[MLAS] Add int8 extreme-value coverage for ARM64 SymmQgemm - #31606
Open
kjg0724 wants to merge 1 commit into
Open
[MLAS] Add int8 extreme-value coverage for ARM64 SymmQgemm#31606kjg0724 wants to merge 1 commit into
kjg0724 wants to merge 1 commit into
Conversation
MlasSymmQgemmTest only ever fills A and B via MatrixGuardBuffer's default fill, which is confined to [21, 64) -- even ExecuteLong's large M/N/K sweep never exercises an int8 extreme. Add SymmQgemmS8SignedInputTest, mirroring the QgemmS8U8SignedInputTest pattern from microsoft#29787, with A and B explicitly filled with int8 extremes across a K/M/N/offa grid sized around the kernel's block structure (PackedK=16, StrideM=4, N aligned to 16). Test-only change, nothing under core/mlas/lib/ is touched.
|
Azure Pipelines: There may be pipelines that require an authorized user to comment /azp run to run. |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
The ARM64 symmetric quantized GEMM tests (
MlasSymmQgemmTest) only ever feed A and B withMatrixGuardBuffer::GetBuffer()'s default fill, which is confined to[21, 64)— evenExecuteLong's large M/N/K sweep never touches an int8 extreme. This addsSymmQgemmS8SignedInputTest, mirroring theQgemmS8U8SignedInputTestpattern from #29787, with A and B both explicitly filled with int8 extremes (-128, -1, 0, 1, 127, ...) across a K/M/N/offa grid sized around the kernel's actual block structure (PackedK=16,StrideM=4, N aligned to 16).This is a test-only change; nothing under
onnxruntime/core/mlas/lib/is touched.While adding these tests, I found a correctness bug in the plain-NEON (non-dotprod)
MlasSymQgemmS8KernelNeonkernel and filed it separately as #31573. The new tests here pass because they exercise the SDOT dispatch (default on both machines I verified this on); the NEON dispatch only gets exercised on ARM64 cores without dot-product support.Testing
onnxruntime_mlas_test, full suite, no regressions:*SymmQGemmS8_Int32_SignedInput*: 1248/1248 passed on both (624 cases x SingleThread/Threaded).