This patch is a following for the previous patch https://reviews.llvm.org/D151519.
With this patch, vector.load op with narrow bitwidth (e.g., i4) can be converted to
supported wider bitwidth (e.g., i8).
Details
Details
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Comment Actions
overall looks good to me, just few comments
mlir/lib/Dialect/Vector/Transforms/VectorEmulateNarrowType.cpp | ||
---|---|---|
53 | This should be renamed to *VectorLoad? We have a very similar implementation for memref.load, can we move the static method to mlir/Dialect/MemRef/Utils/MemRefUtils.h, and reuse it here? Maybe can rename it to getLinearizedMemRefAndOffset(...). | |
165–167 | We should just return a failure if this is not implemented yet. | |
201 | add a new line at end of file |
This should be renamed to *VectorLoad? We have a very similar implementation for memref.load, can we move the static method to mlir/Dialect/MemRef/Utils/MemRefUtils.h, and reuse it here? Maybe can rename it to getLinearizedMemRefAndOffset(...).