GPU Kernel Benchmarking
workflow:gpu-kernel-benchmarking
Workflowworkflows/workflows/workflows-gpu.yaml·Open in Graph → {
"id": "workflow:gpu-kernel-benchmarking",
"_kind": "Workflow",
"_file": "workflows/workflows/workflows-gpu.yaml",
"_cluster": "workflows",
"attributes": {
"displayName": "GPU Kernel Benchmarking",
"workflowKind": "development",
"triggerType": "event-driven",
"typicalCadence": "per-optimization",
"complexity": "single-team",
"description": "Benchmarks GPU compute kernels across hardware targets — measuring\nthroughput, latency, memory bandwidth, and occupancy; comparing\nagainst baseline; profiling hotspots; and publishing results to\na performance dashboard. Excludes kernel development.\n"
},
"outgoingEdges": [
{
"from": "workflow:gpu-kernel-benchmarking",
"to": "role:performance-profiler",
"kind": "involves_role",
"attributes": {}
},
{
"from": "workflow:gpu-kernel-benchmarking",
"to": "role:implementer",
"kind": "involves_role",
"attributes": {}
},
{
"from": "workflow:gpu-kernel-benchmarking",
"to": "role:data-scientist",
"kind": "involves_role",
"attributes": {}
},
{
"from": "workflow:gpu-kernel-benchmarking",
"to": "skill-area:observability-pipeline",
"kind": "requires_skill_area",
"attributes": {}
},
{
"from": "workflow:gpu-kernel-benchmarking",
"to": "domain:scientific-computing",
"kind": "applies_to_domain",
"attributes": {}
},
{
"from": "workflow:gpu-kernel-benchmarking",
"to": "domain:ml-ops",
"kind": "applies_to_domain",
"attributes": {}
},
{
"from": "workflow:gpu-kernel-benchmarking",
"to": "responsibility:performance-budget-tracking",
"kind": "triggers_responsibility",
"attributes": {}
},
{
"from": "workflow:gpu-kernel-benchmarking",
"to": "org-unit:research-engineering",
"kind": "performed_by_org_unit",
"attributes": {}
},
{
"from": "workflow:gpu-kernel-benchmarking",
"to": "org-unit:ml-team",
"kind": "performed_by_org_unit",
"attributes": {}
}
],
"incomingEdges": []
}