Skip to content
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

Allow Terra to initialize when LLVM is not build with NVPTX backend #446

Closed
Tracked by #492
elliottslaughter opened this issue May 26, 2020 · 3 comments · Fixed by #541
Closed
Tracked by #492

Allow Terra to initialize when LLVM is not build with NVPTX backend #446

elliottslaughter opened this issue May 26, 2020 · 3 comments · Fixed by #541

Comments

@elliottslaughter
Copy link
Member

Right now we have a hard dependency on the NVPTX backend of LLVM even when CUDA is not enabled in Terra. This may potentially limit what LLVM binaries we can compile against. If we make the initialization conditional on CUDA support being enabled then we could build with more pre-built LLVM binaries.

Failure looks like:

terra/src/terralib.lua:747: failed to initialize target for LLVM Triple: nvptx64-nvidia-cuda (No available targets are compatible with this triple.)
@astrolemonade
Copy link

astrolemonade commented Dec 20, 2021

Can we make this dependency optional before version 1 is released? A lot of people will consider using it and will have this problem when trying terra.

@elliottslaughter
Copy link
Member Author

As an update, we only initialize the target when terralib.cudahome is set, but that is currently unconditional on Linux. We should really make it smarter, because it's not as if we'll use this target if Terra isn't even built with CUDA.

I've added this to #492.

@elliottslaughter
Copy link
Member Author

This was fixed in #541.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging a pull request may close this issue.

2 participants