[dev.simd] simd, cmd/compile: implement ToMask, unexport asMask.

This CL defines the mask semantic better:
When converting from vector to mask, its element is set to true iff
the corresponding vector element is non zero.

Change-Id: I331c1c7992dc9e81c211bdc6d73e5eb3b8414506
Reviewed-on: https://go-review.googlesource.com/c/go/+/697056
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

1 files changed, 12 insertions, 12 deletions

diff --git a/src/cmd/compile/internal/ssagen/simdintrinsics.go b/src/cmd/compile/internal/ssagen/simdintrinsics.go
index be3d917f8f..90149300b2 100644
--- a/src/cmd/compile/internal/ssagen/simdintrinsics.go
+++ b/src/cmd/compile/internal/ssagen/simdintrinsics.go
@@ -2299,7 +2299,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "LoadMaskedMask64x8", simdMaskedLoad(ssa.OpLoadMasked64), sys.AMD64)
 	addF(simdPackage, "Mask64x8.StoreMasked", simdMaskedStore(ssa.OpStoreMasked64), sys.AMD64)
 	addF(simdPackage, "Mask8x16.AsInt8x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int8x16.AsMask8x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int8x16.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask8x16.And", opLen2(ssa.OpAndInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Mask8x16.Or", opLen2(ssa.OpOrInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "LoadMask8x16FromBits", simdLoadMask(8, 16), sys.AMD64)
@@ -2307,7 +2307,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask8x16FromBits", simdCvtVToMask(8, 16), sys.AMD64)
 	addF(simdPackage, "Mask8x16.ToBits", simdCvtMaskToV(8, 16), sys.AMD64)
 	addF(simdPackage, "Mask8x32.AsInt8x32", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int8x32.AsMask8x32", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int8x32.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask8x32.And", opLen2(ssa.OpAndInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Mask8x32.Or", opLen2(ssa.OpOrInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "LoadMask8x32FromBits", simdLoadMask(8, 32), sys.AMD64)
@@ -2315,7 +2315,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask8x32FromBits", simdCvtVToMask(8, 32), sys.AMD64)
 	addF(simdPackage, "Mask8x32.ToBits", simdCvtMaskToV(8, 32), sys.AMD64)
 	addF(simdPackage, "Mask8x64.AsInt8x64", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int8x64.AsMask8x64", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int8x64.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask8x64.And", opLen2(ssa.OpAndInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Mask8x64.Or", opLen2(ssa.OpOrInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "LoadMask8x64FromBits", simdLoadMask(8, 64), sys.AMD64)
@@ -2323,7 +2323,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask8x64FromBits", simdCvtVToMask(8, 64), sys.AMD64)
 	addF(simdPackage, "Mask8x64.ToBits", simdCvtMaskToV(8, 64), sys.AMD64)
 	addF(simdPackage, "Mask16x8.AsInt16x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int16x8.AsMask16x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int16x8.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask16x8.And", opLen2(ssa.OpAndInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Mask16x8.Or", opLen2(ssa.OpOrInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "LoadMask16x8FromBits", simdLoadMask(16, 8), sys.AMD64)
@@ -2331,7 +2331,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask16x8FromBits", simdCvtVToMask(16, 8), sys.AMD64)
 	addF(simdPackage, "Mask16x8.ToBits", simdCvtMaskToV(16, 8), sys.AMD64)
 	addF(simdPackage, "Mask16x16.AsInt16x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int16x16.AsMask16x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int16x16.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask16x16.And", opLen2(ssa.OpAndInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Mask16x16.Or", opLen2(ssa.OpOrInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "LoadMask16x16FromBits", simdLoadMask(16, 16), sys.AMD64)
@@ -2339,7 +2339,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask16x16FromBits", simdCvtVToMask(16, 16), sys.AMD64)
 	addF(simdPackage, "Mask16x16.ToBits", simdCvtMaskToV(16, 16), sys.AMD64)
 	addF(simdPackage, "Mask16x32.AsInt16x32", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int16x32.AsMask16x32", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int16x32.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask16x32.And", opLen2(ssa.OpAndInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Mask16x32.Or", opLen2(ssa.OpOrInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "LoadMask16x32FromBits", simdLoadMask(16, 32), sys.AMD64)
@@ -2347,7 +2347,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask16x32FromBits", simdCvtVToMask(16, 32), sys.AMD64)
 	addF(simdPackage, "Mask16x32.ToBits", simdCvtMaskToV(16, 32), sys.AMD64)
 	addF(simdPackage, "Mask32x4.AsInt32x4", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int32x4.AsMask32x4", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int32x4.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask32x4.And", opLen2(ssa.OpAndInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Mask32x4.Or", opLen2(ssa.OpOrInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "LoadMask32x4FromBits", simdLoadMask(32, 4), sys.AMD64)
@@ -2355,7 +2355,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask32x4FromBits", simdCvtVToMask(32, 4), sys.AMD64)
 	addF(simdPackage, "Mask32x4.ToBits", simdCvtMaskToV(32, 4), sys.AMD64)
 	addF(simdPackage, "Mask32x8.AsInt32x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int32x8.AsMask32x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int32x8.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask32x8.And", opLen2(ssa.OpAndInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Mask32x8.Or", opLen2(ssa.OpOrInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "LoadMask32x8FromBits", simdLoadMask(32, 8), sys.AMD64)
@@ -2363,7 +2363,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask32x8FromBits", simdCvtVToMask(32, 8), sys.AMD64)
 	addF(simdPackage, "Mask32x8.ToBits", simdCvtMaskToV(32, 8), sys.AMD64)
 	addF(simdPackage, "Mask32x16.AsInt32x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int32x16.AsMask32x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int32x16.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask32x16.And", opLen2(ssa.OpAndInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Mask32x16.Or", opLen2(ssa.OpOrInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "LoadMask32x16FromBits", simdLoadMask(32, 16), sys.AMD64)
@@ -2371,7 +2371,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask32x16FromBits", simdCvtVToMask(32, 16), sys.AMD64)
 	addF(simdPackage, "Mask32x16.ToBits", simdCvtMaskToV(32, 16), sys.AMD64)
 	addF(simdPackage, "Mask64x2.AsInt64x2", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int64x2.AsMask64x2", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int64x2.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask64x2.And", opLen2(ssa.OpAndInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Mask64x2.Or", opLen2(ssa.OpOrInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "LoadMask64x2FromBits", simdLoadMask(64, 2), sys.AMD64)
@@ -2379,7 +2379,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask64x2FromBits", simdCvtVToMask(64, 2), sys.AMD64)
 	addF(simdPackage, "Mask64x2.ToBits", simdCvtMaskToV(64, 2), sys.AMD64)
 	addF(simdPackage, "Mask64x4.AsInt64x4", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int64x4.AsMask64x4", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int64x4.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask64x4.And", opLen2(ssa.OpAndInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Mask64x4.Or", opLen2(ssa.OpOrInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "LoadMask64x4FromBits", simdLoadMask(64, 4), sys.AMD64)
@@ -2387,7 +2387,7 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Mask64x4FromBits", simdCvtVToMask(64, 4), sys.AMD64)
 	addF(simdPackage, "Mask64x4.ToBits", simdCvtMaskToV(64, 4), sys.AMD64)
 	addF(simdPackage, "Mask64x8.AsInt64x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
-	addF(simdPackage, "Int64x8.AsMask64x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
+	addF(simdPackage, "Int64x8.asMask", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask64x8.And", opLen2(ssa.OpAndInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Mask64x8.Or", opLen2(ssa.OpOrInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "LoadMask64x8FromBits", simdLoadMask(64, 8), sys.AMD64)