Aref Automatic Warp Specialization [AutoWS] Implementation #6689
+12,093
−172
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
Jokeren
reviewed
May 2, 2025
| auto barrierOp = mlir::insertBarrier(*builder, op->getLoc()); | ||
| } | ||
|
|
||
| bool MembarAnalysis::isBarrier(Operation *op) { |
There was a problem hiding this comment.
return isa<gpu::BarrierOp, NVVM::BarrierOp>(op);
Jokeren
reviewed
May 2, 2025
| @@ -25,7 +26,28 @@ struct MakeRangeOpConversion | |||
| auto elemTy = ty.getElementType(); | |||
| assert(elemTy.isInteger(32)); | |||
| Value start = createIndexAttrConstant(rewriter, loc, elemTy, op.getStart()); | |||
| auto idxs = emitIndices(loc, rewriter, targetInfo, layout, ty, true); | |||
| std::optional<int> warpGroupStart; | |||
| if (!getWarpGroupStart(rewriter.getInsertionBlock())) { | |||
There was a problem hiding this comment.
This is work-around for limited capability of our partitioner that doesn't put make_range op into a warp_group region. We actually like Meta's partitioner, and hopefully we can transition to it, and this issue get resolved, and this work-around won't be needed.
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Outdated suggestions cannot be applied.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.
Context: In addition to AutoWS implementations in the
release/*andmainbranches, we (see list of contributors below) have been working on our own implementation of AutoWS usingarefabstractions, and we would like to share this with the community in a separate brancharef_auto_ws.Goal: We’re sharing this implementation to facilitate discussion on what abstractions are helpful for automatic warp specialization. We will continue improving this branch to prove out the performance, ergonomics and flexibility for complex workloads that the abstractions here enable.
Simultaneously, we plan to (and have been) taking portions of this branch and porting them into
maininthird_party/nvidiaas they mature and prove their utility on various workloads.Note: This pull request is primarily for information and visibility, but feedback is appreciated and will be considered as we continue development on this branch.
Contributors to this implementation:
@3gx, @acollins3, @binarybana, @BinFan, @chhzh123, @CliveUnger, @csullivan, @masahi, @mbrookhart, @vinodgro