-
Notifications
You must be signed in to change notification settings - Fork 316
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
nvptx target missing in 0.33 when installed using pip #601
Comments
Thanks for the report. Do you have a reproducer for this please? Also, where did you get In [1]: import llvmlite
In [2]: llvmlite.__version__
Out[2]: '0.33.0'
In [3]: import llvmlite.binding as ll
In [4]: ll.initialize()
In [5]: ll.initialize_all_targets()
In [6]: from llvmlite.binding.targets import Target
In [7]: Target.from_triple('nvptx64-nvidia-cuda')
Out[7]: <llvmlite.binding.targets.Target at 0x7facb55b7730> |
Also, the
|
The above example doesn't work when llvmlite is installed from pip:
|
Thanks for confirming, llvmlite 0.33 moved to llvm 9, I think the additional targets didn't get added to the build script in the wheel build path. Conda version: https://github.com/numba/llvmlite/blob/master/conda-recipes/llvmdev/build.sh#L31-L32 Wheel: https://github.com/numba/llvmlite/blob/master/conda-recipes/llvmdev_manylinux1/build.sh#L31 |
Fixes numba#601: add the missing GPU hosts to the wheel build for llvmdev.
I prepped a PR applying the suggested fix here: #602 |
Fixes #601: add the missing GPU hosts to the wheel build for llvmdev.
Hi, has the PTX target been removed in 0.33?
Trying to use target triple
nvptx64-nvidia-cuda
results in:Reverting to 0.32.1 fixes the issue and my program runs without issues.
Rebuilding 0.33.0 from source using system LLVM also works without any changes to the original program.
The text was updated successfully, but these errors were encountered: