cmd/asm: add support for bdnz/bdz extended mnemonics on PPC64

Support BDNZ and BDZ mnemonics, they are commonly used
POWER instructions. The raw BC mnemonic is not easy
to read.

Likewise, cleanup code surrounding these changes.

Change-Id: I72f1dad5013f7856bd0dd320bfb17b5a9f3c69ee
Reviewed-on: https://go-review.googlesource.com/c/go/+/390696
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Trust: Paul Murphy <murp@ibm.com>
Run-TryBot: Paul Murphy <murp@ibm.com>
TryBot-Result: Gopher Robot <gobot@golang.org>

src/cmd/asm/internal/arch/ppc64.go

@@ -15,7 +15,7 @@ import (
 
 func jumpPPC64(word string) bool {
 	switch word {
-	case "BC", "BCL", "BEQ", "BGE", "BGT", "BL", "BLE", "BLT", "BNE", "BR", "BVC", "BVS", "CALL", "JMP":
+	case "BC", "BCL", "BEQ", "BGE", "BGT", "BL", "BLE", "BLT", "BNE", "BR", "BVC", "BVS", "BDNZ", "BDZ", "CALL", "JMP":
 		return true
 	}
 	return false

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

@@ -764,4 +764,10 @@ TEXT asmtest(SB),DUPOK|NOSPLIT,$0
 	MOVFL R1, $128                  // 7c380120
 	MOVFL R1, $3                    // 7c203120
 
+	// Verify supported bdnz/bdz encodings.
+	BC 16,0,0(PC)                   // BC $16,R0,0(PC) // 42000000
+	BDNZ 0(PC)                      // 42000000
+	BDZ 0(PC)                       // 42400000
+	BC 18,0,0(PC)                   // BC $18,R0,0(PC) // 42400000
+
 	RET

src/cmd/internal/obj/ppc64/a.out.go

@@ -362,13 +362,14 @@ const (
 	BI_LT  = 0
 	BI_GT  = 1
 	BI_EQ  = 2
-	BI_OVF = 3
+	BI_FU  = 3
 )
 
 // Common values for the BO field.
 
 const (
 	BO_BCTR    = 16 // decrement ctr, branch on ctr != 0
+	BO_NOTBCTR = 18 // decrement ctr, branch on ctr == 0
 	BO_BCR     = 12 // branch on cr value
 	BO_BCRBCTR = 8  // decrement ctr, branch on ctr != 0 and cr value
 	BO_NOTBCR  = 4  // branch on not cr value
@@ -481,8 +482,10 @@ const (
 	ABLE // not GT = L/E/U
 	ABLT
 	ABNE  // not EQ = L/G/U
-	ABVC // Unordered-clear
+	ABVC  // Branch if float not unordered (also branch on not summary overflow)
-	ABVS // Unordered-set
+	ABVS  // Branch if float unordered (also branch on summary overflow)
+	ABDNZ // Decrement CTR, and branch if CTR != 0
+	ABDZ  // Decrement CTR, and branch if CTR == 0
 	ACMP
 	ACMPU
 	ACMPEQB

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

@@ -42,6 +42,8 @@ var Anames = []string{
 	"BNE",
 	"BVC",
 	"BVS",
+	"BDNZ",
+	"BDZ",
 	"CMP",
 	"CMPU",
 	"CMPEQB",

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

@@ -305,6 +305,7 @@ var optab = []Optab{
 	{as: ABC, a1: C_SCON, a2: C_REG, a6: C_LR, type_: 18, size: 4},
 	{as: ABC, a1: C_SCON, a2: C_REG, a6: C_CTR, type_: 18, size: 4},
 	{as: ABC, a6: C_ZOREG, type_: 15, size: 8},
+	{as: ABDNZ, a6: C_SBRA, type_: 16, size: 4},
 	{as: ASYNC, type_: 46, size: 4},
 	{as: AWORD, a1: C_LCON, type_: 40, size: 4},
 	{as: ADWORD, a1: C_64CON, type_: 31, size: 8},
@@ -1778,6 +1779,9 @@ func buildop(ctxt *obj.Link) {
 		case ABC:
 			opset(ABCL, r0)
 
+		case ABDNZ:
+			opset(ABDZ, r0)
+
 		case AEXTSB: /* op Rs, Ra */
 			opset(AEXTSBCC, r0)
 
@@ -4875,21 +4879,25 @@ func (c *ctxt9) opirr(a obj.As) uint32 {
 		return OPVCC(16, 0, 0, 0) | 1
 
 	case ABEQ:
-		return AOP_RRR(16<<26, 12, 2, 0)
+		return AOP_RRR(16<<26, BO_BCR, BI_EQ, 0)
 	case ABGE:
-		return AOP_RRR(16<<26, 4, 0, 0)
+		return AOP_RRR(16<<26, BO_NOTBCR, BI_LT, 0)
 	case ABGT:
-		return AOP_RRR(16<<26, 12, 1, 0)
+		return AOP_RRR(16<<26, BO_BCR, BI_GT, 0)
 	case ABLE:
-		return AOP_RRR(16<<26, 4, 1, 0)
+		return AOP_RRR(16<<26, BO_NOTBCR, BI_GT, 0)
 	case ABLT:
-		return AOP_RRR(16<<26, 12, 0, 0)
+		return AOP_RRR(16<<26, BO_BCR, BI_LT, 0)
 	case ABNE:
-		return AOP_RRR(16<<26, 4, 2, 0)
+		return AOP_RRR(16<<26, BO_NOTBCR, BI_EQ, 0)
 	case ABVC:
-		return AOP_RRR(16<<26, 4, 3, 0) // apparently unordered-clear
+		return AOP_RRR(16<<26, BO_NOTBCR, BI_FU, 0)
 	case ABVS:
-		return AOP_RRR(16<<26, 12, 3, 0) // apparently unordered-set
+		return AOP_RRR(16<<26, BO_BCR, BI_FU, 0)
+	case ABDZ:
+		return AOP_RRR(16<<26, BO_NOTBCTR, 0, 0)
+	case ABDNZ:
+		return AOP_RRR(16<<26, BO_BCTR, 0, 0)
 
 	case ACMP:
 		return OPVCC(11, 0, 0, 0) | 1<<21 /* L=1 */