AMX-TRANSPOSE is remove
diff --git a/sample/cpuid/cpx.txt b/sample/cpuid/cpx.txt index 707159d..c315b4f 100644 --- a/sample/cpuid/cpx.txt +++ b/sample/cpuid/cpx.txt
@@ -1,2 +1,2 @@ vendor intel - mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt prefetchw enh_rep rdrand adx rdseed smap f16c movbe avx512f avx512dq avx512cd avx512bw avx512vl avx512_vnni avx512_bf16 clflushopt clwb amx_fp8 amx_transpose amx_tf32 amx_avx512 + mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt prefetchw enh_rep rdrand adx rdseed smap f16c movbe avx512f avx512dq avx512cd avx512bw avx512vl avx512_vnni avx512_bf16 clflushopt clwb
diff --git a/sample/cpuid/dmr.txt b/sample/cpuid/dmr.txt index de71868..ed124cc 100644 --- a/sample/cpuid/dmr.txt +++ b/sample/cpuid/dmr.txt
@@ -1,2 +1,2 @@ vendor intel - mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt prefetchw enh_rep rdrand adx rdseed smap sha f16c movbe avx512f avx512dq avx512_ifma avx512cd avx512bw avx512vl avx512_vbmi avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq avx512_bf16 amx(tile) amx(int8) amx(bf16) avx_vnni avx512_fp16 waitpkg clflushopt cldemote clwb movdiri movdir64b uintr serialize amx_fp16 avx_vnni_int8 avx_ne_convert avx_ifma cmpccxadd prefetchiti sha512 sm3 sm4 avx_vnni_int16 apx_f avx10 amx_fp8 amx_transpose amx_tf32 amx_avx512 amx_movrs movrs amx_complex + mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt prefetchw enh_rep rdrand adx rdseed smap sha f16c movbe avx512f avx512dq avx512_ifma avx512cd avx512bw avx512vl avx512_vbmi avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq avx512_bf16 amx(tile) amx(int8) amx(bf16) avx_vnni avx512_fp16 waitpkg clflushopt cldemote clwb movdiri movdir64b uintr serialize amx_fp16 avx_vnni_int8 avx_ne_convert avx_ifma cmpccxadd prefetchiti sha512 sm3 sm4 avx_vnni_int16 apx_f avx10 amx_fp8 amx_tf32 amx_avx512 amx_movrs movrs amx_complex
diff --git a/sample/test_util.cpp b/sample/test_util.cpp index bfa6870..33cc595 100644 --- a/sample/test_util.cpp +++ b/sample/test_util.cpp
@@ -108,7 +108,7 @@ { Cpu::tKEYLOCKER_WIDE, "keylocker_wide" }, { Cpu::tTSXLDTRK, "tsxldtrk" }, { Cpu::tAMX_FP8, "amx_fp8" }, - { Cpu::tAMX_TRANSPOSE, "amx_transpose" }, +// { Cpu::tAMX_TRANSPOSE, "amx_transpose" }, { Cpu::tAMX_TF32, "amx_tf32" }, { Cpu::tAMX_AVX512, "amx_avx512" }, { Cpu::tAMX_MOVRS, "amx_movrs" },
diff --git a/xbyak/xbyak_util.h b/xbyak/xbyak_util.h index a75d3f9..e0d0eb0 100644 --- a/xbyak/xbyak_util.h +++ b/xbyak/xbyak_util.h
@@ -748,10 +748,11 @@ if (edx & (1U << 14)) type_ |= tPREFETCHITI; if (edx & (1U << 19)) type_ |= tAVX10; if (edx & (1U << 21)) type_ |= tAPX_F; - + } + if (maxNum >= 0x1e) { getCpuidEx(0x1e, 1, data); if (eax & (1U << 4)) type_ |= tAMX_FP8; - if (eax & (1U << 5)) type_ |= tAMX_TRANSPOSE; +// if (eax & (1U << 5)) type_ |= tAMX_TRANSPOSE; // removed at 319433-059 if (eax & (1U << 6)) type_ |= tAMX_TF32; if (eax & (1U << 7)) type_ |= tAMX_AVX512; if (eax & (1U << 8)) type_ |= tAMX_MOVRS;