-
Notifications
You must be signed in to change notification settings - Fork 758
check in lib file for now and throw it in the pip install #15713
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/15713
Note: Links to docs will display an error until the docs builds have been completed. ⏳ 16 Pending, 2 Unrelated FailuresAs of commit 0294060 with merge base 234c288 ( BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR needs a
|
| src_dir="backends/cuda/runtime/", | ||
| src_name="aoti_cuda_shims.lib", | ||
| dst="executorch/data/bin/", | ||
| dependent_cmake_flags=[], |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think you need to dependent on EXECUTORCH_BUILD_CUDA
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I dont want to be dependent on that build flag. I want to always put it in the pip package because this is for the AoT flow which is no longer gated by the cmake flag
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I could technically try to gate this behind if we are building cuda, but its just a temporary hack anyway till we have cross compiling so I think its fine to just land.
the .lib file comes from compiling aoti_cuda_shims on windows. Ideally we will not check this in but rather cross compile it on the fly from linux. Thats coming soonish. We need the .lib file in order to lower for cuda windows backend. Cuda windows uses AOTI as the acceleration lib, AOTI generates executable code that calls into the shim. On windows you must provide a reference to all symbols you will link against (unlike linux) so we need the .lib artifact.
the .lib file comes from compiling aoti_cuda_shims on windows. Ideally we will not check this in but rather cross compile it on the fly from linux. Thats coming soonish.
We need the .lib file in order to lower for cuda windows backend. Cuda windows uses AOTI as the acceleration lib, AOTI generates executable code that calls into the shim. On windows you must provide a reference to all symbols you will link against (unlike linux) so we need the .lib artifact.