Closed jmgorius closed 4 years ago
I just followed what was done in commit 1e9bdfc for the GPU dialect. It might be a better fit since it avoids the redundant dialect name in the operation.
Please rebase on 3f37f5278b304fdf79b9f56d625b39e9c9d21b96 and I'll start the integration process.
I just rebased on the latest commit on master c203743.
can you please rebase and make sure to run clang-format? The following 3 files are flagged:
include/mlir/EDSC/Intrinsics.h
lib/Conversion/VectorConversions/VectorToVector.cpp
lib/Transforms/MaterializeVectors.cpp
I just rebased and ran clang-format
, but it does not find anything to format. Could this be a version mismatch? I'm using
clang-format version 9.0.0 (tags/RELEASE_900/final)
Edit: I just tried with a more recent build:
clang-format version 10.0.0 (.../llvm-project.git 020de4a80125ac731d2cc3c4f0b65cc5b548971a)
and it can't find anything do to either. If I run
clang-format mlir/EDSC/Intrinsics.h > tmp.h
diff mlir/EDSC/Intrinsics.h tmp.h
I get an empty diff.
I ran clang-format -i -style=file
on your changes, there was a bunch of those. On the general basis, we are using clang-format compiled from LLVM's HEAD.
@nicolasvasilache Interestingly enough, I cannot manage to get any format changes from the files you pointed out to me.
@ftynse However, if I run
find . -name "*.cpp" | xargs clang-format -i -style=file
find . -name "*.h" | xargs clang-format -i -style=file
in the root of the repository, it changes a lot of files, most of which I did not even touch. A quick git diff
shows that these changes are mostly related to the AlwaysBreakTemplateDeclarations: Yes
in the .clang-format
file. Are we using the same .clang-format
?
Change vector op names from
VectorFooOp
toVector_FooOp
and fromvector::VectorFooOp
tovector::FooOp
.