[dev.simd] simd, cmd/compile: add more element types for Select128FromPair

Also includes a comment cleanup pass.
Fixed NAME processing for additional documentation.

Change-Id: Ide5b60c17ddbf3c6eafd20147981c59493fc8133
Reviewed-on: https://go-review.googlesource.com/c/go/+/722180
Reviewed-by: Junyang Shao <shaojunyang@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

5 files changed, 52 insertions, 0 deletions

diff --git a/src/cmd/compile/internal/ssa/_gen/simdAMD64.rules b/src/cmd/compile/internal/ssa/_gen/simdAMD64.rules
index 283a2e53cd..db426f6615 100644
--- a/src/cmd/compile/internal/ssa/_gen/simdAMD64.rules
+++ b/src/cmd/compile/internal/ssa/_gen/simdAMD64.rules
@@ -941,8 +941,12 @@
 (ScaleFloat64x8 ...) => (VSCALEFPD512 ...)
 (Select128FromPairFloat32x8 ...) => (VPERM2F128256 ...)
 (Select128FromPairFloat64x4 ...) => (VPERM2F128256 ...)
+(Select128FromPairInt8x32 ...) => (VPERM2I128256 ...)
+(Select128FromPairInt16x16 ...) => (VPERM2I128256 ...)
 (Select128FromPairInt32x8 ...) => (VPERM2I128256 ...)
 (Select128FromPairInt64x4 ...) => (VPERM2I128256 ...)
+(Select128FromPairUint8x32 ...) => (VPERM2I128256 ...)
+(Select128FromPairUint16x16 ...) => (VPERM2I128256 ...)
 (Select128FromPairUint32x8 ...) => (VPERM2I128256 ...)
 (Select128FromPairUint64x4 ...) => (VPERM2I128256 ...)
 (SetElemFloat32x4 ...) => (VPINSRD128 ...)
diff --git a/src/cmd/compile/internal/ssa/_gen/simdgenericOps.go b/src/cmd/compile/internal/ssa/_gen/simdgenericOps.go
index 3fae158c0a..5683fcef0d 100644
--- a/src/cmd/compile/internal/ssa/_gen/simdgenericOps.go
+++ b/src/cmd/compile/internal/ssa/_gen/simdgenericOps.go
@@ -1192,8 +1192,12 @@ func simdGenericOps() []opData {
 		{name: "SHA1FourRoundsUint32x4", argLength: 2, commutative: false, aux: "UInt8"},
 		{name: "Select128FromPairFloat32x8", argLength: 2, commutative: false, aux: "UInt8"},
 		{name: "Select128FromPairFloat64x4", argLength: 2, commutative: false, aux: "UInt8"},
+		{name: "Select128FromPairInt8x32", argLength: 2, commutative: false, aux: "UInt8"},
+		{name: "Select128FromPairInt16x16", argLength: 2, commutative: false, aux: "UInt8"},
 		{name: "Select128FromPairInt32x8", argLength: 2, commutative: false, aux: "UInt8"},
 		{name: "Select128FromPairInt64x4", argLength: 2, commutative: false, aux: "UInt8"},
+		{name: "Select128FromPairUint8x32", argLength: 2, commutative: false, aux: "UInt8"},
+		{name: "Select128FromPairUint16x16", argLength: 2, commutative: false, aux: "UInt8"},
 		{name: "Select128FromPairUint32x8", argLength: 2, commutative: false, aux: "UInt8"},
 		{name: "Select128FromPairUint64x4", argLength: 2, commutative: false, aux: "UInt8"},
 		{name: "SetElemFloat32x4", argLength: 2, commutative: false, aux: "UInt8"},
diff --git a/src/cmd/compile/internal/ssa/opGen.go b/src/cmd/compile/internal/ssa/opGen.go
index fa94dfbbd5..bb40ff4117 100644
--- a/src/cmd/compile/internal/ssa/opGen.go
+++ b/src/cmd/compile/internal/ssa/opGen.go
@@ -7151,8 +7151,12 @@ const (
 	OpSHA1FourRoundsUint32x4
 	OpSelect128FromPairFloat32x8
 	OpSelect128FromPairFloat64x4
+	OpSelect128FromPairInt8x32
+	OpSelect128FromPairInt16x16
 	OpSelect128FromPairInt32x8
 	OpSelect128FromPairInt64x4
+	OpSelect128FromPairUint8x32
+	OpSelect128FromPairUint16x16
 	OpSelect128FromPairUint32x8
 	OpSelect128FromPairUint64x4
 	OpSetElemFloat32x4
@@ -92251,6 +92255,18 @@ var opcodeTable = [...]opInfo{
 		generic: true,
 	},
 	{
+		name:    "Select128FromPairInt8x32",
+		auxType: auxUInt8,
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "Select128FromPairInt16x16",
+		auxType: auxUInt8,
+		argLen:  2,
+		generic: true,
+	},
+	{
 		name:    "Select128FromPairInt32x8",
 		auxType: auxUInt8,
 		argLen:  2,
@@ -92263,6 +92279,18 @@ var opcodeTable = [...]opInfo{
 		generic: true,
 	},
 	{
+		name:    "Select128FromPairUint8x32",
+		auxType: auxUInt8,
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "Select128FromPairUint16x16",
+		auxType: auxUInt8,
+		argLen:  2,
+		generic: true,
+	},
+	{
 		name:    "Select128FromPairUint32x8",
 		auxType: auxUInt8,
 		argLen:  2,
diff --git a/src/cmd/compile/internal/ssa/rewriteAMD64.go b/src/cmd/compile/internal/ssa/rewriteAMD64.go
index 5ad2ed3f96..c7995c5c9e 100644
--- a/src/cmd/compile/internal/ssa/rewriteAMD64.go
+++ b/src/cmd/compile/internal/ssa/rewriteAMD64.go
@@ -5017,18 +5017,30 @@ func rewriteValueAMD64(v *Value) bool {
 	case OpSelect128FromPairFloat64x4:
 		v.Op = OpAMD64VPERM2F128256
 		return true
+	case OpSelect128FromPairInt16x16:
+		v.Op = OpAMD64VPERM2I128256
+		return true
 	case OpSelect128FromPairInt32x8:
 		v.Op = OpAMD64VPERM2I128256
 		return true
 	case OpSelect128FromPairInt64x4:
 		v.Op = OpAMD64VPERM2I128256
 		return true
+	case OpSelect128FromPairInt8x32:
+		v.Op = OpAMD64VPERM2I128256
+		return true
+	case OpSelect128FromPairUint16x16:
+		v.Op = OpAMD64VPERM2I128256
+		return true
 	case OpSelect128FromPairUint32x8:
 		v.Op = OpAMD64VPERM2I128256
 		return true
 	case OpSelect128FromPairUint64x4:
 		v.Op = OpAMD64VPERM2I128256
 		return true
+	case OpSelect128FromPairUint8x32:
+		v.Op = OpAMD64VPERM2I128256
+		return true
 	case OpSelectN:
 		return rewriteValueAMD64_OpSelectN(v)
 	case OpSetElemFloat32x4:
diff --git a/src/cmd/compile/internal/ssagen/simdintrinsics.go b/src/cmd/compile/internal/ssagen/simdintrinsics.go
index 34e491371e..413cf92c88 100644
--- a/src/cmd/compile/internal/ssagen/simdintrinsics.go
+++ b/src/cmd/compile/internal/ssagen/simdintrinsics.go
@@ -953,8 +953,12 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Float64x8.Scale", opLen2(ssa.OpScaleFloat64x8, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Float32x8.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairFloat32x8, types.TypeVec256, 0), sys.AMD64)
 	addF(simdPackage, "Float64x4.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairFloat64x4, types.TypeVec256, 0), sys.AMD64)
+	addF(simdPackage, "Int8x32.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairInt8x32, types.TypeVec256, 0), sys.AMD64)
+	addF(simdPackage, "Int16x16.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairInt16x16, types.TypeVec256, 0), sys.AMD64)
 	addF(simdPackage, "Int32x8.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairInt32x8, types.TypeVec256, 0), sys.AMD64)
 	addF(simdPackage, "Int64x4.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairInt64x4, types.TypeVec256, 0), sys.AMD64)
+	addF(simdPackage, "Uint8x32.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairUint8x32, types.TypeVec256, 0), sys.AMD64)
+	addF(simdPackage, "Uint16x16.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairUint16x16, types.TypeVec256, 0), sys.AMD64)
 	addF(simdPackage, "Uint32x8.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairUint32x8, types.TypeVec256, 0), sys.AMD64)
 	addF(simdPackage, "Uint64x4.Select128FromPair", opLen2Imm8_II(ssa.OpSelect128FromPairUint64x4, types.TypeVec256, 0), sys.AMD64)
 	addF(simdPackage, "Float32x4.SetElem", opLen2Imm8(ssa.OpSetElemFloat32x4, types.TypeVec128, 0), sys.AMD64)