[Bugfix][Kernel][SM70] Reconcile GLM KDA GEMV with latest main - #396
Merged
yangzhuxinyzx merged 5 commits intoAug 28, 2026
Merged
Conversation
Signed-off-by: yangzhuxinyzx <153831768+yangzhuxinyzx@users.noreply.github.com>
Signed-off-by: yangzhuxinyzx <153831768+yangzhuxinyzx@users.noreply.github.com>
…92-latest-main-20260828-090550 Signed-off-by: yangzhuxinyzx <153831768+yangzhuxinyzx@users.noreply.github.com> # Conflicts: # docs/design/sm70_v100_migration_control.md
Signed-off-by: yangzhuxinyzx <153831768+yangzhuxinyzx@users.noreply.github.com>
Signed-off-by: yangzhuxinyzx <153831768+yangzhuxinyzx@users.noreply.github.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Purpose
Reconcile #392 with the latest main while retaining the exact SM70 GLM KDA GEMV optimization and all newer QSA/NVFP4 release work. The only latest-main conflict was the migration-control document; both evidence sections are preserved.
This repair also adds a default-on rollback gate:
VLLM_SM70_GLM53_EXACT_KDA_GEMV=0restores the existing cuBLAS linear path. Admission remains structural: SM70, TP4, B1 decode, FP16 contiguous[1,4096] x [6416,4096]; prefill and every incompatible shape/dtype use the existing path.Test Plan
maincontaining [Kernel][SM70] Group QSA Page4 prefill #387, [Bugfix][QSA][SM70] Make sparse block selection deterministic #394, and [Kernel][SM70] Avoid NVFP4 MoE prefill input expansion #390.Test Result
18 passed, 6 skipped(GPU-native cases already covered by [Kernel][SM70] Add exact GLM KDA decode GEMV #392 artifacts).git diff --check: pass.true,1true,0false.Signed-off-by: yangzhuxinyzx 153831768+yangzhuxinyzx@users.noreply.github.com