feat(asm): add vpcmp compare, full opmask set, legacy sse integers and bswap

Assisted-by: GLM 5.3

asm/encode.go

@@ -82,8 +82,12 @@ func (e *enc) encode(mnem string, ops []Operand) error {
 	if m, ok := sseShufTable[upper]; ok {
 		return e.encodeSSEShuf(m, ops)
 	}
-	// Legacy SSE packed binary ops and imm8 shuffles (ADDPS/MULPS/
-	// SHUFPS/PSHUFD/...): no size suffix, dispatch on the full name.
+	// Legacy SSE packed binaries dispatch on the full name: the packed
+	// integer mnemonics carry real width suffixes (PADDB/PCMPGTW/...),
+	// which the size split must not eat.
+	if m, ok := sseBinTable[upper]; ok {
+		return e.encodeSSEBin(m, ops)
+	}
 	if m, ok := sseBinTable[base]; ok {
 		return e.encodeSSEBin(m, ops)
 	}
@@ -108,6 +112,8 @@ func (e *enc) encode(mnem string, ops []Operand) error {
 		return e.encodePushPop(ops, false)
 	case "BSF", "BSR", "LZCNT", "TZCNT", "POPCNT":
 		return e.encodeCount(base, ops, size)
+	case "BSWAP":
+		return e.encodeBswap(ops, size)
 	case "MOVBLZX", "MOVBQZX", "MOVWLZX", "MOVWQZX", "MOVWLSX", "MOVLQSX":
 		return e.encodeMovExtend(base, ops)
 	case "CVTSL2SD", "CVTSQ2SD":