Some rewriters take more iterations to converge,
add a parameter to overwrite the built-in maximum iteration count.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Thanks for the patch! Can you please run clang-format?
mlir/include/mlir/Transforms/GreedyPatternRewriteDriver.h | ||
---|---|---|
19 | Please avoid the use of macros for things like this. | |
37 | Can you just add an additional overload that accepts a max iteration count? The current overload could just call that one with the default. That would remove the need for exposing the default publicly. | |
39 | MLIR uses camelCase for variable names. |
mlir/include/mlir/Transforms/GreedyPatternRewriteDriver.h | ||
---|---|---|
19 | clang-format does not work for me: AlwaysBreakTemplateDeclarations: Yes |
- Updating D91553: Optional argument for pattern rewriter max iteration count. #
- Enter a brief description of the changes included in this update.
- The first line is used as subject, next lines as comment. #
- If you intended to create a new revision, use:
- $ arc diff --create
Do not expose default max iteration count
mlir/include/mlir/Transforms/GreedyPatternRewriteDriver.h | ||
---|---|---|
19 | Seems like you have an old version of clang-format. |
I tweaked slightly the doc and the commit message:
We like to tag commit messages with "NFC" when they don't modify an existing behavior and don't include tests (this change is a bit borderline, it *could* be plumbed and tested with a pass)
Please avoid the use of macros for things like this.