@@ -879,6 +879,8 @@ get_available_features (struct __processor_model *cpu_model,
{
if (eax & bit_AMX_FP16)
set_feature (FEATURE_AMX_FP16);
+ if (edx & bit_AMX_COMPLEX)
+ set_feature (FEATURE_AMX_COMPLEX);
}
}
@@ -117,6 +117,8 @@ along with GCC; see the file COPYING3. If not see
(OPTION_MASK_ISA2_AMX_TILE | OPTION_MASK_ISA2_AMX_FP16)
#define OPTION_MASK_ISA2_PREFETCHI_SET OPTION_MASK_ISA2_PREFETCHI
#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)
/* SSE4 includes both SSE4.1 and SSE4.2. -msse4 should be the same
as -msse4.2. */
@@ -282,7 +284,8 @@ along with GCC; see the file COPYING3. If not see
#define OPTION_MASK_ISA2_TSXLDTRK_UNSET OPTION_MASK_ISA2_TSXLDTRK
#define OPTION_MASK_ISA2_AMX_TILE_UNSET \
(OPTION_MASK_ISA2_AMX_TILE | OPTION_MASK_ISA2_AMX_INT8_UNSET \
- | OPTION_MASK_ISA2_AMX_BF16_UNSET | OPTION_MASK_ISA2_AMX_FP16_UNSET)
+ | OPTION_MASK_ISA2_AMX_BF16_UNSET | OPTION_MASK_ISA2_AMX_FP16_UNSET \
+ | OPTION_MASK_ISA2_AMX_COMPLEX_UNSET)
#define OPTION_MASK_ISA2_AMX_INT8_UNSET OPTION_MASK_ISA2_AMX_INT8
#define OPTION_MASK_ISA2_AMX_BF16_UNSET OPTION_MASK_ISA2_AMX_BF16
#define OPTION_MASK_ISA2_UINTR_UNSET OPTION_MASK_ISA2_UINTR
@@ -296,6 +299,7 @@ along with GCC; see the file COPYING3. If not see
#define OPTION_MASK_ISA2_AMX_FP16_UNSET OPTION_MASK_ISA2_AMX_FP16
#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
/* SSE4 includes both SSE4.1 and SSE4.2. -mno-sse4 should the same
as -mno-sse4.1. */
@@ -1246,6 +1250,19 @@ ix86_handle_option (struct gcc_options *opts,
}
return true;
+ case OPT_mamx_complex:
+ if (value)
+ {
+ opts->x_ix86_isa_flags2 |= OPTION_MASK_ISA2_AMX_COMPLEX_SET;
+ opts->x_ix86_isa_flags2_explicit |= OPTION_MASK_ISA2_AMX_COMPLEX_SET;
+ }
+ else
+ {
+ opts->x_ix86_isa_flags2 &= ~OPTION_MASK_ISA2_AMX_COMPLEX_UNSET;
+ opts->x_ix86_isa_flags2_explicit |= OPTION_MASK_ISA2_AMX_COMPLEX_UNSET;
+ }
+ return true;
+
case OPT_mfma:
if (value)
{
@@ -253,6 +253,7 @@ enum processor_features
FEATURE_AMX_FP16,
FEATURE_PREFETCHI,
FEATURE_RAOINT,
+ FEATURE_AMX_COMPLEX,
CPU_FEATURE_MAX
};
@@ -184,4 +184,6 @@ ISA_NAMES_TABLE_START
ISA_NAMES_TABLE_ENTRY("amx-fp16", FEATURE_AMX_FP16, P_NONE, "-mamx-fp16")
ISA_NAMES_TABLE_ENTRY("prefetchi", FEATURE_PREFETCHI, P_NONE, "-mprefetchi")
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_END
@@ -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"
+ raointintrin.h amxcomplexintrin.h"
;;
ia64-*-*)
extra_headers=ia64intrin.h
new file mode 100644
@@ -0,0 +1,59 @@
+/* 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 <amxcomplexintrin.h> directly; include <immintrin.h> instead."
+#endif
+
+#ifndef _AMXCOMPLEXINTRIN_H_INCLUDED
+#define _AMXCOMPLEXINTRIN_H_INCLUDED
+
+#if !defined(__AMX_COMPLEX__)
+#pragma GCC push_options
+#pragma GCC target("amx-complex")
+#define __DISABLE_AMX_COMPLEX__
+#endif /* __AMX_COMPLEX__ */
+
+#if defined(__x86_64__)
+#define _tile_cmmimfp16ps_internal(src1_dst,src2,src3) \
+ __asm__ volatile\
+ ("{tcmmimfp16ps\t%%tmm"#src3", %%tmm"#src2", %%tmm"#src1_dst"|tcmmimfp16ps\t%%tmm"#src1_dst", %%tmm"#src2", %%tmm"#src3"}" ::)
+
+#define _tile_cmmrlfp16ps_internal(src1_dst,src2,src3) \
+ __asm__ volatile\
+ ("{tcmmrlfp16ps\t%%tmm"#src3", %%tmm"#src2", %%tmm"#src1_dst"|tcmmrlfp16ps\t%%tmm"#src1_dst", %%tmm"#src2", %%tmm"#src3"}" ::)
+
+#define _tile_cmmimfp16ps(src1_dst,src2,src3) \
+ _tile_cmmimfp16ps_internal (src1_dst, src2, src3)
+
+#define _tile_cmmrlfp16ps(src1_dst,src2,src3) \
+ _tile_cmmrlfp16ps_internal (src1_dst, src2, src3)
+
+#endif
+
+#ifdef __DISABLE_AMX_COMPLEX__
+#undef __DISABLE_AMX_COMPLEX__
+#pragma GCC pop_options
+#endif /* __DISABLE_AMX_COMPLEX__ */
+
+#endif /* _AMXCOMPLEXINTRIN_H_INCLUDED */
@@ -141,6 +141,7 @@
#define bit_AMX_BF16 (1 << 22)
#define bit_AMX_TILE (1 << 24)
#define bit_AMX_INT8 (1 << 25)
+#define bit_AMX_COMPLEX (1 << 8)
/* Extended State Enumeration Sub-leaf (%eax == 0xd, %ecx == 1) */
#define bit_XSAVEOPT (1 << 0)
@@ -675,6 +675,8 @@ ix86_target_macros_internal (HOST_WIDE_INT isa_flag,
def_or_undef (parse_in, "__PREFETCHI__");
if (isa_flag2 & OPTION_MASK_ISA2_RAOINT)
def_or_undef (parse_in, "__RAOINT__");
+ if (isa_flag2 & OPTION_MASK_ISA2_AMX_COMPLEX)
+ def_or_undef (parse_in, "__AMX_COMPLEX__");
if (TARGET_IAMCU)
{
def_or_undef (parse_in, "__iamcu");
@@ -116,3 +116,4 @@ DEF_PTA(CMPCCXADD)
DEF_PTA(AMX_FP16)
DEF_PTA(PREFETCHI)
DEF_PTA(RAOINT)
+DEF_PTA(AMX_COMPLEX)
@@ -238,7 +238,8 @@ static struct ix86_target_opts isa2_opts[] =
{ "-mcmpccxadd", OPTION_MASK_ISA2_CMPCCXADD },
{ "-mamx-fp16", OPTION_MASK_ISA2_AMX_FP16 },
{ "-mprefetchi", OPTION_MASK_ISA2_PREFETCHI },
- { "-mraoint", OPTION_MASK_ISA2_RAOINT }
+ { "-mraoint", OPTION_MASK_ISA2_RAOINT },
+ { "-mamx-complex", OPTION_MASK_ISA2_AMX_COMPLEX }
};
static struct ix86_target_opts isa_opts[] =
{
@@ -1089,6 +1090,7 @@ ix86_valid_target_attribute_inner_p (tree fndecl, tree args, char *p_strings[],
IX86_ATTR_ISA ("amx-fp16", OPT_mamx_fp16),
IX86_ATTR_ISA ("prefetchi", OPT_mprefetchi),
IX86_ATTR_ISA ("raoint", OPT_mraoint),
+ IX86_ATTR_ISA ("amx-complex", OPT_mamx_complex),
/* enum options */
IX86_ATTR_ENUM ("fpmath=", OPT_mfpmath_),
@@ -1274,3 +1274,7 @@ Enum(lam_type) String(u48) Value(lam_u48)
EnumValue
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.
@@ -132,6 +132,8 @@
#include <amxbf16intrin.h>
+#include <amxcomplexintrin.h>
+
#include <prfchwintrin.h>
#include <keylockerintrin.h>
@@ -7069,6 +7069,11 @@ Enable/disable the generation of the PREFETCHI instructions.
@itemx no-raoint
Enable/disable the generation of the RAOINT instructions.
+@cindex @code{target("amx-complex")} function attribute, x86
+@item amx-complex
+@itemx no-amx-complex
+Enable/disable the generation of the AMX-COMPLEX instructions.
+
@cindex @code{target("cld")} function attribute, x86
@item cld
@itemx no-cld
@@ -1429,7 +1429,7 @@ 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
+-mprefetchi -mraoint -mamx-complex
-mcldemote -mms-bitfields -mno-align-stringops -minline-all-stringops
-minline-stringops-dynamically -mstringop-strategy=@var{alg}
-mkl -mwidekl
@@ -33333,6 +33333,9 @@ preferred alignment to @option{-mpreferred-stack-boundary=2}.
@need 200
@opindex mraoint
@itemx -mraoint
+@need 200
+@opindex mamx-complex
+@itemx -mamx-complex
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,
@@ -33343,9 +33346,9 @@ XSAVEOPT, XSAVEC, XSAVES, RTM, HLE, TBM, MWAITX, CLZERO, PKU, AVX512VBMI2,
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 or
-CLDEMOTE extended instruction sets. Each has a corresponding @option{-mno-}
-option to disable use of these instructions.
+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.
These extensions are also available as built-in functions: see
@ref{x86 Built-in Functions}, for details of the functions enabled and
@@ -2514,6 +2514,9 @@ Target supports the execution of @code{amx-int8} instructions.
@item amx_bf16
Target supports the execution of @code{amx-bf16} instructions.
+@item amx_complex
+Target supports the execution of @code{amx-complex} instructions.
+
@item amx_fp16
Target supports the execution of @code{amx-fp16} instructions.
@@ -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" } */
+/* { 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" } */
/* 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,
@@ -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" } */
+/* { 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" } */
/* 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,
@@ -216,6 +216,9 @@ main ()
#ifdef AMX_FP16
&& __builtin_cpu_supports ("amx-fp16")
#endif
+#ifdef AMX_COMPLEX
+ && __builtin_cpu_supports ("amx-complex")
+#endif
#ifdef __linux__
&& request_perm_xtile_data ()
#endif
@@ -1,6 +1,6 @@
#ifndef AMX_HELPER_H_INCLUDED
#define AMX_HELPER_H_INCLUDED
-#if defined(AMX_FP16)
+#if defined(AMX_FP16) || defined(AMX_COMPLEX)
#include <immintrin.h>
#include <xmmintrin.h>
#endif
@@ -12,7 +12,7 @@ typedef union
uint16_t u;
} union16f_uw;
-#if defined(AMX_FP16)
+#if defined(AMX_FP16) || defined(AMX_COMPLEX)
/* Transformation functions between fp16/float */
static uint16_t make_f32_fp16 (float f)
{
new file mode 100644
@@ -0,0 +1,15 @@
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2 -mamx-complex" } */
+/* { dg-final { scan-assembler "tcmmimfp16ps\[ \\t]+\[^\n\]*%tmm3+\[^\n\]*%tmm2+\[^\n\]*%tmm1" } } */
+/* { dg-final { scan-assembler "tcmmrlfp16ps\[ \\t]+\[^\n\]*%tmm3+\[^\n\]*%tmm2+\[^\n\]*%tmm1" } } */
+#include <immintrin.h>
+
+#define TMM1 1
+#define TMM2 2
+#define TMM3 3
+
+void TEST()
+{
+ _tile_cmmimfp16ps (TMM1, TMM2, TMM3);
+ _tile_cmmrlfp16ps (TMM1, TMM2, TMM3);
+}
new file mode 100644
@@ -0,0 +1,12 @@
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-require-effective-target masm_intel } */
+/* { dg-options "-O2 -mamx-complex -masm=intel" } */
+/* { dg-final { scan-assembler "tcmmimfp16ps\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2+\[^\n\]*%tmm3" } } */
+/* { dg-final { scan-assembler "tcmmrlfp16ps\[ \\t]+\[^\n\]*%tmm1+\[^\n\]*%tmm2+\[^\n\]*%tmm3" } } */
+#include <immintrin.h>
+
+void TEST()
+{
+ _tile_cmmimfp16ps (1, 2, 3);
+ _tile_cmmrlfp16ps (1, 2, 3);
+}
new file mode 100644
@@ -0,0 +1,53 @@
+/* { dg-do run { target { ! ia32 } } } */
+/* { dg-require-effective-target amx_complex } */
+/* { dg-require-effective-target avx512fp16 } */
+/* { dg-options "-O2 -mamx-complex -mavx512fp16" } */
+#define AMX_COMPLEX
+#define DO_TEST test_amx_complex_cmmimfp16ps
+void test_amx_complex_cmmimfp16ps ();
+#include "amx-helper.h"
+
+void calc_matrix_cmmimfp16ps (__tile *dst, __tile *src1, __tile *src2)
+{
+ uint16_t *src1_buf = (uint16_t *) src1->buf;
+ uint16_t *src2_buf = (uint16_t *) src2->buf;
+ float *dst_buf = (float *) dst->buf;
+
+ int M = src1->rows;
+ int N = src1->colsb / 4;
+ int K = src2->colsb / 4;
+ int i, j, k, t;
+
+ for (i = 0; i < M; i++)
+ for (j = 0; j < N; j++)
+ for (k = 0; k < K; k++)
+ for (t = 0; t < 2; t+=2)
+ dst_buf[i * N + k] +=
+ (make_fp16_f32(src1_buf[i * 2 * N + 2 * j + t]) *
+ make_fp16_f32(src2_buf[j * 2 * K + 2 * k + t + 1])) +
+ (make_fp16_f32(src1_buf[i * 2 * N + 2 * j + t + 1]) *
+ make_fp16_f32(src2_buf[j * 2 * K + 2 * k + t]));
+}
+
+void test_amx_complex_cmmimfp16ps ()
+{
+ __tilecfg_u cfg;
+ __tile dst, dst_ref, src1, src2;
+ uint8_t tmp_dst_buf[1024], tmp_dst_zero_buf[1024];
+
+ init_fp16_max_tile_buffer (tmp_dst_buf);
+ init_fp16_max_tile_zero_buffer (tmp_dst_zero_buf);
+
+ init_tile_config (&cfg);
+ init_tile_reg_and_src_with_buffer (1, dst, tmp_dst_zero_buf);
+ init_tile_reg_and_src_with_buffer (2, src1, tmp_dst_buf);
+ init_tile_reg_and_src_with_buffer (3, src2, tmp_dst_buf);
+
+ calc_matrix_cmmimfp16ps (&dst, &src1, &src2);
+
+ _tile_cmmimfp16ps (1, 2, 3);
+ _tile_stored (1, dst_ref.buf, _STRIDE);
+
+ if (!check_tile_register (&dst_ref, &dst))
+ abort ();
+}
new file mode 100644
@@ -0,0 +1,53 @@
+/* { dg-do run { target { ! ia32 } } } */
+/* { dg-require-effective-target amx_complex } */
+/* { dg-require-effective-target avx512fp16 } */
+/* { dg-options "-O2 -mamx-complex -mavx512fp16" } */
+#define AMX_COMPLEX
+#define DO_TEST test_amx_complex_cmmrlfp16ps
+void test_amx_complex_cmmrlfp16ps();
+#include "amx-helper.h"
+
+void calc_matrix_cmmrlfp16ps (__tile *dst, __tile *src1, __tile *src2)
+{
+ uint16_t *src1_buf = (uint16_t *) src1->buf;
+ uint16_t *src2_buf = (uint16_t *) src2->buf;
+ float *dst_buf = (float *) dst->buf;
+
+ int M = src1->rows;
+ int N = src1->colsb / 4;
+ int K = src2->colsb / 4;
+ int i, j, k, t;
+
+ for (i = 0; i < M; i++)
+ for (j = 0; j < N; j++)
+ for (k = 0; k < K; k++)
+ for (t = 0; t < 2; t+=2)
+ dst_buf[i * N + k] +=
+ (make_fp16_f32 (src1_buf[i * 2 * N + 2 * j + t]) *
+ make_fp16_f32 (src2_buf[j * 2 * K + 2 * k + t])) -
+ (make_fp16_f32 (src1_buf[i * 2 * N + 2 * j + t + 1]) *
+ make_fp16_f32 (src2_buf[j * 2 * K + 2 * k + t + 1]));
+}
+
+void test_amx_complex_cmmrlfp16ps ()
+{
+ __tilecfg_u cfg;
+ __tile dst, dst_ref, src1, src2;
+ uint8_t tmp_dst_buf[1024], tmp_dst_zero_buf[1024];
+
+ init_fp16_max_tile_buffer (tmp_dst_buf);
+ init_fp16_max_tile_zero_buffer (tmp_dst_zero_buf);
+
+ init_tile_config (&cfg);
+ init_tile_reg_and_src_with_buffer (1, dst, tmp_dst_zero_buf);
+ init_tile_reg_and_src_with_buffer (2, src1, tmp_dst_buf);
+ init_tile_reg_and_src_with_buffer (3, src2, tmp_dst_buf);
+
+ calc_matrix_cmmrlfp16ps (&dst, &src1, &src2);
+
+ _tile_cmmrlfp16ps (1, 2, 3);
+ _tile_stored (1, dst_ref.buf, _STRIDE);
+
+ if (!check_tile_register (&dst_ref, &dst))
+ abort ();
+}
@@ -87,6 +87,7 @@ extern void test_cmpccxadd (void) __attribute__((__target__("cmpccxadd")));
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_no_sgx (void) __attribute__((__target__("no-sgx")));
extern void test_no_avx5124fmaps(void) __attribute__((__target__("no-avx5124fmaps")));
@@ -175,6 +176,7 @@ extern void test_no_cmpccxadd (void) __attribute__((__target__("no-cm
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_arch_nocona (void) __attribute__((__target__("arch=nocona")));
extern void test_arch_core2 (void) __attribute__((__target__("arch=core2")));
@@ -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" } */
+/* { 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" } */
#include <x86intrin.h>
@@ -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" } */
+/* { 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-add-options bind_pic_locally } */
#include <mm_malloc.h>
@@ -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" } */
+/* { 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-add-options bind_pic_locally } */
#include <mm_malloc.h>
@@ -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")
+#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")
#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")
+#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")
#endif
#include <immintrin.h>
test_1 (_cvtss_sh, unsigned short, float, 1)
@@ -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")
+#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")
#include <x86intrin.h>
@@ -9750,6 +9750,17 @@ proc check_effective_target_raoint { } {
} "-mraoint" ]
}
+# Return 1 if amx-complex instructions can be compiled.
+proc check_effective_target_amx_complex { } {
+ return [check_no_compiler_messages amx_complex object {
+ void
+ foo ()
+ {
+ __asm__ volatile ("tcmmimfp16ps\t%%tmm1, %%tmm2, %%tmm3" ::);
+ }
+ } "-mamx-complex" ]
+}
+
# Return 1 if sse instructions can be compiled.
proc check_effective_target_sse { } {
return [check_no_compiler_messages sse object {