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
Revert "Fix CUDA with LLVM9" #606
Revert "Fix CUDA with LLVM9" #606
Conversation
This reverts commit 3f66129. Once numba/numba#6030 is merged, it will no longer be necessary to disable the autoupgrade of atomic intrinsics for NVPTX, because LLVM from llvmlite will not be used to optimize the IR before sending it to NVVM.
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.
Thanks for the patch.
@esc probably best to consume this into the LLVM 10 series work? |
Cherry 🍒 picked as: 0fc90da |
So I ran this through the build farm last night, and received the following errors:
|
That looks like Numba without numba/numba#6030 merged - could that be what happened? |
This happens across all {Linux, Windows} x {Cuda8,9,10} x {Python36,38} tests. |
I believe the branch I tested this on contains #6030:
|
Discussion here: https://gitter.im/numba/numba-dev?at=5f2a7a78a8636a6f1690eb68 accepted solution is to "trick" the auto-upgrader into not upgrading, @gmarkall is producing a patch to this effect. |
#6080 fixed it. |
This reverts commit 3f66129.
Once numba/numba#6030 is merged, it will no longer be necessary to disable the autoupgrade of atomic intrinsics for NVPTX, because LLVM from llvmlite will not be used to optimize the IR before sending it to NVVM.