Skip to content
This repository was archived by the owner on Jan 23, 2023. It is now read-only.

Fix SSE4.1 encoding with containment form #16346

Merged
tannergooding merged 4 commits into from
Feb 13, 2018
Merged

Fix SSE4.1 encoding with containment form #16346

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
76c018b
reenable AVX2 compareEqual.Int64
FeiPengIntel Feb 12, 2018
5eec799
Implement SSE4.1 CompareEqual
FeiPengIntel Feb 12, 2018
2878baa
Implement SSE4.2 CompareGreaterThan
FeiPengIntel Feb 12, 2018
28fccbc
Fix SSE4.1 encoding
FeiPengIntel Feb 12, 2018
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
34 changes: 24 additions & 10 deletions src/jit/emitxarch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12072,21 +12072,35 @@ size_t emitter::emitOutputInstr(insGroup* ig, instrDesc* id, BYTE** dp)
case IF_RWR_ARD:
case IF_RRW_ARD:
case IF_RWR_RRD_ARD:
code = insCodeRM(ins);
code = AddVexPrefixIfNeeded(ins, code, size);
regcode = (insEncodeReg345(ins, id->idReg1(), size, &code) << 8);
dst = emitOutputAM(dst, id, code | regcode);
sz = emitSizeOfInsDsc(id);
code = insCodeRM(ins);
if (Is4ByteSSE4OrAVXInstruction(ins))
{
dst = emitOutputAM(dst, id, code);
}
else
{
code = AddVexPrefixIfNeeded(ins, code, size);
regcode = (insEncodeReg345(ins, id->idReg1(), size, &code) << 8);
dst = emitOutputAM(dst, id, code | regcode);
}
sz = emitSizeOfInsDsc(id);
break;

case IF_RWR_RRD_ARD_CNS:
{
emitGetInsAmdCns(id, &cnsVal);
code = insCodeRM(ins);
code = AddVexPrefixIfNeeded(ins, code, size);
regcode = (insEncodeReg345(ins, id->idReg1(), size, &code) << 8);
dst = emitOutputAM(dst, id, code | regcode, &cnsVal);
sz = emitSizeOfInsDsc(id);
code = insCodeRM(ins);
if (Is4ByteSSE4OrAVXInstruction(ins))
{
dst = emitOutputAM(dst, id, code, &cnsVal);
}
else
{
code = AddVexPrefixIfNeeded(ins, code, size);
regcode = (insEncodeReg345(ins, id->idReg1(), size, &code) << 8);
dst = emitOutputAM(dst, id, code | regcode, &cnsVal);
}
sz = emitSizeOfInsDsc(id);
break;
}

Expand Down
4 changes: 3 additions & 1 deletion src/jit/hwintrinsiclistxarch.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -189,12 +189,14 @@ HARDWARE_INTRINSIC(SSSE3_IsSupported, "get_IsSupp

// SSE41 Intrinsics
HARDWARE_INTRINSIC(SSE41_IsSupported, "get_IsSupported", SSE41, -1, 0, 0, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_IsSupportedProperty, HW_Flag_NoFlag)
HARDWARE_INTRINSIC(SSE41_Multiply, "Multiply", SSE41, -1, 16, 2, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_pmuldq, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SimpleSIMD, HW_Flag_Commutative)
HARDWARE_INTRINSIC(SSE41_BlendVariable, "BlendVariable", SSE41, -1, 16, 3, {INS_pblendvb, INS_pblendvb, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_blendvps, INS_blendvpd}, HW_Category_SimpleSIMD, HW_Flag_NoFlag)
HARDWARE_INTRINSIC(SSE41_CompareEqual, "CompareEqual", SSE41, -1, 16, 2, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_pcmpeqq, INS_pcmpeqq, INS_invalid, INS_invalid}, HW_Category_SimpleSIMD, HW_Flag_Commutative)
HARDWARE_INTRINSIC(SSE41_Multiply, "Multiply", SSE41, -1, 16, 2, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_pmuldq, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SimpleSIMD, HW_Flag_Commutative)

// SSE42 Intrinsics
HARDWARE_INTRINSIC(SSE42_IsSupported, "get_IsSupported", SSE42, -1, 0, 0, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_IsSupportedProperty, HW_Flag_NoFlag)
HARDWARE_INTRINSIC(SSE42_Crc32, "Crc32", SSE42, -1, 0, 2, {INS_invalid, INS_crc32, INS_invalid, INS_crc32, INS_invalid, INS_crc32, INS_invalid, INS_crc32, INS_invalid, INS_invalid}, HW_Category_Scalar, HW_Flag_NoFloatingPointUsed)
HARDWARE_INTRINSIC(SSE42_CompareGreaterThan, "CompareGreaterThan", SSE42, -1, 16, 2, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_pcmpgtq, INS_invalid, INS_invalid, INS_invalid}, HW_Category_SimpleSIMD, HW_Flag_NoFlag)

// ************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************
// Intrinsic ID Function name ISA ival SIMD size NumArg instructions Category Flags
Expand Down
6 changes: 3 additions & 3 deletions tests/src/JIT/HardwareIntrinsics/X86/Avx/BlendVariable.Double.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,7 +97,7 @@ static SimpleTernaryOpTest__BlendVariableDouble()
{
var random = new Random();

for (var i = 0; i < ElementCount; i++) { _data1[i] = (double)(random.NextDouble()); _data2[i] = (double)(random.NextDouble()); _data3[i] = (double)(((i % 2) == 0) ? -0.0E0 : 1.0E0); }
for (var i = 0; i < ElementCount; i++) { _data1[i] = (double)(random.NextDouble()); _data2[i] = (double)(random.NextDouble()); _data3[i] = (double)(((i % 2) == 0) ? -0.0 : 1.0); }
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Double>, byte>(ref _clsVar1), ref Unsafe.As<Double, byte>(ref _data2[0]), VectorSize);
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Double>, byte>(ref _clsVar2), ref Unsafe.As<Double, byte>(ref _data1[0]), VectorSize);
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Double>, byte>(ref _clsVar3), ref Unsafe.As<Double, byte>(ref _data3[0]), VectorSize);
Expand All @@ -109,12 +109,12 @@ public SimpleTernaryOpTest__BlendVariableDouble()

var random = new Random();

for (var i = 0; i < ElementCount; i++) { _data1[i] = (double)(random.NextDouble()); _data2[i] = (double)(random.NextDouble()); _data3[i] = (double)(((i % 2) == 0) ? -0.0E0 : 1.0E0); }
for (var i = 0; i < ElementCount; i++) { _data1[i] = (double)(random.NextDouble()); _data2[i] = (double)(random.NextDouble()); _data3[i] = (double)(((i % 2) == 0) ? -0.0 : 1.0); }
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Double>, byte>(ref _fld1), ref Unsafe.As<Double, byte>(ref _data1[0]), VectorSize);
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Double>, byte>(ref _fld2), ref Unsafe.As<Double, byte>(ref _data2[0]), VectorSize);
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Double>, byte>(ref _fld3), ref Unsafe.As<Double, byte>(ref _data3[0]), VectorSize);

for (var i = 0; i < ElementCount; i++) { _data1[i] = (double)(random.NextDouble()); _data2[i] = (double)(random.NextDouble()); _data3[i] = (double)(((i % 2) == 0) ? -0.0E0 : 1.0E0); }
for (var i = 0; i < ElementCount; i++) { _data1[i] = (double)(random.NextDouble()); _data2[i] = (double)(random.NextDouble()); _data3[i] = (double)(((i % 2) == 0) ? -0.0 : 1.0); }
_dataTable = new SimpleTernaryOpTest__DataTable<Double>(_data1, _data2, _data3, new Double[ElementCount], VectorSize);
}

Expand Down
6 changes: 3 additions & 3 deletions tests/src/JIT/HardwareIntrinsics/X86/Avx/BlendVariable.Single.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,7 +97,7 @@ static SimpleTernaryOpTest__BlendVariableSingle()
{
var random = new Random();

for (var i = 0; i < ElementCount; i++) { _data1[i] = (float)(random.NextDouble()); _data2[i] = (float)(random.NextDouble()); _data3[i] = (float)(((i % 2) == 0) ? -0.0E0 : 1.0E0); }
for (var i = 0; i < ElementCount; i++) { _data1[i] = (float)(random.NextDouble()); _data2[i] = (float)(random.NextDouble()); _data3[i] = (float)(((i % 2) == 0) ? -0.0 : 1.0); }
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Single>, byte>(ref _clsVar1), ref Unsafe.As<Single, byte>(ref _data2[0]), VectorSize);
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Single>, byte>(ref _clsVar2), ref Unsafe.As<Single, byte>(ref _data1[0]), VectorSize);
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Single>, byte>(ref _clsVar3), ref Unsafe.As<Single, byte>(ref _data3[0]), VectorSize);
Expand All @@ -109,12 +109,12 @@ public SimpleTernaryOpTest__BlendVariableSingle()

var random = new Random();

for (var i = 0; i < ElementCount; i++) { _data1[i] = (float)(random.NextDouble()); _data2[i] = (float)(random.NextDouble()); _data3[i] = (float)(((i % 2) == 0) ? -0.0E0 : 1.0E0); }
for (var i = 0; i < ElementCount; i++) { _data1[i] = (float)(random.NextDouble()); _data2[i] = (float)(random.NextDouble()); _data3[i] = (float)(((i % 2) == 0) ? -0.0 : 1.0); }
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Single>, byte>(ref _fld1), ref Unsafe.As<Single, byte>(ref _data1[0]), VectorSize);
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Single>, byte>(ref _fld2), ref Unsafe.As<Single, byte>(ref _data2[0]), VectorSize);
Unsafe.CopyBlockUnaligned(ref Unsafe.As<Vector256<Single>, byte>(ref _fld3), ref Unsafe.As<Single, byte>(ref _data3[0]), VectorSize);

for (var i = 0; i < ElementCount; i++) { _data1[i] = (float)(random.NextDouble()); _data2[i] = (float)(random.NextDouble()); _data3[i] = (float)(((i % 2) == 0) ? -0.0E0 : 1.0E0); }
for (var i = 0; i < ElementCount; i++) { _data1[i] = (float)(random.NextDouble()); _data2[i] = (float)(random.NextDouble()); _data3[i] = (float)(((i % 2) == 0) ? -0.0 : 1.0); }
_dataTable = new SimpleTernaryOpTest__DataTable<Single>(_data1, _data2, _data3, new Single[ElementCount], VectorSize);
}

Expand Down
2 changes: 2 additions & 0 deletions tests/src/JIT/HardwareIntrinsics/X86/Avx2/Avx2_r.csproj
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,9 +58,11 @@
<Compile Include="CompareEqual.Byte.cs" />
<Compile Include="CompareEqual.Int16.cs" />
<Compile Include="CompareEqual.Int32.cs" />
<Compile Include="CompareEqual.Int64.cs" />
<Compile Include="CompareEqual.SByte.cs" />
<Compile Include="CompareEqual.UInt16.cs" />
<Compile Include="CompareEqual.UInt32.cs" />
<Compile Include="CompareEqual.UInt64.cs" />
<Compile Include="CompareGreaterThan.Int16.cs" />
<Compile Include="CompareGreaterThan.Int32.cs" />
<Compile Include="CompareGreaterThan.Int64.cs" />
Expand Down
4 changes: 3 additions & 1 deletion tests/src/JIT/HardwareIntrinsics/X86/Avx2/Avx2_ro.csproj
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@
<Service Include="{82A7F48D-3B50-4B1E-B82E-3ADA8210C358}" />
</ItemGroup>
<ItemGroup>
<Compile Include="Add.Byte.cs" />
<Compile Include="Add.Byte.cs" />
<Compile Include="Add.Int16.cs" />
<Compile Include="Add.Int32.cs" />
<Compile Include="Add.Int64.cs" />
Expand Down Expand Up @@ -58,9 +58,11 @@
<Compile Include="CompareEqual.Byte.cs" />
<Compile Include="CompareEqual.Int16.cs" />
<Compile Include="CompareEqual.Int32.cs" />
<Compile Include="CompareEqual.Int64.cs" />
<Compile Include="CompareEqual.SByte.cs" />
<Compile Include="CompareEqual.UInt16.cs" />
<Compile Include="CompareEqual.UInt32.cs" />
<Compile Include="CompareEqual.UInt64.cs" />
<Compile Include="CompareGreaterThan.Int16.cs" />
<Compile Include="CompareGreaterThan.Int32.cs" />
<Compile Include="CompareGreaterThan.Int64.cs" />
Expand Down
Loading