cmd/compile: use a different register for updated value in AtomicAnd8/Or8 on ARM64

ARM64 manual says it is "constrained unpredictable" if the src
and dst registers of STLXRB are same, although it doesn't seem
to cause any problem on real hardwares so far. Fix by allocating
a different register to hold the updated value for
AtomicAnd8/Or8. We do this by making the ops returns <val,mem>
like AtomicAdd, although val will not be used elsewhere.

Fixes #25823.

Change-Id: I735b9822f99877b3c7aee67a65e62b7278dc40df
Reviewed-on: https://go-review.googlesource.com/117976
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Wei Xiao <Wei.Xiao@arm.com>

src/cmd/compile/internal/arm64/ssa.go

@@ -603,25 +603,26 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		gc.Patch(p2, p5)
 	case ssa.OpARM64LoweredAtomicAnd8,
 		ssa.OpARM64LoweredAtomicOr8:
-		// LDAXRB	(Rarg0), Rtmp
-		// AND/OR	Rarg1, Rtmp
-		// STLXRB	Rtmp, (Rarg0), Rtmp
+		// LDAXRB	(Rarg0), Rout
+		// AND/OR	Rarg1, Rout
+		// STLXRB	Rout, (Rarg0), Rtmp
 		// CBNZ		Rtmp, -3(PC)
 		r0 := v.Args[0].Reg()
 		r1 := v.Args[1].Reg()
+		out := v.Reg0()
 		p := s.Prog(arm64.ALDAXRB)
 		p.From.Type = obj.TYPE_MEM
 		p.From.Reg = r0
 		p.To.Type = obj.TYPE_REG
-		p.To.Reg = arm64.REGTMP
+		p.To.Reg = out
 		p1 := s.Prog(v.Op.Asm())
 		p1.From.Type = obj.TYPE_REG
 		p1.From.Reg = r1
 		p1.To.Type = obj.TYPE_REG
-		p1.To.Reg = arm64.REGTMP
+		p1.To.Reg = out
 		p2 := s.Prog(arm64.ASTLXRB)
 		p2.From.Type = obj.TYPE_REG
-		p2.From.Reg = arm64.REGTMP
+		p2.From.Reg = out
 		p2.To.Type = obj.TYPE_MEM
 		p2.To.Reg = r0
 		p2.RegTo2 = arm64.REGTMP

src/cmd/compile/internal/ssa/gen/ARM64.rules

@@ -540,8 +540,9 @@
 (AtomicCompareAndSwap32 ptr old new_ mem) -> (LoweredAtomicCas32 ptr old new_ mem)
 (AtomicCompareAndSwap64 ptr old new_ mem) -> (LoweredAtomicCas64 ptr old new_ mem)
 
-(AtomicAnd8 ptr val mem) -> (LoweredAtomicAnd8 ptr val mem)
-(AtomicOr8  ptr val mem) -> (LoweredAtomicOr8  ptr val mem)
+// Currently the updated value is not used, but we need a register to temporarily hold it.
+(AtomicAnd8 ptr val mem) -> (Select1 (LoweredAtomicAnd8 ptr val mem))
+(AtomicOr8  ptr val mem) -> (Select1 (LoweredAtomicOr8  ptr val mem))
 
 // Write barrier.
 (WB {fn} destptr srcptr mem) -> (LoweredWB {fn} destptr srcptr mem)

src/cmd/compile/internal/ssa/gen/ARM64Ops.go

@@ -596,13 +596,13 @@ func init() {
 		{name: "LoweredAtomicCas32", argLength: 4, reg: gpcas, resultNotInArgs: true, clobberFlags: true, faultOnNilArg0: true, hasSideEffects: true},
 
 		// atomic and/or.
-		// *arg0 &= (|=) arg1. arg2=mem. returns memory. auxint must be zero.
-		// LDAXRB	(Rarg0), Rtmp
-		// AND/OR	Rarg1, Rtmp
-		// STLXRB	Rtmp, (Rarg0), Rtmp
+		// *arg0 &= (|=) arg1. arg2=mem. returns <new content of *arg0, memory>. auxint must be zero.
+		// LDAXRB	(Rarg0), Rout
+		// AND/OR	Rarg1, Rout
+		// STLXRB	Rout, (Rarg0), Rtmp
 		// CBNZ		Rtmp, -3(PC)
-		{name: "LoweredAtomicAnd8", argLength: 3, reg: gpstore, asm: "AND", faultOnNilArg0: true, hasSideEffects: true},
-		{name: "LoweredAtomicOr8", argLength: 3, reg: gpstore, asm: "ORR", faultOnNilArg0: true, hasSideEffects: true},
+		{name: "LoweredAtomicAnd8", argLength: 3, reg: gpxchg, resultNotInArgs: true, asm: "AND", typ: "(UInt8,Mem)", faultOnNilArg0: true, hasSideEffects: true},
+		{name: "LoweredAtomicOr8", argLength: 3, reg: gpxchg, resultNotInArgs: true, asm: "ORR", typ: "(UInt8,Mem)", faultOnNilArg0: true, hasSideEffects: true},
 
 		// LoweredWB invokes runtime.gcWriteBarrier. arg0=destptr, arg1=srcptr, arg2=mem, aux=runtime.gcWriteBarrier
 		// It saves all GP registers if necessary,

src/cmd/compile/internal/ssa/opGen.go

@@ -16761,6 +16761,7 @@ var opcodeTable = [...]opInfo{
 	{
 		name:            "LoweredAtomicAnd8",
 		argLen:          3,
+		resultNotInArgs: true,
 		faultOnNilArg0:  true,
 		hasSideEffects:  true,
 		asm:             arm64.AAND,
@@ -16769,11 +16770,15 @@ var opcodeTable = [...]opInfo{
 				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
 				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
 			},
+			outputs: []outputInfo{
+				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
+			},
 		},
 	},
 	{
 		name:            "LoweredAtomicOr8",
 		argLen:          3,
+		resultNotInArgs: true,
 		faultOnNilArg0:  true,
 		hasSideEffects:  true,
 		asm:             arm64.AORR,
@@ -16782,6 +16787,9 @@ var opcodeTable = [...]opInfo{
 				{1, 805044223},           // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30
 				{0, 9223372038733561855}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 g R30 SP SB
 			},
+			outputs: []outputInfo{
+				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
+			},
 		},
 	},
 	{

src/cmd/compile/internal/ssa/rewriteARM64.go

@@ -25925,18 +25925,24 @@ func rewriteValueARM64_OpAtomicAdd64_0(v *Value) bool {
 	}
 }
 func rewriteValueARM64_OpAtomicAnd8_0(v *Value) bool {
+	b := v.Block
+	_ = b
+	typ := &b.Func.Config.Types
+	_ = typ
 	// match: (AtomicAnd8 ptr val mem)
 	// cond:
-	// result: (LoweredAtomicAnd8 ptr val mem)
+	// result: (Select1 (LoweredAtomicAnd8 ptr val mem))
 	for {
 		_ = v.Args[2]
 		ptr := v.Args[0]
 		val := v.Args[1]
 		mem := v.Args[2]
-		v.reset(OpARM64LoweredAtomicAnd8)
-		v.AddArg(ptr)
-		v.AddArg(val)
-		v.AddArg(mem)
+		v.reset(OpSelect1)
+		v0 := b.NewValue0(v.Pos, OpARM64LoweredAtomicAnd8, types.NewTuple(typ.UInt8, types.TypeMem))
+		v0.AddArg(ptr)
+		v0.AddArg(val)
+		v0.AddArg(mem)
+		v.AddArg(v0)
 		return true
 	}
 }
@@ -26051,18 +26057,24 @@ func rewriteValueARM64_OpAtomicLoadPtr_0(v *Value) bool {
 	}
 }
 func rewriteValueARM64_OpAtomicOr8_0(v *Value) bool {
+	b := v.Block
+	_ = b
+	typ := &b.Func.Config.Types
+	_ = typ
 	// match: (AtomicOr8 ptr val mem)
 	// cond:
-	// result: (LoweredAtomicOr8 ptr val mem)
+	// result: (Select1 (LoweredAtomicOr8 ptr val mem))
 	for {
 		_ = v.Args[2]
 		ptr := v.Args[0]
 		val := v.Args[1]
 		mem := v.Args[2]
-		v.reset(OpARM64LoweredAtomicOr8)
-		v.AddArg(ptr)
-		v.AddArg(val)
-		v.AddArg(mem)
+		v.reset(OpSelect1)
+		v0 := b.NewValue0(v.Pos, OpARM64LoweredAtomicOr8, types.NewTuple(typ.UInt8, types.TypeMem))
+		v0.AddArg(ptr)
+		v0.AddArg(val)
+		v0.AddArg(mem)
+		v.AddArg(v0)
 		return true
 	}
 }