cmd/internal/obj/arm64: fix assemble LDXP bug

The current code calculates register number incorrectly.

The fix corrects the register number calculation.

Add cases created by decoder to test assembler.

Fixes #20697
Fixes #20723

Change-Id: I73ac153df9ea9f51c43a5104828d7a5389551c92
Reviewed-on: https://go-review.googlesource.com/45850
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>

src/cmd/asm/internal/asm/endtoend_test.go

@@ -381,6 +381,10 @@ func TestARM64EndToEnd(t *testing.T) {
 	testEndToEnd(t, "arm64", "arm64")
 }
 
+func TestARM64Encoder(t *testing.T) {
+	testEndToEnd(t, "arm64", "arm64enc")
+}
+
 func TestAMD64EndToEnd(t *testing.T) {
 	testEndToEnd(t, "amd64", "amd64")
 }

src/cmd/internal/obj/arm64/asm7.go

@@ -2232,7 +2232,7 @@ func (c *ctxt7) asmout(p *obj.Prog, o *Optab, out []uint32) {
 			o1 = 0
 		}
 
-	case 15: /* mul/mneg/umulh/umull r,[r,]r; madd/msub Rm,Rn,Ra,Rd */
+	case 15: /* mul/mneg/umulh/umull r,[r,]r; madd/msub Rm,Ra,Rn,Rd */
 		o1 = c.oprrr(p, p.As)
 
 		rf := int(p.From.Reg)
@@ -2900,9 +2900,9 @@ func (c *ctxt7) asmout(p *obj.Prog, o *Optab, out []uint32) {
 		o1 = c.opload(p, p.As)
 
 		o1 |= 0x1F << 16
-		o1 |= uint32(p.From.Reg) << 5
+		o1 |= uint32(p.From.Reg&31) << 5
 		if p.Reg != 0 {
-			o1 |= uint32(p.Reg) << 10
+			o1 |= uint32(p.Reg&31) << 10
 		} else {
 			o1 |= 0x1F << 10
 		}