feat(arch): add the AVX512-FP16 FMA families to the extension layer

Assisted-by: GLM 5.3 Flash

asm/extension_amd64_test.go

@@ -58,6 +58,24 @@ func TestAmd64ExtensionRegistry(t *testing.T) {
 		{"VRNDSCALEPH", 3},
 		{"VREDUCEPH", 3},
 		{"VGETMANTPH", 3},
+		{"VFMADD132PH", 3},
+		{"VFMADD213PH", 3},
+		{"VFMADD231PH", 3},
+		{"VFMSUB132PH", 3},
+		{"VFMSUB213PH", 3},
+		{"VFMSUB231PH", 3},
+		{"VFMADDSUB132PH", 3},
+		{"VFMADDSUB213PH", 3},
+		{"VFMADDSUB231PH", 3},
+		{"VFMSUBADD132PH", 3},
+		{"VFMSUBADD213PH", 3},
+		{"VFMSUBADD231PH", 3},
+		{"VFMADD132SH", 1},
+		{"VFMADD213SH", 1},
+		{"VFMADD231SH", 1},
+		{"VFMSUB132SH", 1},
+		{"VFMSUB213SH", 1},
+		{"VFMSUB231SH", 1},
 	} {
 		cands, ok := LookupExtension(arch.AMD64, tt.mnem)
 		if !ok {
@@ -71,8 +89,8 @@ func TestAmd64ExtensionRegistry(t *testing.T) {
 			t.Errorf("the %s lookup is not case-insensitive", tt.mnem)
 		}
 	}
-	if got := arch.Extensions(arch.AMD64); len(got) != 121 {
-		t.Errorf("the amd64 layer registers %d instructions, want 121", len(got))
+	if got := arch.Extensions(arch.AMD64); len(got) != 163 {
+		t.Errorf("the amd64 layer registers %d instructions, want 163", len(got))
 	}
 	if _, ok := LookupExtension(arch.AMD64, "NOSUCHINSTR"); ok {
 		t.Error("a non-extended mnemonic resolved")