Skip to content
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

JIT ARM64-SVE: Add CreateWhileLessThan* #100949

Merged
merged 8 commits into from
Apr 25, 2024
Merged

JIT ARM64-SVE: Add CreateWhileLessThan* #100949

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
944467a
JIT ARM64-SVE: Add CreateWhileLessThan*
a74nh Apr 9, 2024
3f62ba9
Set simdBaseJitType to type of input args
a74nh Apr 12, 2024
ddd7168
Hardcode opt in codegen
a74nh Apr 12, 2024
a0b5687
Fix gtNewSimdConvertMaskToVectorNode types
a74nh Apr 15, 2024
941c306
Use HW_Flag_BaseTypeFromFirstArg
a74nh Apr 15, 2024
c6eb008
Merge main
a74nh Apr 16, 2024
1bd25c4
Set base type to return type and auxiliary type to input type
a74nh Apr 24, 2024
e5bd3d8
Merge main
a74nh Apr 25, 2024
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
11 changes: 11 additions & 0 deletions src/coreclr/jit/hwintrinsic.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1539,6 +1539,17 @@ GenTree* Compiler::impHWIntrinsic(NamedIntrinsic intrinsic,
}
break;

case NI_Sve_CreateWhileLessThanMask8Bit:
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't like how we have these target specific switch statements in hwintrinsic.cpp, it makes the code hard to follow. I highly expect the number of these special cases to grow too.

I'd much prefer it if each of these instrinsics (including all the neon and X86 ones) were marked with SpecialImport. The special import cases would then have to duplicate the common get args code, but it's only a few lines (which could go into a helper). Not going to do it for this PR.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👍, we've definitely done that for almost all the x86/x64 code, there's only a couple special handlers left for that platform (like Sse42.Crc32). It would be nice to get both platforms doing this overall consistently here.

case NI_Sve_CreateWhileLessThanOrEqualMask8Bit:
case NI_Sve_CreateWhileLessThanMask16Bit:
case NI_Sve_CreateWhileLessThanOrEqualMask16Bit:
case NI_Sve_CreateWhileLessThanMask32Bit:
case NI_Sve_CreateWhileLessThanOrEqualMask32Bit:
case NI_Sve_CreateWhileLessThanMask64Bit:
case NI_Sve_CreateWhileLessThanOrEqualMask64Bit:
retNode->AsHWIntrinsic()->SetAuxiliaryJitType(sigReader.op1JitType);
break;

default:
break;
}
Expand Down
34 changes: 34 additions & 0 deletions src/coreclr/jit/hwintrinsiccodegenarm64.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1409,6 +1409,40 @@ void CodeGen::genHWIntrinsic(GenTreeHWIntrinsic* node)
GetEmitter()->emitIns_R_PATTERN(ins, emitSize, targetReg, opt, SVE_PATTERN_ALL);
break;

case NI_Sve_CreateWhileLessThanMask8Bit:
case NI_Sve_CreateWhileLessThanMask16Bit:
case NI_Sve_CreateWhileLessThanMask32Bit:
case NI_Sve_CreateWhileLessThanMask64Bit:
{
// Emit size and instruction is based on the scalar operands.
var_types auxType = node->GetAuxiliaryType();
emitSize = emitActualTypeSize(auxType);
if (varTypeIsUnsigned(auxType))
{
ins = INS_sve_whilelo;
}

GetEmitter()->emitIns_R_R_R(ins, emitSize, targetReg, op1Reg, op2Reg, opt);
break;
}

case NI_Sve_CreateWhileLessThanOrEqualMask8Bit:
case NI_Sve_CreateWhileLessThanOrEqualMask16Bit:
case NI_Sve_CreateWhileLessThanOrEqualMask32Bit:
case NI_Sve_CreateWhileLessThanOrEqualMask64Bit:
{
// Emit size and instruction is based on the scalar operands.
var_types auxType = node->GetAuxiliaryType();
emitSize = emitActualTypeSize(auxType);
if (varTypeIsUnsigned(auxType))
{
ins = INS_sve_whilels;
}

GetEmitter()->emitIns_R_R_R(ins, emitSize, targetReg, op1Reg, op2Reg, opt);
break;
}

default:
unreached();
}
Expand Down
10 changes: 8 additions & 2 deletions src/coreclr/jit/hwintrinsiclistarm64sve.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,14 @@ HARDWARE_INTRINSIC(Sve, CreateTrueMaskSingle,
HARDWARE_INTRINSIC(Sve, CreateTrueMaskUInt16, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_EnumPattern, HW_Flag_Scalable|HW_Flag_HasImmediateOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskUInt32, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_EnumPattern, HW_Flag_Scalable|HW_Flag_HasImmediateOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskUInt64, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid, INS_invalid}, HW_Category_EnumPattern, HW_Flag_Scalable|HW_Flag_HasImmediateOperand|HW_Flag_ReturnsPerElementMask)

HARDWARE_INTRINSIC(Sve, CreateWhileLessThanMask16Bit, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_sve_whilelt, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_SpecialCodeGen|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateWhileLessThanMask32Bit, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_whilelt, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_SpecialCodeGen|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateWhileLessThanMask64Bit, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_whilelt, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_SpecialCodeGen|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateWhileLessThanMask8Bit, -1, 2, false, {INS_invalid, INS_sve_whilelt, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_SpecialCodeGen|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateWhileLessThanOrEqualMask16Bit, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_sve_whilele, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_SpecialCodeGen|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateWhileLessThanOrEqualMask32Bit, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_whilele, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_SpecialCodeGen|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateWhileLessThanOrEqualMask64Bit, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_whilele, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_SpecialCodeGen|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateWhileLessThanOrEqualMask8Bit, -1, 2, false, {INS_invalid, INS_sve_whilele, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_SpecialCodeGen|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, LoadVector, -1, 2, true, {INS_sve_ld1b, INS_sve_ld1b, INS_sve_ld1h, INS_sve_ld1h, INS_sve_ld1w, INS_sve_ld1w, INS_sve_ld1d, INS_sve_ld1d, INS_sve_ld1w, INS_sve_ld1d}, HW_Category_MemoryLoad, HW_Flag_Scalable|HW_Flag_ExplicitMaskedOperation|HW_Flag_LowMaskedOperation)
HARDWARE_INTRINSIC(Sve, LoadVectorByteZeroExtendToInt16, -1, 2, false, {INS_invalid, INS_invalid, INS_sve_ld1b, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_MemoryLoad, HW_Flag_Scalable|HW_Flag_ExplicitMaskedOperation|HW_Flag_LowMaskedOperation)
HARDWARE_INTRINSIC(Sve, LoadVectorByteZeroExtendToInt32, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ld1b, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_MemoryLoad, HW_Flag_Scalable|HW_Flag_ExplicitMaskedOperation|HW_Flag_LowMaskedOperation)
Expand Down Expand Up @@ -58,7 +65,6 @@ HARDWARE_INTRINSIC(Sve, LoadVectorUInt32ZeroExtendToInt64,
HARDWARE_INTRINSIC(Sve, LoadVectorUInt32ZeroExtendToUInt64, -1, 2, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ld1w, INS_invalid, INS_invalid}, HW_Category_MemoryLoad, HW_Flag_Scalable|HW_Flag_ExplicitMaskedOperation|HW_Flag_LowMaskedOperation)



// ***************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************
// ISA Function name SIMD size NumArg EncodesExtraTypeArg 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