Stowage/go
2
0
Fork 0
mirror of https://github.com/golang/go.git synced 2025-12-08 06:10:04 +00:00
Code Issues Projects Releases Packages Wiki Activity

[dev.simd] cmd/compile: undoes reorder transform in prior commit, changes names

Browse source 
paired with simdgen CL 684655

Change-Id: I819eb601c07b21747d8a1442eb1efbf9fa5aac1d
Reviewed-on: https://go-review.googlesource.com/c/go/+/684775
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
This commit is contained in:
David Chase 2025-06-28 10:20:53 -04:00
parent 10c9621936
commit 55665e1e37
2 changed files with 164 additions and 184 deletions
Download patch file Download diff file Expand all files Collapse all files

44
src/cmd/compile/internal/ssagen/simdintrinsics.go
View file

@ -262,12 +262,12 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
addF(simdPackage, "Float64x2.FusedMultiplySubAdd", opLen3(ssa.OpFusedMultiplySubAddFloat64x2, types.TypeVec128), sys.AMD64) addF(simdPackage, "Float64x2.FusedMultiplySubAdd", opLen3(ssa.OpFusedMultiplySubAddFloat64x2, types.TypeVec128), sys.AMD64)
addF(simdPackage, "Float64x4.FusedMultiplySubAdd", opLen3(ssa.OpFusedMultiplySubAddFloat64x4, types.TypeVec256), sys.AMD64) addF(simdPackage, "Float64x4.FusedMultiplySubAdd", opLen3(ssa.OpFusedMultiplySubAddFloat64x4, types.TypeVec256), sys.AMD64)
addF(simdPackage, "Float64x8.FusedMultiplySubAdd", opLen3(ssa.OpFusedMultiplySubAddFloat64x8, types.TypeVec512), sys.AMD64) addF(simdPackage, "Float64x8.FusedMultiplySubAdd", opLen3(ssa.OpFusedMultiplySubAddFloat64x8, types.TypeVec512), sys.AMD64)
addF(simdPackage, "Uint8x16.GaloisFieldAffineTransform", opGaloisFieldAffineTransform(ssa.OpGaloisFieldAffineTransformUint8x16, types.TypeVec128), sys.AMD64) addF(simdPackage, "Uint8x16.GaloisFieldAffineTransform", opLen2Imm8(ssa.OpGaloisFieldAffineTransformUint8x16, types.TypeVec128, 0), sys.AMD64)
addF(simdPackage, "Uint8x32.GaloisFieldAffineTransform", opGaloisFieldAffineTransform(ssa.OpGaloisFieldAffineTransformUint8x32, types.TypeVec256), sys.AMD64) addF(simdPackage, "Uint8x32.GaloisFieldAffineTransform", opLen2Imm8(ssa.OpGaloisFieldAffineTransformUint8x32, types.TypeVec256, 0), sys.AMD64)
addF(simdPackage, "Uint8x64.GaloisFieldAffineTransform", opGaloisFieldAffineTransform(ssa.OpGaloisFieldAffineTransformUint8x64, types.TypeVec512), sys.AMD64) addF(simdPackage, "Uint8x64.GaloisFieldAffineTransform", opLen2Imm8(ssa.OpGaloisFieldAffineTransformUint8x64, types.TypeVec512, 0), sys.AMD64)
addF(simdPackage, "Uint8x16.GaloisFieldAffineTransformInversed", opGaloisFieldAffineTransform(ssa.OpGaloisFieldAffineTransformInversedUint8x16, types.TypeVec128), sys.AMD64) addF(simdPackage, "Uint8x16.GaloisFieldAffineTransformInversed", opLen2Imm8(ssa.OpGaloisFieldAffineTransformInversedUint8x16, types.TypeVec128, 0), sys.AMD64)
addF(simdPackage, "Uint8x32.GaloisFieldAffineTransformInversed", opGaloisFieldAffineTransform(ssa.OpGaloisFieldAffineTransformInversedUint8x32, types.TypeVec256), sys.AMD64) addF(simdPackage, "Uint8x32.GaloisFieldAffineTransformInversed", opLen2Imm8(ssa.OpGaloisFieldAffineTransformInversedUint8x32, types.TypeVec256, 0), sys.AMD64)
addF(simdPackage, "Uint8x64.GaloisFieldAffineTransformInversed", opGaloisFieldAffineTransform(ssa.OpGaloisFieldAffineTransformInversedUint8x64, types.TypeVec512), sys.AMD64) addF(simdPackage, "Uint8x64.GaloisFieldAffineTransformInversed", opLen2Imm8(ssa.OpGaloisFieldAffineTransformInversedUint8x64, types.TypeVec512, 0), sys.AMD64)
addF(simdPackage, "Uint8x16.GaloisFieldMul", opLen2(ssa.OpGaloisFieldMulUint8x16, types.TypeVec128), sys.AMD64) addF(simdPackage, "Uint8x16.GaloisFieldMul", opLen2(ssa.OpGaloisFieldMulUint8x16, types.TypeVec128), sys.AMD64)
addF(simdPackage, "Uint8x32.GaloisFieldMul", opLen2(ssa.OpGaloisFieldMulUint8x32, types.TypeVec256), sys.AMD64) addF(simdPackage, "Uint8x32.GaloisFieldMul", opLen2(ssa.OpGaloisFieldMulUint8x32, types.TypeVec256), sys.AMD64)
addF(simdPackage, "Uint8x64.GaloisFieldMul", opLen2(ssa.OpGaloisFieldMulUint8x64, types.TypeVec512), sys.AMD64) addF(simdPackage, "Uint8x64.GaloisFieldMul", opLen2(ssa.OpGaloisFieldMulUint8x64, types.TypeVec512), sys.AMD64)
@ -627,12 +627,12 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
addF(simdPackage, "Float64x2.MaskedFusedMultiplySubAdd", opLen4(ssa.OpMaskedFusedMultiplySubAddFloat64x2, types.TypeVec128), sys.AMD64) addF(simdPackage, "Float64x2.MaskedFusedMultiplySubAdd", opLen4(ssa.OpMaskedFusedMultiplySubAddFloat64x2, types.TypeVec128), sys.AMD64)
addF(simdPackage, "Float64x4.MaskedFusedMultiplySubAdd", opLen4(ssa.OpMaskedFusedMultiplySubAddFloat64x4, types.TypeVec256), sys.AMD64) addF(simdPackage, "Float64x4.MaskedFusedMultiplySubAdd", opLen4(ssa.OpMaskedFusedMultiplySubAddFloat64x4, types.TypeVec256), sys.AMD64)
addF(simdPackage, "Float64x8.MaskedFusedMultiplySubAdd", opLen4(ssa.OpMaskedFusedMultiplySubAddFloat64x8, types.TypeVec512), sys.AMD64) addF(simdPackage, "Float64x8.MaskedFusedMultiplySubAdd", opLen4(ssa.OpMaskedFusedMultiplySubAddFloat64x8, types.TypeVec512), sys.AMD64)
addF(simdPackage, "Uint8x16.MaskedGaloisFieldAffineTransform", opGaloisFieldAffineTransformMasked(ssa.OpMaskedGaloisFieldAffineTransformUint8x16, types.TypeVec128), sys.AMD64) addF(simdPackage, "Uint8x16.MaskedGaloisFieldAffineTransform", opLen3Imm8(ssa.OpMaskedGaloisFieldAffineTransformUint8x16, types.TypeVec128, 0), sys.AMD64)
addF(simdPackage, "Uint8x32.MaskedGaloisFieldAffineTransform", opGaloisFieldAffineTransformMasked(ssa.OpMaskedGaloisFieldAffineTransformUint8x32, types.TypeVec256), sys.AMD64) addF(simdPackage, "Uint8x32.MaskedGaloisFieldAffineTransform", opLen3Imm8(ssa.OpMaskedGaloisFieldAffineTransformUint8x32, types.TypeVec256, 0), sys.AMD64)
addF(simdPackage, "Uint8x64.MaskedGaloisFieldAffineTransform", opGaloisFieldAffineTransformMasked(ssa.OpMaskedGaloisFieldAffineTransformUint8x64, types.TypeVec512), sys.AMD64) addF(simdPackage, "Uint8x64.MaskedGaloisFieldAffineTransform", opLen3Imm8(ssa.OpMaskedGaloisFieldAffineTransformUint8x64, types.TypeVec512, 0), sys.AMD64)
addF(simdPackage, "Uint8x16.MaskedGaloisFieldAffineTransformInversed", opGaloisFieldAffineTransformMasked(ssa.OpMaskedGaloisFieldAffineTransformInversedUint8x16, types.TypeVec128), sys.AMD64) addF(simdPackage, "Uint8x16.MaskedGaloisFieldAffineTransformInversed", opLen3Imm8(ssa.OpMaskedGaloisFieldAffineTransformInversedUint8x16, types.TypeVec128, 0), sys.AMD64)
addF(simdPackage, "Uint8x32.MaskedGaloisFieldAffineTransformInversed", opGaloisFieldAffineTransformMasked(ssa.OpMaskedGaloisFieldAffineTransformInversedUint8x32, types.TypeVec256), sys.AMD64) addF(simdPackage, "Uint8x32.MaskedGaloisFieldAffineTransformInversed", opLen3Imm8(ssa.OpMaskedGaloisFieldAffineTransformInversedUint8x32, types.TypeVec256, 0), sys.AMD64)
addF(simdPackage, "Uint8x64.MaskedGaloisFieldAffineTransformInversed", opGaloisFieldAffineTransformMasked(ssa.OpMaskedGaloisFieldAffineTransformInversedUint8x64, types.TypeVec512), sys.AMD64) addF(simdPackage, "Uint8x64.MaskedGaloisFieldAffineTransformInversed", opLen3Imm8(ssa.OpMaskedGaloisFieldAffineTransformInversedUint8x64, types.TypeVec512, 0), sys.AMD64)
addF(simdPackage, "Uint8x16.MaskedGaloisFieldMul", opLen3(ssa.OpMaskedGaloisFieldMulUint8x16, types.TypeVec128), sys.AMD64) addF(simdPackage, "Uint8x16.MaskedGaloisFieldMul", opLen3(ssa.OpMaskedGaloisFieldMulUint8x16, types.TypeVec128), sys.AMD64)
addF(simdPackage, "Uint8x32.MaskedGaloisFieldMul", opLen3(ssa.OpMaskedGaloisFieldMulUint8x32, types.TypeVec256), sys.AMD64) addF(simdPackage, "Uint8x32.MaskedGaloisFieldMul", opLen3(ssa.OpMaskedGaloisFieldMulUint8x32, types.TypeVec256), sys.AMD64)
addF(simdPackage, "Uint8x64.MaskedGaloisFieldMul", opLen3(ssa.OpMaskedGaloisFieldMulUint8x64, types.TypeVec512), sys.AMD64) addF(simdPackage, "Uint8x64.MaskedGaloisFieldMul", opLen3(ssa.OpMaskedGaloisFieldMulUint8x64, types.TypeVec512), sys.AMD64)
@ -2215,23 +2215,3 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
addF(simdPackage, "Mask64x8.And", opLen2(ssa.OpAndInt32x16, types.TypeVec512), 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, "Mask64x8.Or", opLen2(ssa.OpOrInt32x16, types.TypeVec512), sys.AMD64)
} }
func opGaloisFieldAffineTransform(op ssa.Op, t *types.Type) func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
return func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
if args[0].Op == ssa.OpConst8 {
return s.newValue2I(op, t, args[0].AuxInt, args[0], args[1])
}
plainPanicSimdImm(s)
return s.newValue2I(op, t, 0, args[0], args[1])
}
}
func opGaloisFieldAffineTransformMasked(op ssa.Op, t *types.Type) func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
return func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
if args[0].Op == ssa.OpConst8 {
return s.newValue3I(op, t, args[0].AuxInt, args[0], args[1], args[3])
}
plainPanicSimdImm(s)
return s.newValue3I(op, t, 0, args[0], args[1], args[3])
}
}

304
src/simd/stubs_amd64.go
View file

File diff suppressed because it is too large Load diff