Accelerate additional cross platform hardware intrinsics #61649
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
dotnet-issue-labeler
bot
added
the
area-CodeGen-coreclr
|
Tagging subscribers to this area: @JulieLeeMSFT Issue Detailsnull
|
tannergooding
commented
Nov 17, 2021
Comment on lines
+829
to
+841
| case NI_VectorT128_ConvertToUInt32: | ||
| { | ||
| assert(simdBaseType == TYP_FLOAT); | ||
| return gtNewSimdHWIntrinsicNode(retType, op1, NI_AdvSimd_ConvertToUInt32RoundToZero, | ||
| simdBaseJitType, simdSize, /* isSimdAsHWIntrinsic */ true); | ||
| } | ||
|
|
||
| case NI_VectorT128_ConvertToUInt64: | ||
| { | ||
| assert(simdBaseType == TYP_DOUBLE); | ||
| return gtNewSimdHWIntrinsicNode(retType, op1, NI_AdvSimd_Arm64_ConvertToUInt64RoundToZero, | ||
| simdBaseJitType, simdSize, /* isSimdAsHWIntrinsic */ true); | ||
| } |
There was a problem hiding this comment.
It's worth calling out these weren't accelerated at all on ARM64 before; and now they are and via a single instruction.
tannergooding
commented
Nov 17, 2021
| SIMD_INTRINSIC("ConvertToInt32", false, ConvertToInt32, "ConvertToInt32", TYP_STRUCT, 1, {TYP_STRUCT, TYP_UNDEF, TYP_UNDEF}, {TYP_FLOAT, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF}) | ||
| // Convert double to long | ||
| SIMD_INTRINSIC("ConvertToInt64", false, ConvertToInt64, "ConvertToInt64", TYP_STRUCT, 1, {TYP_STRUCT, TYP_UNDEF, TYP_UNDEF}, {TYP_DOUBLE, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF, TYP_UNDEF}) | ||
|
|
There was a problem hiding this comment.
We're getting near the point that the rest of this "legacy" SIMD intrinsic support can be removed entirely as nearly everything has moved onto SIMDAsHWIntrinsic now.
tannergooding
commented
Nov 17, 2021
| if (Sse2.IsSupported) | ||
| { | ||
| // Based on __m256d int64_to_double_fast_precise(const __m256i v) | ||
| // from https://stackoverflow.com/a/41223013/12860347. CC BY-SA 4.0 |
There was a problem hiding this comment.
The three bits of code here are new algorithms and are significantly faster than the previous.
They are also correct where-as various inputs on the old algorithm would actually return a different result as compared to the scalar versions.
|
Only diffs for
|
|
Test failures are because the current RyuJIT algorithm for This can be repro'd using
|
tannergooding
force-pushed
the
xplat-hwintrin
branch
3 times, most recently
from
ecbf81c
to
073a445
Compare
tannergooding
force-pushed
the
xplat-hwintrin
branch
from
073a445
to
e0b3289
Compare
tannergooding
changed the title
…nstant and return true where supported
…ed versions on x86/x64
…han long.MaxValue
tannergooding
force-pushed
the
xplat-hwintrin
branch
from
e0b3289
to
8202352
Compare
|
Rebased onto dotnet/main; this is still pending review and would be very beneficial to get merged so the PRs adding new SIMD logic can trivially support ARM from the start. |
echesakov
approved these changes
Jan 4, 2022
src/libraries/System.Private.CoreLib/src/ILLink/ILLink.Substitutions.NoArmIntrinsics.xml
Show resolved
Hide resolved
This was referenced Jan 6, 2022
|
Improvement on win-arm64: dotnet/perf-autofiling-issues#2974 |
ghost
locked as resolved and limited conversation to collaborators
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This continues the work on #49397 which started with #53450 and #60094
In particular, this moves
to be implemented using the general SIMDAsHWIntrinsic logic and adding then having the new APIs in Vector64/128/256 use the same shared entry points.
There will likely be one or two more PRs after this one covering:
Sumor approved but NYI such asShiftLeft/ShiftRightOnce this is in, the library side work to switch over to using the xplat APIs can also happen.