feat(arch): add the FP16 complex fused multiply-add to the extension layer

Assisted-by: GLM 5.3 Flash

asm/extension_amd64_test.go

@@ -80,6 +80,10 @@ func TestAmd64ExtensionRegistry(t *testing.T) {
 		{"VFCMULCPH", 3},
 		{"VFMULCSH", 1},
 		{"VFCMULCSH", 1},
+		{"VFMADDCPH", 3},
+		{"VFCMADDCPH", 3},
+		{"VFMADDCSH", 1},
+		{"VFCMADDCSH", 1},
 		{"VMINMAXPH", 3},
 	} {
 		cands, ok := LookupExtension(arch.AMD64, tt.mnem)
@@ -94,8 +98,8 @@ func TestAmd64ExtensionRegistry(t *testing.T) {
 			t.Errorf("the %s lookup is not case-insensitive", tt.mnem)
 		}
 	}
-	if got := arch.Extensions(arch.AMD64); len(got) != 174 {
-		t.Errorf("the amd64 layer registers %d instructions, want 174", len(got))
+	if got := arch.Extensions(arch.AMD64); len(got) != 182 {
+		t.Errorf("the amd64 layer registers %d instructions, want 182", len(got))
 	}
 	if _, ok := LookupExtension(arch.AMD64, "NOSUCHINSTR"); ok {
 		t.Error("a non-extended mnemonic resolved")
@@ -201,6 +205,15 @@ func TestEncodeExtensionAmd64(t *testing.T) {
 		{"complex multiply, conjugating the first source", "VFCMULCPH",
 			[]arch.ExtOperand{arch.ExtZmm(29), arch.ExtZmm(28), arch.ExtZmm(30)},
 			"62061740d6f4"},
+		{"complex multiply-add, conjugating the second source", "VFMADDCPH",
+			[]arch.ExtOperand{arch.ExtZmm(29), arch.ExtZmm(28), arch.ExtZmm(30)},
+			"6206164056f4"},
+		{"complex multiply-add, conjugated first source, rounded", "VFCMADDCPH",
+			[]arch.ExtOperand{arch.ExtZmm(5), arch.ExtZmm(4), arch.ExtRounded(arch.ExtZmm(6), arch.ExtRoundNearest)},
+			"62f6571856f4"},
+		{"scalar complex multiply-add out of memory", "VFMADDCSH",
+			[]arch.ExtOperand{arch.ExtXmm(29), arch.ExtMemory(9, 0), arch.ExtXmm(30)},
+			"624616005731"},
 		{"minimum or maximum under a control byte", "VMINMAXPH",
 			[]arch.ExtOperand{arch.ExtImmediate(0x88), arch.ExtZmm(29), arch.ExtMemory(9, 0), arch.ExtZmm(30)},
 			"62431440523188"},