-
Notifications
You must be signed in to change notification settings - Fork 112
Manual cuda graph creation in load balanced bounds presolve #242
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
Conversation
|
This one is critical for 25.08 |
|
/ok to test 8596529 |
|
/ok to test 10f87ea |
|
@kaatish Should we merge this PR ? |
|
/ok to test 2bb8de9 |
|
/ok to test 18b7202 |
|
/ok to test 5502c8d |
rgsl888prabhu
left a comment
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.
minor suggestion, rest looks good
| { | ||
| using f_t2 = typename type_2<f_t>::type; | ||
| cudaGraph_t cnst_slack_graph; | ||
| cudaGraphCreate(&cnst_slack_graph, 0); |
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.
Are we recreating the graph each time?
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.
We create this structure to add graph nodes to it. Once the structure is complete, we instantiate the actual graph with a call like :
cudaGraphInstantiate(&upd_bnd_exec, upd_graph, NULL, NULL, 0);
This is done once per graph in the setup function inside constructor by calling either create_bounds_update_graph() or create_constraint_slack_graph().
|
/ok to test e6afa2e |
|
/merge |
This PR removes cuda graph capture from load balanced bounds strengthening to address crashes encountered in the constructor. The cuda graph is instead created manually. Fixes #219