[LoadStoreOpToLLVM] Enable the block store for tensor pointer #4666
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
Contributor
Author
|
I am cleaning up the LIT test. Will add it soon. |
chengjunlu
force-pushed
the
chengjun/block_store
branch
from
050b19b to
1b4e6e3
Compare
whitneywhtsang
changed the title
chengjunlu
force-pushed
the
chengjun/block_store
branch
4 times, most recently
from
ca8654c to
fcf3c4b
Compare
Contributor
There was a problem hiding this comment.
Pull Request Overview
This PR enables block store functionality for tensor pointer operations by extending the existing block I/O conversion infrastructure to support store operations in addition to the previously supported prefetch and load operations.
- Extends the
isMemoryRowMajortemplate function to includetriton::StoreOpin addition to existing operations - Adds a comprehensive
getBlockIOTileSizetemplate function to determine optimal tile dimensions for block I/O operations - Implements block store conversion logic in
StoreOpToBlockIOConversionwith hardware constraints validation and 2D block store operation generation
Reviewed Changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.
| File | Description |
|---|---|
| LoadStoreOpToLLVM.cpp | Main implementation adding block store support with tile size calculation and store operation conversion |
| GetEnv.hpp | Adds new environment variable for enabling block store on regular pointers |
third_party/intel/lib/TritonIntelGPUToLLVM/LoadStoreOpToLLVM.cpp
Outdated
Show resolved
Hide resolved
third_party/intel/lib/TritonIntelGPUToLLVM/LoadStoreOpToLLVM.cpp
Outdated
Show resolved
Hide resolved
chengjunlu
force-pushed
the
chengjun/block_store
branch
4 times, most recently
from
1bd8db0 to
df8a253
Compare
chengjunlu
commented
Jul 10, 2025
third_party/intel/lib/TritonIntelGPUToLLVM/LoadStoreOpToLLVM.cpp
Outdated
Show resolved
Hide resolved
chengjunlu
commented
Jul 10, 2025
chengjunlu
force-pushed
the
chengjun/block_store
branch
6 times, most recently
from
8dd3f04 to
70a5b73
Compare
chengjunlu
commented
Jul 10, 2025
chengjunlu
commented
Jul 10, 2025
chengjunlu
force-pushed
the
chengjun/block_store
branch
3 times, most recently
from
493783e to
5889663
Compare
chengjunlu
force-pushed
the
chengjun/block_store
branch
8 times, most recently
from
3e66e80 to
b63da93
Compare
Contributor
Author
chengjunlu
force-pushed
the
chengjun/block_store
branch
4 times, most recently
from
fbefa53 to
faf30a9
Compare
Contributor
Author
|
Some part of changes are under reviewing in a separate PR #4704 |
chengjunlu
force-pushed
the
chengjun/block_store
branch
8 times, most recently
from
2cd07aa to
0dbedde
Compare
Contributor
Author
|
@whitneywhtsang , @etiotto , @LiyangLingIntel |
Signed-off-by: Lu,Chengjun <[email protected]>
whitneywhtsang
approved these changes
Jul 17, 2025
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.
Enable the block store for tensor pointer.