auto offloading in dynamic shape IR #1313
Open
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
qiuxiafei
reviewed
Aug 20, 2024
| disc_ral::createDiscSpecializeFusionWithSpeculationPass( | ||
| gpu_options.sm_count, gpu_options.max_threads_per_sm)); | ||
| // pm.addNestedPass<FuncOp>( | ||
| // disc_ral::createDiscSpecializeFusionWithSpeculationPass( |
Collaborator
Author
There was a problem hiding this comment.
just easier to analysis buffer living range, will remove this comment.
| if (main.getName() == SymbolicDimMgr::getShapeConstraintGraphFunctionName()) | ||
| return; | ||
| mlir::OpBuilder b(main); | ||
| const int64_t memoryLimitation = 21474836480; // 30GB |
| } | ||
| return result; | ||
| } | ||
| void SortBuffersByPrioriy(std::vector<LivingBuffer>& livingBuffers) { |
There was a problem hiding this comment.
Is it possible to consider buffer size here, even with symbolic shape?
| return offloadCond; | ||
| } | ||
| std::vector<LivingBuffer> FilterBuffers( | ||
| std::vector<LivingBuffer> livingBuffers) { |
| << " expr: " << memoryUsageList[instIndex] << "\n"; | ||
| return memoryUsageList[instIndex]; | ||
| } | ||
| std::vector<int64_t> ConcretMemoryUsageSimulator(int64_t concretValue) { |
| offloadIfOp.getElseRegion().front().clear(); | ||
| b.setInsertionPointToStart(&offloadIfOp.getElseRegion().front()); | ||
| auto dummyHostBuffer = cloneBuffer(b, endOp->getLoc(), buffer); | ||
| dummyHostBuffer.getDefiningOp()->setAttr(kRematBufferAttr, |
There was a problem hiding this comment.
will this dummy host buffer cause some unexpected behavior .... if it's used somewhere
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Outdated suggestions cannot be applied.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.
No description provided.