Skip to content

[Athena] Clean up and split CI actions (Issue #504)#505

Merged
syifan merged 1 commit intomainfrom
athena/cleanup-actions
Feb 12, 2026
Merged

[Athena] Clean up and split CI actions (Issue #504)#505
syifan merged 1 commit intomainfrom
athena/cleanup-actions

Conversation

@syifan
Copy link
Contributor

@syifan syifan commented Feb 12, 2026

Summary

  • Reduced from 19 GitHub Actions workflows to 9
  • Removed 10 redundant/broken/never-succeeding workflows
  • Split monolithic accuracy-report.yml (120 min) into 3 focused workflows: microbench (~15 min), polybench (~30 min parallel), embench (~30 min)
  • Consolidated 3 separate calibration workflows into 1 with suite selector
  • Changed rarely-used workflows to dispatch-only triggers

Changes

Removed (14 files deleted)

Workflow Reason
h5-accuracy-report.yml Duplicate of accuracy-report, 0 successes
polybench-sim.yml Superseded by polybench-segmented
polybench-segmented.yml Subsumed into new accuracy-polybench
diana-statistical-validation.yml References non-existent scripts, 0 successes
h4-multicore-accuracy.yml H4 not started, premature
ci-health-monitor.yml Meta-monitoring bloat
ci-metrics-dashboard.yml Never runs successfully, bloated
performance-profiling.yml Rarely needed
performance-regression-monitoring.yml 0 successes, overlaps with performance-regression
cpi-comparison.yml Ran on every push unnecessarily
embench-calibration.yml Merged into calibration.yml
polybench-calibration.yml Merged into calibration.yml
memory-bench-calibration.yml Merged into calibration.yml
accuracy-report.yml Split into 3 smaller workflows

Added (4 files)

  • accuracy-microbench.yml — Fast microbenchmark accuracy (~15 min)
  • accuracy-polybench.yml — PolyBench in 3 parallel groups (~30 min)
  • accuracy-embench.yml — EmBench accuracy (~30 min)
  • calibration.yml — Unified calibration (dispatch-only, suite selector)

Modified (3 files)

  • matmul-calibration.yml — Changed to dispatch-only
  • spec-bench.yml — Removed daily cron (SPEC unavailable on hosted runners)
  • performance-regression.yml — PR-only (removed push trigger)

Final Workflow Inventory (9 total)

Workflow Trigger Purpose
ci.yml push, PR Build, lint, test
accuracy-microbench.yml push (paths), dispatch Microbenchmark accuracy
accuracy-polybench.yml push (paths), dispatch PolyBench accuracy (parallel)
accuracy-embench.yml push (paths), dispatch EmBench accuracy
calibration.yml dispatch Hardware calibration (all suites)
benchmark.yml dispatch M2 hardware benchmarking
matmul-calibration.yml dispatch Fast timing matmul CPI
performance-regression.yml PR Benchstat regression detection
spec-bench.yml dispatch SPEC benchmark validation

Test plan

  • Verify CI workflow runs on this PR (ci.yml)
  • Manually trigger accuracy-microbench to verify it works
  • Check that no critical functionality was lost

Generated with Claude Code

Reduce from 19 workflows to 9:
- Remove 10 redundant/broken workflows (h5-accuracy-report, polybench-sim,
  polybench-segmented, diana-statistical-validation, h4-multicore-accuracy,
  ci-health-monitor, ci-metrics-dashboard, performance-profiling,
  performance-regression-monitoring, cpi-comparison)
- Split monolithic accuracy-report.yml into 3 focused workflows:
  accuracy-microbench (~15 min), accuracy-polybench (~30 min parallel),
  accuracy-embench (~30 min)
- Consolidate 3 calibration workflows into 1 (calibration.yml)
- Change matmul-calibration, spec-bench to workflow_dispatch only
- Change performance-regression to PR-only (no push trigger)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@syifan syifan merged commit 918c23f into main Feb 12, 2026
4 of 7 checks passed
@github-actions
Copy link

Performance Regression Analysis

Performance Benchmark Comparison

Compares PR benchmarks against main branch baseline.
Benchmarks: pipeline tick throughput across ALU, memory, mixed workloads.



---
*Automated benchmark comparison via `go test -bench` + benchstat*

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

Comments