petrbalvin/gasm-sdk
0
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 37 Wiki Activity

feat(arch): add VMINMAXSH to the extension layer

Browse Source 
Assisted-by: GLM 5.3 Flash
This commit is contained in:
petrbalvin committed 2026-10-07 19:49:15 +02:00
1 parent 03f9ef0ac6
commit 27ef71859b
3 files changed
+35 -4
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

arch/amd64_ext.go
+9
View File
@@ -1627,4 +1627,13 @@ var amd64Extensions = []ExtInstr{
{Name: "VMINMAXPH", Summary: "Return the per-lane minimum or maximum of packed FP16 values under an imm8 control",
Bytes: []byte{0x62, 0x03, 0x04, 0x00, 0x52, 0xC0}, Form: ExtFormAmdVec3Imm, Mem: 3, Mask: true, Feature: ExtFeatureFP16,
Ref: "Intel SDM Vol. 2C, VMINMAXPH (EVEX.NDS.128.0F3A.W0 52 /r /ib)"},
// The scalar mirror: VMINMAXSH picks between one pair of FP16 values
// under the same imm8 control, EVEX.NDS.LIG.0F3A.W0 53 /r /ib, the
// control byte riding last as the packed form's does. A scalar
// destination takes no write mask and the m16 memory source no
// broadcast, like the rest of the scalar core.
{Name: "VMINMAXSH", Summary: "Return the minimum or maximum of scalar FP16 values under an imm8 control",
Bytes: []byte{0x62, 0x03, 0x04, 0x00, 0x53, 0xC0}, Form: ExtFormAmdVec3Imm, Mem: 3, Feature: ExtFeatureFP16,
Ref: "Intel SDM Vol. 2C, VMINMAXSH (EVEX.NDS.LIG.0F3A.W0 53 /r /ib)"},
}