[Loop Specialization]: Specialize loops containing masked operations with loop invariant mask #3586
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
Signed-off-by: Tiotto, Ettore <[email protected]>
Signed-off-by: Tiotto, Ettore <[email protected]>
Signed-off-by: Tiotto, Ettore <[email protected]>
Signed-off-by: Tiotto, Ettore <[email protected]>
…ith loop invariant masks Signed-off-by: Tiotto, Ettore <[email protected]>
etiotto
commented
Mar 3, 2025
| @@ -3,9 +3,9 @@ | |||
| #include "mlir/Dialect/Arith/IR/Arith.h" | |||
| #include "mlir/Dialect/SCF/IR/SCF.h" | |||
| #include "mlir/IR/Verifier.h" | |||
| // #include "mlir/Pass/Pass.h" | |||
etiotto
commented
Mar 3, 2025
| class VersioningCondition { | ||
| // Abstract base class for mask validators. | ||
| // Mask validators are used to check whether a given mask has an expected form. | ||
| // Concreate subclasses define the expected form. |
There was a problem hiding this comment.
Typo: Concreate -> Concrete
Signed-off-by: Tiotto, Ettore <[email protected]>
| static bool version(scf::ForOp &forOp, | ||
| MaskedOpsCollector<CanonicalMaskValidator> &collector) { | ||
| // Limitation | ||
| // Currently we can version the loop only is it doesn't have downward |
There was a problem hiding this comment.
Suggested change
| // Currently we can version the loop only is it doesn't have downward | |
| // Currently we can version the loop only if it doesn't have downward |
Comment on lines
+379
to
+390
| // Combine the versioning conditions. | ||
| OpBuilder builder(forOp); | ||
| Location loc = forOp.getLoc(); | ||
| auto it = maskConds.begin(); | ||
| Value firstCond = (*it++)->getResult(0); | ||
| auto maskValidator = collector.getMaskValidator(); | ||
| Value verCond = maskValidator.getVersioningCond(forOp, firstCond); | ||
| for (; it != maskConds.end(); ++it) { | ||
| Value nextCond = (*it)->getResult(0); | ||
| Value cond = maskValidator.getVersioningCond(forOp, nextCond); | ||
| verCond = builder.create<arith::AndIOp>(loc, verCond, cond); | ||
| } |
There was a problem hiding this comment.
Would that make sense to create a third (and fourth?) branch if only load (or store) masks can be removed?
I think that it is still interesting to use 2D block loads even if store operations cannot not use 2D block store.
WDYT?
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.
Version loops containing masked operations (e.g. tt.load with a mask) where the mask is loop invariant.