feat(backend-cpu): Panama SIMD Q6_K SPI kernel; route Q6_K via the registry#720
Merged
Conversation
…gistry Closes #718. Brings Q6_K in line with Q4_K/Q5: it now resolves a SIMD kernel through the KernelProvider registry instead of the legacy DefaultCpuOpsJvm intercept. - New PanamaVectorQ6_KMatmulKernel (jvmMain, block-major): reuses the existing SIMD Q6_K block dequant (JvmQuantizedVectorKernels.dequantQ6_KBlock, now internal) into a 256-element scratch + a Vector-API FMA dot. Wired into PanamaVectorKernelProvider.matmulQ6K(). - Retire the JVM-only legacy intercept: removed the `is Q6_KTensorData` matmul branch and the Q6_K lazy-transpose branch from DefaultCpuOpsJvm, so Q6_K flows through the unified DefaultCpuOpsBase registry dispatch (block-major, shared with Native). The legacy JvmQuantizedVectorKernels.matmulQ6_KVec is kept only as a tested reference (Q6KMatmulTest); dequantQ6_KBlock is now the shared helper. - docs/kernel-support-matrix.md regenerates to `Q6_K JVM = panama-vector` (KernelSupportMatrixTest's declared Panama set gains Q6_K). Tests: PanamaVectorQ6KParityTest (Panama vs Scalar Q6_K) + PackedMatmulDispatchTest gains a Q6_K case — green on jvmTest AND linuxX64Test. Full backend-cpu jvmTest passes (no regression from retiring the intercept); apiDump regenerated. 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. |
michalharakal
added a commit
that referenced
this pull request
Jun 9, 2026
docs: refresh eager-kernels mindmap after Panama Q6_K (#720)
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.
Closes #718.
Brings Q6_K in line with Q4_K/Q5: it now resolves a SIMD kernel through the
KernelProviderregistry instead of the legacyDefaultCpuOpsJvmintercept — the last 🚧 from the eager backends & kernels mindmap.Changes
PanamaVectorQ6_KMatmulKernel(jvmMain, block-major): reuses the existing SIMD Q6_K block dequant (JvmQuantizedVectorKernels.dequantQ6_KBlock, nowinternal) into a 256-element scratch + a Vector-API FMA dot. Wired intoPanamaVectorKernelProvider.matmulQ6K().is Q6_KTensorDatamatmul branch and the Q6_K lazy-transpose branch fromDefaultCpuOpsJvm, so Q6_K flows through the unifiedDefaultCpuOpsBaseregistry dispatch (block-major, shared with Native).matmulQ6_KVecis kept only as a tested reference (Q6KMatmulTest);dequantQ6_KBlockis the shared helper.Q6_K JVM = panama-vector(KernelSupportMatrixTest's declared Panama set gains Q6_K) —docs/kernel-support-matrix.mdupdated.Verification
PanamaVectorQ6KParityTest(Panama vs Scalar Q6_K, within FMA tolerance) ✅PackedMatmulDispatchTestgains a Q6_K case — green on jvmTest AND linuxX64Test ✅backend-cpujvmTest suite passes (no regression from retiring the intercept) ✅Native-FFM Q6_K (and Q5) remains the separate gap in #708 / SKaiNET-transformers#170; this PR is the Panama SIMD SPI kernel only.
🤖 Generated with Claude Code