | .fusion_shifted_zextw, | |
| 4186 | .fusion_zexth, | |
| 4187 | .fusion_zextw, | |
| 3771 | 4188 | .i, |
| 3772 | 4189 | .m, |
| 3773 | 4190 | .no_default_unroll, |
| 3774 | 4191 | .sha, |
| 3775 | .shifted_zextw_fusion, | |
| 3776 | 4192 | .smaia, |
| 3777 | 4193 | .smcsrind, |
| 3778 | 4194 | .smdbltrp, |
| ... | ... | @@ -3806,8 +4222,6 @@ pub const cpu = struct { |
| 3806 | 4222 | .zbc, |
| 3807 | 4223 | .zcb, |
| 3808 | 4224 | .zcmop, |
| 3809 | .zexth_fusion, | |
| 3810 | .zextw_fusion, | |
| 3811 | 4225 | .zfa, |
| 3812 | 4226 | .zfh, |
| 3813 | 4227 | .zic64b, |
| ... | ... | @@ -3841,17 +4255,17 @@ pub const cpu = struct { |
| 3841 | 4255 | .a, |
| 3842 | 4256 | .c, |
| 3843 | 4257 | .d, |
| 4258 | .fusion_shifted_zextw, | |
| 4259 | .fusion_zexth, | |
| 4260 | .fusion_zextw, | |
| 3844 | 4261 | .i, |
| 3845 | 4262 | .m, |
| 3846 | 4263 | .no_default_unroll, |
| 3847 | .shifted_zextw_fusion, | |
| 3848 | 4264 | .svinval, |
| 3849 | 4265 | .zba, |
| 3850 | 4266 | .zbb, |
| 3851 | 4267 | .zbc, |
| 3852 | 4268 | .zbs, |
| 3853 | .zexth_fusion, | |
| 3854 | .zextw_fusion, | |
| 3855 | 4269 | .zicbom, |
| 3856 | 4270 | .zicboz, |
| 3857 | 4271 | .zifencei, |
| ... | ... | @@ -3860,4 +4274,129 @@ pub const cpu = struct { |
| 3860 | 4274 | .zksh, |
| 3861 | 4275 | }), |
| 3862 | 4276 | }; |
| 4277 | pub const xt_c910v2: CpuModel = .{ | |
| 4278 | .name = "xt_c910v2", | |
| 4279 | .llvm_name = "xt-c910v2", | |
| 4280 | .features = featureSet(&[_]Feature{ | |
| 4281 | .@"64bit", | |
| 4282 | .a, | |
| 4283 | .b, | |
| 4284 | .c, | |
| 4285 | .i, | |
| 4286 | .m, | |
| 4287 | .ssccptr, | |
| 4288 | .sscofpmf, | |
| 4289 | .sscounterenw, | |
| 4290 | .sstc, | |
| 4291 | .sstvala, | |
| 4292 | .sstvecd, | |
| 4293 | .ssu64xl, | |
| 4294 | .svade, | |
| 4295 | .svbare, | |
| 4296 | .svinval, | |
| 4297 | .svnapot, | |
| 4298 | .svpbmt, | |
| 4299 | .unaligned_scalar_mem, | |
| 4300 | .use_postra_scheduler, | |
| 4301 | .vxrm_pipeline_flush, | |
| 4302 | .xtheadba, | |
| 4303 | .xtheadbb, | |
| 4304 | .xtheadbs, | |
| 4305 | .xtheadcmo, | |
| 4306 | .xtheadcondmov, | |
| 4307 | .xtheadfmemidx, | |
| 4308 | .xtheadmac, | |
| 4309 | .xtheadmemidx, | |
| 4310 | .xtheadmempair, | |
| 4311 | .xtheadsync, | |
| 4312 | .za64rs, | |
| 4313 | .zawrs, | |
| 4314 | .zbc, | |
| 4315 | .zcb, | |
| 4316 | .zcd, | |
| 4317 | .zcmop, | |
| 4318 | .zfa, | |
| 4319 | .zfbfmin, | |
| 4320 | .zfh, | |
| 4321 | .zic64b, | |
| 4322 | .zicbom, | |
| 4323 | .zicbop, | |
| 4324 | .zicboz, | |
| 4325 | .ziccamoa, | |
| 4326 | .ziccif, | |
| 4327 | .zicclsm, | |
| 4328 | .ziccrse, | |
| 4329 | .zicntr, | |
| 4330 | .zicond, | |
| 4331 | .zifencei, | |
| 4332 | .zihintntl, | |
| 4333 | .zihintpause, | |
| 4334 | .zihpm, | |
| 4335 | .zimop, | |
| 4336 | .zkt, | |
| 4337 | }), | |
| 4338 | }; | |
| 4339 | pub const xt_c920v2: CpuModel = .{ | |
| 4340 | .name = "xt_c920v2", | |
| 4341 | .llvm_name = "xt-c920v2", | |
| 4342 | .features = featureSet(&[_]Feature{ | |
| 4343 | .@"64bit", | |
| 4344 | .a, | |
| 4345 | .b, | |
| 4346 | .c, | |
| 4347 | .i, | |
| 4348 | .m, | |
| 4349 | .ssccptr, | |
| 4350 | .sscofpmf, | |
| 4351 | .sscounterenw, | |
| 4352 | .sstc, | |
| 4353 | .sstvala, | |
| 4354 | .sstvecd, | |
| 4355 | .ssu64xl, | |
| 4356 | .svade, | |
| 4357 | .svbare, | |
| 4358 | .svinval, | |
| 4359 | .svnapot, | |
| 4360 | .svpbmt, | |
| 4361 | .unaligned_scalar_mem, | |
| 4362 | .use_postra_scheduler, | |
| 4363 | .vxrm_pipeline_flush, | |
| 4364 | .xtheadba, | |
| 4365 | .xtheadbb, | |
| 4366 | .xtheadbs, | |
| 4367 | .xtheadcmo, | |
| 4368 | .xtheadcondmov, | |
| 4369 | .xtheadfmemidx, | |
| 4370 | .xtheadmac, | |
| 4371 | .xtheadmemidx, | |
| 4372 | .xtheadmempair, | |
| 4373 | .xtheadsync, | |
| 4374 | .xtheadvdot, | |
| 4375 | .za64rs, | |
| 4376 | .zawrs, | |
| 4377 | .zbc, | |
| 4378 | .zcb, | |
| 4379 | .zcd, | |
| 4380 | .zcmop, | |
| 4381 | .zfa, | |
| 4382 | .zfh, | |
| 4383 | .zic64b, | |
| 4384 | .zicbom, | |
| 4385 | .zicbop, | |
| 4386 | .zicboz, | |
| 4387 | .ziccamoa, | |
| 4388 | .ziccif, | |
| 4389 | .ziccrse, | |
| 4390 | .zicntr, | |
| 4391 | .zicond, | |
| 4392 | .zifencei, | |
| 4393 | .zihintntl, | |
| 4394 | .zihintpause, | |
| 4395 | .zihpm, | |
| 4396 | .zimop, | |
| 4397 | .zkt, | |
| 4398 | .zvfbfwma, | |
| 4399 | .zvfh, | |
| 4400 | }), | |
| 4401 | }; | |
| 3863 | 4402 | }; |
lib/std/Target/sparc.zig+1| ... | ... | @@ -462,6 +462,7 @@ pub const cpu = struct { |
| 462 | 462 | .name = "leon3", |
| 463 | 463 | .llvm_name = "leon3", |
| 464 | 464 | .features = featureSet(&[_]Feature{ |
| 465 | .hasleoncasa, | |
| 465 | 466 | .hasumacsmac, |
| 466 | 467 | .leon, |
| 467 | 468 | }), |
lib/std/Target/wasm.zig+13-4| ... | ... | @@ -19,6 +19,7 @@ pub const Feature = enum { |
| 19 | 19 | nontrapping_bulk_memory_len0, |
| 20 | 20 | nontrapping_fptoint, |
| 21 | 21 | reference_types, |
| 22 | relaxed_atomics, | |
| 22 | 23 | relaxed_simd, |
| 23 | 24 | sign_ext, |
| 24 | 25 | simd128, |
| ... | ... | @@ -60,7 +61,10 @@ pub const all_features = blk: { |
| 60 | 61 | result[@backingInt(Feature.exception_handling)] = .{ |
| 61 | 62 | .llvm_name = "exception-handling", |
| 62 | 63 | .description = "Enable Wasm exception handling", |
| 63 | .dependencies = featureSet(&[_]Feature{}), | |
| 64 | .dependencies = featureSet(&[_]Feature{ | |
| 65 | .multivalue, | |
| 66 | .reference_types, | |
| 67 | }), | |
| 64 | 68 | }; |
| 65 | 69 | result[@backingInt(Feature.extended_const)] = .{ |
| 66 | 70 | .llvm_name = "extended-const", |
| ... | ... | @@ -75,7 +79,9 @@ pub const all_features = blk: { |
| 75 | 79 | result[@backingInt(Feature.gc)] = .{ |
| 76 | 80 | .llvm_name = "gc", |
| 77 | 81 | .description = "Enable wasm gc", |
| 78 | .dependencies = featureSet(&[_]Feature{}), | |
| 82 | .dependencies = featureSet(&[_]Feature{ | |
| 83 | .reference_types, | |
| 84 | }), | |
| 79 | 85 | }; |
| 80 | 86 | result[@backingInt(Feature.multimemory)] = .{ |
| 81 | 87 | .llvm_name = "multimemory", |
| ... | ... | @@ -111,6 +117,11 @@ pub const all_features = blk: { |
| 111 | 117 | .call_indirect_overlong, |
| 112 | 118 | }), |
| 113 | 119 | }; |
| 120 | result[@backingInt(Feature.relaxed_atomics)] = .{ | |
| 121 | .llvm_name = "relaxed-atomics", | |
| 122 | .description = "Enable relaxed-atomics proposal", | |
| 123 | .dependencies = featureSet(&[_]Feature{}), | |
| 124 | }; | |
| 114 | 125 | result[@backingInt(Feature.relaxed_simd)] = .{ |
| 115 | 126 | .llvm_name = "relaxed-simd", |
| 116 | 127 | .description = "Enable relaxed-simd instructions", |
| ... | ... | @@ -156,10 +167,8 @@ pub const cpu = struct { |
| 156 | 167 | .fp16, |
| 157 | 168 | .gc, |
| 158 | 169 | .multimemory, |
| 159 | .multivalue, | |
| 160 | 170 | .mutable_globals, |
| 161 | 171 | .nontrapping_fptoint, |
| 162 | .reference_types, | |
| 163 | 172 | .relaxed_simd, |
| 164 | 173 | .sign_ext, |
| 165 | 174 | .simd128, |
lib/std/Target/x86.zig+533-16| ... | ... | @@ -20,7 +20,6 @@ pub const Feature = enum { |
| 20 | 20 | amx_fp8, |
| 21 | 21 | amx_int8, |
| 22 | 22 | amx_movrs, |
| 23 | amx_tf32, | |
| 24 | 23 | amx_tile, |
| 25 | 24 | avx, |
| 26 | 25 | avx10_1, |
| ... | ... | @@ -28,6 +27,7 @@ pub const Feature = enum { |
| 28 | 27 | avx2, |
| 29 | 28 | avx512bf16, |
| 30 | 29 | avx512bitalg, |
| 30 | avx512bmm, | |
| 31 | 31 | avx512bw, |
| 32 | 32 | avx512cd, |
| 33 | 33 | avx512dq, |
| ... | ... | @@ -67,13 +67,17 @@ pub const Feature = enum { |
| 67 | 67 | enqcmd, |
| 68 | 68 | ermsb, |
| 69 | 69 | f16c, |
| 70 | false_deps_bls, | |
| 71 | false_deps_compress, | |
| 72 | false_deps_expand, | |
| 70 | 73 | false_deps_getmant, |
| 71 | false_deps_lzcnt_tzcnt, | |
| 74 | false_deps_lzcnt, | |
| 72 | 75 | false_deps_mulc, |
| 73 | 76 | false_deps_mullq, |
| 74 | 77 | false_deps_perm, |
| 75 | 78 | false_deps_popcnt, |
| 76 | 79 | false_deps_range, |
| 80 | false_deps_tzcnt, | |
| 77 | 81 | fast_11bytenop, |
| 78 | 82 | fast_15bytenop, |
| 79 | 83 | fast_7bytenop, |
| ... | ... | @@ -105,6 +109,7 @@ pub const Feature = enum { |
| 105 | 109 | idivq_to_divl, |
| 106 | 110 | inline_asm_use_gpr32, |
| 107 | 111 | invpcid, |
| 112 | jmpabs, | |
| 108 | 113 | kl, |
| 109 | 114 | lea_sp, |
| 110 | 115 | lea_uses_ag, |
| ... | ... | @@ -137,6 +142,7 @@ pub const Feature = enum { |
| 137 | 142 | prefer_legacy_setcc, |
| 138 | 143 | prefer_mask_registers, |
| 139 | 144 | prefer_movmsk_over_vtest, |
| 145 | prefer_ndd_mem, | |
| 140 | 146 | prefer_no_gather, |
| 141 | 147 | prefer_no_scatter, |
| 142 | 148 | prefetchi, |
| ... | ... | @@ -149,6 +155,31 @@ pub const Feature = enum { |
| 149 | 155 | rdpru, |
| 150 | 156 | rdrnd, |
| 151 | 157 | rdseed, |
| 158 | reserve_edi, | |
| 159 | reserve_r10, | |
| 160 | reserve_r11, | |
| 161 | reserve_r12, | |
| 162 | reserve_r13, | |
| 163 | reserve_r14, | |
| 164 | reserve_r15, | |
| 165 | reserve_r16, | |
| 166 | reserve_r17, | |
| 167 | reserve_r18, | |
| 168 | reserve_r19, | |
| 169 | reserve_r20, | |
| 170 | reserve_r21, | |
| 171 | reserve_r22, | |
| 172 | reserve_r23, | |
| 173 | reserve_r24, | |
| 174 | reserve_r25, | |
| 175 | reserve_r26, | |
| 176 | reserve_r27, | |
| 177 | reserve_r28, | |
| 178 | reserve_r29, | |
| 179 | reserve_r30, | |
| 180 | reserve_r31, | |
| 181 | reserve_r8, | |
| 182 | reserve_r9, | |
| 152 | 183 | retpoline, |
| 153 | 184 | retpoline_external_thunk, |
| 154 | 185 | retpoline_indirect_branches, |
| ... | ... | @@ -164,6 +195,7 @@ pub const Feature = enum { |
| 164 | 195 | shstk, |
| 165 | 196 | slow_3ops_lea, |
| 166 | 197 | slow_incdec, |
| 198 | slow_indirect_call, | |
| 167 | 199 | slow_lea, |
| 168 | 200 | slow_pmaddwd, |
| 169 | 201 | slow_pmulld, |
| ... | ... | @@ -313,13 +345,6 @@ pub const all_features = blk: { |
| 313 | 345 | .amx_tile, |
| 314 | 346 | }), |
| 315 | 347 | }; |
| 316 | result[@backingInt(Feature.amx_tf32)] = .{ | |
| 317 | .llvm_name = "amx-tf32", | |
| 318 | .description = "Support AMX-TF32 instructions", | |
| 319 | .dependencies = featureSet(&[_]Feature{ | |
| 320 | .amx_tile, | |
| 321 | }), | |
| 322 | }; | |
| 323 | 348 | result[@backingInt(Feature.amx_tile)] = .{ |
| 324 | 349 | .llvm_name = "amx-tile", |
| 325 | 350 | .description = "Support AMX-TILE instructions", |
| ... | ... | @@ -377,6 +402,13 @@ pub const all_features = blk: { |
| 377 | 402 | .avx512bw, |
| 378 | 403 | }), |
| 379 | 404 | }; |
| 405 | result[@backingInt(Feature.avx512bmm)] = .{ | |
| 406 | .llvm_name = "avx512bmm", | |
| 407 | .description = "Enable AVX512 Bit Matrix Multiply", | |
| 408 | .dependencies = featureSet(&[_]Feature{ | |
| 409 | .avx512bw, | |
| 410 | }), | |
| 411 | }; | |
| 380 | 412 | result[@backingInt(Feature.avx512bw)] = .{ |
| 381 | 413 | .llvm_name = "avx512bw", |
| 382 | 414 | .description = "Enable AVX-512 Byte and Word Instructions", |
| ... | ... | @@ -616,14 +648,29 @@ pub const all_features = blk: { |
| 616 | 648 | .avx, |
| 617 | 649 | }), |
| 618 | 650 | }; |
| 651 | result[@backingInt(Feature.false_deps_bls)] = .{ | |
| 652 | .llvm_name = "false-deps-bls", | |
| 653 | .description = "BLSR/BLSI/BLSMSK have a false dependency on dest register", | |
| 654 | .dependencies = featureSet(&[_]Feature{}), | |
| 655 | }; | |
| 656 | result[@backingInt(Feature.false_deps_compress)] = .{ | |
| 657 | .llvm_name = "false-deps-compress", | |
| 658 | .description = "V[P]COMPRESSPD/PS/Q/D/W/B with zero-masking has a false dependency on dest register", | |
| 659 | .dependencies = featureSet(&[_]Feature{}), | |
| 660 | }; | |
| 661 | result[@backingInt(Feature.false_deps_expand)] = .{ | |
| 662 | .llvm_name = "false-deps-expand", | |
| 663 | .description = "V[P]EXPANDPD/PS/Q/D/W/B with zero-masking has a false dependency on dest register", | |
| 664 | .dependencies = featureSet(&[_]Feature{}), | |
| 665 | }; | |
| 619 | 666 | result[@backingInt(Feature.false_deps_getmant)] = .{ |
| 620 | 667 | .llvm_name = "false-deps-getmant", |
| 621 | 668 | .description = "VGETMANTSS/SD/SH and VGETMANDPS/PD(memory version) has a false dependency on dest register", |
| 622 | 669 | .dependencies = featureSet(&[_]Feature{}), |
| 623 | 670 | }; |
| 624 | result[@backingInt(Feature.false_deps_lzcnt_tzcnt)] = .{ | |
| 625 | .llvm_name = "false-deps-lzcnt-tzcnt", | |
| 626 | .description = "LZCNT/TZCNT have a false dependency on dest register", | |
| 671 | result[@backingInt(Feature.false_deps_lzcnt)] = .{ | |
| 672 | .llvm_name = "false-deps-lzcnt", | |
| 673 | .description = "LZCNT has a false dependency on dest register", | |
| 627 | 674 | .dependencies = featureSet(&[_]Feature{}), |
| 628 | 675 | }; |
| 629 | 676 | result[@backingInt(Feature.false_deps_mulc)] = .{ |
| ... | ... | @@ -651,6 +698,11 @@ pub const all_features = blk: { |
| 651 | 698 | .description = "VRANGEPD/PS/SD/SS has a false dependency on dest register", |
| 652 | 699 | .dependencies = featureSet(&[_]Feature{}), |
| 653 | 700 | }; |
| 701 | result[@backingInt(Feature.false_deps_tzcnt)] = .{ | |
| 702 | .llvm_name = "false-deps-tzcnt", | |
| 703 | .description = "TZCNT has a false dependency on dest register", | |
| 704 | .dependencies = featureSet(&[_]Feature{}), | |
| 705 | }; | |
| 654 | 706 | result[@backingInt(Feature.fast_11bytenop)] = .{ |
| 655 | 707 | .llvm_name = "fast-11bytenop", |
| 656 | 708 | .description = "Target can quickly decode up to 11 byte NOPs", |
| ... | ... | @@ -813,6 +865,11 @@ pub const all_features = blk: { |
| 813 | 865 | .description = "Invalidate Process-Context Identifier", |
| 814 | 866 | .dependencies = featureSet(&[_]Feature{}), |
| 815 | 867 | }; |
| 868 | result[@backingInt(Feature.jmpabs)] = .{ | |
| 869 | .llvm_name = "jmpabs", | |
| 870 | .description = "Support 64-bit absolute JMP", | |
| 871 | .dependencies = featureSet(&[_]Feature{}), | |
| 872 | }; | |
| 816 | 873 | result[@backingInt(Feature.kl)] = .{ |
| 817 | 874 | .llvm_name = "kl", |
| 818 | 875 | .description = "Support Key Locker kl Instructions", |
| ... | ... | @@ -977,6 +1034,11 @@ pub const all_features = blk: { |
| 977 | 1034 | .description = "Prefer movmsk over vtest instruction", |
| 978 | 1035 | .dependencies = featureSet(&[_]Feature{}), |
| 979 | 1036 | }; |
| 1037 | result[@backingInt(Feature.prefer_ndd_mem)] = .{ | |
| 1038 | .llvm_name = "prefer-ndd-mem", | |
| 1039 | .description = "Prefer NDD memory addressing", | |
| 1040 | .dependencies = featureSet(&[_]Feature{}), | |
| 1041 | }; | |
| 980 | 1042 | result[@backingInt(Feature.prefer_no_gather)] = .{ |
| 981 | 1043 | .llvm_name = "prefer-no-gather", |
| 982 | 1044 | .description = "Prefer no gather instructions", |
| ... | ... | @@ -1037,6 +1099,131 @@ pub const all_features = blk: { |
| 1037 | 1099 | .description = "Support RDSEED instruction", |
| 1038 | 1100 | .dependencies = featureSet(&[_]Feature{}), |
| 1039 | 1101 | }; |
| 1102 | result[@backingInt(Feature.reserve_edi)] = .{ | |
| 1103 | .llvm_name = "reserve-edi", | |
| 1104 | .description = "Reserve EDI, making it unavailable as a GPR", | |
| 1105 | .dependencies = featureSet(&[_]Feature{}), | |
| 1106 | }; | |
| 1107 | result[@backingInt(Feature.reserve_r10)] = .{ | |
| 1108 | .llvm_name = "reserve-r10", | |
| 1109 | .description = "Reserve R10, making it unavailable as a GPR", | |
| 1110 | .dependencies = featureSet(&[_]Feature{}), | |
| 1111 | }; | |
| 1112 | result[@backingInt(Feature.reserve_r11)] = .{ | |
| 1113 | .llvm_name = "reserve-r11", | |
| 1114 | .description = "Reserve R11, making it unavailable as a GPR", | |
| 1115 | .dependencies = featureSet(&[_]Feature{}), | |
| 1116 | }; | |
| 1117 | result[@backingInt(Feature.reserve_r12)] = .{ | |
| 1118 | .llvm_name = "reserve-r12", | |
| 1119 | .description = "Reserve R12, making it unavailable as a GPR", | |
| 1120 | .dependencies = featureSet(&[_]Feature{}), | |
| 1121 | }; | |
| 1122 | result[@backingInt(Feature.reserve_r13)] = .{ | |
| 1123 | .llvm_name = "reserve-r13", | |
| 1124 | .description = "Reserve R13, making it unavailable as a GPR", | |
| 1125 | .dependencies = featureSet(&[_]Feature{}), | |
| 1126 | }; | |
| 1127 | result[@backingInt(Feature.reserve_r14)] = .{ | |
| 1128 | .llvm_name = "reserve-r14", | |
| 1129 | .description = "Reserve R14, making it unavailable as a GPR", | |
| 1130 | .dependencies = featureSet(&[_]Feature{}), | |
| 1131 | }; | |
| 1132 | result[@backingInt(Feature.reserve_r15)] = .{ | |
| 1133 | .llvm_name = "reserve-r15", | |
| 1134 | .description = "Reserve R15, making it unavailable as a GPR", | |
| 1135 | .dependencies = featureSet(&[_]Feature{}), | |
| 1136 | }; | |
| 1137 | result[@backingInt(Feature.reserve_r16)] = .{ | |
| 1138 | .llvm_name = "reserve-r16", | |
| 1139 | .description = "Reserve R16, making it unavailable as a GPR", | |
| 1140 | .dependencies = featureSet(&[_]Feature{}), | |
| 1141 | }; | |
| 1142 | result[@backingInt(Feature.reserve_r17)] = .{ | |
| 1143 | .llvm_name = "reserve-r17", | |
| 1144 | .description = "Reserve R17, making it unavailable as a GPR", | |
| 1145 | .dependencies = featureSet(&[_]Feature{}), | |
| 1146 | }; | |
| 1147 | result[@backingInt(Feature.reserve_r18)] = .{ | |
| 1148 | .llvm_name = "reserve-r18", | |
| 1149 | .description = "Reserve R18, making it unavailable as a GPR", | |
| 1150 | .dependencies = featureSet(&[_]Feature{}), | |
| 1151 | }; | |
| 1152 | result[@backingInt(Feature.reserve_r19)] = .{ | |
| 1153 | .llvm_name = "reserve-r19", | |
| 1154 | .description = "Reserve R19, making it unavailable as a GPR", | |
| 1155 | .dependencies = featureSet(&[_]Feature{}), | |
| 1156 | }; | |
| 1157 | result[@backingInt(Feature.reserve_r20)] = .{ | |
| 1158 | .llvm_name = "reserve-r20", | |
| 1159 | .description = "Reserve R20, making it unavailable as a GPR", | |
| 1160 | .dependencies = featureSet(&[_]Feature{}), | |
| 1161 | }; | |
| 1162 | result[@backingInt(Feature.reserve_r21)] = .{ | |
| 1163 | .llvm_name = "reserve-r21", | |
| 1164 | .description = "Reserve R21, making it unavailable as a GPR", | |
| 1165 | .dependencies = featureSet(&[_]Feature{}), | |
| 1166 | }; | |
| 1167 | result[@backingInt(Feature.reserve_r22)] = .{ | |
| 1168 | .llvm_name = "reserve-r22", | |
| 1169 | .description = "Reserve R22, making it unavailable as a GPR", | |
| 1170 | .dependencies = featureSet(&[_]Feature{}), | |
| 1171 | }; | |
| 1172 | result[@backingInt(Feature.reserve_r23)] = .{ | |
| 1173 | .llvm_name = "reserve-r23", | |
| 1174 | .description = "Reserve R23, making it unavailable as a GPR", | |
| 1175 | .dependencies = featureSet(&[_]Feature{}), | |
| 1176 | }; | |
| 1177 | result[@backingInt(Feature.reserve_r24)] = .{ | |
| 1178 | .llvm_name = "reserve-r24", | |
| 1179 | .description = "Reserve R24, making it unavailable as a GPR", | |
| 1180 | .dependencies = featureSet(&[_]Feature{}), | |
| 1181 | }; | |
| 1182 | result[@backingInt(Feature.reserve_r25)] = .{ | |
| 1183 | .llvm_name = "reserve-r25", | |
| 1184 | .description = "Reserve R25, making it unavailable as a GPR", | |
| 1185 | .dependencies = featureSet(&[_]Feature{}), | |
| 1186 | }; | |
| 1187 | result[@backingInt(Feature.reserve_r26)] = .{ | |
| 1188 | .llvm_name = "reserve-r26", | |
| 1189 | .description = "Reserve R26, making it unavailable as a GPR", | |
| 1190 | .dependencies = featureSet(&[_]Feature{}), | |
| 1191 | }; | |
| 1192 | result[@backingInt(Feature.reserve_r27)] = .{ | |
| 1193 | .llvm_name = "reserve-r27", | |
| 1194 | .description = "Reserve R27, making it unavailable as a GPR", | |
| 1195 | .dependencies = featureSet(&[_]Feature{}), | |
| 1196 | }; | |
| 1197 | result[@backingInt(Feature.reserve_r28)] = .{ | |
| 1198 | .llvm_name = "reserve-r28", | |
| 1199 | .description = "Reserve R28, making it unavailable as a GPR", | |
| 1200 | .dependencies = featureSet(&[_]Feature{}), | |
| 1201 | }; | |
| 1202 | result[@backingInt(Feature.reserve_r29)] = .{ | |
| 1203 | .llvm_name = "reserve-r29", | |
| 1204 | .description = "Reserve R29, making it unavailable as a GPR", | |
| 1205 | .dependencies = featureSet(&[_]Feature{}), | |
| 1206 | }; | |
| 1207 | result[@backingInt(Feature.reserve_r30)] = .{ | |
| 1208 | .llvm_name = "reserve-r30", | |
| 1209 | .description = "Reserve R30, making it unavailable as a GPR", | |
| 1210 | .dependencies = featureSet(&[_]Feature{}), | |
| 1211 | }; | |
| 1212 | result[@backingInt(Feature.reserve_r31)] = .{ | |
| 1213 | .llvm_name = "reserve-r31", | |
| 1214 | .description = "Reserve R31, making it unavailable as a GPR", | |
| 1215 | .dependencies = featureSet(&[_]Feature{}), | |
| 1216 | }; | |
| 1217 | result[@backingInt(Feature.reserve_r8)] = .{ | |
| 1218 | .llvm_name = "reserve-r8", | |
| 1219 | .description = "Reserve R8, making it unavailable as a GPR", | |
| 1220 | .dependencies = featureSet(&[_]Feature{}), | |
| 1221 | }; | |
| 1222 | result[@backingInt(Feature.reserve_r9)] = .{ | |
| 1223 | .llvm_name = "reserve-r9", | |
| 1224 | .description = "Reserve R9, making it unavailable as a GPR", | |
| 1225 | .dependencies = featureSet(&[_]Feature{}), | |
| 1226 | }; | |
| 1040 | 1227 | result[@backingInt(Feature.retpoline)] = .{ |
| 1041 | 1228 | .llvm_name = "retpoline", |
| 1042 | 1229 | .description = "Remove speculation of indirect branches from the generated code, either by avoiding them entirely or lowering them with a speculation blocking construct", |
| ... | ... | @@ -1123,6 +1310,11 @@ pub const all_features = blk: { |
| 1123 | 1310 | .description = "INC and DEC instructions are slower than ADD and SUB", |
| 1124 | 1311 | .dependencies = featureSet(&[_]Feature{}), |
| 1125 | 1312 | }; |
| 1313 | result[@backingInt(Feature.slow_indirect_call)] = .{ | |
| 1314 | .llvm_name = "slow-indirect-call", | |
| 1315 | .description = "Indirect calls from memory are slow", | |
| 1316 | .dependencies = featureSet(&[_]Feature{}), | |
| 1317 | }; | |
| 1126 | 1318 | result[@backingInt(Feature.slow_lea)] = .{ |
| 1127 | 1319 | .llvm_name = "slow-lea", |
| 1128 | 1320 | .description = "LEA instruction with certain arguments is slow", |
| ... | ... | @@ -2039,8 +2231,9 @@ pub const cpu = struct { |
| 2039 | 2231 | .cx16, |
| 2040 | 2232 | .ermsb, |
| 2041 | 2233 | .f16c, |
| 2042 | .false_deps_lzcnt_tzcnt, | |
| 2234 | .false_deps_lzcnt, | |
| 2043 | 2235 | .false_deps_popcnt, |
| 2236 | .false_deps_tzcnt, | |
| 2044 | 2237 | .fast_15bytenop, |
| 2045 | 2238 | .fast_scalar_fsqrt, |
| 2046 | 2239 | .fast_shld_rotate, |
| ... | ... | @@ -2159,6 +2352,239 @@ pub const cpu = struct { |
| 2159 | 2352 | .x87, |
| 2160 | 2353 | }), |
| 2161 | 2354 | }; |
| 2355 | pub const c86_4g_m4: CpuModel = .{ | |
| 2356 | .name = "c86_4g_m4", | |
| 2357 | .llvm_name = "c86-4g-m4", | |
| 2358 | .features = featureSet(&[_]Feature{ | |
| 2359 | .@"64bit", | |
| 2360 | .adx, | |
| 2361 | .aes, | |
| 2362 | .allow_light_256_bit, | |
| 2363 | .avx2, | |
| 2364 | .bmi, | |
| 2365 | .bmi2, | |
| 2366 | .clflushopt, | |
| 2367 | .clzero, | |
| 2368 | .cmov, | |
| 2369 | .cx16, | |
| 2370 | .f16c, | |
| 2371 | .fast_15bytenop, | |
| 2372 | .fast_bextr, | |
| 2373 | .fast_imm16, | |
| 2374 | .fast_lzcnt, | |
| 2375 | .fast_movbe, | |
| 2376 | .fast_scalar_fsqrt, | |
| 2377 | .fast_scalar_shift_masks, | |
| 2378 | .fast_variable_perlane_shuffle, | |
| 2379 | .fast_vector_fsqrt, | |
| 2380 | .fma, | |
| 2381 | .fsgsbase, | |
| 2382 | .fxsr, | |
| 2383 | .idivq_to_divl, | |
| 2384 | .lzcnt, | |
| 2385 | .mmx, | |
| 2386 | .movbe, | |
| 2387 | .mwaitx, | |
| 2388 | .nopl, | |
| 2389 | .pclmul, | |
| 2390 | .popcnt, | |
| 2391 | .prfchw, | |
| 2392 | .rdrnd, | |
| 2393 | .rdseed, | |
| 2394 | .sahf, | |
| 2395 | .sbb_dep_breaking, | |
| 2396 | .sha, | |
| 2397 | .slow_shld, | |
| 2398 | .sse4a, | |
| 2399 | .vzeroupper, | |
| 2400 | .x87, | |
| 2401 | .xsavec, | |
| 2402 | .xsaveopt, | |
| 2403 | .xsaves, | |
| 2404 | }), | |
| 2405 | }; | |
| 2406 | pub const c86_4g_m6: CpuModel = .{ | |
| 2407 | .name = "c86_4g_m6", | |
| 2408 | .llvm_name = "c86-4g-m6", | |
| 2409 | .features = featureSet(&[_]Feature{ | |
| 2410 | .@"64bit", | |
| 2411 | .adx, | |
| 2412 | .aes, | |
| 2413 | .allow_light_256_bit, | |
| 2414 | .avx2, | |
| 2415 | .bmi, | |
| 2416 | .bmi2, | |
| 2417 | .clflushopt, | |
| 2418 | .clzero, | |
| 2419 | .cmov, | |
| 2420 | .cx16, | |
| 2421 | .f16c, | |
| 2422 | .fast_15bytenop, | |
| 2423 | .fast_bextr, | |
| 2424 | .fast_imm16, | |
| 2425 | .fast_lzcnt, | |
| 2426 | .fast_movbe, | |
| 2427 | .fast_scalar_fsqrt, | |
| 2428 | .fast_scalar_shift_masks, | |
| 2429 | .fast_variable_perlane_shuffle, | |
| 2430 | .fast_vector_fsqrt, | |
| 2431 | .fma, | |
| 2432 | .fsgsbase, | |
| 2433 | .fxsr, | |
| 2434 | .idivq_to_divl, | |
| 2435 | .lzcnt, | |
| 2436 | .mmx, | |
| 2437 | .movbe, | |
| 2438 | .mwaitx, | |
| 2439 | .nopl, | |
| 2440 | .pclmul, | |
| 2441 | .popcnt, | |
| 2442 | .prfchw, | |
| 2443 | .rdrnd, | |
| 2444 | .rdseed, | |
| 2445 | .sahf, | |
| 2446 | .sbb_dep_breaking, | |
| 2447 | .sha, | |
| 2448 | .slow_shld, | |
| 2449 | .sse4a, | |
| 2450 | .vzeroupper, | |
| 2451 | .x87, | |
| 2452 | .xsavec, | |
| 2453 | .xsaveopt, | |
| 2454 | .xsaves, | |
| 2455 | }), | |
| 2456 | }; | |
| 2457 | pub const c86_4g_m7: CpuModel = .{ | |
| 2458 | .name = "c86_4g_m7", | |
| 2459 | .llvm_name = "c86-4g-m7", | |
| 2460 | .features = featureSet(&[_]Feature{ | |
| 2461 | .@"64bit", | |
| 2462 | .adx, | |
| 2463 | .allow_light_256_bit, | |
| 2464 | .avx512bf16, | |
| 2465 | .avx512bitalg, | |
| 2466 | .avx512cd, | |
| 2467 | .avx512dq, | |
| 2468 | .avx512ifma, | |
| 2469 | .avx512vbmi, | |
| 2470 | .avx512vbmi2, | |
| 2471 | .avx512vl, | |
| 2472 | .avx512vnni, | |
| 2473 | .avx512vpopcntdq, | |
| 2474 | .bmi, | |
| 2475 | .bmi2, | |
| 2476 | .branchfusion, | |
| 2477 | .clflushopt, | |
| 2478 | .clwb, | |
| 2479 | .clzero, | |
| 2480 | .cmov, | |
| 2481 | .cx16, | |
| 2482 | .fast_15bytenop, | |
| 2483 | .fast_bextr, | |
| 2484 | .fast_imm16, | |
| 2485 | .fast_lzcnt, | |
| 2486 | .fast_movbe, | |
| 2487 | .fast_scalar_fsqrt, | |
| 2488 | .fast_scalar_shift_masks, | |
| 2489 | .fast_variable_perlane_shuffle, | |
| 2490 | .fast_vector_fsqrt, | |
| 2491 | .fsgsbase, | |
| 2492 | .fxsr, | |
| 2493 | .gfni, | |
| 2494 | .idivq_to_divl, | |
| 2495 | .lzcnt, | |
| 2496 | .mmx, | |
| 2497 | .movbe, | |
| 2498 | .mwaitx, | |
| 2499 | .nopl, | |
| 2500 | .popcnt, | |
| 2501 | .prefer_256_bit, | |
| 2502 | .prefer_no_gather, | |
| 2503 | .prefer_no_scatter, | |
| 2504 | .prfchw, | |
| 2505 | .rdrnd, | |
| 2506 | .rdseed, | |
| 2507 | .sahf, | |
| 2508 | .sbb_dep_breaking, | |
| 2509 | .sha, | |
| 2510 | .slow_shld, | |
| 2511 | .sse4a, | |
| 2512 | .vaes, | |
| 2513 | .vpclmulqdq, | |
| 2514 | .vzeroupper, | |
| 2515 | .wbnoinvd, | |
| 2516 | .x87, | |
| 2517 | .xsavec, | |
| 2518 | .xsaveopt, | |
| 2519 | .xsaves, | |
| 2520 | }), | |
| 2521 | }; | |
| 2522 | pub const c86_4g_m8: CpuModel = .{ | |
| 2523 | .name = "c86_4g_m8", | |
| 2524 | .llvm_name = "c86-4g-m8", | |
| 2525 | .features = featureSet(&[_]Feature{ | |
| 2526 | .@"64bit", | |
| 2527 | .adx, | |
| 2528 | .allow_light_256_bit, | |
| 2529 | .avx512bf16, | |
| 2530 | .avx512bitalg, | |
| 2531 | .avx512cd, | |
| 2532 | .avx512dq, | |
| 2533 | .avx512ifma, | |
| 2534 | .avx512vbmi, | |
| 2535 | .avx512vbmi2, | |
| 2536 | .avx512vl, | |
| 2537 | .avx512vnni, | |
| 2538 | .avx512vpopcntdq, | |
| 2539 | .bmi, | |
| 2540 | .bmi2, | |
| 2541 | .branchfusion, | |
| 2542 | .clflushopt, | |
| 2543 | .clwb, | |
| 2544 | .clzero, | |
| 2545 | .cmov, | |
| 2546 | .cx16, | |
| 2547 | .fast_15bytenop, | |
| 2548 | .fast_bextr, | |
| 2549 | .fast_imm16, | |
| 2550 | .fast_lzcnt, | |
| 2551 | .fast_movbe, | |
| 2552 | .fast_scalar_fsqrt, | |
| 2553 | .fast_scalar_shift_masks, | |
| 2554 | .fast_variable_perlane_shuffle, | |
| 2555 | .fast_vector_fsqrt, | |
| 2556 | .fsgsbase, | |
| 2557 | .fxsr, | |
| 2558 | .gfni, | |
| 2559 | .idivq_to_divl, | |
| 2560 | .lzcnt, | |
| 2561 | .mmx, | |
| 2562 | .movbe, | |
| 2563 | .mwaitx, | |
| 2564 | .nopl, | |
| 2565 | .popcnt, | |
| 2566 | .prefer_256_bit, | |
| 2567 | .prefer_no_gather, | |
| 2568 | .prefer_no_scatter, | |
| 2569 | .prfchw, | |
| 2570 | .rdrnd, | |
| 2571 | .rdseed, | |
| 2572 | .sahf, | |
| 2573 | .sbb_dep_breaking, | |
| 2574 | .sha, | |
| 2575 | .shstk, | |
| 2576 | .slow_shld, | |
| 2577 | .sse4a, | |
| 2578 | .vaes, | |
| 2579 | .vpclmulqdq, | |
| 2580 | .vzeroupper, | |
| 2581 | .wbnoinvd, | |
| 2582 | .x87, | |
| 2583 | .xsavec, | |
| 2584 | .xsaveopt, | |
| 2585 | .xsaves, | |
| 2586 | }), | |
| 2587 | }; | |
| 2162 | 2588 | pub const cannonlake: CpuModel = .{ |
| 2163 | 2589 | .name = "cannonlake", |
| 2164 | 2590 | .llvm_name = "cannonlake", |
| ... | ... | @@ -2467,7 +2893,6 @@ pub const cpu = struct { |
| 2467 | 2893 | .amx_fp8, |
| 2468 | 2894 | .amx_int8, |
| 2469 | 2895 | .amx_movrs, |
| 2470 | .amx_tf32, | |
| 2471 | 2896 | .avx10_2, |
| 2472 | 2897 | .avxifma, |
| 2473 | 2898 | .avxneconvert, |
| ... | ... | @@ -2505,6 +2930,7 @@ pub const cpu = struct { |
| 2505 | 2930 | .gfni, |
| 2506 | 2931 | .idivq_to_divl, |
| 2507 | 2932 | .invpcid, |
| 2933 | .jmpabs, | |
| 2508 | 2934 | .lzcnt, |
| 2509 | 2935 | .macrofusion, |
| 2510 | 2936 | .mmx, |
| ... | ... | @@ -2523,6 +2949,7 @@ pub const cpu = struct { |
| 2523 | 2949 | .popcnt, |
| 2524 | 2950 | .ppx, |
| 2525 | 2951 | .prefer_256_bit, |
| 2952 | .prefer_legacy_setcc, | |
| 2526 | 2953 | .prefetchi, |
| 2527 | 2954 | .prfchw, |
| 2528 | 2955 | .ptwrite, |
| ... | ... | @@ -3056,8 +3483,9 @@ pub const cpu = struct { |
| 3056 | 3483 | .cx16, |
| 3057 | 3484 | .ermsb, |
| 3058 | 3485 | .f16c, |
| 3059 | .false_deps_lzcnt_tzcnt, | |
| 3486 | .false_deps_lzcnt, | |
| 3060 | 3487 | .false_deps_popcnt, |
| 3488 | .false_deps_tzcnt, | |
| 3061 | 3489 | .fast_15bytenop, |
| 3062 | 3490 | .fast_scalar_fsqrt, |
| 3063 | 3491 | .fast_shld_rotate, |
| ... | ... | @@ -3679,6 +4107,7 @@ pub const cpu = struct { |
| 3679 | 4107 | .hreset, |
| 3680 | 4108 | .idivq_to_divl, |
| 3681 | 4109 | .invpcid, |
| 4110 | .jmpabs, | |
| 3682 | 4111 | .lzcnt, |
| 3683 | 4112 | .macrofusion, |
| 3684 | 4113 | .mmx, |
| ... | ... | @@ -3696,6 +4125,7 @@ pub const cpu = struct { |
| 3696 | 4125 | .pku, |
| 3697 | 4126 | .popcnt, |
| 3698 | 4127 | .ppx, |
| 4128 | .prefer_legacy_setcc, | |
| 3699 | 4129 | .prefer_movmsk_over_vtest, |
| 3700 | 4130 | .prefetchi, |
| 3701 | 4131 | .prfchw, |
| ... | ... | @@ -4813,8 +5243,9 @@ pub const cpu = struct { |
| 4813 | 5243 | .cmov, |
| 4814 | 5244 | .cx16, |
| 4815 | 5245 | .f16c, |
| 4816 | .false_deps_lzcnt_tzcnt, | |
| 5246 | .false_deps_lzcnt, | |
| 4817 | 5247 | .false_deps_popcnt, |
| 5248 | .false_deps_tzcnt, | |
| 4818 | 5249 | .fast_15bytenop, |
| 4819 | 5250 | .fast_scalar_fsqrt, |
| 4820 | 5251 | .fast_shld_rotate, |
| ... | ... | @@ -5086,6 +5517,8 @@ pub const cpu = struct { |
| 5086 | 5517 | .clzero, |
| 5087 | 5518 | .cmov, |
| 5088 | 5519 | .cx16, |
| 5520 | .false_deps_compress, | |
| 5521 | .false_deps_expand, | |
| 5089 | 5522 | .fast_15bytenop, |
| 5090 | 5523 | .fast_bextr, |
| 5091 | 5524 | .fast_dpwssd, |
| ... | ... | @@ -5159,6 +5592,10 @@ pub const cpu = struct { |
| 5159 | 5592 | .clzero, |
| 5160 | 5593 | .cmov, |
| 5161 | 5594 | .cx16, |
| 5595 | .false_deps_bls, | |
| 5596 | .false_deps_compress, | |
| 5597 | .false_deps_expand, | |
| 5598 | .false_deps_tzcnt, | |
| 5162 | 5599 | .fast_15bytenop, |
| 5163 | 5600 | .fast_bextr, |
| 5164 | 5601 | .fast_dpwssd, |
| ... | ... | @@ -5195,6 +5632,7 @@ pub const cpu = struct { |
| 5195 | 5632 | .sbb_dep_breaking, |
| 5196 | 5633 | .sha, |
| 5197 | 5634 | .shstk, |
| 5635 | .slow_indirect_call, | |
| 5198 | 5636 | .smap, |
| 5199 | 5637 | .smep, |
| 5200 | 5638 | .sse4a, |
| ... | ... | @@ -5208,4 +5646,83 @@ pub const cpu = struct { |
| 5208 | 5646 | .xsaves, |
| 5209 | 5647 | }), |
| 5210 | 5648 | }; |
| 5649 | pub const znver6: CpuModel = .{ | |
| 5650 | .name = "znver6", | |
| 5651 | .llvm_name = "znver6", | |
| 5652 | .features = featureSet(&[_]Feature{ | |
| 5653 | .@"64bit", | |
| 5654 | .adx, | |
| 5655 | .allow_light_256_bit, | |
| 5656 | .avx512bf16, | |
| 5657 | .avx512bitalg, | |
| 5658 | .avx512bmm, | |
| 5659 | .avx512cd, | |
| 5660 | .avx512dq, | |
| 5661 | .avx512fp16, | |
| 5662 | .avx512ifma, | |
| 5663 | .avx512vbmi, | |
| 5664 | .avx512vbmi2, | |
| 5665 | .avx512vl, | |
| 5666 | .avx512vnni, | |
| 5667 | .avx512vp2intersect, | |
| 5668 | .avx512vpopcntdq, | |
| 5669 | .avxifma, | |
| 5670 | .avxneconvert, | |
| 5671 | .avxvnni, | |
| 5672 | .avxvnniint8, | |
| 5673 | .bmi, | |
| 5674 | .bmi2, | |
| 5675 | .branchfusion, | |
| 5676 | .clflushopt, | |
| 5677 | .clwb, | |
| 5678 | .clzero, | |
| 5679 | .cmov, | |
| 5680 | .cx16, | |
| 5681 | .fast_15bytenop, | |
| 5682 | .fast_bextr, | |
| 5683 | .fast_dpwssd, | |
| 5684 | .fast_imm16, | |
| 5685 | .fast_lzcnt, | |
| 5686 | .fast_movbe, | |
| 5687 | .fast_scalar_fsqrt, | |
| 5688 | .fast_scalar_shift_masks, | |
| 5689 | .fast_variable_perlane_shuffle, | |
| 5690 | .fast_vector_fsqrt, | |
| 5691 | .fsgsbase, | |
| 5692 | .fsrm, | |
| 5693 | .fxsr, | |
| 5694 | .gfni, | |
| 5695 | .idivq_to_divl, | |
| 5696 | .invpcid, | |
| 5697 | .lzcnt, | |
| 5698 | .macrofusion, | |
| 5699 | .mmx, | |
| 5700 | .movbe, | |
| 5701 | .movdir64b, | |
| 5702 | .movdiri, | |
| 5703 | .mwaitx, | |
| 5704 | .nopl, | |
| 5705 | .pku, | |
| 5706 | .popcnt, | |
| 5707 | .prefetchi, | |
| 5708 | .prfchw, | |
| 5709 | .rdpid, | |
| 5710 | .rdpru, | |
| 5711 | .rdrnd, | |
| 5712 | .rdseed, | |
| 5713 | .sahf, | |
| 5714 | .sbb_dep_breaking, | |
| 5715 | .sha, | |
| 5716 | .shstk, | |
| 5717 | .sse4a, | |
| 5718 | .vaes, | |
| 5719 | .vpclmulqdq, | |
| 5720 | .vzeroupper, | |
| 5721 | .wbnoinvd, | |
| 5722 | .x87, | |
| 5723 | .xsavec, | |
| 5724 | .xsaveopt, | |
| 5725 | .xsaves, | |
| 5726 | }), | |
| 5727 | }; | |
| 5211 | 5728 | }; |
lib/std/Target/xtensa.zig+35| ... | ... | @@ -13,6 +13,7 @@ pub const Feature = enum { |
| 13 | 13 | density, |
| 14 | 14 | dfpaccel, |
| 15 | 15 | div32, |
| 16 | esp32s2ops, | |
| 16 | 17 | exception, |
| 17 | 18 | extendedl32r, |
| 18 | 19 | forced_atomics, |
| ... | ... | @@ -93,6 +94,11 @@ pub const all_features = blk: { |
| 93 | 94 | .description = "Enable Xtensa Div32 option", |
| 94 | 95 | .dependencies = featureSet(&[_]Feature{}), |
| 95 | 96 | }; |
| 97 | result[@backingInt(Feature.esp32s2ops)] = .{ | |
| 98 | .llvm_name = "esp32s2ops", | |
| 99 | .description = "Support Xtensa esp32-s2 ISA extension", | |
| 100 | .dependencies = featureSet(&[_]Feature{}), | |
| 101 | }; | |
| 96 | 102 | result[@backingInt(Feature.exception)] = .{ |
| 97 | 103 | .llvm_name = "exception", |
| 98 | 104 | .description = "Enable Xtensa Exception option", |
| ... | ... | @@ -291,6 +297,35 @@ pub const cpu = struct { |
| 291 | 297 | .windowed, |
| 292 | 298 | }), |
| 293 | 299 | }; |
| 300 | pub const esp32s2: CpuModel = .{ | |
| 301 | .name = "esp32s2", | |
| 302 | .llvm_name = "esp32s2", | |
| 303 | .features = featureSet(&[_]Feature{ | |
| 304 | .clamps, | |
| 305 | .coprocessor, | |
| 306 | .dcache, | |
| 307 | .debug, | |
| 308 | .density, | |
| 309 | .div32, | |
| 310 | .esp32s2ops, | |
| 311 | .exception, | |
| 312 | .highpriinterrupts_level7, | |
| 313 | .interrupt, | |
| 314 | .minmax, | |
| 315 | .miscsr, | |
| 316 | .mul16, | |
| 317 | .mul32, | |
| 318 | .mul32high, | |
| 319 | .nsa, | |
| 320 | .prid, | |
| 321 | .regprotect, | |
| 322 | .rvector, | |
| 323 | .sext, | |
| 324 | .threadptr, | |
| 325 | .timers3, | |
| 326 | .windowed, | |
| 327 | }), | |
| 328 | }; | |
| 294 | 329 | pub const esp8266: CpuModel = .{ |
| 295 | 330 | .name = "esp8266", |
| 296 | 331 | .llvm_name = "esp8266", |
src/codegen/x86_64/CodeGen.zig+19-18| ... | ... | @@ -67564,7 +67564,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 67564 | 67564 | .{ ._, ._, .sub, .dst0b, .sia(32, .src0, .sub_bit_size), ._, ._ }, |
| 67565 | 67565 | } }, |
| 67566 | 67566 | }, .{ |
| 67567 | .required_features = .{ .false_deps_lzcnt_tzcnt, .lzcnt, null, null }, | |
| 67567 | .required_features = .{ .false_deps_lzcnt, .lzcnt, null, null }, | |
| 67568 | 67568 | .src_constraints = .{ .{ .exact_int = 16 }, .any, .any }, |
| 67569 | 67569 | .patterns = &.{ |
| 67570 | 67570 | .{ .src = .{ .to_mut_gpr, .none, .none } }, |
| ... | ... | @@ -67600,7 +67600,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 67600 | 67600 | .{ ._, ._, .sub, .dst0b, .sia(16, .src0, .sub_bit_size), ._, ._ }, |
| 67601 | 67601 | } }, |
| 67602 | 67602 | }, .{ |
| 67603 | .required_features = .{ .false_deps_lzcnt_tzcnt, .lzcnt, null, null }, | |
| 67603 | .required_features = .{ .false_deps_lzcnt, .lzcnt, null, null }, | |
| 67604 | 67604 | .src_constraints = .{ .{ .unsigned_int = .word }, .any, .any }, |
| 67605 | 67605 | .patterns = &.{ |
| 67606 | 67606 | .{ .src = .{ .to_mut_gpr, .none, .none } }, |
| ... | ... | @@ -67625,7 +67625,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 67625 | 67625 | .{ ._, ._, .sub, .dst0b, .sia(16, .src0, .sub_bit_size), ._, ._ }, |
| 67626 | 67626 | } }, |
| 67627 | 67627 | }, .{ |
| 67628 | .required_features = .{ .false_deps_lzcnt_tzcnt, .lzcnt, null, null }, | |
| 67628 | .required_features = .{ .false_deps_lzcnt, .lzcnt, null, null }, | |
| 67629 | 67629 | .src_constraints = .{ .{ .exact_int = 32 }, .any, .any }, |
| 67630 | 67630 | .patterns = &.{ |
| 67631 | 67631 | .{ .src = .{ .to_mut_gpr, .none, .none } }, |
| ... | ... | @@ -67661,7 +67661,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 67661 | 67661 | .{ ._, ._, .sub, .dst0b, .sia(32, .src0, .sub_bit_size), ._, ._ }, |
| 67662 | 67662 | } }, |
| 67663 | 67663 | }, .{ |
| 67664 | .required_features = .{ .false_deps_lzcnt_tzcnt, .lzcnt, null, null }, | |
| 67664 | .required_features = .{ .false_deps_lzcnt, .lzcnt, null, null }, | |
| 67665 | 67665 | .src_constraints = .{ .{ .unsigned_int = .dword }, .any, .any }, |
| 67666 | 67666 | .patterns = &.{ |
| 67667 | 67667 | .{ .src = .{ .to_mut_gpr, .none, .none } }, |
| ... | ... | @@ -67686,7 +67686,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 67686 | 67686 | .{ ._, ._, .sub, .dst0b, .sia(32, .src0, .sub_bit_size), ._, ._ }, |
| 67687 | 67687 | } }, |
| 67688 | 67688 | }, .{ |
| 67689 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 67689 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 67690 | 67690 | .src_constraints = .{ .{ .exact_int = 64 }, .any, .any }, |
| 67691 | 67691 | .patterns = &.{ |
| 67692 | 67692 | .{ .src = .{ .to_mut_gpr, .none, .none } }, |
| ... | ... | @@ -67724,7 +67724,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 67724 | 67724 | .{ ._, ._, .sub, .dst0b, .sia(64, .src0, .sub_bit_size), ._, ._ }, |
| 67725 | 67725 | } }, |
| 67726 | 67726 | }, .{ |
| 67727 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 67727 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 67728 | 67728 | .src_constraints = .{ .{ .unsigned_int = .qword }, .any, .any }, |
| 67729 | 67729 | .patterns = &.{ |
| 67730 | 67730 | .{ .src = .{ .to_mut_gpr, .none, .none } }, |
| ... | ... | @@ -68581,7 +68581,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 68581 | 68581 | .{ ._, ._, .sub, .dst0b, .tmp0b, ._, ._ }, |
| 68582 | 68582 | } }, |
| 68583 | 68583 | }, .{ |
| 68584 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 68584 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 68585 | 68585 | .src_constraints = .{ .{ .unsigned_or_exact_remainder_int = .{ .of = .xword, .is = .qword } }, .any, .any }, |
| 68586 | 68586 | .patterns = &.{ |
| 68587 | 68587 | .{ .src = .{ .to_mem, .none, .none } }, |
| ... | ... | @@ -68709,7 +68709,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 68709 | 68709 | .{ ._, ._, .neg, .dst0d, ._, ._, ._ }, |
| 68710 | 68710 | } }, |
| 68711 | 68711 | }, .{ |
| 68712 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 68712 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 68713 | 68713 | .src_constraints = .{ .{ .unsigned_or_exact_remainder_int = .{ .of = .xword, .is = .xword } }, .any, .any }, |
| 68714 | 68714 | .patterns = &.{ |
| 68715 | 68715 | .{ .src = .{ .to_mem, .none, .none } }, |
| ... | ... | @@ -68837,7 +68837,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 68837 | 68837 | .{ ._, ._, .neg, .dst0d, ._, ._, ._ }, |
| 68838 | 68838 | } }, |
| 68839 | 68839 | }, .{ |
| 68840 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 68840 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 68841 | 68841 | .src_constraints = .{ .{ .remainder_int = .{ .of = .xword, .is = .qword } }, .any, .any }, |
| 68842 | 68842 | .patterns = &.{ |
| 68843 | 68843 | .{ .src = .{ .to_mem, .none, .none } }, |
| ... | ... | @@ -68940,7 +68940,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 68940 | 68940 | .{ ._, ._, .neg, .dst0d, ._, ._, ._ }, |
| 68941 | 68941 | } }, |
| 68942 | 68942 | }, .{ |
| 68943 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 68943 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 68944 | 68944 | .src_constraints = .{ .{ .remainder_int = .{ .of = .xword, .is = .xword } }, .any, .any }, |
| 68945 | 68945 | .patterns = &.{ |
| 68946 | 68946 | .{ .src = .{ .to_mem, .none, .none } }, |
| ... | ... | @@ -70096,7 +70096,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 70096 | 70096 | .{ ._, ._nz, .j, .@"0b", ._, ._, ._ }, |
| 70097 | 70097 | } }, |
| 70098 | 70098 | }, .{ |
| 70099 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 70099 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 70100 | 70100 | .src_constraints = .{ .{ .scalar_remainder_int = .{ .of = .xword, .is = .qword } }, .any, .any }, |
| 70101 | 70101 | .dst_constraints = .{ .{ .scalar_int_is = .byte }, .any }, |
| 70102 | 70102 | .patterns = &.{ |
| ... | ... | @@ -70220,7 +70220,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 70220 | 70220 | .{ ._, ._nc, .j, .@"0b", ._, ._, ._ }, |
| 70221 | 70221 | } }, |
| 70222 | 70222 | }, .{ |
| 70223 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 70223 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 70224 | 70224 | .src_constraints = .{ .{ .scalar_remainder_int = .{ .of = .xword, .is = .xword } }, .any, .any }, |
| 70225 | 70225 | .dst_constraints = .{ .{ .scalar_int_is = .byte }, .any }, |
| 70226 | 70226 | .patterns = &.{ |
| ... | ... | @@ -70344,7 +70344,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 70344 | 70344 | .{ ._, ._nc, .j, .@"0b", ._, ._, ._ }, |
| 70345 | 70345 | } }, |
| 70346 | 70346 | }, .{ |
| 70347 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 70347 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 70348 | 70348 | .src_constraints = .{ .{ .scalar_remainder_int = .{ .of = .xword, .is = .qword } }, .any, .any }, |
| 70349 | 70349 | .dst_constraints = .{ .{ .scalar_int_is = .word }, .any }, |
| 70350 | 70350 | .patterns = &.{ |
| ... | ... | @@ -70468,7 +70468,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 70468 | 70468 | .{ ._, ._nc, .j, .@"0b", ._, ._, ._ }, |
| 70469 | 70469 | } }, |
| 70470 | 70470 | }, .{ |
| 70471 | .required_features = .{ .@"64bit", .false_deps_lzcnt_tzcnt, .lzcnt, null }, | |
| 70471 | .required_features = .{ .@"64bit", .false_deps_lzcnt, .lzcnt, null }, | |
| 70472 | 70472 | .src_constraints = .{ .{ .scalar_remainder_int = .{ .of = .xword, .is = .xword } }, .any, .any }, |
| 70473 | 70473 | .dst_constraints = .{ .{ .scalar_int_is = .word }, .any }, |
| 70474 | 70474 | .patterns = &.{ |
| ... | ... | @@ -70653,7 +70653,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 70653 | 70653 | .{ ._, ._, .tzcnt, .dst0w, .src0w, ._, ._ }, |
| 70654 | 70654 | } }, |
| 70655 | 70655 | }, .{ |
| 70656 | .required_features = .{ .bmi, .false_deps_lzcnt_tzcnt, null, null }, | |
| 70656 | .required_features = .{ .bmi, .false_deps_tzcnt, null, null }, | |
| 70657 | 70657 | .src_constraints = .{ .{ .exact_int = 32 }, .any, .any }, |
| 70658 | 70658 | .patterns = &.{ |
| 70659 | 70659 | .{ .src = .{ .mem, .none, .none } }, |
| ... | ... | @@ -70690,7 +70690,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 70690 | 70690 | .{ ._, ._, .tzcnt, .dst0d, .src0d, ._, ._ }, |
| 70691 | 70691 | } }, |
| 70692 | 70692 | }, .{ |
| 70693 | .required_features = .{ .@"64bit", .bmi, .false_deps_lzcnt_tzcnt, null }, | |
| 70693 | .required_features = .{ .@"64bit", .bmi, .false_deps_tzcnt, null }, | |
| 70694 | 70694 | .src_constraints = .{ .{ .exact_int = 64 }, .any, .any }, |
| 70695 | 70695 | .patterns = &.{ |
| 70696 | 70696 | .{ .src = .{ .mem, .none, .none } }, |
| ... | ... | @@ -70867,7 +70867,7 @@ fn genBody(cg: *CodeGen, body: []const Air.Inst.Index) InnerError!void { |
| 70867 | 70867 | .{ ._, ._f, .bs, .dst0q, .src0q, ._, ._ }, |
| 70868 | 70868 | } }, |
| 70869 | 70869 | }, .{ |
| 70870 | .required_features = .{ .@"64bit", .bmi, .false_deps_lzcnt_tzcnt, null }, | |
| 70870 | .required_features = .{ .@"64bit", .bmi, .false_deps_tzcnt, null }, | |
| 70871 | 70871 | .src_constraints = .{ .{ .remainder_int = .{ .of = .qword, .is = .qword } }, .any, .any }, |
| 70872 | 70872 | .patterns = &.{ |
| 70873 | 70873 | .{ .src = .{ .to_mem, .none, .none } }, |
| ... | ... | @@ -182371,12 +182371,13 @@ fn hasFeature(cg: *CodeGen, feature: std.Target.x86.Feature) bool { |
| 182371 | 182371 | .x86_64 => true, |
| 182372 | 182372 | }, |
| 182373 | 182373 | .false_deps_getmant, |
| 182374 | .false_deps_lzcnt_tzcnt, | |
| 182374 | .false_deps_lzcnt, | |
| 182375 | 182375 | .false_deps_mulc, |
| 182376 | 182376 | .false_deps_mullq, |
| 182377 | 182377 | .false_deps_perm, |
| 182378 | 182378 | .false_deps_popcnt, |
| 182379 | 182379 | .false_deps_range, |
| 182380 | .false_deps_tzcnt, | |
| 182380 | 182381 | .slow_3ops_lea, |
| 182381 | 182382 | .slow_incdec, |
| 182382 | 182383 | .slow_lea, |
src/link/Wasm.zig+3| ... | ... | @@ -3254,6 +3254,7 @@ pub const Feature = packed struct(u8) { |
| 3254 | 3254 | @"tail-call", |
| 3255 | 3255 | @"shared-mem", |
| 3256 | 3256 | @"wide-arithmetic", |
| 3257 | @"relaxed-atomics", | |
| 3257 | 3258 | |
| 3258 | 3259 | pub fn fromCpuFeature(feature: std.Target.wasm.Feature) Tag { |
| 3259 | 3260 | return switch (feature) { |
| ... | ... | @@ -3271,6 +3272,7 @@ pub const Feature = packed struct(u8) { |
| 3271 | 3272 | .nontrapping_bulk_memory_len0 => .@"nontrapping-bulk-memory-len0", // Zig extension. |
| 3272 | 3273 | .nontrapping_fptoint => .@"nontrapping-fptoint", |
| 3273 | 3274 | .reference_types => .@"reference-types", |
| 3275 | .relaxed_atomics => .@"relaxed-atomics", | |
| 3274 | 3276 | .relaxed_simd => .@"relaxed-simd", |
| 3275 | 3277 | .sign_ext => .@"sign-ext", |
| 3276 | 3278 | .simd128 => .simd128, |
| ... | ... | @@ -3296,6 +3298,7 @@ pub const Feature = packed struct(u8) { |
| 3296 | 3298 | .@"nontrapping-bulk-memory-len0" => .nontrapping_bulk_memory_len0, // Zig extension. |
| 3297 | 3299 | .@"nontrapping-fptoint" => .nontrapping_fptoint, |
| 3298 | 3300 | .@"reference-types" => .reference_types, |
| 3301 | .@"relaxed-atomics" => .relaxed_atomics, | |
| 3299 | 3302 | .@"relaxed-simd" => .relaxed_simd, |
| 3300 | 3303 | .@"sign-ext" => .sign_ext, |
| 3301 | 3304 | .simd128 => .simd128, |
src/target.zig+3| ... | ... | @@ -415,6 +415,9 @@ pub fn hasDebugInfo(target: *const std.Target) bool { |
| 415 | 415 | .ptx87, |
| 416 | 416 | .ptx88, |
| 417 | 417 | .ptx90, |
| 418 | .ptx91, | |
| 419 | .ptx92, | |
| 420 | .ptx93, | |
| 418 | 421 | }), |
| 419 | 422 | .bpfel, .bpfeb => false, |
| 420 | 423 | else => true, |
test/tests.zig+12-1| ... | ... | @@ -2357,13 +2357,17 @@ pub fn isNative(actual_target: *const std.Build.ResolvedTarget, host: *const std |
| 2357 | 2357 | .@"16bit_mode", |
| 2358 | 2358 | .@"32bit_mode", |
| 2359 | 2359 | .@"64bit", |
| 2360 | .false_deps_bls, | |
| 2361 | .false_deps_compress, | |
| 2362 | .false_deps_expand, | |
| 2360 | 2363 | .false_deps_getmant, |
| 2361 | .false_deps_lzcnt_tzcnt, | |
| 2364 | .false_deps_lzcnt, | |
| 2362 | 2365 | .false_deps_mulc, |
| 2363 | 2366 | .false_deps_mullq, |
| 2364 | 2367 | .false_deps_perm, |
| 2365 | 2368 | .false_deps_popcnt, |
| 2366 | 2369 | .false_deps_range, |
| 2370 | .false_deps_tzcnt, | |
| 2367 | 2371 | .fast_11bytenop, |
| 2368 | 2372 | .fast_15bytenop, |
| 2369 | 2373 | .fast_7bytenop, |
| ... | ... | @@ -2391,10 +2395,12 @@ pub fn isNative(actual_target: *const std.Build.ResolvedTarget, host: *const std |
| 2391 | 2395 | .prefer_legacy_setcc, |
| 2392 | 2396 | .prefer_mask_registers, |
| 2393 | 2397 | .prefer_movmsk_over_vtest, |
| 2398 | .prefer_ndd_mem, | |
| 2394 | 2399 | .prefer_no_gather, |
| 2395 | 2400 | .prefer_no_scatter, |
| 2396 | 2401 | .slow_3ops_lea, |
| 2397 | 2402 | .slow_incdec, |
| 2403 | .slow_indirect_call, | |
| 2398 | 2404 | .slow_lea, |
| 2399 | 2405 | .slow_pmaddwd, |
| 2400 | 2406 | .slow_pmulld, |
| ... | ... | @@ -2406,10 +2412,13 @@ pub fn isNative(actual_target: *const std.Build.ResolvedTarget, host: *const std |
| 2406 | 2412 | }), |
| 2407 | 2413 | .aarch64, .aarch64_be => std.Target.aarch64.featureSet(&.{ |
| 2408 | 2414 | .addr_lsl_slow_14, |
| 2415 | .align_cmp_csel_pairs, | |
| 2409 | 2416 | .alu_lsl_fast, |
| 2410 | 2417 | .avoid_ldapur, |
| 2411 | 2418 | .disable_fast_inc_vl, |
| 2412 | 2419 | .exynos_cheap_as_move, |
| 2420 | .fast_ld1_single, | |
| 2421 | .fixed_load_latency_4, | |
| 2413 | 2422 | .fuse_address, |
| 2414 | 2423 | .fuse_addsub_2reg_const1, |
| 2415 | 2424 | .fuse_adrp_add, |
| ... | ... | @@ -2418,7 +2427,9 @@ pub fn isNative(actual_target: *const std.Build.ResolvedTarget, host: *const std |
| 2418 | 2427 | .fuse_crypto_eor, |
| 2419 | 2428 | .fuse_csel, |
| 2420 | 2429 | .fuse_cset, |
| 2430 | .fuse_fcsel, | |
| 2421 | 2431 | .fuse_literals, |
| 2432 | .has_limited_64bit_vector_mul_bandwidth, | |
| 2422 | 2433 | .predictable_select_expensive, |
| 2423 | 2434 | .slow_misaligned_128store, |
| 2424 | 2435 | .slow_paired_128, |
tools/update_cpu_features.zig+10-99| ... | ... | @@ -131,6 +131,14 @@ const targets = [_]ArchTarget{ |
| 131 | 131 | .zig_name = "contextidr_el2", |
| 132 | 132 | .desc = "Enable RW operand Context ID Register (EL2)", |
| 133 | 133 | }, |
| 134 | .{ | |
| 135 | .llvm_name = "hip12", | |
| 136 | .flatten = true, | |
| 137 | }, | |
| 138 | .{ | |
| 139 | .llvm_name = "mca-streaming-sched", | |
| 140 | .omit = true, | |
| 141 | }, | |
| 134 | 142 | .{ |
| 135 | 143 | .llvm_name = "neoversee1", |
| 136 | 144 | .flatten = true, |
| ... | ... | @@ -1553,103 +1561,6 @@ const targets = [_]ArchTarget{ |
| 1553 | 1561 | .zig_name = "baseline_rv64", |
| 1554 | 1562 | .features = &.{ "64bit", "a", "c", "d", "f", "i", "m" }, |
| 1555 | 1563 | }, |
| 1556 | .{ | |
| 1557 | .llvm_name = null, | |
| 1558 | .zig_name = "spacemit_a100", | |
| 1559 | .features = &.{ | |
| 1560 | "64bit", | |
| 1561 | "a", | |
| 1562 | "b", | |
| 1563 | "c", | |
| 1564 | "dlen_factor_2", | |
| 1565 | "i", | |
| 1566 | "m", | |
| 1567 | "optimized_nf2_segment_load_store", | |
| 1568 | "optimized_nf3_segment_load_store", | |
| 1569 | "optimized_nf4_segment_load_store", | |
| 1570 | "smepmp", | |
| 1571 | "smnpm", | |
| 1572 | "smstateen", | |
| 1573 | "ssccptr", | |
| 1574 | "sscofpmf", | |
| 1575 | "sscounterenw", | |
| 1576 | "ssnpm", | |
| 1577 | "sspm", | |
| 1578 | "sstc", | |
| 1579 | "sstvala", | |
| 1580 | "sstvecd", | |
| 1581 | "ssu64xl", | |
| 1582 | "supm", | |
| 1583 | "svade", | |
| 1584 | "svbare", | |
| 1585 | "svinval", | |
| 1586 | "svnapot", | |
| 1587 | "svpbmt", | |
| 1588 | "unaligned_scalar_mem", | |
| 1589 | "v", | |
| 1590 | "vxrm_pipeline_flush", | |
| 1591 | "za64rs", | |
| 1592 | "zawrs", | |
| 1593 | "zbc", | |
| 1594 | "zbkc", | |
| 1595 | "zcb", | |
| 1596 | "zcmop", | |
| 1597 | "zfa", | |
| 1598 | "zfh", | |
| 1599 | "zic64b", | |
| 1600 | "zicbom", | |
| 1601 | "zicbop", | |
| 1602 | "zicboz", | |
| 1603 | "ziccamoa", | |
| 1604 | "ziccif", | |
| 1605 | "zicclsm", | |
| 1606 | "ziccrse", | |
| 1607 | "zicntr", | |
| 1608 | "zicond", | |
| 1609 | "zifencei", | |
| 1610 | "zihintntl", | |
| 1611 | "zihintpause", | |
| 1612 | "zihpm", | |
| 1613 | "zimop", | |
| 1614 | "zkt", | |
| 1615 | "zvbb", | |
| 1616 | "zvfbfwma", | |
| 1617 | "zvfh", | |
| 1618 | "zvkng", | |
| 1619 | "zvknha", | |
| 1620 | "zvksc", | |
| 1621 | "zvksg", | |
| 1622 | "zvl1024b", | |
| 1623 | }, | |
| 1624 | }, | |
| 1625 | .{ | |
| 1626 | .llvm_name = null, | |
| 1627 | .zig_name = "spacemit_x100", | |
| 1628 | .features = &.{ | |
| 1629 | "dlen_factor_2", | |
| 1630 | "optimized_nf2_segment_load_store", | |
| 1631 | "optimized_nf3_segment_load_store", | |
| 1632 | "optimized_nf4_segment_load_store", | |
| 1633 | "rva23s64", | |
| 1634 | "smepmp", | |
| 1635 | "smnpm", | |
| 1636 | "smstateen", | |
| 1637 | "sspm", | |
| 1638 | "unaligned_scalar_mem", | |
| 1639 | "vxrm_pipeline_flush", | |
| 1640 | "xsmtvdot", | |
| 1641 | "zbc", | |
| 1642 | "zbkc", | |
| 1643 | "zfh", | |
| 1644 | "zvfbfwma", | |
| 1645 | "zvfh", | |
| 1646 | "zvkng", | |
| 1647 | "zvknha", | |
| 1648 | "zvksc", | |
| 1649 | "zvksg", | |
| 1650 | "zvl256b", | |
| 1651 | }, | |
| 1652 | }, | |
| 1653 | 1564 | }, |
| 1654 | 1565 | }, |
| 1655 | 1566 | .{ |
| ... | ... | @@ -2450,7 +2361,7 @@ fn processOneTargetInner(io: Io, job: Job) !void { |
| 2450 | 2361 | for (all_features.items) |feature| { |
| 2451 | 2362 | if (feature.llvm_name) |llvm_name| { |
| 2452 | 2363 | try w.print( |
| 2453 | \\ result[@intFromEnum(Feature.{f})] = .{{ | |
| 2364 | \\ result[@backingInt(Feature.{f})] = .{{ | |
| 2454 | 2365 | \\ .llvm_name = "{f}", |
| 2455 | 2366 | \\ .description = "{f}", |
| 2456 | 2367 | \\ .dependencies = featureSet(&[_]Feature{{ |
| ... | ... | @@ -2463,7 +2374,7 @@ fn processOneTargetInner(io: Io, job: Job) !void { |
| 2463 | 2374 | ); |
| 2464 | 2375 | } else { |
| 2465 | 2376 | try w.print( |
| 2466 | \\ result[@intFromEnum(Feature.{f})] = .{{ | |
| 2377 | \\ result[@backingInt(Feature.{f})] = .{{ | |
| 2467 | 2378 | \\ .llvm_name = null, |
| 2468 | 2379 | \\ .description = "{f}", |
| 2469 | 2380 | \\ .dependencies = featureSet(&[_]Feature{{ |