Skip to content

Workaround for CUDA failure that can occur for optimized code #201

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.

Sign up for GitHub

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

Merged
liu15 merged 4 commits into from
Mar 31, 2026
Merged

Workaround for CUDA failure that can occur for optimized code #201

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9b267e5
Workaround for CUDA failure that can occur for optimized code
liu15 Mar 27, 2026
9c5d86e
Reformat
liu15 Mar 28, 2026
5bbd027
Add comment
liu15 Mar 30, 2026
fe36738
More style fix
liu15 Mar 30, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion src/tribol/search/InterfacePairFinder.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -126,7 +126,9 @@ class CartesianProduct : public SearchBase {
auto& contactPairs = m_coupling_scheme->getInterfacePairs();
contactPairs.resize( countArray_host[0] );

countArray.fill( 0 );
int zero = 0;
Copy link
Copy Markdown
Member

@ebchin ebchin Mar 28, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
int zero = 0;
// Workaround for axom::Array::fill() issue on rzvector (CUDA). See Axom #1833.
int zero = 0;

Yes, the description is clear, I just meant something like the above, so the rationale is clear in the code and so we can track the issue and it's resolution.

But, as you mention, an axom reducer is a better fit for what the code is trying to do here. I'm OK merging in what you have here to fix the build issue and I can do a follow-on PR that uses axom::ReduceSum here instead.

// Workaround for axom::Array::fill() issue for optimized CUDA code, see Axom #1833.
axom::copy( countArray.data(), &zero, sizeof( int ) );
auto pairs_view = m_coupling_scheme->getInterfacePairs().view();
// fill proximate pairs array
forAllExec(
Expand Down
Loading