cmd/compile, cmd/internal/obj/ppc64: make math.Round an intrinsic on ppc64x

This change implements math.Round as an intrinsic on ppc64x so it can be
done using a single instruction.

benchmark                   old ns/op     new ns/op     delta
BenchmarkRound-16           2.60          0.69          -73.46%

Change-Id: I9408363e96201abdfc73ced7bcd5f0c29db006a8
Reviewed-on: https://go-review.googlesource.com/109395
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>

10 files changed, 43 insertions, 2 deletions

diff --git a/src/cmd/compile/internal/gc/ssa.go b/src/cmd/compile/internal/gc/ssa.go
index a3d2230964..6d19e47d38 100644
--- a/src/cmd/compile/internal/gc/ssa.go
+++ b/src/cmd/compile/internal/gc/ssa.go
@@ -3005,7 +3005,7 @@ func init() {
 		func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
 			return s.newValue1(ssa.OpRound, types.Types[TFLOAT64], args[0])
 		},
-		sys.ARM64, sys.S390X)
+		sys.ARM64, sys.PPC64, sys.S390X)
 	addF("math", "RoundToEven",
 		func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
 			return s.newValue1(ssa.OpRoundToEven, types.Types[TFLOAT64], args[0])
diff --git a/src/cmd/compile/internal/ppc64/ssa.go b/src/cmd/compile/internal/ppc64/ssa.go
index 3c10149eab..4c91a04a03 100644
--- a/src/cmd/compile/internal/ppc64/ssa.go
+++ b/src/cmd/compile/internal/ppc64/ssa.go
@@ -581,7 +581,7 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		p.To.Type = obj.TYPE_REG
 		p.To.Reg = ppc64.REGTMP // Ignored; this is for the carry effect.
 
-	case ssa.OpPPC64NEG, ssa.OpPPC64FNEG, ssa.OpPPC64FSQRT, ssa.OpPPC64FSQRTS, ssa.OpPPC64FFLOOR, ssa.OpPPC64FTRUNC, ssa.OpPPC64FCEIL, ssa.OpPPC64FCTIDZ, ssa.OpPPC64FCTIWZ, ssa.OpPPC64FCFID, ssa.OpPPC64FCFIDS, ssa.OpPPC64FRSP, ssa.OpPPC64CNTLZD, ssa.OpPPC64CNTLZW, ssa.OpPPC64POPCNTD, ssa.OpPPC64POPCNTW, ssa.OpPPC64POPCNTB, ssa.OpPPC64MFVSRD, ssa.OpPPC64MTVSRD, ssa.OpPPC64FABS, ssa.OpPPC64FNABS:
+	case ssa.OpPPC64NEG, ssa.OpPPC64FNEG, ssa.OpPPC64FSQRT, ssa.OpPPC64FSQRTS, ssa.OpPPC64FFLOOR, ssa.OpPPC64FTRUNC, ssa.OpPPC64FCEIL, ssa.OpPPC64FCTIDZ, ssa.OpPPC64FCTIWZ, ssa.OpPPC64FCFID, ssa.OpPPC64FCFIDS, ssa.OpPPC64FRSP, ssa.OpPPC64CNTLZD, ssa.OpPPC64CNTLZW, ssa.OpPPC64POPCNTD, ssa.OpPPC64POPCNTW, ssa.OpPPC64POPCNTB, ssa.OpPPC64MFVSRD, ssa.OpPPC64MTVSRD, ssa.OpPPC64FABS, ssa.OpPPC64FNABS, ssa.OpPPC64FROUND:
 		r := v.Reg()
 		p := s.Prog(v.Op.Asm())
 		p.To.Type = obj.TYPE_REG
diff --git a/src/cmd/compile/internal/ssa/gen/PPC64.rules b/src/cmd/compile/internal/ssa/gen/PPC64.rules
index 6f3e893d8d..5d416151ee 100644
--- a/src/cmd/compile/internal/ssa/gen/PPC64.rules
+++ b/src/cmd/compile/internal/ssa/gen/PPC64.rules
@@ -63,6 +63,7 @@
 (Floor x) -> (FFLOOR x)
 (Ceil x) -> (FCEIL x)
 (Trunc x) -> (FTRUNC x)
+(Round x) -> (FROUND x)
 (Copysign x y) -> (FCPSGN y x)
 (Abs x) -> (FABS x)
 
diff --git a/src/cmd/compile/internal/ssa/gen/PPC64Ops.go b/src/cmd/compile/internal/ssa/gen/PPC64Ops.go
index 567e34ec2a..ad68794de0 100644
--- a/src/cmd/compile/internal/ssa/gen/PPC64Ops.go
+++ b/src/cmd/compile/internal/ssa/gen/PPC64Ops.go
@@ -249,6 +249,7 @@ func init() {
 		{name: "FFLOOR", argLength: 1, reg: fp11, asm: "FRIM"},                              // floor(arg0), float64
 		{name: "FCEIL", argLength: 1, reg: fp11, asm: "FRIP"},                               // ceil(arg0), float64
 		{name: "FTRUNC", argLength: 1, reg: fp11, asm: "FRIZ"},                              // trunc(arg0), float64
+		{name: "FROUND", argLength: 1, reg: fp11, asm: "FRIN"},                              // round(arg0), float64
 		{name: "FABS", argLength: 1, reg: fp11, asm: "FABS"},                                // abs(arg0), float64
 		{name: "FNABS", argLength: 1, reg: fp11, asm: "FNABS"},                              // -abs(arg0), float64
 		{name: "FCPSGN", argLength: 2, reg: fp21, asm: "FCPSGN"},                            // copysign arg0 -> arg1, float64
diff --git a/src/cmd/compile/internal/ssa/opGen.go b/src/cmd/compile/internal/ssa/opGen.go
index aea2246e84..de04eacfa8 100644
--- a/src/cmd/compile/internal/ssa/opGen.go
+++ b/src/cmd/compile/internal/ssa/opGen.go
@@ -1541,6 +1541,7 @@ const (
 	OpPPC64FFLOOR
 	OpPPC64FCEIL
 	OpPPC64FTRUNC
+	OpPPC64FROUND
 	OpPPC64FABS
 	OpPPC64FNABS
 	OpPPC64FCPSGN
@@ -20297,6 +20298,19 @@ var opcodeTable = [...]opInfo{
 		},
 	},
 	{
+		name:   "FROUND",
+		argLen: 1,
+		asm:    ppc64.AFRIN,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 576460743713488896}, // F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12 F13 F14 F15 F16 F17 F18 F19 F20 F21 F22 F23 F24 F25 F26
+			},
+			outputs: []outputInfo{
+				{0, 576460743713488896}, // F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12 F13 F14 F15 F16 F17 F18 F19 F20 F21 F22 F23 F24 F25 F26
+			},
+		},
+	},
+	{
 		name:   "FABS",
 		argLen: 1,
 		asm:    ppc64.AFABS,
diff --git a/src/cmd/compile/internal/ssa/rewritePPC64.go b/src/cmd/compile/internal/ssa/rewritePPC64.go
index 331a8c9232..8610f08e7f 100644
--- a/src/cmd/compile/internal/ssa/rewritePPC64.go
+++ b/src/cmd/compile/internal/ssa/rewritePPC64.go
@@ -505,6 +505,8 @@ func rewriteValuePPC64(v *Value) bool {
 		return rewriteValuePPC64_OpPopCount64_0(v)
 	case OpPopCount8:
 		return rewriteValuePPC64_OpPopCount8_0(v)
+	case OpRound:
+		return rewriteValuePPC64_OpRound_0(v)
 	case OpRound32F:
 		return rewriteValuePPC64_OpRound32F_0(v)
 	case OpRound64F:
@@ -13466,6 +13468,15 @@ func rewriteValuePPC64_OpPopCount8_0(v *Value) bool {
 		return true
 	}
 }
+func rewriteValuePPC64_OpRound_0(v *Value) bool {
+	// match: (Round x)
+	// cond:
+	// result: (FROUND x)
+	x := v.Args[0]
+	v.reset(OpPPC64FROUND)
+	v.AddArg(x)
+	return true
+}
 func rewriteValuePPC64_OpRound32F_0(v *Value) bool {
 	// match: (Round32F x)
 	// cond:
diff --git a/src/cmd/internal/obj/ppc64/a.out.go b/src/cmd/internal/obj/ppc64/a.out.go
index 8f56da0763..3c374579ec 100644
--- a/src/cmd/internal/obj/ppc64/a.out.go
+++ b/src/cmd/internal/obj/ppc64/a.out.go
@@ -643,6 +643,8 @@ const (
 	AFRIPCC
 	AFRIZ
 	AFRIZCC
+	AFRIN
+	AFRINCC
 	AFRSQRTE
 	AFRSQRTECC
 	AFSEL
diff --git a/src/cmd/internal/obj/ppc64/anames.go b/src/cmd/internal/obj/ppc64/anames.go
index 6006f15ad6..16a27591c7 100644
--- a/src/cmd/internal/obj/ppc64/anames.go
+++ b/src/cmd/internal/obj/ppc64/anames.go
@@ -245,6 +245,8 @@ var Anames = []string{
 	"FRIPCC",
 	"FRIZ",
 	"FRIZCC",
+	"FRIN",
+	"FRINCC",
 	"FRSQRTE",
 	"FRSQRTECC",
 	"FSEL",
diff --git a/src/cmd/internal/obj/ppc64/asm9.go b/src/cmd/internal/obj/ppc64/asm9.go
index 4dcb52f9ee..72738f1967 100644
--- a/src/cmd/internal/obj/ppc64/asm9.go
+++ b/src/cmd/internal/obj/ppc64/asm9.go
@@ -1671,6 +1671,8 @@ func buildop(ctxt *obj.Link) {
 			opset(AFRIPCC, r0)
 			opset(AFRIZ, r0)
 			opset(AFRIZCC, r0)
+			opset(AFRIN, r0)
+			opset(AFRINCC, r0)
 			opset(AFRSQRTE, r0)
 			opset(AFRSQRTECC, r0)
 			opset(AFSQRT, r0)
@@ -3898,6 +3900,10 @@ func (c *ctxt9) oprrr(a obj.As) uint32 {
 		return OPVCC(63, 424, 0, 0)
 	case AFRIZCC:
 		return OPVCC(63, 424, 0, 1)
+	case AFRIN:
+		return OPVCC(63, 392, 0, 0)
+	case AFRINCC:
+		return OPVCC(63, 392, 0, 1)
 	case AFRSP:
 		return OPVCC(63, 12, 0, 0)
 	case AFRSPCC:
diff --git a/test/codegen/math.go b/test/codegen/math.go
index efa3a2bc8f..f73321200b 100644
--- a/test/codegen/math.go
+++ b/test/codegen/math.go
@@ -13,18 +13,22 @@ var sink64 [8]float64
 func approx(x float64) {
 	// s390x:"FIDBR\t[$]6"
 	// arm64:"FRINTPD"
+	// ppc64le:"FRIP"
 	sink64[0] = math.Ceil(x)
 
 	// s390x:"FIDBR\t[$]7"
 	// arm64:"FRINTMD"
+	// ppc64le:"FRIM"
 	sink64[1] = math.Floor(x)
 
 	// s390x:"FIDBR\t[$]1"
 	// arm64:"FRINTAD"
+	// ppc64le:"FRIN"
 	sink64[2] = math.Round(x)
 
 	// s390x:"FIDBR\t[$]5"
 	// arm64:"FRINTZD"
+	// ppc64le:"FRIZ"
 	sink64[3] = math.Trunc(x)
 
 	// s390x:"FIDBR\t[$]4"