Skip to content

Feature/framework ceiling microbenchmarks - #2

Open
chengronglai wants to merge 5 commits into
NVIDIA-ISAAC-ROS:rollingfrom
chengronglai:feature/framework-ceiling-microbenchmarks
Open

Feature/framework ceiling microbenchmarks#2
chengronglai wants to merge 5 commits into
NVIDIA-ISAAC-ROS:rollingfrom
chengronglai:feature/framework-ceiling-microbenchmarks

Conversation

@chengronglai

Copy link
Copy Markdown

Summary

Add focused CPU ceiling microbenchmarks that measure EventsCBGExecutor independently of application-graph instrumentation:

  • Intra-process std_msgs/msg/Int64 message throughput and callback latency
  • Minimal executor scheduler-dispatch throughput
  • Continuously ready, guard-condition-backed rclcpp::Waitable workloads
  • YAML-driven matrix runner with repeated executions
  • Result summarization with completeness and executor-dispatch invariant validation
  • Methodology, metrics, and limitations documentation

Validation

  • Completed the default 55-run matrix on ROS 2 Rolling
  • All runs completed successfully
  • All waitable invariants passed
  • Result-set validation passed
  • Python tooling tests: 13 passed

Provide minimal waitable-driven message-passing and scheduler baselines so EventsCBGExecutor limits can be measured independently of application-graph instrumentation.

Generated-by: Cursor
Signed-off-by: Cheng-Rong Lai <chengrongl@nvidia.com>
Keep the first microbenchmark change focused while existing repository tests are made compatible with current Rolling separately.

Generated-by: Cursor
Signed-off-by: Cheng-Rong Lai <chengrongl@nvidia.com>
Describe the always-ready workload directly without relying on historical benchmark context.

Generated-by: Cursor
Signed-off-by: Cheng-Rong Lai <chengrongl@nvidia.com>
- Track effective run configurations
- Reject incomplete or unexpected results
- Avoid misleading medians on failures
- Remove redundant RMW metadata
@chengronglai
chengronglai requested a review from hemalshahNV July 23, 2026 07:36
@chengronglai chengronglai self-assigned this Jul 23, 2026
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 this pull request may close these issues.

1 participant