Fix and improve the gemm quant pipeline infrastructure #3245
+88
−283
Conversation
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
ThomasNing
requested review from
a team,
afagaj,
andriy-ca,
aosewski,
asleepzzz,
bartekxk,
carlushuang,
cgmillette,
coderfeli,
ddembeckAMD,
geyyer,
illsilin,
poyenc,
qianfengz,
shumway,
tenpercent and
vidyasagar-amd
as code owners
spolifroni-amd
previously approved these changes
Nov 20, 2025
Refactor `GetBlockLoopTailNum` and `TailHandler` logic. When `PrefetchStages == 2` in BaseGemmPipelineAgBgCrCompV3, the tail_number can only be 0 or 1. Updated implementation to handle only these two cases.
The Y and X should be the same order as before
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Proposed changes
This PR has fix and improved the gemm quant pipeline infrastructure with the 3 following aspects:
(1) Fix the dram reading of bquant with the correct sequence aligned with b matrix with the column major pattern
(2) Improve the base pipeline structure and increase the performance for the base pipeline comp v3 for quant gemm.
(3) Improve the instruction use case of the new compiler "and_or" mechanism on using the bit operator instead of the builtin instruction.
Checklist
Please put an
xinto the boxes that apply. You can also fill these out after creating the PR. If you're not sure, please don't hesitate to ask.clang-formaton all changed filesDiscussion
If this is a relatively large or complex change, feel free to start a discussion by explaining why you chose the solution you did and what alternatives you considered