cmd/internal/obj/ppc64: fix wrong register encoding in XX1-Form instructions

A bug in the encoding of XX1-Form is flipping bit 31 of such instructions.
This may result in register clobering when using VSX instructions.

This was not exposed before because we currently don't generate these
instructions in SSA, and the asm files in which they are present aren't
affected by register clobbering.

This change fixes the bug and adds a testcase for the problem.

Fixes #30112

Change-Id: I77b606159ae1efea33d2ba3e1c74b7fae8d5d2e7
Reviewed-on: https://go-review.googlesource.com/c/go/+/163759
Reviewed-by: Bryan C. Mills <[email protected]>
Reviewed-by: Lynn Boger <[email protected]>
Run-TryBot: Bryan C. Mills <[email protected]>
TryBot-Result: Gobot Gobot <[email protected]>

Author: ceseo

src/cmd/asm/internal/asm/testdata/ppc64.s

@@ -1021,18 +1021,24 @@ label1:
 //	VSX move from VSR, XX1-form
 //	<MNEMONIC> XS,RA produces
 //	<mnemonic> RA,XS
+//	Extended mnemonics accept VMX and FP registers as sources
 	MFVSRD	    VS0, R1
 	MFVSRWZ	    VS33, R1
 	MFVSRLD	    VS63, R1
+	MFVRD       V0, R1
+	MFFPRD      F0, R1
 
 //	VSX move to VSR, XX1-form
 //	<MNEMONIC> RA,XT produces
 //	<mnemonic> XT,RA
+//	Extended mnemonics accept VMX and FP registers as targets
 	MTVSRD	    R1, VS0
 	MTVSRWA	    R1, VS31
 	MTVSRWZ	    R1, VS63
 	MTVSRDD	    R1, R2, VS0
 	MTVSRWS	    R1, VS32
+	MTVRD       R1, V13
+	MTFPRD      R1, F24
 
 //	VSX AND, XX3-form
 //	<MNEMONIC> XA,XB,XT produces

src/cmd/internal/obj/ppc64/asm9.go

@@ -3555,22 +3555,22 @@ func (c *ctxt9) asmout(p *obj.Prog, o *Optab, out []uint32) {
 		if REG_V0 <= xt && xt <= REG_V31 {
 			/* Convert V0-V31 to VS32-VS63 */
 			xt = xt + 64
-			o1 = AOP_XX1(c.oprrr(p.As), uint32(p.To.Reg), uint32(p.From.Reg), uint32(p.Reg))
+			o1 = AOP_XX1(c.oprrr(p.As), uint32(xt), uint32(p.From.Reg), uint32(p.Reg))
 		} else if REG_F0 <= xt && xt <= REG_F31 {
 			/* Convert F0-F31 to VS0-VS31 */
 			xt = xt + 64
-			o1 = AOP_XX1(c.oprrr(p.As), uint32(p.To.Reg), uint32(p.From.Reg), uint32(p.Reg))
+			o1 = AOP_XX1(c.oprrr(p.As), uint32(xt), uint32(p.From.Reg), uint32(p.Reg))
 		} else if REG_VS0 <= xt && xt <= REG_VS63 {
-			o1 = AOP_XX1(c.oprrr(p.As), uint32(p.To.Reg), uint32(p.From.Reg), uint32(p.Reg))
+			o1 = AOP_XX1(c.oprrr(p.As), uint32(xt), uint32(p.From.Reg), uint32(p.Reg))
 		} else if REG_V0 <= xs && xs <= REG_V31 {
 			/* Likewise for XS */
 			xs = xs + 64
-			o1 = AOP_XX1(c.oprrr(p.As), uint32(p.From.Reg), uint32(p.To.Reg), uint32(p.Reg))
+			o1 = AOP_XX1(c.oprrr(p.As), uint32(xs), uint32(p.To.Reg), uint32(p.Reg))
 		} else if REG_F0 <= xs && xs <= REG_F31 {
 			xs = xs + 64
-			o1 = AOP_XX1(c.oprrr(p.As), uint32(p.From.Reg), uint32(p.To.Reg), uint32(p.Reg))
+			o1 = AOP_XX1(c.oprrr(p.As), uint32(xs), uint32(p.To.Reg), uint32(p.Reg))
 		} else if REG_VS0 <= xs && xs <= REG_VS63 {
-			o1 = AOP_XX1(c.oprrr(p.As), uint32(p.From.Reg), uint32(p.To.Reg), uint32(p.Reg))
+			o1 = AOP_XX1(c.oprrr(p.As), uint32(xs), uint32(p.To.Reg), uint32(p.Reg))
 		}
 
 	case 89: /* VSX instructions, XX2-form */