feat(asm): add the wider EVEX set and the rounding, SAE and broadcast suffixes

Assisted-by: Qwen 3.8 Max Preview

asm/encode.go

@@ -51,16 +51,17 @@ func (e *enc) encode(mnem string, ops []Operand) error {
 
 	// VEX (AVX/AVX2) and EVEX (AVX-512) instructions: the trailing
 	// B/W/L/Q/D is part of the mnemonic, not a size suffix, so dispatch
-	// before splitSize.  A ".Z" suffix requests EVEX zeroing.
-	base, zeroing, err := stripEvexSuffix(upper)
+	// before splitSize.  EVEX suffixes (.Z, .SAE, rounding, .BCST) split
+	// off the mnemonic too.
+	base, sfx, err := parseEvexSuffix(upper)
 	if err != nil {
 		return err
 	}
-	if isVex(base) || isEvex(base) || base == "KMOVW" {
-		return e.encodeVec(base, ops, zeroing)
+	if isVex(base) || isEvex(base) || isKOp(base) || base == "KMOVW" || base == "KMOVQ" {
+		return e.encodeVec(base, ops, sfx)
 	}
-	if zeroing {
-		return fmt.Errorf("%s: the .Z suffix requires an EVEX instruction", mnem)
+	if sfx.any() {
+		return fmt.Errorf("%s: the suffix requires an EVEX instruction", mnem)
 	}
 
 	// CMOVcc and SETcc carry the condition in the mnemonic (CMOVLGT, SETNE).
@@ -128,20 +129,26 @@ func splitSize(upper string) (base string, size int) {
 // its own direction-dependent opcodes; KTESTW is always VEX; everything else
 // takes EVEX when an operand demands it (a ZMM or K register, or an
 // EVEX-only mnemonic) and VEX otherwise.
-func (e *enc) encodeVec(upper string, ops []Operand, zeroing bool) error {
-	if upper == "KMOVW" {
-		if zeroing {
-			return fmt.Errorf("KMOVW takes no .Z suffix")
+func (e *enc) encodeVec(upper string, ops []Operand, sfx evexSuffix) error {
+	if upper == "KMOVW" || upper == "KMOVQ" {
+		if sfx.any() {
+			return fmt.Errorf("%s takes no EVEX suffixes", upper)
 		}
-		return e.encodeKmovw(ops)
+		return e.encodeKmov(upper, ops)
 	}
-	if upper == "KTESTW" || !evexRequired(upper, ops) {
-		if zeroing {
+	if isKOp(upper) {
+		if sfx.any() {
+			return fmt.Errorf("%s takes no EVEX suffixes", upper)
+		}
+		return e.encodeKOp(upper, ops)
+	}
+	if upper == "KTESTW" || (!evexRequired(upper, ops) && !sfx.evexOnly()) {
+		if sfx.any() {
 			return fmt.Errorf("%s: the .Z suffix requires an EVEX instruction", upper)
 		}
 		return e.encodeVex(upper, ops)
 	}
-	return e.encodeEvex(upper, ops, zeroing)
+	return e.encodeEvex(upper, ops, sfx)
 }
 
 // --- instruction components -------------------------------------------------