Open antiagainst opened 7 months ago
@llvm/issue-subscribers-mlir-spirv
Author: Lei Zhang (antiagainst)
Hi!
This issue may be a good introductory issue for people new to working on LLVM. If you would like to work on this issue, your first steps are:
1) In the comments of the issue, request for it to be assigned to you.
2) Fix the issue locally.
3) Run the test suite locally.
3.1) Remember that the subdirectories under test/
create fine-grained testing targets, so you can
e.g. use make check-clang-ast
to only run Clang's AST tests.
4) Create a Git commit.
5) Run git clang-format HEAD~1
to format your changes.
6) Open a pull request to the upstream repository on GitHub.
6.1) Detailed instructions can be found here.
If you have any further questions about this issue, don't hesitate to ask via a comment on this Github issue.
@llvm/issue-subscribers-good-first-issue
Author: Lei Zhang (antiagainst)
Hi. I would like to work on this. I'll try to take ⚙ D98396 [mlir] Remove mlir-cuda-runner as reference.
Hi @bhaskar1001101, sorry I missed your reply previously. Are you still interested to push this forward? If so I'll assign you to the issue. :)
Hello, I am new to LLVM and would like to work on this. @antiagainst , can you please assign me this issue.
@bhaskar1001101 and @Sh0g0-1758, you both have shown interest, so I assigned both of you. Can you work together on this?
To be clear, the idea is to remove mlir-vulkan-runner
, moving the logic inside mlir-cpu-runner
(like CUDA did) and then renaming mlir-cpu-runner
to just mlir-runner
.
Yeah. Note that I've marked this as good first issue but it's a relative large effort than normal, and may need some reading and understanding of mlir runners and vulkan specficially. Please let me know if you have questions. There are also other smaller good first issues if you are interested, just search with label "mlir:spirv" and "good first issues" to find them.
yes sure thing @antiagainst . I was getting familiar with mlir and will update you when a question of which I can't answer on the discourse arise.
Hey @bhaskar1001101 and @Sh0g0-1758, is this something you are still interested? Have you able to make progress on it?
@antiagainst Any particular insights that you would like to give apart from the comment links in the issue description?
I'll take a look at this as well.
We added mlir-vulkan-runner in way early days of MLIR. Recently various MLIR client API runners (e.g., mlir-cuda-runner) were removed in favor of performing translation using
mlir-opt
and then leveragemlir-cpu-runner
as the host coordnation mechanism. See @joker-eph's https://github.com/llvm/llvm-project/pull/65539#issuecomment-1710872236 for more context. We should migrate mlir-vulkan-runner to follow there. This would unify the runner story in MLIR to have one single mlir-runner, as @Jianhui-Li's https://github.com/llvm/llvm-project/pull/65539#issuecomment-1712414848 here.