cmd/compile: add MIPS64 optimizations, SSA on by default

Add the following optimizations:
- fold constants
- fold address into load/store
- simplify extensions and conditional branches
- remove nil checks

Turn on SSA on MIPS64 by default, and toggle the tests.

Fixes #16359.

Change-Id: I7f1e38c2509e22e42cd024e712990ebbe47176bd
Reviewed-on: https://go-review.googlesource.com/27870
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Reviewed-by: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

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

@@ -40,7 +40,7 @@ func shouldssa(fn *Node) bool {
 		if os.Getenv("SSATEST") == "" {
 			return false
 		}
-	case "amd64", "amd64p32", "arm", "386", "arm64", "ppc64le":
+	case "amd64", "amd64p32", "arm", "386", "arm64", "ppc64le", "mips64", "mips64le":
 		// Generally available.
 	}
 	if !ssaEnabled {

src/cmd/compile/internal/mips64/ssa.go

@@ -421,7 +421,32 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		ssa.OpMIPS64MOVHUreg,
 		ssa.OpMIPS64MOVWreg,
 		ssa.OpMIPS64MOVWUreg:
-		// TODO: remove extension if after proper load
+		a := v.Args[0]
+		for a.Op == ssa.OpCopy || a.Op == ssa.OpMIPS64MOVVreg {
+			a = a.Args[0]
+		}
+		if a.Op == ssa.OpLoadReg {
+			t := a.Type
+			switch {
+			case v.Op == ssa.OpMIPS64MOVBreg && t.Size() == 1 && t.IsSigned(),
+				v.Op == ssa.OpMIPS64MOVBUreg && t.Size() == 1 && !t.IsSigned(),
+				v.Op == ssa.OpMIPS64MOVHreg && t.Size() == 2 && t.IsSigned(),
+				v.Op == ssa.OpMIPS64MOVHUreg && t.Size() == 2 && !t.IsSigned(),
+				v.Op == ssa.OpMIPS64MOVWreg && t.Size() == 4 && t.IsSigned(),
+				v.Op == ssa.OpMIPS64MOVWUreg && t.Size() == 4 && !t.IsSigned():
+				// arg is a proper-typed load, already zero/sign-extended, don't extend again
+				if gc.SSARegNum(v) == gc.SSARegNum(v.Args[0]) {
+					return
+				}
+				p := gc.Prog(mips.AMOVV)
+				p.From.Type = obj.TYPE_REG
+				p.From.Reg = gc.SSARegNum(v.Args[0])
+				p.To.Type = obj.TYPE_REG
+				p.To.Reg = gc.SSARegNum(v)
+				return
+			default:
+			}
+		}
 		fallthrough
 	case ssa.OpMIPS64MOVWF,
 		ssa.OpMIPS64MOVWD,
@@ -613,7 +638,64 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 			gc.Maxarg = v.AuxInt
 		}
 	case ssa.OpMIPS64LoweredNilCheck:
-		// TODO: optimization
+		// Optimization - if the subsequent block has a load or store
+		// at the same address, we don't need to issue this instruction.
+		mem := v.Args[1]
+		for _, w := range v.Block.Succs[0].Block().Values {
+			if w.Op == ssa.OpPhi {
+				if w.Type.IsMemory() {
+					mem = w
+				}
+				continue
+			}
+			if len(w.Args) == 0 || !w.Args[len(w.Args)-1].Type.IsMemory() {
+				// w doesn't use a store - can't be a memory op.
+				continue
+			}
+			if w.Args[len(w.Args)-1] != mem {
+				v.Fatalf("wrong store after nilcheck v=%s w=%s", v, w)
+			}
+			switch w.Op {
+			case ssa.OpMIPS64MOVBload, ssa.OpMIPS64MOVBUload, ssa.OpMIPS64MOVHload, ssa.OpMIPS64MOVHUload,
+				ssa.OpMIPS64MOVWload, ssa.OpMIPS64MOVWUload, ssa.OpMIPS64MOVVload,
+				ssa.OpMIPS64MOVFload, ssa.OpMIPS64MOVDload,
+				ssa.OpMIPS64MOVBstore, ssa.OpMIPS64MOVHstore, ssa.OpMIPS64MOVWstore, ssa.OpMIPS64MOVVstore,
+				ssa.OpMIPS64MOVFstore, ssa.OpMIPS64MOVDstore:
+				// arg0 is ptr, auxint is offset
+				if w.Args[0] == v.Args[0] && w.Aux == nil && w.AuxInt >= 0 && w.AuxInt < minZeroPage {
+					if gc.Debug_checknil != 0 && int(v.Line) > 1 {
+						gc.Warnl(v.Line, "removed nil check")
+					}
+					return
+				}
+			case ssa.OpMIPS64DUFFZERO, ssa.OpMIPS64LoweredZero:
+				// arg0 is ptr
+				if w.Args[0] == v.Args[0] {
+					if gc.Debug_checknil != 0 && int(v.Line) > 1 {
+						gc.Warnl(v.Line, "removed nil check")
+					}
+					return
+				}
+			case ssa.OpMIPS64LoweredMove:
+				// arg0 is dst ptr, arg1 is src ptr
+				if w.Args[0] == v.Args[0] || w.Args[1] == v.Args[0] {
+					if gc.Debug_checknil != 0 && int(v.Line) > 1 {
+						gc.Warnl(v.Line, "removed nil check")
+					}
+					return
+				}
+			default:
+			}
+			if w.Type.IsMemory() {
+				if w.Op == ssa.OpVarDef || w.Op == ssa.OpVarKill || w.Op == ssa.OpVarLive {
+					// these ops are OK
+					mem = w
+					continue
+				}
+				// We can't delay the nil check past the next store.
+				break
+			}
+		}
 		// Issue a load which will fault if arg is nil.
 		p := gc.Prog(mips.AMOVB)
 		p.From.Type = obj.TYPE_MEM

src/cmd/compile/internal/ssa/gen/MIPS64Ops.go

@@ -201,9 +201,9 @@ func init() {
 
 		// comparisons
 		{name: "SGT", argLength: 2, reg: gp21, asm: "SGT", typ: "Bool"},                      // 1 if arg0 > arg1 (signed), 0 otherwise
-		{name: "SGTconst", argLength: 2, reg: gp21, asm: "SGT", aux: "Int64", typ: "Bool"},   // 1 if arg0 > auxInt (signed), 0 otherwise
+		{name: "SGTconst", argLength: 1, reg: gp11, asm: "SGT", aux: "Int64", typ: "Bool"},   // 1 if auxInt > arg0 (signed), 0 otherwise
 		{name: "SGTU", argLength: 2, reg: gp21, asm: "SGTU", typ: "Bool"},                    // 1 if arg0 > arg1 (unsigned), 0 otherwise
-		{name: "SGTUconst", argLength: 2, reg: gp21, asm: "SGTU", aux: "Int64", typ: "Bool"}, // 1 if arg0 > auxInt (unsigned), 0 otherwise
+		{name: "SGTUconst", argLength: 1, reg: gp11, asm: "SGTU", aux: "Int64", typ: "Bool"}, // 1 if auxInt > arg0 (unsigned), 0 otherwise
 
 		{name: "CMPEQF", argLength: 2, reg: fp2flags, asm: "CMPEQF", typ: "Flags"}, // flags=true if arg0 = arg1, float32
 		{name: "CMPEQD", argLength: 2, reg: fp2flags, asm: "CMPEQD", typ: "Flags"}, // flags=true if arg0 = arg1, float64

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

@@ -12421,12 +12421,11 @@ var opcodeTable = [...]opInfo{
 	{
 		name:    "SGTconst",
 		auxType: auxInt64,
-		argLen:  2,
+		argLen:  1,
 		asm:     mips.ASGT,
 		reg: regInfo{
 			inputs: []inputInfo{
 				{0, 100663294}, // R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R24 R25 g
-				{1, 100663294}, // R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R24 R25 g
 			},
 			outputs: []outputInfo{
 				{0, 33554430}, // R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R24 R25
@@ -12450,12 +12449,11 @@ var opcodeTable = [...]opInfo{
 	{
 		name:    "SGTUconst",
 		auxType: auxInt64,
-		argLen:  2,
+		argLen:  1,
 		asm:     mips.ASGTU,
 		reg: regInfo{
 			inputs: []inputInfo{
 				{0, 100663294}, // R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R24 R25 g
-				{1, 100663294}, // R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R24 R25 g
 			},
 			outputs: []outputInfo{
 				{0, 33554430}, // R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R24 R25

test/live.go

-// +build !amd64,!arm,!amd64p32,!386,!arm64,!ppc64le
+// +build !amd64,!arm,!amd64p32,!386,!arm64,!ppc64le,!mips64,!mips64le
 // errorcheck -0 -l -live -wb=0
 
 // Copyright 2014 The Go Authors. All rights reserved.

test/live_ssa.go

-// +build amd64 arm amd64p32 386 arm64
+// +build amd64 arm amd64p32 386 arm64 mips64 mips64le
 // errorcheck -0 -l -live -wb=0
 
 // Copyright 2014 The Go Authors. All rights reserved.

test/nilptr3_ssa.go

 // errorcheck -0 -d=nil
-// +build amd64 arm amd64p32 386 arm64
+// +build amd64 arm amd64p32 386 arm64 mips64 mips64le
 
 // Copyright 2013 The Go Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style

test/sliceopt.go

-// +build !amd64,!arm,!amd64p32,!386,!arm64,!ppc64le
+// +build !amd64,!arm,!amd64p32,!386,!arm64,!ppc64le,!mips64,!mips64le
 // errorcheck -0 -d=append,slice
 
 // Copyright 2015 The Go Authors. All rights reserved.