diff options
34 files changed, 735 insertions, 15 deletions
diff --git a/gcc/common/config/i386/cpuinfo.h b/gcc/common/config/i386/cpuinfo.h index 7c2565c..3599f9d 100644 --- a/gcc/common/config/i386/cpuinfo.h +++ b/gcc/common/config/i386/cpuinfo.h @@ -875,6 +875,8 @@ get_available_features (struct __processor_model *cpu_model, set_feature (FEATURE_AVXVNNIINT8); if (edx & bit_AVXNECONVERT) set_feature (FEATURE_AVXNECONVERT); + if (edx & bit_AVXVNNIINT16) + set_feature (FEATURE_AVXVNNIINT16); } if (avx512_usable) { diff --git a/gcc/common/config/i386/i386-common.cc b/gcc/common/config/i386/i386-common.cc index 8cea366..32c6d00 100644 --- a/gcc/common/config/i386/i386-common.cc +++ b/gcc/common/config/i386/i386-common.cc @@ -119,6 +119,7 @@ along with GCC; see the file COPYING3. If not see #define OPTION_MASK_ISA2_RAOINT_SET OPTION_MASK_ISA2_RAOINT #define OPTION_MASK_ISA2_AMX_COMPLEX_SET \ (OPTION_MASK_ISA2_AMX_TILE | OPTION_MASK_ISA2_AMX_COMPLEX) +#define OPTION_MASK_ISA2_AVXVNNIINT16_SET OPTION_MASK_ISA2_AVXVNNIINT16 /* SSE4 includes both SSE4.1 and SSE4.2. -msse4 should be the same as -msse4.2. */ @@ -228,7 +229,7 @@ along with GCC; see the file COPYING3. If not see #define OPTION_MASK_ISA2_AVX2_UNSET \ (OPTION_MASK_ISA2_AVXIFMA_UNSET | OPTION_MASK_ISA2_AVXVNNI_UNSET \ | OPTION_MASK_ISA2_AVXVNNIINT8_UNSET | OPTION_MASK_ISA2_AVXNECONVERT_UNSET \ - | OPTION_MASK_ISA2_AVX512F_UNSET) + | OPTION_MASK_ISA2_AVXVNNIINT16_UNSET | OPTION_MASK_ISA2_AVX512F_UNSET) #define OPTION_MASK_ISA_AVX512F_UNSET \ (OPTION_MASK_ISA_AVX512F | OPTION_MASK_ISA_AVX512CD_UNSET \ | OPTION_MASK_ISA_AVX512PF_UNSET | OPTION_MASK_ISA_AVX512ER_UNSET \ @@ -301,6 +302,7 @@ along with GCC; see the file COPYING3. If not see #define OPTION_MASK_ISA2_PREFETCHI_UNSET OPTION_MASK_ISA2_PREFETCHI #define OPTION_MASK_ISA2_RAOINT_UNSET OPTION_MASK_ISA2_RAOINT #define OPTION_MASK_ISA2_AMX_COMPLEX_UNSET OPTION_MASK_ISA2_AMX_COMPLEX +#define OPTION_MASK_ISA2_AVXVNNIINT16_UNSET OPTION_MASK_ISA2_AVXVNNIINT16 /* SSE4 includes both SSE4.1 and SSE4.2. -mno-sse4 should the same as -mno-sse4.1. */ @@ -1268,6 +1270,24 @@ ix86_handle_option (struct gcc_options *opts, } return true; + case OPT_mavxvnniint16: + if (value) + { + opts->x_ix86_isa_flags2 |= OPTION_MASK_ISA2_AVXVNNIINT16_SET; + opts->x_ix86_isa_flags2_explicit |= + OPTION_MASK_ISA2_AVXVNNIINT16_SET; + opts->x_ix86_isa_flags |= OPTION_MASK_ISA_AVX2_SET; + opts->x_ix86_isa_flags_explicit |= OPTION_MASK_ISA_AVX2_SET; + } + else + { + opts->x_ix86_isa_flags2 &= + ~OPTION_MASK_ISA2_AVXVNNIINT16_UNSET; + opts->x_ix86_isa_flags2_explicit |= + OPTION_MASK_ISA2_AVXVNNIINT16_UNSET; + } + return true; + case OPT_mfma: if (value) { diff --git a/gcc/common/config/i386/i386-cpuinfo.h b/gcc/common/config/i386/i386-cpuinfo.h index 254dfec..ae4e6a0 100644 --- a/gcc/common/config/i386/i386-cpuinfo.h +++ b/gcc/common/config/i386/i386-cpuinfo.h @@ -255,6 +255,7 @@ enum processor_features FEATURE_PREFETCHI, FEATURE_RAOINT, FEATURE_AMX_COMPLEX, + FEATURE_AVXVNNIINT16, CPU_FEATURE_MAX }; diff --git a/gcc/common/config/i386/i386-isas.h b/gcc/common/config/i386/i386-isas.h index d4b0b23..fc6abde 100644 --- a/gcc/common/config/i386/i386-isas.h +++ b/gcc/common/config/i386/i386-isas.h @@ -186,4 +186,6 @@ ISA_NAMES_TABLE_START ISA_NAMES_TABLE_ENTRY("raoint", FEATURE_RAOINT, P_NONE, "-mraoint") ISA_NAMES_TABLE_ENTRY("amx-complex", FEATURE_AMX_COMPLEX, P_NONE, "-mamx-complex") + ISA_NAMES_TABLE_ENTRY("avxvnniint16", FEATURE_AVXVNNIINT16, + P_NONE, "-mavxvnniint16") ISA_NAMES_TABLE_END diff --git a/gcc/config.gcc b/gcc/config.gcc index 1446eb2..fc74d77 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -435,7 +435,7 @@ i[34567]86-*-* | x86_64-*-*) mwaitintrin.h avx512fp16intrin.h avx512fp16vlintrin.h avxifmaintrin.h avxvnniint8intrin.h avxneconvertintrin.h cmpccxaddintrin.h amxfp16intrin.h prfchiintrin.h - raointintrin.h amxcomplexintrin.h" + raointintrin.h amxcomplexintrin.h avxvnniint16intrin.h" ;; ia64-*-*) extra_headers=ia64intrin.h diff --git a/gcc/config/i386/avxvnniint16intrin.h b/gcc/config/i386/avxvnniint16intrin.h new file mode 100644 index 0000000..f87d76c --- /dev/null +++ b/gcc/config/i386/avxvnniint16intrin.h @@ -0,0 +1,138 @@ +/* Copyright (C) 2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3, or (at your option) + any later version. + + GCC is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#if !defined _IMMINTRIN_H_INCLUDED +#error "Never use <avxvnniint16intrin.h> directly; include <immintrin.h> instead." +#endif + +#ifndef _AVXVNNIINT16INTRIN_H_INCLUDED +#define _AVXVNNIINT16INTRIN_H_INCLUDED + +#if !defined(__AVXVNNIINT16__) +#pragma GCC push_options +#pragma GCC target("avxvnniint16") +#define __DISABLE_AVXVNNIINT16__ +#endif /* __AVXVNNIINT16__ */ + +extern __inline __m128i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm_dpwsud_avx_epi32 (__m128i __W, __m128i __A, __m128i __B) +{ + return (__m128i) + __builtin_ia32_vpdpwsud128 ((__v4si) __W, (__v4si) __A, (__v4si) __B); +} + +extern __inline __m128i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm_dpwsuds_avx_epi32 (__m128i __W, __m128i __A, __m128i __B) +{ + return (__m128i) + __builtin_ia32_vpdpwsuds128 ((__v4si) __W, (__v4si) __A, (__v4si) __B); +} + +extern __inline __m128i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm_dpwusd_avx_epi32 (__m128i __W, __m128i __A, __m128i __B) +{ + return (__m128i) + __builtin_ia32_vpdpwusd128 ((__v4si) __W, (__v4si) __A, (__v4si) __B); +} + +extern __inline __m128i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm_dpwusds_avx_epi32 (__m128i __W, __m128i __A, __m128i __B) +{ + return (__m128i) + __builtin_ia32_vpdpwusds128 ((__v4si) __W, (__v4si) __A, (__v4si) __B); +} + +extern __inline __m128i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm_dpwuud_avx_epi32 (__m128i __W, __m128i __A, __m128i __B) +{ + return (__m128i) + __builtin_ia32_vpdpwuud128 ((__v4si) __W, (__v4si) __A, (__v4si) __B); +} + +extern __inline __m128i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm_dpwuuds_avx_epi32 (__m128i __W, __m128i __A, __m128i __B) +{ + return (__m128i) + __builtin_ia32_vpdpwuuds128 ((__v4si) __W, (__v4si) __A, (__v4si) __B); +} + +extern __inline __m256i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm256_dpwsud_avx_epi32 (__m256i __W, __m256i __A, __m256i __B) +{ + return (__m256i) + __builtin_ia32_vpdpwsud256 ((__v8si) __W, (__v8si) __A, (__v8si) __B); +} + +extern __inline __m256i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm256_dpwsuds_avx_epi32 (__m256i __W, __m256i __A, __m256i __B) +{ + return (__m256i) + __builtin_ia32_vpdpwsuds256 ((__v8si) __W, (__v8si) __A, (__v8si) __B); +} + +extern __inline __m256i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm256_dpwusd_avx_epi32 (__m256i __W, __m256i __A, __m256i __B) +{ + return (__m256i) + __builtin_ia32_vpdpwusd256 ((__v8si) __W, (__v8si) __A, (__v8si) __B); +} + +extern __inline __m256i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm256_dpwusds_avx_epi32 (__m256i __W, __m256i __A, __m256i __B) +{ + return (__m256i) + __builtin_ia32_vpdpwusds256 ((__v8si) __W, (__v8si) __A, (__v8si) __B); +} + +extern __inline __m256i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm256_dpwuud_avx_epi32 (__m256i __W, __m256i __A, __m256i __B) +{ + return (__m256i) + __builtin_ia32_vpdpwuud256 ((__v8si) __W, (__v8si) __A, (__v8si) __B); +} + +extern __inline __m256i +__attribute__((__gnu_inline__, __always_inline__, __artificial__)) +_mm256_dpwuuds_avx_epi32 (__m256i __W, __m256i __A, __m256i __B) +{ + return (__m256i) + __builtin_ia32_vpdpwuuds256 ((__v8si) __W, (__v8si) __A, (__v8si) __B); +} + +#ifdef __DISABLE_AVXVNNIINT16__ +#undef __DISABLE_AVXVNNIINT16__ +#pragma GCC pop_options +#endif /* __DISABLE_AVXVNNIINT16__ */ + +#endif /* __AVXVNNIINT16INTRIN_H_INCLUDED */ diff --git a/gcc/config/i386/cpuid.h b/gcc/config/i386/cpuid.h index 4cc4461..98d0f19 100644 --- a/gcc/config/i386/cpuid.h +++ b/gcc/config/i386/cpuid.h @@ -144,6 +144,7 @@ /* %edx */ #define bit_AVXVNNIINT8 (1 << 4) #define bit_AVXNECONVERT (1 << 5) +#define bit_AVXVNNIINT16 (1 << 10) #define bit_PREFETCHI (1 << 14) /* Extended State Enumeration Sub-leaf (%eax == 0xd, %ecx == 1) */ diff --git a/gcc/config/i386/i386-builtin.def b/gcc/config/i386/i386-builtin.def index 7ba5b6a..ff5b3dc 100644 --- a/gcc/config/i386/i386-builtin.def +++ b/gcc/config/i386/i386-builtin.def @@ -2740,6 +2740,20 @@ BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT8, CODE_FOR_vpdpbsuds_v4si, "__builtin_ia32 BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT8, CODE_FOR_vpdpbuud_v4si, "__builtin_ia32_vpdpbuud128", IX86_BUILTIN_VPDPBUUDV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI) BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT8, CODE_FOR_vpdpbuuds_v4si, "__builtin_ia32_vpdpbuuds128", IX86_BUILTIN_VPDPBUUDSV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI) +/* AVXVNNIINT16 */ +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwusd_v8si, "__builtin_ia32_vpdpwusd256", IX86_BUILTIN_VPDPWUSDV8SI, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI_V8SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwusds_v8si, "__builtin_ia32_vpdpwusds256", IX86_BUILTIN_VPDPWUSDSV8SI, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI_V8SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwsud_v8si, "__builtin_ia32_vpdpwsud256", IX86_BUILTIN_VPDPWSUDV8SI, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI_V8SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwsuds_v8si, "__builtin_ia32_vpdpwsuds256", IX86_BUILTIN_VPDPWSUDSV8SI, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI_V8SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwuud_v8si, "__builtin_ia32_vpdpwuud256", IX86_BUILTIN_VPDPWUUDV8SI, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI_V8SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwuuds_v8si, "__builtin_ia32_vpdpwuuds256", IX86_BUILTIN_VPDPWUUDSV8SI, UNKNOWN, (int) V8SI_FTYPE_V8SI_V8SI_V8SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwusd_v4si, "__builtin_ia32_vpdpwusd128", IX86_BUILTIN_VPDPWUSDV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwusds_v4si, "__builtin_ia32_vpdpwusds128", IX86_BUILTIN_VPDPWUSDSV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwsud_v4si, "__builtin_ia32_vpdpwsud128", IX86_BUILTIN_VPDPWSUDV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwsuds_v4si, "__builtin_ia32_vpdpwsuds128", IX86_BUILTIN_VPDPWSUDSV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwuud_v4si, "__builtin_ia32_vpdpwuud128", IX86_BUILTIN_VPDPWUUDV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI) +BDESC (0, OPTION_MASK_ISA2_AVXVNNIINT16, CODE_FOR_vpdpwuuds_v4si, "__builtin_ia32_vpdpwuuds128", IX86_BUILTIN_VPDPWUUDSV4SI, UNKNOWN, (int) V4SI_FTYPE_V4SI_V4SI_V4SI) + /* VPCLMULQDQ */ BDESC (OPTION_MASK_ISA_VPCLMULQDQ | OPTION_MASK_ISA_AVX512VL, 0, CODE_FOR_vpclmulqdq_v2di, "__builtin_ia32_vpclmulqdq_v2di", IX86_BUILTIN_VPCLMULQDQ2, UNKNOWN, (int) V2DI_FTYPE_V2DI_V2DI_INT) BDESC (OPTION_MASK_ISA_VPCLMULQDQ | OPTION_MASK_ISA_AVX, 0, CODE_FOR_vpclmulqdq_v4di, "__builtin_ia32_vpclmulqdq_v4di", IX86_BUILTIN_VPCLMULQDQ4, UNKNOWN, (int) V4DI_FTYPE_V4DI_V4DI_INT) diff --git a/gcc/config/i386/i386-c.cc b/gcc/config/i386/i386-c.cc index e7bd7cc..d3514dd 100644 --- a/gcc/config/i386/i386-c.cc +++ b/gcc/config/i386/i386-c.cc @@ -677,6 +677,8 @@ ix86_target_macros_internal (HOST_WIDE_INT isa_flag, def_or_undef (parse_in, "__RAOINT__"); if (isa_flag2 & OPTION_MASK_ISA2_AMX_COMPLEX) def_or_undef (parse_in, "__AMX_COMPLEX__"); + if (isa_flag2 & OPTION_MASK_ISA2_AVXVNNIINT16) + def_or_undef (parse_in, "__AVXVNNIINT16__"); if (TARGET_IAMCU) { def_or_undef (parse_in, "__iamcu"); diff --git a/gcc/config/i386/i386-isa.def b/gcc/config/i386/i386-isa.def index 0634c6f..fbf22f7 100644 --- a/gcc/config/i386/i386-isa.def +++ b/gcc/config/i386/i386-isa.def @@ -117,3 +117,4 @@ DEF_PTA(AMX_FP16) DEF_PTA(PREFETCHI) DEF_PTA(RAOINT) DEF_PTA(AMX_COMPLEX) +DEF_PTA(AVXVNNIINT16) diff --git a/gcc/config/i386/i386-options.cc b/gcc/config/i386/i386-options.cc index 37cb5a0..d981666 100644 --- a/gcc/config/i386/i386-options.cc +++ b/gcc/config/i386/i386-options.cc @@ -239,7 +239,8 @@ static struct ix86_target_opts isa2_opts[] = { "-mamx-fp16", OPTION_MASK_ISA2_AMX_FP16 }, { "-mprefetchi", OPTION_MASK_ISA2_PREFETCHI }, { "-mraoint", OPTION_MASK_ISA2_RAOINT }, - { "-mamx-complex", OPTION_MASK_ISA2_AMX_COMPLEX } + { "-mamx-complex", OPTION_MASK_ISA2_AMX_COMPLEX }, + { "-mavxvnniint16", OPTION_MASK_ISA2_AVXVNNIINT16 }, }; static struct ix86_target_opts isa_opts[] = { @@ -1091,6 +1092,7 @@ ix86_valid_target_attribute_inner_p (tree fndecl, tree args, char *p_strings[], IX86_ATTR_ISA ("prefetchi", OPT_mprefetchi), IX86_ATTR_ISA ("raoint", OPT_mraoint), IX86_ATTR_ISA ("amx-complex", OPT_mamx_complex), + IX86_ATTR_ISA ("avxvnniint16", OPT_mavxvnniint16), /* enum options */ IX86_ATTR_ENUM ("fpmath=", OPT_mfpmath_), diff --git a/gcc/config/i386/i386.opt b/gcc/config/i386/i386.opt index d74f6b1..618d713 100644 --- a/gcc/config/i386/i386.opt +++ b/gcc/config/i386/i386.opt @@ -1278,3 +1278,8 @@ Enum(lam_type) String(u57) Value(lam_u57) mamx-complex Target Mask(ISA2_AMX_COMPLEX) Var(ix86_isa_flags2) Save Support AMX-COMPLEX built-in functions and code generation. + +mavxvnniint16 +Target Mask(ISA2_AVXVNNIINT16) Var(ix86_isa_flags2) Save +Support MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, AVX, AVX2 and +AVXVNNIINT16 built-in functions and code generation. diff --git a/gcc/config/i386/immintrin.h b/gcc/config/i386/immintrin.h index b220d87..52dc35d 100644 --- a/gcc/config/i386/immintrin.h +++ b/gcc/config/i386/immintrin.h @@ -48,6 +48,8 @@ #include <avxvnniint8intrin.h> +#include <avxvnniint16intrin.h> + #include <avx2intrin.h> #include <avx512fintrin.h> diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md index 6bf9c99..85a5f80 100644 --- a/gcc/config/i386/sse.md +++ b/gcc/config/i386/sse.md @@ -204,6 +204,14 @@ UNSPEC_VPDPBSUDS UNSPEC_VPDPBUUD UNSPEC_VPDPBUUDS + + ;; For AVX-VNNI-INT16 support + UNSPEC_VPDPWUSD + UNSPEC_VPDPWUSDS + UNSPEC_VPDPWSUD + UNSPEC_VPDPWSUDS + UNSPEC_VPDPWUUD + UNSPEC_VPDPWUUDS ]) (define_c_enum "unspecv" [ @@ -30209,3 +30217,27 @@ "vcvtneo<bf16_ph>2ps\t{%1, %0|%0, %1}" [(set_attr "prefix" "vex") (set_attr "mode" "<sseinsnmode>")]) + +(define_int_iterator VPDPWPROD + [UNSPEC_VPDPWUSD + UNSPEC_VPDPWUSDS + UNSPEC_VPDPWSUD + UNSPEC_VPDPWSUDS + UNSPEC_VPDPWUUD + UNSPEC_VPDPWUUDS]) + +(define_int_attr vpdpwprodtype + [(UNSPEC_VPDPWUSD "wusd") (UNSPEC_VPDPWUSDS "wusds") + (UNSPEC_VPDPWSUD "wsud") (UNSPEC_VPDPWSUDS "wsuds") + (UNSPEC_VPDPWUUD "wuud") (UNSPEC_VPDPWUUDS "wuuds")]) + +(define_insn "vpdp<vpdpwprodtype>_<mode>" + [(set (match_operand:VI4_AVX 0 "register_operand" "=x") + (unspec:VI4_AVX + [(match_operand:VI4_AVX 1 "register_operand" "0") + (match_operand:VI4_AVX 2 "register_operand" "x") + (match_operand:VI4_AVX 3 "nonimmediate_operand" "xm")] + VPDPWPROD))] + "TARGET_AVXVNNIINT16" + "vpdp<vpdpwprodtype>\t{%3, %2, %0|%0, %2, %3}" + [(set_attr "prefix" "vex")]) diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi index dda3535..2646dd5 100644 --- a/gcc/doc/extend.texi +++ b/gcc/doc/extend.texi @@ -7163,6 +7163,11 @@ Enable/disable the generation of the RAOINT instructions. @itemx no-amx-complex Enable/disable the generation of the AMX-COMPLEX instructions. +@cindex @code{target("avxvnniint16")} function attribute, x86 +@item avxvnniint16 +@itemx no-avxvnniint16 +Enable/disable the generation of the AVXVNNIINT16 instructions. + @cindex @code{target("cld")} function attribute, x86 @item cld @itemx no-cld diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi index cbc1282..359887db 100644 --- a/gcc/doc/invoke.texi +++ b/gcc/doc/invoke.texi @@ -1433,7 +1433,7 @@ See RS/6000 and PowerPC Options. -mrdseed -msgx -mavx512vp2intersect -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -muintr -mhreset -mavxvnni -mavx512fp16 -mavxifma -mavxvnniint8 -mavxneconvert -mcmpccxadd -mamx-fp16 --mprefetchi -mraoint -mamx-complex +-mprefetchi -mraoint -mamx-complex -mavxvnniint16 -mcldemote -mms-bitfields -mno-align-stringops -minline-all-stringops -minline-stringops-dynamically -mstringop-strategy=@var{alg} -mkl -mwidekl @@ -33552,8 +33552,10 @@ preferred alignment to @option{-mpreferred-stack-boundary=2}. @need 200 @opindex mamx-complex @itemx -mamx-complex +@need 200 +@opindex mavxvnniint16 +@itemx -mavxvnniint16 These switches enable the use of instructions in the MMX, SSE, -SSE2, SSE3, SSSE3, SSE4, SSE4A, SSE4.1, SSE4.2, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD, AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, SHA, AES, PCLMUL, CLFLUSHOPT, CLWB, FSGSBASE, PTWRITE, RDRND, F16C, FMA, PCONFIG, WBNOINVD, FMA4, PREFETCHW, RDPID, PREFETCHWT1, RDSEED, SGX, XOP, LWP, @@ -33563,8 +33565,8 @@ GFNI, VAES, WAITPKG, VPCLMULQDQ, AVX512BITALG, MOVDIRI, MOVDIR64B, AVX512BF16, ENQCMD, AVX512VPOPCNTDQ, AVX5124FMAPS, AVX512VNNI, AVX5124VNNIW, SERIALIZE, UINTR, HRESET, AMXTILE, AMXINT8, AMXBF16, KL, WIDEKL, AVXVNNI, AVX512-FP16, AVXIFMA, AVXVNNIINT8, AVXNECONVERT, CMPCCXADD, AMX-FP16, PREFETCHI, RAOINT, -AMX-COMPLEX or CLDEMOTE extended instruction sets. Each has a corresponding -@option{-mno-} option to disable use of these instructions. +AMX-COMPLEX, AVXVNNIINT16 or CLDEMOTE extended instruction sets. Each has a +corresponding @option{-mno-} option to disable use of these instructions. These extensions are also available as built-in functions: see @ref{x86 Built-in Functions}, for details of the functions enabled and diff --git a/gcc/doc/sourcebuild.texi b/gcc/doc/sourcebuild.texi index ffb6eb1..40919b3 100644 --- a/gcc/doc/sourcebuild.texi +++ b/gcc/doc/sourcebuild.texi @@ -2511,6 +2511,9 @@ Target supports the execution of @code{avxneconvert} instructions. @item avxvnniint8 Target supports the execution of @code{avxvnniint8} instructions. +@item avxvnniint16 +Target supports the execution of @code{avxvnniint16} instructions. + @item amx_tile Target supports the execution of @code{amx-tile} instructions. diff --git a/gcc/testsuite/g++.dg/other/i386-2.C b/gcc/testsuite/g++.dg/other/i386-2.C index 6fe07e1..53622df 100644 --- a/gcc/testsuite/g++.dg/other/i386-2.C +++ b/gcc/testsuite/g++.dg/other/i386-2.C @@ -1,5 +1,5 @@ /* { dg-do compile { target i?86-*-* x86_64-*-* } } */ -/* { dg-options "-O -pedantic-errors -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512dq -mavx512bw -mavx512vl -mavx512ifma -mavx512vbmi -mavx512vbmi2 -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mavx512bitalg -mpconfig -mwbnoinvd -mavx512bf16 -menqcmd -mavx512vp2intersect -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavx512fp16 -mavxifma -mavxvnniint8 -mavxneconvert -mcmpccxadd -mamx-fp16 -mprefetchi -mraoint -mamx-complex" } */ +/* { dg-options "-O -pedantic-errors -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512dq -mavx512bw -mavx512vl -mavx512ifma -mavx512vbmi -mavx512vbmi2 -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mavx512bitalg -mpconfig -mwbnoinvd -mavx512bf16 -menqcmd -mavx512vp2intersect -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavx512fp16 -mavxifma -mavxvnniint8 -mavxneconvert -mcmpccxadd -mamx-fp16 -mprefetchi -mraoint -mamx-complex -mavxvnniint16" } */ /* Test that {,x,e,p,t,s,w,a,b,i}mmintrin.h, mm3dnow.h, fma4intrin.h, xopintrin.h, abmintrin.h, bmiintrin.h, tbmintrin.h, lwpintrin.h, diff --git a/gcc/testsuite/g++.dg/other/i386-3.C b/gcc/testsuite/g++.dg/other/i386-3.C index 55c8167..3b76cee 100644 --- a/gcc/testsuite/g++.dg/other/i386-3.C +++ b/gcc/testsuite/g++.dg/other/i386-3.C @@ -1,5 +1,5 @@ /* { dg-do compile { target i?86-*-* x86_64-*-* } } */ -/* { dg-options "-O -fkeep-inline-functions -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512dq -mavx512bw -mavx512vl -mavx512ifma -mavx512vbmi -mavx512vbmi2 -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mavx512bitalg -mpconfig -mwbnoinvd -mavx512bf16 -menqcmd -mavx512vp2intersect -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavx512fp16 -mavxifma -mavxvnniint8 -mavxneconvert -mcmpccxadd -mamx-fp16 -mprefetchi -mraoint -mamx-complex" } */ +/* { dg-options "-O -fkeep-inline-functions -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512dq -mavx512bw -mavx512vl -mavx512ifma -mavx512vbmi -mavx512vbmi2 -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mavx512bitalg -mpconfig -mwbnoinvd -mavx512bf16 -menqcmd -mavx512vp2intersect -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavx512fp16 -mavxifma -mavxvnniint8 -mavxneconvert -mcmpccxadd -mamx-fp16 -mprefetchi -mraoint -mamx-complex -mavxvnniint16" } */ /* Test that {,x,e,p,t,s,w,a,b,i}mmintrin.h, mm3dnow.h, fma4intrin.h, xopintrin.h, abmintrin.h, bmiintrin.h, tbmintrin.h, lwpintrin.h, diff --git a/gcc/testsuite/gcc.target/i386/avx-check.h b/gcc/testsuite/gcc.target/i386/avx-check.h index 666eff5..3d417ea1 100644 --- a/gcc/testsuite/gcc.target/i386/avx-check.h +++ b/gcc/testsuite/gcc.target/i386/avx-check.h @@ -32,6 +32,9 @@ main () #ifdef AVXNECONVERT && __builtin_cpu_supports ("avxneconvert") #endif +#ifdef AVXVNNIINT16 + && __builtin_cpu_supports ("avxvnniint16") +#endif ) { do_test (); diff --git a/gcc/testsuite/gcc.target/i386/avxvnniint16-1.c b/gcc/testsuite/gcc.target/i386/avxvnniint16-1.c new file mode 100644 index 0000000..6ae57b1 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avxvnniint16-1.c @@ -0,0 +1,43 @@ +/* { dg-do compile } */ +/* { dg-options "-mavxvnniint16 -O2" } */ +/* { dg-final { scan-assembler-times "vpdpwusd\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwusd\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwusds\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwusds\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwsud\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwsud\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwsuds\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwsuds\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwuud\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwuud\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwuuds\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+\[^\n\r]*%ymm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ +/* { dg-final { scan-assembler-times "vpdpwuuds\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */ + + +#include <immintrin.h> + +volatile __m256i x,y,z; +volatile __m128i x_,y_,z_; +volatile __mmask8 m; + +void extern +avxvnniint16_test (void) +{ + x = _mm256_dpwusd_avx_epi32 (x, y, z); + x_ = _mm_dpwusd_avx_epi32 (x_, y_, z_); + + x = _mm256_dpwusds_avx_epi32 (x, y, z); + x_ = _mm_dpwusds_avx_epi32 (x_, y_, z_); + + x = _mm256_dpwsud_avx_epi32 (x, y, z); + x_ = _mm_dpwsud_avx_epi32 (x_, y_, z_); + + x = _mm256_dpwsuds_avx_epi32 (x, y, z); + x_ = _mm_dpwsuds_avx_epi32 (x_, y_, z_); + + x = _mm256_dpwuud_avx_epi32 (x, y, z); + x_ = _mm_dpwuud_avx_epi32 (x_, y_, z_); + + x = _mm256_dpwuuds_avx_epi32 (x, y, z); + x_ = _mm_dpwuuds_avx_epi32 (x_, y_, z_); +} diff --git a/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwsud-2.c b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwsud-2.c new file mode 100644 index 0000000..bc57a8a --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwsud-2.c @@ -0,0 +1,71 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -mavxvnniint16" } */ +/* { dg-require-effective-target avxvnniint16 } */ +#define AVXVNNIINT16 +#ifndef CHECK +#define CHECK "avx-check.h" +#endif + +#ifndef TEST +#define TEST avx_test +#endif + +#include CHECK + +static void +CALC (int *r, int *dst, short *s1, unsigned short *s2, int size) +{ + int tempres[16]; + for (int i = 0; i < size; i++) + tempres[i] = (int) s1[i] * (unsigned int) s2[i]; + for (int i = 0; i < size / 2; i++) + { + int test = (int) dst[i] + tempres[i * 2] + tempres[i * 2 + 1]; + r[i] = test; + } +} + +void +TEST (void) +{ + int i; + union256i_d res_256; + union256i_w src1_256; + union256i_uw src2_256; + int res_ref_256[8]; + + for (i = 0; i < 16; i++) + { + int sign = i % 2 ? 1 : -1; + src1_256.a[i] = 10 + 3 * i + sign; + src2_256.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 8; i++) + res_256.a[i] = 0x7fffffff; + + CALC (res_ref_256, res_256.a, src1_256.a, src2_256.a, 16); + res_256.x = _mm256_dpwsud_avx_epi32 (res_256.x, src1_256.x, src2_256.x); + if (check_union256i_d (res_256, res_ref_256)) + abort (); + + union128i_d res_128; + union128i_w src1_128; + union128i_uw src2_128; + int res_ref_128[4]; + + for (i = 0; i < 8; i++) + { + int sign = i % 2 ? 1 : -1; + src1_128.a[i] = 10 + 3 * i * i + sign; + src2_128.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 4; i++) + res_128.a[i] = 0x7fffffff; + + CALC (res_ref_128, res_128.a, src1_128.a, src2_128.a, 8); + res_128.x = _mm_dpwsud_avx_epi32 (res_128.x, src1_128.x, src2_128.x); + if (check_union128i_d (res_128, res_ref_128)) + abort (); +} diff --git a/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwsuds-2.c b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwsuds-2.c new file mode 100644 index 0000000..fbcf46a --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwsuds-2.c @@ -0,0 +1,72 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -mavxvnniint16" } */ +/* { dg-require-effective-target avxvnniint16 } */ +#define AVXVNNIINT16 +#ifndef CHECK +#define CHECK "avx-check.h" +#endif + +#ifndef TEST +#define TEST avx_test +#endif + +#include CHECK + +static void +CALC (int *r, int *dst, short *s1, unsigned short *s2, int size) +{ + int tempres[16]; + for (int i = 0; i < size; i++) + tempres[i] = (int) s1[i] * (unsigned int) s2[i]; + for (int i = 0; i < size / 2; i++) + { + long long test = (long long) dst[i] + tempres[i * 2] + tempres[i * 2 + 1]; + r[i] = test > 0x7FFFFFFF ? 0x7FFFFFFF : + (test < 0xffffffff80000000LL ? 0x80000000 : test); + } +} + +void +TEST (void) +{ + int i; + union256i_d res_256; + union256i_w src1_256; + union256i_uw src2_256; + int res_ref_256[8]; + + for (i = 0; i < 16; i++) + { + int sign = i % 2 ? 1 : -1; + src1_256.a[i] = 10 + 3 * i + sign; + src2_256.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 8; i++) + res_256.a[i] = 0x7fffffff; + + CALC (res_ref_256, res_256.a, src1_256.a, src2_256.a, 16); + res_256.x = _mm256_dpwsuds_avx_epi32 (res_256.x, src1_256.x, src2_256.x); + if (check_union256i_d (res_256, res_ref_256)) + abort (); + + union128i_d res_128; + union128i_w src1_128; + union128i_uw src2_128; + int res_ref_128[4]; + + for (i = 0; i < 8; i++) + { + int sign = i % 2 ? 1 : -1; + src1_128.a[i] = 10 + 3 * i * i + sign; + src2_128.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 4; i++) + res_128.a[i] = 0x7fffffff; + + CALC (res_ref_128, res_128.a, src1_128.a, src2_128.a, 8); + res_128.x = _mm_dpwsuds_avx_epi32 (res_128.x, src1_128.x, src2_128.x); + if (check_union128i_d (res_128, res_ref_128)) + abort (); +} diff --git a/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwusd-2.c b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwusd-2.c new file mode 100644 index 0000000..54cf271 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwusd-2.c @@ -0,0 +1,71 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -mavxvnniint16" } */ +/* { dg-require-effective-target avxvnniint16 } */ +#define AVXVNNIINT16 +#ifndef CHECK +#define CHECK "avx-check.h" +#endif + +#ifndef TEST +#define TEST avx_test +#endif + +#include CHECK + +static void +CALC (int *r, int *dst, unsigned short *s1, short *s2, int size) +{ + int tempres[16]; + for (int i = 0; i < size; i++) + tempres[i] = (unsigned int) s1[i] * (int) s2[i]; + for (int i = 0; i < size / 2; i++) + { + int test = (int) dst[i] + tempres[i * 2] + tempres[i * 2 + 1]; + r[i] = test; + } +} + +void +TEST (void) +{ + int i; + union256i_d res_256; + union256i_uw src1_256; + union256i_w src2_256; + int res_ref_256[8]; + + for (i = 0; i < 16; i++) + { + int sign = i % 2 ? 1 : -1; + src1_256.a[i] = 10 + 3 * i + sign; + src2_256.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 8; i++) + res_256.a[i] = 0x7fffffff; + + CALC (res_ref_256, res_256.a, src1_256.a, src2_256.a, 16); + res_256.x = _mm256_dpwusd_avx_epi32 (res_256.x, src1_256.x, src2_256.x); + if (check_union256i_d (res_256, res_ref_256)) + abort (); + + union128i_d res_128; + union128i_uw src1_128; + union128i_w src2_128; + int res_ref_128[4]; + + for (i = 0; i < 8; i++) + { + int sign = i % 2 ? 1 : -1; + src1_128.a[i] = 10 + 3 * i * i + sign; + src2_128.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 4; i++) + res_128.a[i] = 0x7fffffff; + + CALC (res_ref_128, res_128.a, src1_128.a, src2_128.a, 8); + res_128.x = _mm_dpwusd_avx_epi32 (res_128.x, src1_128.x, src2_128.x); + if (check_union128i_d (res_128, res_ref_128)) + abort (); +} diff --git a/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwusds-2.c b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwusds-2.c new file mode 100644 index 0000000..ed9594c --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwusds-2.c @@ -0,0 +1,72 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -mavxvnniint16" } */ +/* { dg-require-effective-target avxvnniint16 } */ +#define AVXVNNIINT16 +#ifndef CHECK +#define CHECK "avx-check.h" +#endif + +#ifndef TEST +#define TEST avx_test +#endif + +#include CHECK + +static void +CALC (int *r, int *dst, unsigned short *s1, short *s2, int size) +{ + int tempres[16]; + for (int i = 0; i < size; i++) + tempres[i] = (unsigned int) s1[i] * (int) s2[i]; + for (int i = 0; i < size / 2; i++) + { + long long test = (long long) dst[i] + tempres[i * 2] + tempres[i * 2 + 1]; + r[i] = test > 0x7FFFFFFF ? 0x7FFFFFFF : + (test < 0xffffffff80000000LL ? 0x80000000 : test); + } +} + +void +TEST (void) +{ + int i; + union256i_d res_256; + union256i_uw src1_256; + union256i_w src2_256; + int res_ref_256[8]; + + for (i = 0; i < 16; i++) + { + int sign = i % 2 ? 1 : -1; + src1_256.a[i] = 10 + 3 * i + sign; + src2_256.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 8; i++) + res_256.a[i] = 0x7fffffff; + + CALC (res_ref_256, res_256.a, src1_256.a, src2_256.a, 16); + res_256.x = _mm256_dpwusds_avx_epi32 (res_256.x, src1_256.x, src2_256.x); + if (check_union256i_d (res_256, res_ref_256)) + abort (); + + union128i_d res_128; + union128i_uw src1_128; + union128i_w src2_128; + int res_ref_128[4]; + + for (i = 0; i < 8; i++) + { + int sign = i % 2 ? 1 : -1; + src1_128.a[i] = 10 + 3 * i * i + sign; + src2_128.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 4; i++) + res_128.a[i] = 0x7fffffff; + + CALC (res_ref_128, res_128.a, src1_128.a, src2_128.a, 8); + res_128.x = _mm_dpwusds_avx_epi32 (res_128.x, src1_128.x, src2_128.x); + if (check_union128i_d (res_128, res_ref_128)) + abort(); +} diff --git a/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwuud-2.c b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwuud-2.c new file mode 100644 index 0000000..8bdc433 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwuud-2.c @@ -0,0 +1,71 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -mavxvnniint16" } */ +/* { dg-require-effective-target avxvnniint16 } */ +#define AVXVNNIINT16 +#ifndef CHECK +#define CHECK "avx-check.h" +#endif + +#ifndef TEST +#define TEST avx_test +#endif + +#include CHECK + +static void +CALC (unsigned int *r, unsigned int *dst, unsigned short *s1, unsigned short *s2, int size) +{ + unsigned int tempres[16]; + for (int i = 0; i < size; i++) + tempres[i] = (unsigned int) s1[i] * (unsigned int) s2[i]; + for (int i = 0; i < size / 2; i++) + { + unsigned int test = (unsigned) dst[i] + tempres[i * 2] + tempres[i * 2 + 1]; + r[i] = test; + } +} + +void +TEST (void) +{ + int i; + union256i_ud res_256; + union256i_uw src1_256; + union256i_uw src2_256; + unsigned int res_ref_256[8]; + + for (i = 0; i < 16; i++) + { + int sign = i % 2 ? 1 : -1; + src1_256.a[i] = 10 + 3 * i + sign; + src2_256.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 8; i++) + res_256.a[i] = 0x7fffffff; + + CALC (res_ref_256, res_256.a, src1_256.a, src2_256.a, 16); + res_256.x = _mm256_dpwuud_avx_epi32 (res_256.x, src1_256.x, src2_256.x); + if (check_union256i_ud (res_256, res_ref_256)) + abort (); + + union128i_ud res_128; + union128i_uw src2_128; + union128i_uw src1_128; + unsigned int res_ref_128[4]; + + for (i = 0; i < 8; i++) + { + int sign = i % 2 ? 1 : -1; + src1_128.a[i] = 10 + 3 * i * i + sign; + src2_128.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 4; i++) + res_128.a[i] = 0x7fffffff; + + CALC (res_ref_128, res_128.a, src1_128.a, src2_128.a, 8); + res_128.x = _mm_dpwuud_avx_epi32 (res_128.x, src1_128.x, src2_128.x); + if (check_union128i_ud (res_128, res_ref_128)) + abort (); +} diff --git a/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwuuds-2.c b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwuuds-2.c new file mode 100644 index 0000000..3220412 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/avxvnniint16-vpdpwuuds-2.c @@ -0,0 +1,71 @@ +/* { dg-do run } */ +/* { dg-options "-O2 -mavxvnniint16" } */ +/* { dg-require-effective-target avxvnniint16 } */ +#define AVXVNNIINT16 +#ifndef CHECK +#define CHECK "avx-check.h" +#endif + +#ifndef TEST +#define TEST avx_test +#endif + +#include CHECK + +static void +CALC (unsigned int *r, unsigned int *dst, unsigned short *s1, unsigned short *s2, int size) +{ + unsigned int tempres[16]; + for (int i = 0; i < size; i++) + tempres[i] = (unsigned int) s1[i] * (unsigned int) s2[i]; + for (int i = 0; i < size / 2; i++) + { + unsigned int test = (unsigned) dst[i] + tempres[i * 2] + tempres[i * 2 + 1]; + r[i] = test > 0xFFFFFFFF ? 0xFFFFFFFF : test; + } +} + +void +TEST (void) +{ + int i; + union256i_ud res_256; + union256i_uw src1_256; + union256i_uw src2_256; + unsigned int res_ref_256[8]; + + for (i = 0; i < 16; i++) + { + int sign = i % 2 ? 1 : -1; + src1_256.a[i] = 10 + 3 * i + sign; + src2_256.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 8; i++) + res_256.a[i] = 0x7fffffff; + + CALC (res_ref_256, res_256.a, src1_256.a, src2_256.a, 16); + res_256.x = _mm256_dpwuuds_avx_epi32 (res_256.x, src1_256.x, src2_256.x); + if (check_union256i_ud (res_256, res_ref_256)) + abort (); + + union128i_ud res_128; + union128i_uw src2_128; + union128i_uw src1_128; + unsigned int res_ref_128[4]; + + for (i = 0; i < 8; i++) + { + int sign = i % 2 ? 1 : -1; + src1_128.a[i] = 10 + 3 * i * i + sign; + src2_128.a[i] = sign * 10 * i * i; + } + + for (i = 0; i < 4; i++) + res_128.a[i] = 0x7fffffff; + + CALC (res_ref_128, res_128.a, src1_128.a, src2_128.a, 8); + res_128.x = _mm_dpwuuds_avx_epi32 (res_128.x, src1_128.x, src2_128.x); + if (check_union128i_ud (res_128, res_ref_128)) + abort (); +} diff --git a/gcc/testsuite/gcc.target/i386/funcspec-56.inc b/gcc/testsuite/gcc.target/i386/funcspec-56.inc index f466962..bba0fa3 100644 --- a/gcc/testsuite/gcc.target/i386/funcspec-56.inc +++ b/gcc/testsuite/gcc.target/i386/funcspec-56.inc @@ -88,6 +88,7 @@ extern void test_amx_fp16 (void) __attribute__((__target__("amx-fp16"))); extern void test_prefetchi (void) __attribute__((__target__("prefetchi"))); extern void test_raoint (void) __attribute__((__target__("raoint"))); extern void test_amx_complex (void) __attribute__((__target__("amx-complex"))); +extern void test_avxvnniint16 (void) __attribute__((__target__("avxvnniint16"))); extern void test_no_sgx (void) __attribute__((__target__("no-sgx"))); extern void test_no_avx5124fmaps(void) __attribute__((__target__("no-avx5124fmaps"))); @@ -177,6 +178,7 @@ extern void test_no_amx_fp16 (void) __attribute__((__target__("no-amx-fp16"))); extern void test_no_prefetchi (void) __attribute__((__target__("no-prefetchi"))); extern void test_no_raoint (void) __attribute__((__target__("no-raoint"))); extern void test_no_amx_complex (void) __attribute__((__target__("no-amx-complex"))); +extern void test_no_avxvnniint16 (void) __attribute__((__target__("no-avxvnniint16"))); extern void test_arch_nocona (void) __attribute__((__target__("arch=nocona"))); extern void test_arch_core2 (void) __attribute__((__target__("arch=core2"))); diff --git a/gcc/testsuite/gcc.target/i386/sse-12.c b/gcc/testsuite/gcc.target/i386/sse-12.c index ae4ffd1..2b7d78c 100644 --- a/gcc/testsuite/gcc.target/i386/sse-12.c +++ b/gcc/testsuite/gcc.target/i386/sse-12.c @@ -3,7 +3,7 @@ popcntintrin.h gfniintrin.h and mm_malloc.h are usable with -O -std=c89 -pedantic-errors. */ /* { dg-do compile } */ -/* { dg-options "-O -std=c89 -pedantic-errors -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512bw -mavx512dq -mavx512vl -mavx512vbmi -mavx512vbmi2 -mavx512ifma -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mavx512bitalg -mpconfig -mwbnoinvd -mavx512bf16 -menqcmd -mavx512vp2intersect -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavxifma -mavxvnniint8 -mavxneconvert -mamx-fp16 -mraoint -mamx-complex" } */ +/* { dg-options "-O -std=c89 -pedantic-errors -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512bw -mavx512dq -mavx512vl -mavx512vbmi -mavx512vbmi2 -mavx512ifma -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mavx512bitalg -mpconfig -mwbnoinvd -mavx512bf16 -menqcmd -mavx512vp2intersect -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavxifma -mavxvnniint8 -mavxneconvert -mamx-fp16 -mraoint -mamx-complex -mavxvnniint16" } */ #include <x86intrin.h> diff --git a/gcc/testsuite/gcc.target/i386/sse-13.c b/gcc/testsuite/gcc.target/i386/sse-13.c index f046a68..3369348 100644 --- a/gcc/testsuite/gcc.target/i386/sse-13.c +++ b/gcc/testsuite/gcc.target/i386/sse-13.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-O2 -Werror-implicit-function-declaration -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512vl -mavx512dq -mavx512bw -mavx512vbmi -mavx512vbmi2 -mavx512ifma -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mavx512vp2intersect -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mavx512bitalg -mpconfig -mwbnoinvd -mavx512bf16 -menqcmd -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavx512fp16 -mavxifma -mavxvnniint8 -mavxneconvert -mcmpccxadd -mamx-fp16 -mprefetchi -mraoint -mamx-complex" } */ +/* { dg-options "-O2 -Werror-implicit-function-declaration -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512vl -mavx512dq -mavx512bw -mavx512vbmi -mavx512vbmi2 -mavx512ifma -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mavx512vp2intersect -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mavx512bitalg -mpconfig -mwbnoinvd -mavx512bf16 -menqcmd -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavx512fp16 -mavxifma -mavxvnniint8 -mavxneconvert -mcmpccxadd -mamx-fp16 -mprefetchi -mraoint -mamx-complex -mavxvnniint16" } */ /* { dg-add-options bind_pic_locally } */ #include <mm_malloc.h> diff --git a/gcc/testsuite/gcc.target/i386/sse-14.c b/gcc/testsuite/gcc.target/i386/sse-14.c index 05322f7..51c2946 100644 --- a/gcc/testsuite/gcc.target/i386/sse-14.c +++ b/gcc/testsuite/gcc.target/i386/sse-14.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-O0 -Werror-implicit-function-declaration -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512dq -mavx512bw -mavx512vl -mavx512ifma -mavx512vbmi -mavx512vbmi2 -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mpconfig -mwbnoinvd -mavx512vl -mavx512bf16 -menqcmd -mavx512vp2intersect -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavx512fp16 -mavxifma -mavxvnniint8 -mavxneconvert -mamx-fp16 -mraoint -mamx-complex" } */ +/* { dg-options "-O0 -Werror-implicit-function-declaration -march=k8 -msse4a -m3dnow -mavx -mavx2 -mfma4 -mxop -maes -mpclmul -mpopcnt -mabm -mlzcnt -mbmi -mbmi2 -mtbm -mlwp -mfsgsbase -mrdrnd -mf16c -mfma -mrtm -mrdseed -mprfchw -madx -mfxsr -mxsaveopt -mavx512f -mavx512er -mavx512cd -mavx512pf -msha -mprefetchwt1 -mxsavec -mxsaves -mclflushopt -mavx512dq -mavx512bw -mavx512vl -mavx512ifma -mavx512vbmi -mavx512vbmi2 -mavx5124fmaps -mavx5124vnniw -mavx512vpopcntdq -mclwb -mmwaitx -mclzero -mpku -msgx -mrdpid -mgfni -mpconfig -mwbnoinvd -mavx512vl -mavx512bf16 -menqcmd -mavx512vp2intersect -mserialize -mtsxldtrk -mamx-tile -mamx-int8 -mamx-bf16 -mkl -mwidekl -mavxvnni -mavx512fp16 -mavxifma -mavxvnniint8 -mavxneconvert -mamx-fp16 -mraoint -mamx-complex -mavxvnniint16" } */ /* { dg-add-options bind_pic_locally } */ #include <mm_malloc.h> diff --git a/gcc/testsuite/gcc.target/i386/sse-22.c b/gcc/testsuite/gcc.target/i386/sse-22.c index 53c38b7..4982fde 100644 --- a/gcc/testsuite/gcc.target/i386/sse-22.c +++ b/gcc/testsuite/gcc.target/i386/sse-22.c @@ -103,7 +103,7 @@ #ifndef DIFFERENT_PRAGMAS -#pragma GCC target ("sse4a,3dnow,avx,avx2,fma4,xop,aes,pclmul,popcnt,abm,lzcnt,bmi,bmi2,tbm,lwp,fsgsbase,rdrnd,f16c,rtm,rdseed,prfchw,adx,fxsr,xsaveopt,avx512f,avx512er,avx512cd,avx512pf,sha,prefetchwt1,avx512vl,avx512bw,avx512dq,avx512vbmi,avx512vbmi2,avx512ifma,avx5124fmaps,avx5124vnniw,avx512vpopcntdq,gfni,avx512bitalg,avx512bf16,avx512vp2intersect,serialize,tsxldtrk,amx-tile,amx-int8,amx-bf16,kl,widekl,avxvnni,avx512fp16,avxifma,avxvnniint8,avxneconvert,amx-fp16,raoint,amx-complex") +#pragma GCC target ("sse4a,3dnow,avx,avx2,fma4,xop,aes,pclmul,popcnt,abm,lzcnt,bmi,bmi2,tbm,lwp,fsgsbase,rdrnd,f16c,rtm,rdseed,prfchw,adx,fxsr,xsaveopt,avx512f,avx512er,avx512cd,avx512pf,sha,prefetchwt1,avx512vl,avx512bw,avx512dq,avx512vbmi,avx512vbmi2,avx512ifma,avx5124fmaps,avx5124vnniw,avx512vpopcntdq,gfni,avx512bitalg,avx512bf16,avx512vp2intersect,serialize,tsxldtrk,amx-tile,amx-int8,amx-bf16,kl,widekl,avxvnni,avx512fp16,avxifma,avxvnniint8,avxneconvert,amx-fp16,raoint,amx-complex,avxvnniint16") #endif /* Following intrinsics require immediate arguments. They @@ -220,7 +220,7 @@ test_4 (_mm_cmpestrz, int, __m128i, int, __m128i, int, 1) /* immintrin.h (AVX/AVX2/RDRND/FSGSBASE/F16C/RTM/AVX512F/SHA) */ #ifdef DIFFERENT_PRAGMAS -#pragma GCC target ("avx,avx2,rdrnd,fsgsbase,f16c,rtm,avx512f,avx512er,avx512cd,avx512pf,sha,avx512vl,avx512bw,avx512dq,avx512ifma,avx512vbmi,avx512vbmi2,avx5124fmaps,avx5124vnniw,avx512vpopcntdq,gfni,avx512bitalg,avx512bf16,avx512vp2intersect,serialize,tsxldtrk,amx-tile,amx-int8,amx-bf16,kl,widekl,avxvnni,avx512fp16,avxifma,avxvnniint8,avxneconvert,amx-fp16,raoint,amx-complex") +#pragma GCC target ("avx,avx2,rdrnd,fsgsbase,f16c,rtm,avx512f,avx512er,avx512cd,avx512pf,sha,avx512vl,avx512bw,avx512dq,avx512ifma,avx512vbmi,avx512vbmi2,avx5124fmaps,avx5124vnniw,avx512vpopcntdq,gfni,avx512bitalg,avx512bf16,avx512vp2intersect,serialize,tsxldtrk,amx-tile,amx-int8,amx-bf16,kl,widekl,avxvnni,avx512fp16,avxifma,avxvnniint8,avxneconvert,amx-fp16,raoint,amx-complex,avxvnniint16") #endif #include <immintrin.h> test_1 (_cvtss_sh, unsigned short, float, 1) diff --git a/gcc/testsuite/gcc.target/i386/sse-23.c b/gcc/testsuite/gcc.target/i386/sse-23.c index 50bf85a..7e9c9f2 100644 --- a/gcc/testsuite/gcc.target/i386/sse-23.c +++ b/gcc/testsuite/gcc.target/i386/sse-23.c @@ -847,6 +847,6 @@ #define __builtin_ia32_cmpccxadd(A, B, C, D) __builtin_ia32_cmpccxadd(A, B, C, 1) #define __builtin_ia32_cmpccxadd64(A, B, C, D) __builtin_ia32_cmpccxadd64(A, B, C, 1) -#pragma GCC target ("sse4a,3dnow,avx,avx2,fma4,xop,aes,pclmul,popcnt,abm,lzcnt,bmi,bmi2,tbm,lwp,fsgsbase,rdrnd,f16c,fma,rtm,rdseed,prfchw,adx,fxsr,xsaveopt,avx512f,avx512er,avx512cd,avx512pf,sha,prefetchwt1,xsavec,xsaves,clflushopt,avx512bw,avx512dq,avx512vl,avx512vbmi,avx512ifma,avx5124fmaps,avx5124vnniw,avx512vpopcntdq,clwb,mwaitx,clzero,pku,sgx,rdpid,gfni,avx512vbmi2,vpclmulqdq,avx512bitalg,pconfig,wbnoinvd,avx512bf16,enqcmd,avx512vp2intersect,serialize,tsxldtrk,amx-tile,amx-int8,amx-bf16,kl,widekl,avxvnni,avx512fp16,avxifma,avxvnniint8,avxneconvert,cmpccxadd,amx-fp16,prefetchi,raoint,amx-complex") +#pragma GCC target ("sse4a,3dnow,avx,avx2,fma4,xop,aes,pclmul,popcnt,abm,lzcnt,bmi,bmi2,tbm,lwp,fsgsbase,rdrnd,f16c,fma,rtm,rdseed,prfchw,adx,fxsr,xsaveopt,avx512f,avx512er,avx512cd,avx512pf,sha,prefetchwt1,xsavec,xsaves,clflushopt,avx512bw,avx512dq,avx512vl,avx512vbmi,avx512ifma,avx5124fmaps,avx5124vnniw,avx512vpopcntdq,clwb,mwaitx,clzero,pku,sgx,rdpid,gfni,avx512vbmi2,vpclmulqdq,avx512bitalg,pconfig,wbnoinvd,avx512bf16,enqcmd,avx512vp2intersect,serialize,tsxldtrk,amx-tile,amx-int8,amx-bf16,kl,widekl,avxvnni,avx512fp16,avxifma,avxvnniint8,avxneconvert,cmpccxadd,amx-fp16,prefetchi,raoint,amx-complex,avxvnniint16") #include <x86intrin.h> diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 33482b2..60de239 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -9856,6 +9856,18 @@ proc check_effective_target_amx_complex { } { } "-mamx-complex" ] } +# Return 1 if avxvnniint16 instructions can be compiled. +proc check_effective_target_avxvnniint16 { } { + return [check_no_compiler_messages avxvnniint16 object { + typedef int __v8si __attribute__ ((__vector_size__ (32))); + __v8si + _mm256_dpwsud_avx_epi32 (__v8si __A, __v8si __B, __v8si __C) + { + return __builtin_ia32_vpdpwsud256 (__A, __B, __C); + } + } "-O0 -mavxvnniint16" ] +} + # Return 1 if sse instructions can be compiled. proc check_effective_target_sse { } { return [check_no_compiler_messages sse object { |