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

Jump to bottom

JIT ARM64-SVE: Add Count*BitElements #101188

Merged
merged 9 commits into from
Apr 26, 2024
Merged

JIT ARM64-SVE: Add Count*BitElements #101188

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
bd8ecde
JIT ARM64-SVE: Add Count*BitElements
a74nh Apr 16, 2024
2766611
Generic ValidateResult testing
a74nh Apr 22, 2024
901c200
Fix formatting
a74nh Apr 23, 2024
2937557
Add option to change SVE vector length for current and children proce…
SwapnilGaikwad Apr 16, 2024
3dbad56
Merge main
a74nh Apr 25, 2024
7ffe49b
Revert "Add option to change SVE vector length for current and childr…
a74nh Apr 25, 2024
18d80bf
Merge main
a74nh Apr 26, 2024
3e82dd8
fix CreateTrueMask
a74nh Apr 26, 2024
0dded3e
Merge main
a74nh Apr 26, 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
4 changes: 4 additions & 0 deletions src/coreclr/inc/clrconfigvalues.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -790,6 +790,10 @@ RETAIL_CONFIG_DWORD_INFO(EXTERNAL_EnableArm64Sha256, W("EnableArm64Sh
RETAIL_CONFIG_DWORD_INFO(EXTERNAL_EnableArm64Rcpc, W("EnableArm64Rcpc"), 1, "Allows Arm64 Rcpc+ hardware intrinsics to be disabled")
RETAIL_CONFIG_DWORD_INFO(EXTERNAL_EnableArm64Rcpc2, W("EnableArm64Rcpc2"), 1, "Allows Arm64 Rcpc2+ hardware intrinsics to be disabled")
RETAIL_CONFIG_DWORD_INFO(EXTERNAL_EnableArm64Sve, W("EnableArm64Sve"), 1, "Allows Arm64 SVE hardware intrinsics to be disabled")
#if defined(TARGET_LINUX) && (defined(DEBUG) || defined(_DEBUG))
// TODO-SVE: Once coreclr supports vector lengths >16bytes, the default should be -1.
RETAIL_CONFIG_DWORD_INFO_EX(EXTERNAL_MaxVectorLength, W("MaxVectorLength"), 16, "Specifies maximum size of the vector length in bytes while using SVE on Arm64", CLRConfig::LookupOptions::ParseIntegerAsBase10)
a74nh marked this conversation as resolved.
Show resolved Hide resolved
#endif //defined(TARGET_LINUX) && (defined(DEBUG) || defined(_DEBUG))
#endif

///
Expand Down
14 changes: 12 additions & 2 deletions src/coreclr/jit/hwintrinsic.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,6 @@ enum HWIntrinsicCategory : uint8_t
HW_Category_ShiftLeftByImmediate,
HW_Category_ShiftRightByImmediate,
HW_Category_SIMDByIndexedElement,
HW_Category_EnumPattern,

// Helper intrinsics
// - do not directly correspond to a instruction, such as Vector64.AllBitsSet
Expand Down Expand Up @@ -232,6 +231,11 @@ enum HWIntrinsicFlag : unsigned int

// The intrinsic is an embedded masking incompatible intrinsic
HW_Flag_EmbMaskingIncompatible = 0x20000000,
#elif defined(TARGET_ARM64)

// The intrinsic has an enum operand. Using this implies HW_Flag_HasImmediateOperand.
HW_Flag_HasEnumOperand = 0x1000000,

#endif // TARGET_XARCH

HW_Flag_CanBenefitFromConstantProp = 0x80000000,
Expand Down Expand Up @@ -867,7 +871,7 @@ struct HWIntrinsicInfo
static bool HasImmediateOperand(NamedIntrinsic id)
{
const HWIntrinsicFlag flags = lookupFlags(id);
return (flags & HW_Flag_HasImmediateOperand) != 0;
return ((flags & HW_Flag_HasImmediateOperand) != 0) || HasEnumOperand(id);
}

static bool IsScalable(NamedIntrinsic id)
Expand Down Expand Up @@ -906,6 +910,12 @@ struct HWIntrinsicInfo
return (flags & HW_Flag_ExplicitMaskedOperation) != 0;
}

static bool HasEnumOperand(NamedIntrinsic id)
{
const HWIntrinsicFlag flags = lookupFlags(id);
return (flags & HW_Flag_HasEnumOperand) != 0;
}

#endif // TARGET_ARM64

static bool HasSpecialSideEffect(NamedIntrinsic id)
Expand Down
4 changes: 4 additions & 0 deletions src/coreclr/jit/hwintrinsicarm64.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -290,6 +290,10 @@ void HWIntrinsicInfo::lookupImmBounds(
case NI_Sve_CreateTrueMaskUInt16:
case NI_Sve_CreateTrueMaskUInt32:
case NI_Sve_CreateTrueMaskUInt64:
case NI_Sve_Count16BitElements:
case NI_Sve_Count32BitElements:
case NI_Sve_Count64BitElements:
case NI_Sve_Count8BitElements:
immLowerBound = (int)SVE_PATTERN_POW2;
immUpperBound = (int)SVE_PATTERN_ALL;
break;
Expand Down
18 changes: 17 additions & 1 deletion src/coreclr/jit/hwintrinsiccodegenarm64.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -380,7 +380,7 @@ void CodeGen::genHWIntrinsic(GenTreeHWIntrinsic* node)
emitShift(intrin.op2, op1Reg);
}
}
else if (intrin.category == HW_Category_EnumPattern)
else if (HWIntrinsicInfo::HasEnumOperand(intrin.id))
{
assert(hasImmediateOperand);

Expand Down Expand Up @@ -1404,6 +1404,22 @@ void CodeGen::genHWIntrinsic(GenTreeHWIntrinsic* node)
GetEmitter()->emitIns_R_R_R_I(ins, emitSize, targetReg, op1Reg, op2Reg, 0, opt);
break;

case NI_Sve_Count16BitElements:
case NI_Sve_Count32BitElements:
case NI_Sve_Count64BitElements:
case NI_Sve_Count8BitElements:
{
// Instruction has an additional immediate to multiply the result by. Use 1.
kunalspathak marked this conversation as resolved.
Show resolved Hide resolved
assert(hasImmediateOperand);
HWIntrinsicImmOpHelper helper(this, intrin.op1, node);
for (helper.EmitBegin(); !helper.Done(); helper.EmitCaseEnd())
{
const insSvePattern pattern = (insSvePattern)helper.ImmValue();
GetEmitter()->emitIns_R_PATTERN_I(ins, emitSize, targetReg, pattern, 1, opt);
}
break;
}

case NI_Sve_CreateTrueMaskAll:
// Must use the pattern variant, as the non-pattern varient is SVE2.1.
GetEmitter()->emitIns_R_PATTERN(ins, emitSize, targetReg, opt, SVE_PATTERN_ALL);
Expand Down
25 changes: 14 additions & 11 deletions src/coreclr/jit/hwintrinsiclistarm64sve.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,17 +20,20 @@
HARDWARE_INTRINSIC(Sve, Abs, -1, -1, false, {INS_sve_abs, INS_invalid, INS_sve_abs, INS_invalid, INS_sve_abs, INS_invalid, INS_sve_abs, INS_invalid, INS_sve_fabs, INS_sve_fabs}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_EmbeddedMaskedOperation)
HARDWARE_INTRINSIC(Sve, Add, -1, -1, false, {INS_sve_add, INS_sve_add, INS_sve_add, INS_sve_add, INS_sve_add, INS_sve_add, INS_sve_add, INS_sve_add, INS_sve_fadd, INS_sve_fadd}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_OptionalEmbeddedMaskedOperation|HW_Flag_HasRMWSemantics|HW_Flag_LowMaskedOperation)
HARDWARE_INTRINSIC(Sve, ConditionalSelect, -1, 3, true, {INS_sve_sel, INS_sve_sel, INS_sve_sel, INS_sve_sel, INS_sve_sel, INS_sve_sel, INS_sve_sel, INS_sve_sel, INS_sve_sel, INS_sve_sel}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_ExplicitMaskedOperation|HW_Flag_SupportsContainment)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskByte, -1, 1, false, {INS_invalid, INS_sve_ptrue, INS_invalid, INS_invalid, 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, CreateTrueMaskDouble, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue}, HW_Category_EnumPattern, HW_Flag_Scalable|HW_Flag_HasImmediateOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskInt16, -1, 1, false, {INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid, 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, CreateTrueMaskInt32, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue, 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, CreateTrueMaskInt64, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid, INS_invalid, INS_invalid}, HW_Category_EnumPattern, HW_Flag_Scalable|HW_Flag_HasImmediateOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskSByte, -1, 1, false, {INS_sve_ptrue, INS_invalid, INS_invalid, INS_invalid, 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, CreateTrueMaskSingle, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid}, HW_Category_EnumPattern, HW_Flag_Scalable|HW_Flag_HasImmediateOperand|HW_Flag_ReturnsPerElementMask)
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, Count16BitElements, 0, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_cnth, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Scalar, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_SpecialCodeGen|HW_Flag_NoFloatingPointUsed)
HARDWARE_INTRINSIC(Sve, Count32BitElements, 0, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_cntw, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Scalar, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_SpecialCodeGen|HW_Flag_NoFloatingPointUsed)
HARDWARE_INTRINSIC(Sve, Count64BitElements, 0, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_cntd, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Scalar, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_SpecialCodeGen|HW_Flag_NoFloatingPointUsed)
HARDWARE_INTRINSIC(Sve, Count8BitElements, 0, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_cntb, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Scalar, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_SpecialCodeGen|HW_Flag_NoFloatingPointUsed)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskByte, -1, 1, false, {INS_invalid, INS_sve_ptrue, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskDouble, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskInt16, -1, 1, false, {INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskInt32, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskInt64, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskSByte, -1, 1, false, {INS_sve_ptrue, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_ReturnsPerElementMask)
HARDWARE_INTRINSIC(Sve, CreateTrueMaskSingle, -1, 1, false, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_sve_ptrue, INS_invalid}, HW_Category_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|HW_Flag_ReturnsPerElementMask)
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_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|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_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|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_SIMD, HW_Flag_Scalable|HW_Flag_HasEnumOperand|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
4 changes: 4 additions & 0 deletions src/coreclr/jit/lowerarmarch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3295,6 +3295,10 @@ void Lowering::ContainCheckHWIntrinsic(GenTreeHWIntrinsic* node)
case NI_Sve_CreateTrueMaskUInt16:
case NI_Sve_CreateTrueMaskUInt32:
case NI_Sve_CreateTrueMaskUInt64:
case NI_Sve_Count16BitElements:
case NI_Sve_Count32BitElements:
case NI_Sve_Count64BitElements:
case NI_Sve_Count8BitElements:
assert(hasImmediateOperand);
assert(varTypeIsIntegral(intrin.op1));
if (intrin.op1->IsCnsIntOrI())
Expand Down
4 changes: 4 additions & 0 deletions src/coreclr/jit/lsraarm64.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1464,6 +1464,10 @@ int LinearScan::BuildHWIntrinsic(GenTreeHWIntrinsic* intrinsicTree, int* pDstCou
case NI_Sve_CreateTrueMaskUInt16:
case NI_Sve_CreateTrueMaskUInt32:
case NI_Sve_CreateTrueMaskUInt64:
case NI_Sve_Count16BitElements:
case NI_Sve_Count32BitElements:
case NI_Sve_Count64BitElements:
case NI_Sve_Count8BitElements:
needBranchTargetReg = !intrin.op1->isContainedIntOrIImmed();
break;

Expand Down
17 changes: 16 additions & 1 deletion src/coreclr/vm/codeman.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,10 @@
//
// codeman.cpp - a managment class for handling multiple code managers
//

#if defined(TARGET_LINUX) && (defined(DEBUG) || defined(_DEBUG))
#include <sys/prctl.h>
#include <sys/syscall.h>
#endif // defined(TARGET_LINUX) && (defined(DEBUG) || defined(_DEBUG))
#include "common.h"
#include "jitinterface.h"
#include "corjit.h"
Expand Down Expand Up @@ -1525,6 +1528,18 @@ void EEJitManager::SetCpuInfo()

if (((cpuFeatures & ARM64IntrinsicConstants_Sve) != 0) && CLRConfig::GetConfigValue(CLRConfig::EXTERNAL_EnableArm64Sve))
{
#if defined(TARGET_LINUX) && (defined(DEBUG) || defined(_DEBUG))
int maxVectorLength = CLRConfig::GetConfigValue(CLRConfig::EXTERNAL_MaxVectorLength);

// Limit the SVE vector length to 'maxVectorLength' if the underlying hardware offers longer vectors.
if ((prctl(PR_SVE_GET_VL, 0,0,0,0) & PR_SVE_VL_LEN_MASK) > maxVectorLength)
{
if (prctl(PR_SVE_SET_VL, (maxVectorLength | PR_SVE_VL_INHERIT), 0, 0, 0) == -1)
{
LogErrorToHost("LoadAndInitializeJIT: prctl() FAILED - unable to set maxVectorLength to %d", maxVectorLength);
}
}
#endif // defined(TARGET_LINUX) && (defined(DEBUG) || defined(_DEBUG))
CPUCompileFlags.Set(InstructionSet_Sve);
}

Expand Down
Loading
Loading