cmd/asm, cmd/internal/obj/s390x: add new s390x instructions for SSA

This commit adds the following instructions to support the new SSA
backend for s390x:

32-bit operations:
ADDW
SUBW
NEGW
FNEGS

Conditional moves:
MOVDEQ
MOVDGE
MOVDGT
MOVDLE
MOVDLT
MOVDNE

Unordered branches (for floating point comparisons):
BLEU
BLTU

Modulo operations:
MODW
MODWU
MODD
MODDU

The modulo operations might be removed in a future commit because
I'd like to change DIV to produce a tuple once the old backend is
removed.

This commit also removes uses of REGZERO from the assembler. They
aren't necessary and R0 will be used as a GPR by SSA.

Change-Id: I05756c1cbb74bf4a35fc492f8f0cd34b50763dc9
Reviewed-on: https://go-review.googlesource.com/29075
Run-TryBot: Michael Munday <munday@ca.ibm.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

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

@@ -13,6 +13,13 @@ TEXT main·foo(SB),7,$16-0 // TEXT main.foo(SB), 7, $16-0
 	MOVDBR	R1, R2                // b90f0021
 	MOVWBR	R3, R4                // b91f0043
 
+	MOVDEQ	R0, R1                // b9e28010
+	MOVDGE	R2, R3                // b9e2a032
+	MOVDGT	R4, R5                // b9e22054
+	MOVDLE	R6, R7                // b9e2c076
+	MOVDLT	R8, R9                // b9e24098
+	MOVDNE	R10, R11              // b9e270ba
+
 	MOVD	(R15), R1             // e310f0000004
 	MOVW	(R15), R2             // e320f0000014
 	MOVH	(R15), R3             // e330f0000015
@@ -50,6 +57,10 @@ TEXT main·foo(SB),7,$16-0 // TEXT main.foo(SB), 7, $16-0
 	ADDC	R1, R2                // b9ea1022
 	ADDC	$1, R1, R2            // b9040021c22a00000001
 	ADDC	R1, R2, R3            // b9ea1032
+	ADDW	R1, R2                // 1a21
+	ADDW	R1, R2, R3            // b9f81032
+	ADDW	$8192, R1             // c21900002000
+	ADDW	$8192, R1, R2         // ec21200000d8
 	SUB	R3, R4                // b9090043
 	SUB	R3, R4, R5            // b9e93054
 	SUB	$8192, R3             // c238ffffe000
@@ -57,6 +68,10 @@ TEXT main·foo(SB),7,$16-0 // TEXT main.foo(SB), 7, $16-0
 	SUBC	R1, R2                // b90b0021
 	SUBC	$1, R1, R2            // b9040021c22affffffff
 	SUBC	R2, R3, R4            // b9eb2043
+	SUBW	R3, R4                // 1b43
+	SUBW	R3, R4, R5            // b9f93054
+	SUBW	$8192, R1             // c21500002000
+	SUBW	$8192, R1, R2         // 1821c22500002000
 	MULLW	R6, R7                // b91c0076
 	MULLW	R6, R7, R8            // b9040087b91c0086
 	MULLW	$8192, R6             // a76d2000
@@ -73,10 +88,22 @@ TEXT main·foo(SB),7,$16-0 // TEXT main.foo(SB), 7, $16-0
 	DIVD	R1, R2, R3            // b90400b2b90d00a1b904003b
 	DIVW	R4, R5                // b90400b5b91d00a4b904005b
 	DIVW	R4, R5, R6            // b90400b5b91d00a4b904006b
-	DIVDU	R7, R8                // b90400a0b90400b8b98700a7b904008b
-	DIVDU	R7, R8, R9            // b90400a0b90400b8b98700a7b904009b
-	DIVWU	R1, R2                // b90400a0b90400b2b99700a1b904002b
-	DIVWU	R1, R2, R3            // b90400a0b90400b2b99700a1b904003b
+	DIVDU	R7, R8                // a7a90000b90400b8b98700a7b904008b
+	DIVDU	R7, R8, R9            // a7a90000b90400b8b98700a7b904009b
+	DIVWU	R1, R2                // a7a90000b90400b2b99700a1b904002b
+	DIVWU	R1, R2, R3            // a7a90000b90400b2b99700a1b904003b
+	MODD	R1, R2                // b90400b2b90d00a1b904002a
+	MODD	R1, R2, R3            // b90400b2b90d00a1b904003a
+	MODW	R4, R5                // b90400b5b91d00a4b904005a
+	MODW	R4, R5, R6            // b90400b5b91d00a4b904006a
+	MODDU	R7, R8                // a7a90000b90400b8b98700a7b904008a
+	MODDU	R7, R8, R9            // a7a90000b90400b8b98700a7b904009a
+	MODWU	R1, R2                // a7a90000b90400b2b99700a1b904002a
+	MODWU	R1, R2, R3            // a7a90000b90400b2b99700a1b904003a
+	NEG	R1                    // b9030011
+	NEG	R1, R2                // b9030021
+	NEGW	R1                    // b9130011
+	NEGW	R1, R2                // b9130021
 
 	LAA	R1, R2, 524287(R3)    // eb213fff7ff8
 	LAAG	R4, R5, -524288(R6)   // eb54600080e8

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

@@ -212,10 +212,15 @@ const (
 	AADDME
 	AADDE
 	AADDZE
+	AADDW
 	ADIVW
 	ADIVWU
 	ADIVD
 	ADIVDU
+	AMODW
+	AMODWU
+	AMODD
+	AMODDU
 	AMULLW
 	AMULLD
 	AMULHD
@@ -226,7 +231,9 @@ const (
 	ASUBV
 	ASUBE
 	ASUBZE
+	ASUBW
 	ANEG
+	ANEGW
 
 	// integer moves
 	AMOVWBR
@@ -240,6 +247,14 @@ const (
 	AMOVD
 	AMOVDBR
 
+	// conditional moves
+	AMOVDEQ
+	AMOVDGE
+	AMOVDGT
+	AMOVDLE
+	AMOVDLT
+	AMOVDNE
+
 	// integer bitwise
 	AAND
 	AANDN
@@ -276,6 +291,7 @@ const (
 	AFMULS
 	AFNABS
 	AFNEG
+	AFNEGS
 	AFNMADD
 	AFNMADDS
 	AFNMSUB
@@ -334,6 +350,8 @@ const (
 	ABGT
 	ABLE
 	ABLT
+	ABLEU
+	ABLTU
 	ABNE
 	ABVC
 	ABVS

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

@@ -11,10 +11,15 @@ var Anames = []string{
 	"ADDME",
 	"ADDE",
 	"ADDZE",
+	"ADDW",
 	"DIVW",
 	"DIVWU",
 	"DIVD",
 	"DIVDU",
+	"MODW",
+	"MODWU",
+	"MODD",
+	"MODDU",
 	"MULLW",
 	"MULLD",
 	"MULHD",
@@ -25,7 +30,9 @@ var Anames = []string{
 	"SUBV",
 	"SUBE",
 	"SUBZE",
+	"SUBW",
 	"NEG",
+	"NEGW",
 	"MOVWBR",
 	"MOVB",
 	"MOVBZ",
@@ -36,6 +43,12 @@ var Anames = []string{
 	"MOVWZ",
 	"MOVD",
 	"MOVDBR",
+	"MOVDEQ",
+	"MOVDGE",
+	"MOVDGT",
+	"MOVDLE",
+	"MOVDLT",
+	"MOVDNE",
 	"AND",
 	"ANDN",
 	"NAND",
@@ -69,6 +82,7 @@ var Anames = []string{
 	"FMULS",
 	"FNABS",
 	"FNEG",
+	"FNEGS",
 	"FNMADD",
 	"FNMADDS",
 	"FNMSUB",
@@ -111,6 +125,8 @@ var Anames = []string{
 	"BGT",
 	"BLE",
 	"BLT",
+	"BLEU",
+	"BLTU",
 	"BNE",
 	"BVC",
 	"BVS",

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

@@ -109,13 +109,13 @@ func progedit(ctxt *obj.Link, p *obj.Prog) {
 	// Rewrite SUB constants into ADD.
 	switch p.As {
 	case ASUBC:
-		if p.From.Type == obj.TYPE_CONST {
+		if p.From.Type == obj.TYPE_CONST && isint32(-p.From.Offset) {
 			p.From.Offset = -p.From.Offset
 			p.As = AADDC
 		}
 
 	case ASUB:
-		if p.From.Type == obj.TYPE_CONST {
+		if p.From.Type == obj.TYPE_CONST && isint32(-p.From.Offset) {
 			p.From.Offset = -p.From.Offset
 			p.As = AADD
 		}
@@ -258,14 +258,6 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym) {
 				p.Link.Mark |= LABEL
 			}
 
-		case ANOR:
-			q = p
-			if p.To.Type == obj.TYPE_REG {
-				if p.To.Reg == REGZERO {
-					p.Mark |= LABEL | SYNC
-				}
-			}
-
 		case ASYNC,
 			AWORD:
 			q = p
@@ -312,6 +304,8 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym) {
 			ABGT,
 			ABLE,
 			ABLT,
+			ABLEU,
+			ABLTU,
 			ABNE,
 			ABR,
 			ABVC,
@@ -401,7 +395,7 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym) {
 
 			q = p
 
-			if p.From3.Offset&obj.NOSPLIT == 0 {
+			if p.From3.Offset&obj.NOSPLIT == 0 && p.From3.Offset&obj.NOFRAME == 0 {
 				p, pPreempt = stacksplitPre(ctxt, p, autosize) // emit pre part of split check
 				pPre = p
 				wasSplit = true //need post part of split
@@ -440,7 +434,7 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym) {
 				// if(g->panic != nil && g->panic->argp == FP) g->panic->argp = bottom-of-frame
 				//
 				//	MOVD g_panic(g), R3
-				//	CMP R0, R3
+				//	CMP R3, $0
 				//	BEQ end
 				//	MOVD panic_argp(R3), R4
 				//	ADD $(autosize+8), R1, R5
@@ -466,9 +460,9 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym) {
 				q = obj.Appendp(ctxt, q)
 				q.As = ACMP
 				q.From.Type = obj.TYPE_REG
-				q.From.Reg = REG_R0
-				q.To.Type = obj.TYPE_REG
-				q.To.Reg = REG_R3
+				q.From.Reg = REG_R3
+				q.To.Type = obj.TYPE_CONST
+				q.To.Offset = 0
 
 				q = obj.Appendp(ctxt, q)
 				q.As = ABEQ
@@ -993,6 +987,7 @@ var unaryDst = map[obj.As]bool{
 	ASTCKE: true,
 	ASTCKF: true,
 	ANEG:   true,
+	ANEGW:  true,
 	AVONE:  true,
 	AVZERO: true,
 }