Concurrent Frontier CI: parallel test+bench, consolidated SLURM jobs#1147

Closed

sbryngelson wants to merge 17 commits intoMFlowCode:masterfrom

sbryngelson:concurrent-frontier

Commits on Feb 14, 2026

Merge benchmarks into test.yml for parallel CI execution

sbryngelson
and
claude
committed
Remove cancel-on-bench-failure to avoid killing tests on bench flakes

sbryngelson
and
claude
committed
Consolidate Frontier SLURM jobs into multi-node allocations

sbryngelson
and
claude
committed
Fix cp-into-self error in consolidated Frontier test script

sbryngelson
and
claude
committed
Fix review issues: rocm-smi gating, subshell env, job ID parsing

sbryngelson
and
claude
committed
Move concurrency from workflow-level to per-job

sbryngelson
and
claude
committed
Parallelize Frontier builds on login node

sbryngelson
and
claude
committed
Restore workflow-level concurrency for full run cancellation

sbryngelson
and
claude
committed
Add build progress heartbeat and collapsible GHA log groups

sbryngelson
and
claude
committed
Fix heartbeat kill guard and hardcoded node count in log message

sbryngelson
and
claude
committed
Clean up stale dirs before creating source copies

sbryngelson
and
claude
committed
Harden Frontier CI orchestration scripts

sbryngelson
and
claude
committed
Limit parallel builds to 2 concurrent on login node

sbryngelson
and
claude
committed
Set SETUPTOOLS_SCM_PRETEND_VERSION for hardlink source copies

sbryngelson
and
claude
committed
Add build heartbeat during throttled build phase

sbryngelson
and
claude
committed
Use real copies instead of hardlinks and increase parallel builds to 3

sbryngelson
and
claude
committed
Isolate build temp dirs to prevent Cray compiler conflicts

sbryngelson
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Concurrent Frontier CI: parallel test+bench, consolidated SLURM jobs#1147

Concurrent Frontier CI: parallel test+bench, consolidated SLURM jobs#1147
sbryngelson wants to merge 17 commits intoMFlowCode:masterfrom
sbryngelson:concurrent-frontier

Commits on Feb 14, 2026

Merge benchmarks into test.yml for parallel CI execution

Remove cancel-on-bench-failure to avoid killing tests on bench flakes

Consolidate Frontier SLURM jobs into multi-node allocations

Fix cp-into-self error in consolidated Frontier test script

Fix review issues: rocm-smi gating, subshell env, job ID parsing

Move concurrency from workflow-level to per-job

Parallelize Frontier builds on login node

Restore workflow-level concurrency for full run cancellation

Add build progress heartbeat and collapsible GHA log groups

Fix heartbeat kill guard and hardcoded node count in log message

Clean up stale dirs before creating source copies

Harden Frontier CI orchestration scripts

Limit parallel builds to 2 concurrent on login node

Set SETUPTOOLS_SCM_PRETEND_VERSION for hardlink source copies

Add build heartbeat during throttled build phase

Use real copies instead of hardlinks and increase parallel builds to 3

Isolate build temp dirs to prevent Cray compiler conflicts