cmd/asm: add s390x VMSLG instruction

This instruction was introduced on the z14 to accelerate "limbified"
multiplications for certain cryptographic algorithms. This change allows
it to be used in Go assembly.

Change-Id: Ic93dae7fec1756f662874c08a5abc435bce9dd9e
Reviewed-on: https://go-review.googlesource.com/109695
Run-TryBot: Michael Munday <[email protected]>
TryBot-Result: Gobot Gobot <[email protected]>
Reviewed-by: Brad Fitzpatrick <[email protected]>

Author: wgo

src/cmd/asm/internal/asm/testdata/s390x.s

@@ -367,6 +367,7 @@ TEXT main·foo(SB),DUPOK|NOSPLIT,$16-0 // TEXT main.foo(SB), DUPOK|NOSPLIT, $16-
 	VSTEF	$3, V2, (R9)            // e7209000300b
 	VSTEH	$7, V31, (R2)           // e7f020007809
 	VSTEB	$15, V29, 4094(R12)     // e7d0cffef808
+	VMSLG	V21, V22, V23, V24      // e78563007fb8
 
 	RET
 	RET	foo(SB)

src/cmd/internal/obj/s390x/a.out.go

@@ -941,6 +941,7 @@ const (
 	AVUPLB
 	AVUPLHW
 	AVUPLF
+	AVMSLG
 
 	// binary
 	ABYTE

src/cmd/internal/obj/s390x/anames.go

@@ -677,6 +677,7 @@ var Anames = []string{
 	"VUPLB",
 	"VUPLHW",
 	"VUPLF",
+	"VMSLG",
 	"BYTE",
 	"WORD",
 	"DWORD",

src/cmd/internal/obj/s390x/asmz.go

@@ -1376,6 +1376,7 @@ func buildop(ctxt *obj.Link) {
 			opset(AVSTRCZFS, r)
 			opset(AVSBCBIQ, r)
 			opset(AVSBIQ, r)
+			opset(AVMSLG, r)
 		case AVSEL:
 			opset(AVFMADB, r)
 			opset(AWFMADB, r)
@@ -2529,6 +2530,7 @@ const (
 	op_VUPLH  uint32 = 0xE7D5 // 	VRR-a	VECTOR UNPACK LOGICAL HIGH
 	op_VUPLL  uint32 = 0xE7D4 // 	VRR-a	VECTOR UNPACK LOGICAL LOW
 	op_VUPL   uint32 = 0xE7D6 // 	VRR-a	VECTOR UNPACK LOW
+	op_VMSL   uint32 = 0xE7B8 // 	VRR-d	VECTOR MULTIPLY SUM LOGICAL
 )
 
 func oclass(a *obj.Addr) int {

src/cmd/internal/obj/s390x/vector.go

@@ -45,6 +45,8 @@ func vop(as obj.As) (opcode, es, cs uint32) {
 		return op_VAC, 0, 0
 	case AVACQ:
 		return op_VAC, 4, 0
+	case AVMSLG:
+		return op_VMSL, 3, 0
 	case AVACCC:
 		return op_VACCC, 0, 0
 	case AVACCCQ: