Skip to content

Accelerate Vector512 ConvertToSingle and ConvertToInt32. #84932

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

Merged
tannergooding merged 2 commits into from
Apr 20, 2023
Merged

Accelerate Vector512 ConvertToSingle and ConvertToInt32. #84932

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bc17f36
Accelerate Vector512 `ConvertToSingle` and `ConvertToInt32`.
anthonycanino Apr 17, 2023
62ebb42
Moving intrinsics to proper place.
anthonycanino Apr 18, 2023
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: 2 additions & 0 deletions src/coreclr/jit/hwintrinsiclistxarch.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -263,6 +263,8 @@ HARDWARE_INTRINSIC(Vector512, BitwiseOr,
HARDWARE_INTRINSIC(Vector512, Create, 64, -1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Helper, HW_Flag_SpecialImport|HW_Flag_NoCodeGen)
HARDWARE_INTRINSIC(Vector512, CreateScalar, 64, -1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Helper, HW_Flag_SpecialImport|HW_Flag_NoCodeGen)
HARDWARE_INTRINSIC(Vector512, CreateScalarUnsafe, 64, 1, {INS_movd, INS_movd, INS_movd, INS_movd, INS_movd, INS_movd, INS_movd, INS_movd, INS_movss, INS_movsd_simd}, HW_Category_SIMDScalar, HW_Flag_SpecialImport|HW_Flag_SpecialCodeGen|HW_Flag_NoRMWSemantics)
HARDWARE_INTRINSIC(Vector512, ConvertToSingle, 64, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Helper, HW_Flag_SpecialImport|HW_Flag_NoCodeGen|HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(Vector512, ConvertToInt32, 64, 1, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Helper, HW_Flag_SpecialImport|HW_Flag_NoCodeGen|HW_Flag_BaseTypeFromFirstArg)
HARDWARE_INTRINSIC(Vector512, Equals, 64, 2, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Helper, HW_Flag_SpecialImport|HW_Flag_BaseTypeFromFirstArg|HW_Flag_NoCodeGen)
HARDWARE_INTRINSIC(Vector512, EqualsAll, 64, 2, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Helper, HW_Flag_SpecialImport|HW_Flag_BaseTypeFromFirstArg|HW_Flag_NoCodeGen)
HARDWARE_INTRINSIC(Vector512, EqualsAny, 64, 2, {INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid, INS_invalid}, HW_Category_Helper, HW_Flag_SpecialImport|HW_Flag_BaseTypeFromFirstArg|HW_Flag_NoCodeGen)
Expand Down
33 changes: 30 additions & 3 deletions src/coreclr/jit/hwintrinsicxarch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1004,12 +1004,25 @@ GenTree* Compiler::impSpecialIntrinsic(NamedIntrinsic intrinsic,

case NI_Vector128_ConvertToInt32:
case NI_Vector256_ConvertToInt32:
case NI_Vector512_ConvertToInt32:
{
assert(sig->numArgs == 1);
assert(simdBaseType == TYP_FLOAT);

intrinsic = (simdSize == 32) ? NI_AVX_ConvertToVector256Int32WithTruncation
: NI_SSE2_ConvertToVector128Int32WithTruncation;
switch (simdSize)
{
case 16:
intrinsic = NI_SSE2_ConvertToVector128Int32WithTruncation;
break;
case 32:
intrinsic = NI_AVX_ConvertToVector256Int32WithTruncation;
break;
case 64:
intrinsic = NI_AVX512F_ConvertToVector512Int32WithTruncation;
break;
default:
unreached();
}
Comment on lines +1012 to +1025
Copy link
Copy Markdown
Member

@tannergooding tannergooding Apr 18, 2023

Choose a reason for hiding this comment

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

nit: We might not want to switch on the size because we sometimes see simdSize == 8 and simdSize == 12 for some nodes.

I don't think we currently have any for ConvertTo, but its probably better to be "safe" and do if (simdSize == 64) { } else if (simdSize == 32) { } else { } like the other paths are doing.

Copy link
Copy Markdown
Contributor

@kunalspathak kunalspathak Apr 20, 2023

Choose a reason for hiding this comment

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

@tannergooding - so for simdSize == 12 and simdSize == 8 we will end up with NI_SSE2_ConvertToVector128Int32WithTruncation which is ok?

@anthonycanino - can you address this?

Copy link
Copy Markdown
Member

@tannergooding tannergooding Apr 20, 2023

Choose a reason for hiding this comment

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

Right. We won't see it today so its not required to fix it "right now". But there are some optimizations we'd like to do in the future that would need it so we'd want to at least fix it at that time


op1 = impSIMDPopStack(retType);
retNode = gtNewSimdHWIntrinsicNode(retType, op1, intrinsic, simdBaseJitType, simdSize);
Expand All @@ -1018,12 +1031,26 @@ GenTree* Compiler::impSpecialIntrinsic(NamedIntrinsic intrinsic,

case NI_Vector128_ConvertToSingle:
case NI_Vector256_ConvertToSingle:
case NI_Vector512_ConvertToSingle:
{
assert(sig->numArgs == 1);

if (simdBaseType == TYP_INT)
{
intrinsic = (simdSize == 32) ? NI_AVX_ConvertToVector256Single : NI_SSE2_ConvertToVector128Single;
switch (simdSize)
{
case 16:
intrinsic = NI_SSE2_ConvertToVector128Single;
break;
case 32:
intrinsic = NI_AVX_ConvertToVector256Single;
break;
case 64:
intrinsic = NI_AVX512F_ConvertToVector512Single;
break;
default:
unreached();
}

op1 = impSIMDPopStack(retType);
retNode = gtNewSimdHWIntrinsicNode(retType, op1, intrinsic, simdBaseJitType, simdSize);
Expand Down