Add CODEOWNERS file

[tstellar]

No description provided.

[MaskRay]

While others backend maintainers haven't requested yet, you can create all backend teams for these patterns

/llvm/include/llvm/IR/Intrinsics$backend.td
/llvm/lib/Target/$backend/
/llvm/test/CodeGen/$backend/
/clang/lib/Basic/Targets/$backend*
/clang/lib/Driver/ToolChains/Arch/$backend.*
/clang/lib/CodeGen/Targets/$backend.cpp
/clang/include/clang/Basic/Builtins$backend*

as a starter, and let backend maintainers add individual files as needed.

@efriedma-quic

[smithp35]

I noticed on the Discourse post that subscribers for backend teams wouldn't be created automatically.

On behalf of Arm we'd like to follow MaskRays suggestion and add two subscription groups
$backend = ARM , pr-subscribers-arm
$backend = AArch64, pr-subscribers-aarch64

Will add this to the Discourse post to.

[ldionne]

This is not how I thought we'd set up the CODEOWNERS file. I thought the pr-subscribers-FOO groups were created for notification purposes only. Here it seems like these groups will be given the ability to approve PRs.

Is there a reason why we don't have separate groups for code owners?

Edit: Ah, I see that folks are actually asking for the teams they want. Sorry, I looked too quickly. I'll add my request as a comment on the patch so it can be marked as resolved once done. Sorry for the noise :-).

[clementval]

Can we add one team for OpenACC with these folder/files.

@pr-subscribers-openacc

./flang/**/OpenACC/
./flang/include/flang/Lower/OpenACC.h
./flang/docs/OpenACC.md
./flang/lib/Parser/openacc-parsers.cpp
./flang/lib/Lower/OpenACC.cpp
./llvm/**/Frontend/OpenACC/
./llvm/unittests/Frontend/OpenACCTest.cpp
./mlir/test/Target/LLVMIR/openacc-llvm.mlir
./mlir/test/Dialect/OpenACC
./mlir/**/Dialect/OpenACC/
./mlir/**/Conversion/OpenACCToSCF/
./mlir/lib/Target/LLVMIR/Dialect/OpenACC/
./mlir/lib/Conversion/OpenACCToSCF
./mlir/lib/Conversion/OpenACCToSCF/OpenACCToSCF.cpp
./mlir/lib/Dialect/OpenACC/

[joker-eph]

Can we add one team for OpenACC with these folder/files.

@pr-subscribers-openacc

./flang/**/OpenACC/
./flang/include/flang/Lower/OpenACC.h
./flang/docs/OpenACC.md
./flang/lib/Parser/openacc-parsers.cpp
./flang/lib/Lower/OpenACC.cpp
./llvm/**/Frontend/OpenACC/
./llvm/unittests/Frontend/OpenACCTest.cpp
./mlir/test/Target/LLVMIR/openacc-llvm.mlir
./mlir/test/Dialect/OpenACC
./mlir/**/Dialect/OpenACC/
./mlir/**/Conversion/OpenACCToSCF/
./mlir/lib/Target/LLVMIR/Dialect/OpenACC/
./mlir/lib/Conversion/OpenACCToSCF
./mlir/lib/Conversion/OpenACCToSCF/OpenACCToSCF.cpp
./mlir/lib/Dialect/OpenACC/

Likely better to extend the pattern more generically: mlir/**/*OpenACC like for the other generic MLIR patterns.
(I don't know about the flang/llvm side, but I'd think it applies equally well)

[joker-eph]

@tstellar I'm concerned about the maintainability, can we have it grouped by list?

<empty line>
pattern1 <list name>
pattern2 <list name>
<empty line>

(seems scriptable to convert your file to this. I can do it if needed)

[tstellar]

@joker-eph How does this look?

[joker-eph]

Looks great, thanks!
Not that GitHub says Escaping a pattern starting with # using \ so it is treated as a pattern and not a comment ; but I assume that this just wouldn't match anything so it seems just fine.

[joker-eph]

Thanks Tom!

[efriedma-quic]

Other teams I'd like:

pr-subscribers-polly:
polly/

pr-subscribers-scev:
llvm/lib/Analysis/ScalarEvolution*
llvm/test/Analysis/ScalarEvolution/
llvm/include/llvm/Analysis/ScalarEvolution*
llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h

For pr-subscribers-AArch64 can we also add:

clang/utils/TableGen/SveEmitter.cpp
clang/utils/TableGen/NeonEmitter.cpp
clang/include/clang/Basic/BuiltinsSVE.def
clang/include/clang/Basic/BuiltinsSME.def
llvm/include/llvm/TargetParser/AArch64TargetParser.h
llvm/lib/TargetParser/AArch64TargetParser.cpp
llvm/docs/AArch64SME.rst
llvm/docs/BigEndianNEON.rst

And for pr-subscribers-arm:

clang/utils/TableGen/MveEmitter.cpp
llvm/include/llvm/TargetParser/ARMTargetParser*
llvm/lib/TargetParser/ARMTargetParser*
llvm/docs/HowToBuildOnARM.rst

[clementval]

I added these but I cannot create the associated teams.

[kiranchandramohan]

Can we add one for flang-openmp with @pr-subscribers-flang-openmp ?

flang/test/**/OpenMP
flang/lib/Lower/OpenMP.cpp
flang/lib/Semantics/resolve-directives.cpp
flang/lib/Semantics/check-omp-structure.cpp
flang/lib/Optimizer/Transforms/OMP*
flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
flang/test/Lower/OpenMP
flang/test/Transforms/omp*
mlir/**/*OpenMP*
mlir/test/Target/LLVMIR/openmp*
llvm/lib/Frontend/OpenMP
llvm/include/llvm/Frontend/OpenMP
llvm/unittests/Frontend/OpenMP*

[tstellar]

I think I have all the requested teams added now plus the other suggestions implemented. I'm planning to push this tonight, just so we have something that works We can commit follow up changes whenever we want.

[clementval]

Thanks for all the work @tstellar

[MaskRay]

This is not how I thought we'd set up the CODEOWNERS file. I thought the pr-subscribers-FOO groups were created for notification purposes only. Here it seems like these groups will be given the ability to approve PRs.

Is there a reason why we don't have separate groups for code owners?

Edit: Ah, I see that folks are actually asking for the teams they want. Sorry, I looked too quickly. I'll add my request as a comment on the patch so it can be marked as resolved once done. Sorry for the noise :-).

Bump this request from @ldionne

[tstellar]

This is not how I thought we'd set up the CODEOWNERS file. I thought the pr-subscribers-FOO groups were created for notification purposes only. Here it seems like these groups will be given the ability to approve PRs.
Is there a reason why we don't have separate groups for code owners?
Edit: Ah, I see that folks are actually asking for the teams they want. Sorry, I looked too quickly. I'll add my request as a comment on the patch so it can be marked as resolved once done. Sorry for the noise :-).

Bump this request from @ldionne

As far as I understand putting a group or a person in the CODEOWNERS file does not give them any special privileges to approve PRs.

[boomanaiden154]

As far as I understand putting a group or a person in the CODEOWNERS file does not give them any special privileges to approve PRs.

It depends upon how exactly we want to use the CODEOWNERS file. There was discussion early today in #clang on Discord about adding blocking reviewers. That is possible using a CODEOWNERS file (along with some branch protection rules), but difficult to combine with this method of doing pull request notifications.

I also believe that the teams specified in the CODEOWNERS file only get notifications for PRs for paths they're subscribed to as they get added as reviewers, which probably isn't the behavior that we want in all cases (although probably is fine for now).

[MaskRay]

analysis is incorrectly lowercase.

Consider llvm/test/Analysis as well?

[kiranchandramohan]

If this can be /mlir/**/*OpenMP*/ then it will cover the OpenMPToLLVM and SCFToOpenMP conversions.

[HighCommander4]

Could we add a pr-subscribers-clangd for /clang-tools-extra/clangd please?

[joker-eph]

Shall we land and iterate in-tree to refine? Everyone can adjust the exact pattern anytime in the future.

[bogner]

Can we also get SelectionDAG?

llvm/include/llvm/CodeGen/DAGCombine.h
llvm/include/llvm/CodeGen/SelectionDAG*.h
llvm/include/llvm/CodeGen/SDNodeProperties.td
llvm/include/llvm/Target/TargetSelectionDAG.td
llvm/lib/CodeGen/SelectionDAG/
llvm/utils/TableGen/CodeGenDAG*
llvm/utils/TableGen/DAGISel*

[tstellar]

@bogner I missed your request, can you open a pull request for this?