diff options
author | MITSUNARI Shigeo <[email protected]> | 2024-10-11 14:40:00 +0900 |
---|---|---|
committer | MITSUNARI Shigeo <[email protected]> | 2024-10-13 13:51:06 +0900 |
commit | 842c3cc83f209c0789f15a809ae45cf30a8b2f5f (patch) | |
tree | 7f0ca97100c6eae52d6394681607cae34537675a /test | |
parent | 6dc564185b85d16e240f9f15aad5be08036176d9 (diff) | |
download | xbyak-842c3cc83f209c0789f15a809ae45cf30a8b2f5f.tar.gz xbyak-842c3cc83f209c0789f15a809ae45cf30a8b2f5f.zip |
support all avx10 bf16 instructions
Diffstat (limited to 'test')
-rw-r--r-- | test/avx10/bf16.txt | 99 |
1 files changed, 99 insertions, 0 deletions
diff --git a/test/avx10/bf16.txt b/test/avx10/bf16.txt index f3e44a6..c544e02 100644 --- a/test/avx10/bf16.txt +++ b/test/avx10/bf16.txt @@ -109,3 +109,102 @@ vfpclasspbf16(k5|k5, yword[rax+128], 10); vfpclasspbf16(k6|k5, yword_b[rax+128], 11); vfpclasspbf16(k7|k5, zword[rax+128], 12); vfpclasspbf16(k7|k5, zword_b[rax+128], 13); + +vcomsbf16(xm2, xm3); +vcomsbf16(xm2, ptr[rax+128]); + +vgetexppbf16(xm1|k3, xmm2); +vgetexppbf16(xm1|k3, ptr[rax+128]); +vgetexppbf16(xm1|k3, ptr_b[rax+128]); + +vgetexppbf16(ym1|k3, ymm2); +vgetexppbf16(ym1|k3, ptr[rax+128]); +vgetexppbf16(ym1|k3, ptr_b[rax+128]); + +vgetexppbf16(zm1|k3, zmm2); +vgetexppbf16(zm1|k3, ptr[rax+128]); +vgetexppbf16(zm1|k3, ptr_b[rax+128]); + +vgetmantpbf16(xm1|k3, xmm2, 3); +vgetmantpbf16(xm1|k3, ptr[rax+128], 5); +vgetmantpbf16(xm1|k3, ptr_b[rax+128], 9); + +vgetmantpbf16(ym1|k3, ymm2, 3); +vgetmantpbf16(ym1|k3, ptr[rax+128], 5); +vgetmantpbf16(ym1|k3, ptr_b[rax+128], 9); + +vgetmantpbf16(zm1|k3, zmm2, 3); +vgetmantpbf16(zm1|k3, ptr[rax+128], 5); +vgetmantpbf16(zm1|k3, ptr_b[rax+128], 9); + +vrcppbf16(xm1|k5, xm2); +vrcppbf16(xm1|k5, ptr[rcx+128]); +vrcppbf16(xm1|k5, ptr_b[rcx+128]); + +vrcppbf16(ym1|k5, ym2); +vrcppbf16(ym1|k5, ptr[rcx+128]); +vrcppbf16(ym1|k5, ptr_b[rcx+128]); + +vrcppbf16(zm1|k5, zm2); +vrcppbf16(zm1|k5, ptr[rcx+128]); +vrcppbf16(zm1|k5, ptr_b[rcx+128]); + +vreducenepbf16(xm1|k4, xm2, 1); +vreducenepbf16(xm1|k4, ptr[rax+128], 1); +vreducenepbf16(xm1|k4, ptr_b[rax+128], 1); + +vreducenepbf16(ym1|k4, ym2, 1); +vreducenepbf16(ym1|k4, ptr[rax+128], 1); +vreducenepbf16(ym1|k4, ptr_b[rax+128], 1); + +vreducenepbf16(zm1|k4, zm2, 1); +vreducenepbf16(zm1|k4, ptr[rax+128], 1); +vreducenepbf16(zm1|k4, ptr_b[rax+128], 1); + +vrndscalenepbf16(xm1|k4, xm2, 1); +vrndscalenepbf16(xm1|k4, ptr[rax+128], 1); +vrndscalenepbf16(xm1|k4, ptr_b[rax+128], 1); + +vrndscalenepbf16(ym1|k4, ym2, 1); +vrndscalenepbf16(ym1|k4, ptr[rax+128], 1); +vrndscalenepbf16(ym1|k4, ptr_b[rax+128], 1); + +vrndscalenepbf16(zm1|k4, zm2, 1); +vrndscalenepbf16(zm1|k4, ptr[rax+128], 1); +vrndscalenepbf16(zm1|k4, ptr_b[rax+128], 1); + +vrsqrtpbf16(xm1|k5, xm2); +vrsqrtpbf16(xm1|k5, ptr[rcx+128]); +vrsqrtpbf16(xm1|k5, ptr_b[rcx+128]); + +vrsqrtpbf16(ym1|k5, ym2); +vrsqrtpbf16(ym1|k5, ptr[rcx+128]); +vrsqrtpbf16(ym1|k5, ptr_b[rcx+128]); + +vrsqrtpbf16(zm1|k5, zm2); +vrsqrtpbf16(zm1|k5, ptr[rcx+128]); +vrsqrtpbf16(zm1|k5, ptr_b[rcx+128]); + +vscalefpbf16(xm1|k5, xm5, xm2); +vscalefpbf16(xm1|k5, xm5, ptr[rcx+128]); +vscalefpbf16(xm1|k5, xm5, ptr_b[rcx+128]); + +vscalefpbf16(ym1|k5, ym9, ym2); +vscalefpbf16(ym1|k5, ym9, ptr[rcx+128]); +vscalefpbf16(ym1|k5, ym9, ptr_b[rcx+128]); + +vscalefpbf16(zm1|k5, zm30, zm2); +vscalefpbf16(zm1|k5, zm30, ptr[rcx+128]); +vscalefpbf16(zm1|k5, zm30, ptr_b[rcx+128]); + +vsqrtnepbf16(xm5|k3, xmm4); +vsqrtnepbf16(xm5|k3, ptr[rax+128]); +vsqrtnepbf16(xm5|k3, ptr_b[rax+128]); + +vsqrtnepbf16(ym5|k3, ymm4); +vsqrtnepbf16(ym5|k3, ptr[rax+128]); +vsqrtnepbf16(ym5|k3, ptr_b[rax+128]); + +vsqrtnepbf16(zm5|k3, zmm4); +vsqrtnepbf16(zm5|k3, ptr[rax+128]); +vsqrtnepbf16(zm5|k3, ptr_b[rax+128]); |