Skip to content

Ensure partial Vector256 acceleration is still possible on AVX only hardware #73720

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
Aug 12, 2022
Merged

Ensure partial Vector256 acceleration is still possible on AVX only hardware #73720

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
88fd001
Ensure that Vector256 APIs exposed in .NET Core 3.1 remain accelerate…
tannergooding Aug 10, 2022
35cdcd5
Ensure new Vector256 APIs that are trivially limited to Avx only are …
tannergooding Aug 10, 2022
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: 2 additions & 2 deletions src/coreclr/jit/hwintrinsic.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -383,8 +383,8 @@ NamedIntrinsic HWIntrinsicInfo::lookupId(Compiler* comp,
NamedIntrinsic ni = intrinsicInfo.id;

#if defined(TARGET_XARCH)
// on AVX1-only CPUs we only support NI_Vector256_Create intrinsic in Vector256
if (isLimitedVector256Isa && (ni != NI_Vector256_Create))
// on AVX1-only CPUs we only support a subset of intrinsics in Vector256
if (isLimitedVector256Isa && !AvxOnlyCompatible(ni))
{
return NI_Illegal;
}
Expand Down
14 changes: 13 additions & 1 deletion src/coreclr/jit/hwintrinsic.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -147,7 +147,11 @@ enum HWIntrinsicFlag : unsigned int
// Returns Per-Element Mask
// the intrinsic returns a vector containing elements that are either "all bits set" or "all bits clear"
// this output can be used as a per-element mask
HW_Flag_ReturnsPerElementMask = 0x20000
HW_Flag_ReturnsPerElementMask = 0x20000,

// AvxOnlyCompatible
// the intrinsic can be used on hardware with AVX but not AVX2 support
HW_Flag_AvxOnlyCompatible = 0x40000,

#elif defined(TARGET_ARM64)
// The intrinsic has an immediate operand
Expand Down Expand Up @@ -658,6 +662,14 @@ struct HWIntrinsicInfo
#endif
}

#if defined(TARGET_XARCH)
static bool AvxOnlyCompatible(NamedIntrinsic id)
{
HWIntrinsicFlag flags = lookupFlags(id);
return (flags & HW_Flag_AvxOnlyCompatible) != 0;
}
#endif

static bool BaseTypeFromFirstArg(NamedIntrinsic id)
{
HWIntrinsicFlag flags = lookupFlags(id);
Expand Down
Loading