Repository navigation
Conversation
Collaborator
|
CI pipeline fails to build vt with |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
cuda_lambdasince Kokkos 5 rejects it (still effectively enabled through Kokkos/nvcc_wrapper flags)cuda_constexprto more closely replicate Sandia appsWe also may want to consider adding
+cuda_relocatable_device_codeand/or several other changes to more closely replicate Sandia apps. We could also add more images for these build variants. Here is a branch that more closely replicates this environment: (https://github.com/sandialabs/distBVH/tree/config-testing)I also haven't looked into what the CI is doing on the develop branch and if we need to update anything over there.
Build Fail
cuda_constexprbrings up device compilation errors in vt (so maybe hold off merging until this is resolved, or we could addcuda_constexprlater). There are multiple potential fixes for this, either addressing constexpr in vt (#2526: constexpr breaks CUDA relaxed-constexpr builds DARMA-tasking/vt#2527) and/or adding an opt-in target to magistrate so vt isn't compiled as device code (#414: Base Magistrate target forces Kokkos CUDA Compilation DARMA-tasking/magistrate#415)