docs: auto-generated, CI-gated kernel × platform support matrix#716
Merged
Conversation
Phase 6 of the Native-parity kernel work. Adds KernelSupportMatrixTest (skainet-backend-native-cpu jvmTest, sees all three providers) that renders the kernel × platform matrix and gates drift in the scalar floor: the all-platform baseline coverage is auto-derived from ScalarKernelProvider.supports(...), so adding/removing a scalar packed kernel without updating the docs fails the test (runs under java-tests in CI). The SIMD/native tiers (env-availability-gated) are declared with the source-set→targets map. The rendered matrix is committed at docs/kernel-support-matrix.md and cross-linked with the eager backends & kernels mindmap. Current state: scalar floor covers all 8 formats on every target; Panama adds Q5/Q4_K SIMD on JVM/Android; native-FFM adds FP32/BF16/Q8_0/Q4_0/Q4_K on JVM. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
📖 Documentation Preview The documentation has been built successfully for this PR. Generated Files:
Artifacts:
This comment will be updated automatically when the PR is updated. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Phase 6 of the Native-parity quantized-kernel work (follows the merged #711/#715).
What
KernelSupportMatrixTest(skainet-backend-native-cpujvmTest — the one module whose test classpath sees all three providers: Scalar, Panama-Vector, Native-FFM). It renders the kernel × platform matrix and gates drift in the scalar floor: the all-platform baseline coverage is auto-derived fromScalarKernelProvider.supports("matmul", …), so adding/removing a scalar packed kernel without updating the docs fails the test (it runs underjava-testsin CI). The SIMD/native tiers (whoseisAvailable()is environment-gated) are declared with the source-set→targets map — the single place to edit when a provider gains a kernel.docs/kernel-support-matrix.md— the rendered matrix, committed and cross-linked with the eager backends & kernels mindmap.Current matrix
(Scalar floor covers all 8 formats on every target — the Native-parity outcome of this work; Q5_K/Q2_K/Q3_K/IQ4 remain dequant-to-FP32 only.)
Why this approach
A full build-logic/Antora generation pipeline was the heavier option; the availability-probing of the Panama/native providers (JDK incubator module / loaded
.so) makes a pure-registry capability probe machine-dependent. This test-based generator keeps the load-bearing axis (what runs on Native = the scalar floor) auto-synced and CI-gated, renders the full matrix tobuild/kernel-support-matrix.md+ stdout for refresh, and avoids env-flaky assertions.🤖 Generated with Claude Code