Support any-dimensional memrefs in AllocsToSLM pass
#411
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
Signed-off-by: dchigarev <[email protected]>
Signed-off-by: dchigarev <[email protected]>
Signed-off-by: dchigarev <[email protected]>
Signed-off-by: dchigarev <[email protected]>
dchigarev
commented
Nov 21, 2024
| int64_t newX = | ||
| originalShape[0] * blockSizes[0] * blockSizes[1] * blockSizes[2]; | ||
| SmallVector<int64_t> newShape({newX}); | ||
| newShape.append(originalShape.begin() + 1, originalShape.end()); |
Contributor
Author
There was a problem hiding this comment.
The logic of scaling allocation size by the total number of threads was reworked. We now scale only the zero dimension of the memref:
// num threads in the work group: X=2, Y=3, Z=4
// alloc before the pass
%slm_buff = memref.alloc() : memref<8x16xf16>
// alloc after the pass (only scaled the zero dimension)
%slm_buff_root = memref.alloc() : memref<192x16xf16, 3>
%x_offset = slm_chunk_x_size * (X_thread_idx * Y_block_size * Z_block_size + Y_thread_idx * Z_block_size + Z_thread_idx) = 8 * (X_thread_idx * 12 + Y_thread_idx * 4 + Z_thread_idx)
%slm_buff = memref.subview [%x_offset, 0] : memref<192x16xf16, 3> -> memref<8x16xf16, 3>
AndreyPavlenko
approved these changes
Nov 21, 2024
Signed-off-by: dchigarev <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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.
This PR adds support for memrefs of any dimensions as well as for thread indices of all 3 dimensions (thread_idx_z was not supported before)
This PR fixes the failure of the ROPE module caused by the
AllocsToSLMpass