cudax/stf: cuda_try migration — graph misc (PR6) by andralex · Pull Request #9301 · NVIDIA/cccl

andralex · 2026-06-08T22:40:04Z

Summary

Migrate the three leaf graph/ headers from cuda_safe_call to templated cuda_try<>:
- graph/internal/event_types.cuh — cudaGraphAddEmptyNode
- graph/interfaces/void_interface.cuh — cudaGraphAddEmptyNode
- graph/interfaces/slice.cuh — cudaGraphAddMemcpyNode
Part of the staged STF cuda_safe_call → cuda_try migration (PR6 of 11).

Test plan

Local build: ninja -j 12 -C build/cudax-cpp20 cudax.test.stf.graph.explicit_graph cudax.test.stf.interface.graph_use_device_data
CI

Convert the three remaining leaf graph headers (event_types, void_interface, slice) to templated cuda_try for graph node creation calls.

copy-pr-bot · 2026-06-08T22:40:07Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

andralex · 2026-06-08T22:40:10Z

/ok to test b854c6c

coderabbitai · 2026-06-08T22:45:45Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 11c35592-3ac1-48c3-b445-2398f8fc6516

📥 Commits

Reviewing files that changed from the base of the PR and between 4960c0b and fc48d74.

📒 Files selected for processing (3)

cudax/include/cuda/experimental/__stf/graph/interfaces/slice.cuh
cudax/include/cuda/experimental/__stf/graph/interfaces/void_interface.cuh
cudax/include/cuda/experimental/__stf/graph/internal/event_types.cuh

🚧 Files skipped from review as they are similar to previous changes (3)

cudax/include/cuda/experimental/__stf/graph/interfaces/void_interface.cuh
cudax/include/cuda/experimental/__stf/graph/interfaces/slice.cuh
cudax/include/cuda/experimental/__stf/graph/internal/event_types.cuh

Note: CodeRabbit is enabled on this repository as a convenience for maintainers
and contributors. Use your best judgment when considering its review comments and
suggestions — a suggested change may be inadequate, unnecessary, or safe to ignore.
Contributors are not expected to address every comment. Human reviews are what
ultimately matter for merging.

Overview

This PR continues the staged migration of the STF (Stateful Task Framework) from cuda_safe_call to the templated cuda_try<> for CUDA API error handling. This is part 6 of an 11-PR migration series.

Changes

Three leaf headers in the graph module were updated to use cuda_try<> and now directly include the header that defines cuda_try (instead of relying on transitive includes):

cudax/include/cuda/experimental/__stf/graph/internal/event_types.cuh
- In graph_event_impl::factorize (path for events.size() > 16) the empty-node creation now assigns the node via cuda_try(...) and the cudaGraphNode_t is initialized to nullptr before use (avoids -Werror=maybe-uninitialized).
- Lines changed: +3/-2
cudax/include/cuda/experimental/__stf/graph/interfaces/void_interface.cuh
- In void_graph_interface::graph_data_copy, the empty-node creation now directly returns cuda_try(...) instead of using a temporary cudaGraphNode_t with cuda_safe_call.
- Lines changed: +2/-3
cudax/include/cuda/experimental/__stf/graph/interfaces/slice.cuh
- In slice_graph_interface::graph_data_copy, the CUDA memcpy node addition now returns the result of cuda_try(...) directly, replacing the prior local cudaGraphNode_t + cuda_safe_call pattern.
- Lines changed: +2/-4

No public API signatures were changed.

Commits / Notes

Single commit: ensure each modified graph header directly includes the cuda_try utility and initialize out-parameter handles where needed to prevent -Werror=maybe-uninitialized.

Testing

Local build tested (ninja -j 12 -C build/cudax-cpp20) for:

cudax.test.stf.graph.explicit_graph
cudax.test.stf.interface.graph_use_device_data

CI validation pending.

Repository discussion

Three automated test-trigger comments by the author:
- "/ok to test b854c6c"
- "/ok to test 4960c0b"
- "/ok to test fc48d74"
PR note includes reviewer username context: "Your username is coderabbitai."

suggestion:

Walkthrough

Three CUDA Graph node creation sites switch from the out-parameter cuda_safe_call(cudaGraphAdd*Node(&node,...)) pattern to returning node handles directly via cuda_try<cudaGraphAdd*Node>(...); headers now include cuda_safe_call.cuh and an event node variable is initialized to nullptr before assignment.

Changes

CUDA Graph Node Creation Error Handling

Layer / File(s)	Summary
Migrate cuda_safe_call to cuda_try for graph node creation `cudax/include/cuda/experimental/__stf/graph/interfaces/slice.cuh`, `cudax/include/cuda/experimental/__stf/graph/interfaces/void_interface.cuh`, `cudax/include/cuda/experimental/__stf/graph/internal/event_types.cuh`	`cudaGraphAddMemcpyNode` in slice interface, `cudaGraphAddEmptyNode` in void interface `graph_data_copy`, and `cudaGraphAddEmptyNode` in event factorization now return results directly via `cuda_try<>`; temporary out-parameter node variables were removed, headers now include `cuda_safe_call.cuh`, and the event-node handle is default-initialized to `nullptr`.

Possibly related PRs

NVIDIA/cccl#9165: Similar migration of graph node creation calls from cuda_safe_call(..., &node) to cuda_try<> returning cudaGraphNode_t.

Suggested reviewers

caugonnet

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

coderabbitai

Actionable comments posted: 3

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: fdc337b9-f7f0-421f-9e74-7d7935334e50

📥 Commits

Reviewing files that changed from the base of the PR and between 2ca0829 and b854c6c.

📒 Files selected for processing (3)

cudax/include/cuda/experimental/__stf/graph/interfaces/slice.cuh
cudax/include/cuda/experimental/__stf/graph/interfaces/void_interface.cuh
cudax/include/cuda/experimental/__stf/graph/internal/event_types.cuh

Initialize the out-parameter to nullptr before cuda_try<cudaGraphAddEmptyNode> so GCC -Werror=maybe-uninitialized builds pass.

andralex · 2026-06-08T22:53:13Z

/ok to test 4960c0b

Each header must directly include the utility that defines cuda_try instead of relying on transitive includes.

andralex · 2026-06-08T23:00:12Z

/ok to test fc48d74

andralex · 2026-06-09T00:32:46Z

/ok to test fc48d74

andralex · 2026-06-09T18:51:59Z

/ok to test 77bd508

github-actions · 2026-06-09T20:53:24Z

🥳 CI Workflow Results

🟩 Finished in 1h 58m: Pass: 100%/55 | Total: 1d 00h | Max: 1h 04m | Hits: 15%/153894

See results here.

cudax/stf: migrate graph misc headers from cuda_safe_call to cuda_try

b854c6c

Convert the three remaining leaf graph headers (event_types, void_interface, slice) to templated cuda_try for graph node creation calls.

andralex requested a review from a team as a code owner June 8, 2026 22:40

andralex requested a review from srinivasyadav18 June 8, 2026 22:40

github-project-automation Bot added this to CCCL Jun 8, 2026

github-project-automation Bot moved this to Todo in CCCL Jun 8, 2026

cccl-authenticator-app Bot moved this from Todo to In Review in CCCL Jun 8, 2026

coderabbitai Bot reviewed Jun 8, 2026

View reviewed changes

Comment thread cudax/include/cuda/experimental/__stf/graph/interfaces/slice.cuh

Comment thread cudax/include/cuda/experimental/__stf/graph/interfaces/void_interface.cuh

Comment thread cudax/include/cuda/experimental/__stf/graph/internal/event_types.cuh

andralex added 2 commits June 8, 2026 18:53

Merge branch 'main' into andralex/stf-cuda-try-graph-misc

534012a

Fix maybe-uninitialized cudaGraphNode_t in event_types.cuh

4960c0b

Initialize the out-parameter to nullptr before cuda_try<cudaGraphAddEmptyNode> so GCC -Werror=maybe-uninitialized builds pass.

Include cuda_safe_call.cuh in graph headers using cuda_try

fc48d74

Each header must directly include the utility that defines cuda_try instead of relying on transitive includes.

This comment has been minimized.

Sign in to view

Merge branch 'main' into andralex/stf-cuda-try-graph-misc

cc6af4f

bernhardmgruber approved these changes Jun 9, 2026

View reviewed changes

Merge branch 'main' into andralex/stf-cuda-try-graph-misc

77bd508

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

cudax/stf: cuda_try migration — graph misc (PR6)#9301

cudax/stf: cuda_try migration — graph misc (PR6)#9301
andralex wants to merge 6 commits into
NVIDIA:mainfrom
andralex:andralex/stf-cuda-try-graph-misc

andralex commented Jun 8, 2026

Uh oh!

copy-pr-bot Bot commented Jun 8, 2026

Uh oh!

andralex commented Jun 8, 2026

Uh oh!

coderabbitai Bot commented Jun 8, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

andralex commented Jun 8, 2026

Uh oh!

andralex commented Jun 8, 2026

Uh oh!

This comment has been minimized.

andralex commented Jun 9, 2026

Uh oh!

andralex commented Jun 9, 2026

Uh oh!

github-actions Bot commented Jun 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

andralex commented Jun 8, 2026

Summary

Test plan

Uh oh!

copy-pr-bot Bot commented Jun 8, 2026

Uh oh!

andralex commented Jun 8, 2026

Uh oh!

coderabbitai Bot commented Jun 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Overview

Changes

Commits / Notes

Testing

Repository discussion

Walkthrough

Changes

Possibly related PRs

Suggested reviewers

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

andralex commented Jun 8, 2026

Uh oh!

andralex commented Jun 8, 2026

Uh oh!

This comment has been minimized.

andralex commented Jun 9, 2026

Uh oh!

andralex commented Jun 9, 2026

Uh oh!

github-actions Bot commented Jun 9, 2026

🥳 CI Workflow Results

🟩 Finished in 1h 58m: Pass: 100%/55 | Total: 1d 00h | Max: 1h 04m | Hits: 15%/153894

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

coderabbitai Bot commented Jun 8, 2026 •

edited

Loading