[quidditch_snitch] Implement bufferization for microkernels #38
Conversation
This file contains 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
By making it possible to bufferize the operations we are able to form our micro kernels earlier in the pass pipeline prior to bufferization. This is required for other transformation passes such as explicit tensor promotion to L1 memory and subgroup distribution that we are planning to add later. As a consequence of this change the existing `xdsl_kernel` (now renamed to just `microkernel`) had to be split into two operations, one operating on tensors, the other on memrefs. The reason for this is that the memref version should be isolated from above while the tensor version cannot be for bufferization purposes. The bufferization process itself now converts between the two.
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Outdated suggestions cannot be applied.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.
By making it possible to bufferize the operations we are able to form our micro kernels earlier in the pass pipeline prior to bufferization. This is required for other transformation passes such as explicit tensor promotion to L1 memory and subgroup distribution that we are planning to add later.
As a consequence of this change the existing
xdsl_kernel(now renamed to justmicrokernel) had to be split into two operations, one operating on tensors, the other on memrefs. The reason for this is that the memref version should be isolated from above while the tensor version cannot be for bufferization purposes. The bufferization process itself now converts between the two.