Issue |
159749
|
Summary |
[mlir] Vectorize directly to a named contraction in e2e tests
|
Labels |
mlir
|
Assignees |
banach-space
|
Reporter |
banach-space
|
Following #147296, we should try simplify our lowering pipelines for [e2e tests](https://github.com/llvm/llvm-project/tree/0d989b2aefe8d8f66981f1d39f56ce0214b5de86/mlir/test/Integration/Dialect/Linalg/CPU). Specifically,
* We should lower directly to named contractions, i.e. `vector.contract`, rather than via `vector.multi_reduction`.
Below are links to two examples. Specifically, lowering steps that we should be able to avoid if we lowered directly to `vector.contract`.
https://github.com/llvm/llvm-project/blob/0d989b2aefe8d8f66981f1d39f56ce0214b5de86/mlir/test/Integration/Dialect/Linalg/CPU/mmt4d.mlir#L82-L86
https://github.com/llvm/llvm-project/blob/0d989b2aefe8d8f66981f1d39f56ce0214b5de86/mlir/test/Integration/Dialect/Linalg/CPU/ArmSVE/matmul.mlir#L105-L111
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs