Skip to content

JIT: Move remaining xarch floating->integral cast implementation to lowering #117571

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 11 commits into
base: main
Choose a base branch
from
Open

JIT: Move remaining xarch floating->integral cast implementation to lowering #117571

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
059f82b
move remaining floating->integral cast implementation to lowering
saucecontrol Jul 12, 2025
36d884f
fix linux-x86 build?
saucecontrol Jul 12, 2025
59e01ee
Merge remote-tracking branch 'upstream/main' into lng2flt4
saucecontrol Jul 28, 2025
cae2432
update intrinsic names
saucecontrol Jul 28, 2025
33ceb0a
Merge branch 'main' into lng2flt4
tannergooding Aug 8, 2025
2392806
Merge remote-tracking branch 'upstream/main' into lng2flt4
saucecontrol Aug 21, 2025
7391f12
Merge branch 'main' into lng2flt4
saucecontrol Sep 7, 2025
3bce24e
Merge branch 'main' into lng2flt4
saucecontrol Sep 15, 2025
448c743
Merge branch 'main' into lng2flt4
saucecontrol Oct 9, 2025
9271ede
Merge branch 'main' into lng2flt4
tannergooding Oct 23, 2025
95a95a1
Merge branch 'main' into lng2flt4
saucecontrol Nov 17, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion src/coreclr/jit/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,7 +75,7 @@ function(create_standalone_jit)
endif ()
endfunction()

if (CLR_CMAKE_TARGET_ARCH_AMD64 OR CLR_CMAKE_TARGET_ARCH_ARM64 OR (CLR_CMAKE_TARGET_ARCH_I386 AND NOT CLR_CMAKE_HOST_UNIX))
if (CLR_CMAKE_TARGET_ARCH_AMD64 OR CLR_CMAKE_TARGET_ARCH_ARM64 OR CLR_CMAKE_TARGET_ARCH_I386)
add_compile_definitions($<$<NOT:$<BOOL:$<TARGET_PROPERTY:IGNORE_DEFAULT_TARGET_ARCH>>>:FEATURE_SIMD>)
add_compile_definitions($<$<NOT:$<BOOL:$<TARGET_PROPERTY:IGNORE_DEFAULT_TARGET_ARCH>>>:FEATURE_HW_INTRINSICS>)
add_compile_definitions($<$<NOT:$<BOOL:$<TARGET_PROPERTY:IGNORE_DEFAULT_TARGET_ARCH>>>:FEATURE_MASKED_HW_INTRINSICS>)
Expand Down
5 changes: 5 additions & 0 deletions src/coreclr/jit/codegenlinear.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2277,8 +2277,13 @@ void CodeGen::genCodeForCast(GenTreeOp* tree)
}
else if (varTypeIsFloating(tree->gtOp1))
{
#ifdef TARGET_XARCH
// These casts should have been lowered to HWIntrinsics
unreached();
#else
// Casts float/double --> int32/int64
genFloatToIntCast(tree);
#endif
}
else if (varTypeIsFloating(targetType))
{
Expand Down
68 changes: 0 additions & 68 deletions src/coreclr/jit/codegenxarch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7363,74 +7363,6 @@ void CodeGen::genIntToFloatCast(GenTree* treeNode)
genProduceReg(treeNode);
}

//------------------------------------------------------------------------
// genFloatToIntCast: Generate code to cast float/double to int/long
//
// Arguments:
// treeNode - The GT_CAST node
//
// Return Value:
// None.
//
// Assumptions:
// Cast is a non-overflow conversion.
// The treeNode must have an assigned register.
// SrcType=float/double and DstType= int32/uint32/int64/uint64
//
// TODO-XArch-CQ: (Low-pri) - generate in-line code when DstType = uint64
//
void CodeGen::genFloatToIntCast(GenTree* treeNode)
{
// we don't expect to see overflow detecting float/double --> int type conversions here
// as they should have been converted into helper calls by front-end.
assert(treeNode->OperIs(GT_CAST));
assert(!treeNode->gtOverflow());

regNumber targetReg = treeNode->GetRegNum();
assert(genIsValidIntReg(targetReg));

GenTree* op1 = treeNode->AsOp()->gtOp1;
#ifdef DEBUG
if (op1->isUsedFromReg())
{
assert(genIsValidFloatReg(op1->GetRegNum()));
}
#endif

var_types dstType = treeNode->CastToType();
var_types srcType = op1->TypeGet();
assert(varTypeIsFloating(srcType) && !varTypeIsFloating(dstType));

// We should never be seeing dstType whose size is neither sizeof(TYP_INT) nor sizeof(TYP_LONG).
// For conversions to byte/sbyte/int16/uint16 from float/double, we would expect the
// front-end or lowering phase to have generated two levels of cast. The first one is
// for float or double to int32/uint32 and the second one for narrowing int32/uint32 to
// the required smaller int type.
emitAttr dstSize = EA_ATTR(genTypeSize(dstType));
noway_assert((dstSize == EA_ATTR(genTypeSize(TYP_INT))) || (dstSize == EA_ATTR(genTypeSize(TYP_LONG))));

// We shouldn't be seeing uint64 here as it should have been converted
// into a helper call by either front-end or lowering phase, unless we have AVX512
// accelerated conversions.
assert(!varTypeIsUnsigned(dstType) || (dstSize != EA_ATTR(genTypeSize(TYP_LONG))) ||
compiler->canUseEvexEncodingDebugOnly());

// If the dstType is TYP_UINT, we have 32-bits to encode the
// float number. Any of 33rd or above bits can be the sign bit.
// To achieve it we pretend as if we are converting it to a long.
if (varTypeIsUnsigned(dstType) && (dstSize == EA_ATTR(genTypeSize(TYP_INT))) && !compiler->canUseEvexEncoding())
{
dstType = TYP_LONG;
}

// Note that we need to specify dstType here so that it will determine
// the size of destination integer register and also the rex.w prefix.
genConsumeOperands(treeNode->AsOp());
instruction ins = ins_FloatConv(dstType, srcType);
GetEmitter()->emitInsBinary(ins, emitTypeSize(dstType), treeNode, op1);
genProduceReg(treeNode);
}

//------------------------------------------------------------------------
// genCkfinite: Generate code for ckfinite opcode.
//
Expand Down
8 changes: 4 additions & 4 deletions src/coreclr/jit/gentree.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21884,19 +21884,19 @@ GenTree* Compiler::gtNewSimdCvtNode(var_types type,
switch (simdTargetBaseType)
{
case TYP_INT:
cvtIntrinsic = NI_AVX10v2_ConvertToVectorInt32WithTruncationSaturation;
cvtIntrinsic = NI_AVX10v2_ConvertToVectorInt32WithTruncatedSaturation;
break;

case TYP_UINT:
cvtIntrinsic = NI_AVX10v2_ConvertToVectorUInt32WithTruncationSaturation;
cvtIntrinsic = NI_AVX10v2_ConvertToVectorUInt32WithTruncatedSaturation;
break;

case TYP_LONG:
cvtIntrinsic = NI_AVX10v2_ConvertToVectorInt64WithTruncationSaturation;
cvtIntrinsic = NI_AVX10v2_ConvertToVectorInt64WithTruncatedSaturation;
break;

case TYP_ULONG:
cvtIntrinsic = NI_AVX10v2_ConvertToVectorUInt64WithTruncationSaturation;
cvtIntrinsic = NI_AVX10v2_ConvertToVectorUInt64WithTruncatedSaturation;
break;

default:
Expand Down
2 changes: 1 addition & 1 deletion src/coreclr/jit/hwintrinsic.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -978,7 +978,7 @@ static const HWIntrinsicIsaRange hwintrinsicIsaRangeArray[] = {
{ NI_Illegal, NI_Illegal }, // AVX512v2_X64
{ NI_Illegal, NI_Illegal }, // AVX512v3_X64
{ NI_Illegal, NI_Illegal }, // AVX10v1_X64
{ NI_Illegal, NI_Illegal }, // AVX10v2_X64
{ FIRST_NI_AVX10v2_X64, LAST_NI_AVX10v2_X64 }, // AVX10v2_X64
{ NI_Illegal, NI_Illegal }, // AES_X64
{ NI_Illegal, NI_Illegal }, // AVX512VP2INTERSECT_X64
{ NI_Illegal, NI_Illegal }, // AVXIFMA_X64
Expand Down
8 changes: 7 additions & 1 deletion src/coreclr/jit/hwintrinsiccodegenxarch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1011,6 +1011,8 @@ void CodeGen::genHWIntrinsic(GenTreeHWIntrinsic* node)
case InstructionSet_AVX512:
case InstructionSet_AVX512_X64:
case InstructionSet_AVX512v2:
case InstructionSet_AVX10v2:
case InstructionSet_AVX10v2_X64:
case InstructionSet_AVXVNNIINT:
case InstructionSet_AVXVNNIINT_V512:
{
Expand Down Expand Up @@ -3502,6 +3504,10 @@ void CodeGen::genAvxFamilyIntrinsic(GenTreeHWIntrinsic* node, insOpts instOption
case NI_AVX512_X64_ConvertToInt64:
case NI_AVX512_X64_ConvertToUInt64:
case NI_AVX512_X64_ConvertToUInt64WithTruncation:
case NI_AVX10v2_ConvertToInt32WithTruncatedSaturation:
case NI_AVX10v2_ConvertToUInt32WithTruncatedSaturation:
case NI_AVX10v2_X64_ConvertToInt64WithTruncatedSaturation:
case NI_AVX10v2_X64_ConvertToUInt64WithTruncatedSaturation:
{
assert(baseType == TYP_DOUBLE || baseType == TYP_FLOAT);
emitAttr attr = emitTypeSize(targetType);
Expand All @@ -3512,7 +3518,6 @@ void CodeGen::genAvxFamilyIntrinsic(GenTreeHWIntrinsic* node, insOpts instOption
}

case NI_AVX512_ConvertToVector128UInt32:
case NI_AVX512_ConvertToVector128UInt32WithSaturation:
case NI_AVX512_ConvertToVector256Int32:
case NI_AVX512_ConvertToVector256UInt32:
{
Expand All @@ -3535,6 +3540,7 @@ void CodeGen::genAvxFamilyIntrinsic(GenTreeHWIntrinsic* node, insOpts instOption
case NI_AVX512_ConvertToVector128SByteWithSaturation:
case NI_AVX512_ConvertToVector128UInt16:
case NI_AVX512_ConvertToVector128UInt16WithSaturation:
case NI_AVX512_ConvertToVector128UInt32WithSaturation:
case NI_AVX512_ConvertToVector256Byte:
case NI_AVX512_ConvertToVector256ByteWithSaturation:
case NI_AVX512_ConvertToVector256Int16:
Expand Down
20 changes: 16 additions & 4 deletions src/coreclr/jit/hwintrinsiclistxarch.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1077,19 +1077,31 @@ HARDWARE_INTRINSIC(AVX512v3, ExpandLoad,
#define FIRST_NI_AVX10v2 NI_AVX10v2_ConvertToByteWithSaturationAndZeroExtendToInt32
HARDWARE_INTRINSIC(AVX10v2, ConvertToByteWithSaturationAndZeroExtendToInt32, -1, -1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvtps2iubs, INS_invalid}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg|HW_Flag_EmbRoundingCompatible)
HARDWARE_INTRINSIC(AVX10v2, ConvertToByteWithTruncatedSaturationAndZeroExtendToInt32, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2iubs, INS_invalid}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, ConvertToInt32WithTruncatedSaturation, 16, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttss2sis32, INS_vcvttsd2sis32}, HW_Category_SIMDScalar, HW_Flag_BaseTypeFromFirstArg|HW_Flag_SpecialCodeGen)
HARDWARE_INTRINSIC(AVX10v2, ConvertToSByteWithSaturationAndZeroExtendToInt32, -1, -1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvtps2ibs, INS_invalid}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg|HW_Flag_EmbRoundingCompatible)
HARDWARE_INTRINSIC(AVX10v2, ConvertToSByteWithTruncatedSaturationAndZeroExtendToInt32, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2ibs, INS_invalid}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, ConvertToVectorInt32WithTruncationSaturation, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2dqs, INS_vcvttpd2dqs}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, ConvertToVectorInt64WithTruncationSaturation, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2qqs, INS_vcvttpd2qqs}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, ConvertToVectorUInt32WithTruncationSaturation, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2udqs, INS_vcvttpd2udqs}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, ConvertToVectorUInt64WithTruncationSaturation, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2uqqs, INS_vcvttpd2uqqs}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, ConvertToUInt32WithTruncatedSaturation, 16, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttss2usis32, INS_vcvttsd2usis32}, HW_Category_SIMDScalar, HW_Flag_BaseTypeFromFirstArg|HW_Flag_SpecialCodeGen)
HARDWARE_INTRINSIC(AVX10v2, ConvertToVectorInt32WithTruncatedSaturation, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2dqs, INS_vcvttpd2dqs}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, ConvertToVectorInt64WithTruncatedSaturation, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2qqs, INS_vcvttpd2qqs}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, ConvertToVectorUInt32WithTruncatedSaturation, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2udqs, INS_vcvttpd2udqs}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, ConvertToVectorUInt64WithTruncatedSaturation, -1, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttps2uqqs, INS_vcvttpd2uqqs}, HW_Category_SimpleSIMD, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, MinMax, -1, 3, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vminmaxps, INS_vminmaxpd}, HW_Category_IMM, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, MinMaxScalar, -1, 3, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vminmaxss, INS_vminmaxsd}, HW_Category_IMM, HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(AVX10v2, MoveScalar, 16, -1, {INS_invalid, INS_invalid, INS_vmovw_simd, INS_vmovw_simd, INS_vmovd_simd, INS_vmovd_simd, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMDScalar, HW_Flag_NoContainment)
HARDWARE_INTRINSIC(AVX10v2, MultipleSumAbsoluteDifferences, 64, 3, {INS_invalid, INS_invalid, INS_invalid, INS_vmpsadbw, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_IMM, HW_Flag_FullRangeIMM)
HARDWARE_INTRINSIC(AVX10v2, StoreScalar, 16, 2, {INS_invalid, INS_invalid, INS_vmovw_simd, INS_vmovw_simd, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_MemoryStore, HW_Flag_NoRMWSemantics|HW_Flag_BaseTypeFromSecondArg)
#define LAST_NI_AVX10v2 NI_AVX10v2_StoreScalar

// ***************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************
// ISA Function name SIMD size NumArg Instructions Category Flags
// {TYP_BYTE, TYP_UBYTE, TYP_SHORT, TYP_USHORT, TYP_INT, TYP_UINT, TYP_LONG, TYP_ULONG, TYP_FLOAT, TYP_DOUBLE}
// ***************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************
// 64-bit only intrinsics for AVX10v2
#define FIRST_NI_AVX10v2_X64 NI_AVX10v2_X64_ConvertToInt64WithTruncatedSaturation
HARDWARE_INTRINSIC(AVX10v2_X64, ConvertToInt64WithTruncatedSaturation, 16, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttss2sis64, INS_vcvttsd2sis64}, HW_Category_SIMDScalar, HW_Flag_BaseTypeFromFirstArg|HW_Flag_SpecialCodeGen)
HARDWARE_INTRINSIC(AVX10v2_X64, ConvertToUInt64WithTruncatedSaturation, 16, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_vcvttss2usis64, INS_vcvttsd2usis64}, HW_Category_SIMDScalar, HW_Flag_BaseTypeFromFirstArg|HW_Flag_SpecialCodeGen)
#define LAST_NI_AVX10v2_X64 NI_AVX10v2_X64_ConvertToUInt64WithTruncatedSaturation

// ***************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************
// ISA Function name SIMD size NumArg Instructions Category Flags
// {TYP_BYTE, TYP_UBYTE, TYP_SHORT, TYP_USHORT, TYP_INT, TYP_UINT, TYP_LONG, TYP_ULONG, TYP_FLOAT, TYP_DOUBLE}
Expand Down
Loading
Loading