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

ARM64-SVE: Allow SVE ops to re-use the same registers #107084

Merged
merged 5 commits into from
Sep 13, 2024
Merged

ARM64-SVE: Allow SVE ops to re-use the same registers #107084

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
ba1d58a
ARM64-SVE: Allow SVE ops to re-use the same registers
a74nh Aug 28, 2024
6b64ef8
Add Sve.IsSupported
a74nh Sep 9, 2024
41ef196
Merge main
a74nh Sep 13, 2024
554d139
restore an assert
a74nh Sep 13, 2024
b8b0fb5
better asserts
a74nh Sep 13, 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
6 changes: 2 additions & 4 deletions src/coreclr/jit/hwintrinsiccodegenarm64.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1114,17 +1114,15 @@ void CodeGen::genHWIntrinsic(GenTreeHWIntrinsic* node)
{
if (HWIntrinsicInfo::IsExplicitMaskedOperation(intrin.id))
{
assert((targetReg == op1Reg) || (targetReg != op1Reg));
Copy link
Member

Choose a reason for hiding this comment

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

these are OK to have.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

These ones are hit by the new hwintrinsic tests:

------------------- {'JitStressRegs': '1'} -------------------
Errors:

Assert failure(PID 366410 [0x0005974a], Thread: 366410 [0x5974a]): Assertion failed '(targetReg == op1Reg) || (targetReg != op3Reg)' in 'JIT.HardwareIntrinsics.Arm._Sve.SimpleTernaryOpTest__Sve_ConditionalExtractAfterLastActiveElementAndReplicate_float:RunSameClassFldScenario():this' during 'Generate code' (IL size 92; hash 0xa42df59e; FullOpts)

    File: /home/alahay01/dotnet/runtime_sve_api/src/coreclr/jit/hwintrinsiccodegenarm64.cpp:1118
    Image: /home/alahay01/dotnet/runtime_sve_api/artifacts/tests/coreclr/linux.arm64.Checked/Tests/Core_Root/corerun

Copy link
Member

@tannergooding tannergooding Sep 13, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

We need some kind of assert validating the mov is valid.

Since we're doing mov targetReg, op2Reg, then when targetReg == op1Reg and op1Reg != op2Reg, then the we would overwrite op1 and produce in incorrect results.

So, therefore, we expect that targetReg == op2Reg so no mov is emitted -or- op1/op2 have been marked delayFree

The assert therefore needs to be something like assert((targetReg == op2Reg) || ((targetReg != op1Reg) && (targetReg != op3Reg)). Which validates no move or no overwrite

Copy link
Member

Choose a reason for hiding this comment

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

All paths need some kind of similar assert validating no move or no overwrite, if a move is emitted

Copy link
Member

Choose a reason for hiding this comment

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

All paths need some kind of similar assert validating no move or no overwrite, if a move is emitted

Agree.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Done replaced this assert as suggested, and the next one with a similar assert

Copy link
Contributor Author

Choose a reason for hiding this comment

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

confirmed stress tests passing.

assert((targetReg == op1Reg) || (targetReg != op3Reg));
assert((targetReg == op2Reg) || ((targetReg != op1Reg) && (targetReg != op3Reg)));

GetEmitter()->emitIns_Mov(INS_mov, emitTypeSize(node), targetReg, op2Reg,
/* canSkip */ true);
GetEmitter()->emitIns_R_R_R(ins, emitSize, targetReg, op1Reg, op3Reg, opt);
}
else
{
assert((targetReg == op1Reg) || (targetReg != op2Reg));
a74nh marked this conversation as resolved.
Show resolved Hide resolved
assert((targetReg == op1Reg) || (targetReg != op3Reg));
assert((targetReg == op1Reg) || ((targetReg != op2Reg) && (targetReg != op3Reg)));

GetEmitter()->emitIns_Mov(INS_mov, emitTypeSize(node), targetReg, op1Reg,
/* canSkip */ true);
Expand Down
29 changes: 29 additions & 0 deletions src/tests/JIT/HardwareIntrinsics/Arm/Shared/_SveBinaryOpTestTemplate.template
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,9 +43,15 @@ namespace JIT.HardwareIntrinsics.Arm
// Validates passing a local works, using Unsafe.Read
test.RunLclVarScenario_UnsafeRead();

// Validates using the same local multiple times works, using Unsafe.Read
test.RunSameLclVarScenario_UnsafeRead();

// Validates passing an instance member of a class works
test.RunClassFldScenario();

// Validates using the same instance member of a class multiple times works
test.RunSameClassFldScenario();

// Validates passing the field of a local struct works
test.RunStructLclFldScenario();

Expand Down Expand Up @@ -250,6 +256,19 @@ namespace JIT.HardwareIntrinsics.Arm
ValidateResult(op1, op2, _dataTable.outArrayPtr);
}

public void RunSameLclVarScenario_UnsafeRead()
{
TestLibrary.TestFramework.BeginScenario(nameof(RunSameLclVarScenario_UnsafeRead));

var op = Unsafe.Read<{Op1VectorType}<{Op1BaseType}>>(_dataTable.inArray1Ptr);
var op1 = op;
var op2 = op;
var result = {Isa}.{Method}(op1, op2);

Unsafe.Write(_dataTable.outArrayPtr, result);
ValidateResult(op1, op2, _dataTable.outArrayPtr);
}

public void RunClassFldScenario()
{
TestLibrary.TestFramework.BeginScenario(nameof(RunClassFldScenario));
Expand All @@ -260,6 +279,16 @@ namespace JIT.HardwareIntrinsics.Arm
ValidateResult(_fld1, _fld2, _dataTable.outArrayPtr);
}

public void RunSameClassFldScenario()
{
TestLibrary.TestFramework.BeginScenario(nameof(RunSameClassFldScenario));

var result = {Isa}.{Method}(_fld1, _fld1);

Unsafe.Write(_dataTable.outArrayPtr, result);
ValidateResult(_fld1, _fld1, _dataTable.outArrayPtr);
}

public void RunStructLclFldScenario()
{
TestLibrary.TestFramework.BeginScenario(nameof(RunStructLclFldScenario));
Expand Down
30 changes: 30 additions & 0 deletions src/tests/JIT/HardwareIntrinsics/Arm/Shared/_SveTernOpTestTemplate.template
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,9 +43,15 @@ namespace JIT.HardwareIntrinsics.Arm
// Validates passing a local works, using Unsafe.Read
test.RunLclVarScenario_UnsafeRead();

// Validates using the same local multiple times works, using Unsafe.Read
test.RunSameLclVarScenario_UnsafeRead();

// Validates passing an instance member of a class works
test.RunClassFldScenario();

// Validates using the same instance member of a class multiple times works
test.RunSameClassFldScenario();

// Validates passing the field of a local struct works
test.RunStructLclFldScenario();

Expand Down Expand Up @@ -277,6 +283,20 @@ namespace JIT.HardwareIntrinsics.Arm
ValidateResult(op1, op2, op3, _dataTable.outArrayPtr);
}

public void RunSameLclVarScenario_UnsafeRead()
{
TestLibrary.TestFramework.BeginScenario(nameof(RunSameLclVarScenario_UnsafeRead));

var op = Unsafe.Read<{Op1VectorType}<{Op1BaseType}>>(_dataTable.inArray1Ptr);
var op1 = op;
var op2 = op;
var op3 = op;
var result = {Isa}.{Method}(op1, op2, op3);

Unsafe.Write(_dataTable.outArrayPtr, result);
ValidateResult(op1, op2, op3, _dataTable.outArrayPtr);
}

public void RunClassFldScenario()
{
TestLibrary.TestFramework.BeginScenario(nameof(RunClassFldScenario));
Expand All @@ -287,6 +307,16 @@ namespace JIT.HardwareIntrinsics.Arm
ValidateResult(_fld1, _fld2, _fld3, _dataTable.outArrayPtr);
}

public void RunSameClassFldScenario()
{
TestLibrary.TestFramework.BeginScenario(nameof(RunSameClassFldScenario));

var result = {Isa}.{Method}(_fld1, _fld1, _fld1);

Unsafe.Write(_dataTable.outArrayPtr, result);
ValidateResult(_fld1, _fld1, _fld1, _dataTable.outArrayPtr);
}

public void RunStructLclFldScenario()
{
TestLibrary.TestFramework.BeginScenario(nameof(RunStructLclFldScenario));
Expand Down
50 changes: 50 additions & 0 deletions src/tests/JIT/Regression/JitBlue/Runtime_106866/Runtime_106866.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,50 @@
// Licensed to the .NET Foundation under one or more agreements.
// The .NET Foundation licenses this file to you under the MIT license.

using Xunit;
using System.Runtime.CompilerServices;

// Generated by Fuzzlyn v2.3 on 2024-08-23 10:04:52
// Run on Arm64 Windows
// Seed: 12028719405363964033-vectort,vector64,vector128,armsve
// Reduced from 60.4 KiB to 0.7 KiB in 00:00:33
// Hits JIT assert in Release:
// Assertion failed '(targetReg == op1Reg) || (targetReg != op3Reg)' in 'S0:M3():this' during 'Generate code' (IL size 57; hash 0x4541fc9f; FullOpts)
//
// File: C:\dev\dotnet\runtime2\src\coreclr\jit\hwintrinsiccodegenarm64.cpp Line: 1128
//
using System;
using System.Numerics;
using System.Runtime.Intrinsics;
using System.Runtime.Intrinsics.Arm;

public struct S0
{
public bool F0;
public Vector<sbyte> F2;
public void M3()
{
if (Sve.IsSupported)
{
var vr0 = this.F2;
var vr1 = this.F2;
var vr2 = this.F2;
this.F2 = Sve.Splice(vr0, vr1, vr2);
Consume(this.F0);
}
}

[MethodImpl(MethodImplOptions.NoInlining)]
static void Consume<T>(T val)
{
}
}

public class Runtime_106866
{
[Fact]
public static void TestEntryPoint()
{
new S0().M3();
}
}
9 changes: 9 additions & 0 deletions src/tests/JIT/Regression/JitBlue/Runtime_106866/Runtime_106866.csproj
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,9 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<Optimize>True</Optimize>
<NoWarn>$(NoWarn),SYSLIB5003</NoWarn>
</PropertyGroup>
<ItemGroup>
<Compile Include="$(MSBuildProjectName).cs" />
</ItemGroup>
</Project>
Loading