Skip to content

Add latency benchmark harness - #75

Merged
LimiNode merged 21 commits into
mainfrom
codex/add-latency-and-throughput-benchmarks-for-logging-libraries
Sep 20, 2025
Merged

Add latency benchmark harness#75
LimiNode merged 21 commits into
mainfrom
codex/add-latency-and-throughput-benchmarks-for-logging-libraries

Conversation

@LimiNode

Copy link
Copy Markdown
Owner

Summary

  • add a benchmark runner that exercises the latency/throughput matrix and appends results to bench/results/latency.csv
  • provide adapters for log-it-cpp and spdlog that propagate producer timestamps without string parsing
  • wire the build flags and document how to build/run the benchmarks

Testing

  • cmake --build build --target logit_bench

https://chatgpt.com/codex/tasks/task_e_68cb7658ef80832c8baa9e24a922bc78

Add benchmark harness with adapters for log-it-cpp and spdlog, integrate build options, and document usage.
Run the benchmark configure/build/run steps only when relevant events trigger.
Adopt the revised latency recorder, scenario runner, and README guidance so the benchmark captures timestamps after slot reservation, calculates percentiles via nearest rank, and allows workload overrides via environment variables.
Ensure the Linux workflow runs the built logit_bench executable from the build directory so the step no longer fails with a missing file.
Ensure the CI job finds the logit_bench binary and add timeout-based watchdog logging so hangs surface with context.
Switch the CI workflow to pull_request_target so forked pull requests run without maintainer approval. Explicitly check out the contributor repository when needed and tighten permissions while still allowing checks and artifacts.
Run the workflow on pull_request events for repository branches while keeping pull_request_target for forks, and ensure each job checks out the proper head and skips duplicate runs.
Revert the conditional triggers added earlier so the CI matches the prior configuration without pull_request_target logic.
Reset the CI workflow to commit 728c7a5 so benchmark jobs, timeout, and stable branch trigger return.
Print timestamps for every bench log line, refresh the watchdog when output appears, and fail only after prolonged inactivity so the run no longer times out mid-execution.
Update the benchmark watchdog while producers run so slow scenarios keep reporting progress, bump the default timeout to twenty minutes, and keep the watchdog alive through adapter flushes.
…-libraries' into codex/fix-ci-issue-with-logit_bench-execution-5dkzon
@LimiNode
LimiNode merged commit 2b765b1 into main Sep 20, 2025
8 of 10 checks passed
@LimiNode
LimiNode deleted the codex/add-latency-and-throughput-benchmarks-for-logging-libraries branch September 20, 2025 05:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant