observability: ignore sub-tolerance CPU>wall skew in accounting anomaly
A 2 ms phase whose CPU clock reads 0.1 us over its wall clock is clock granularity, not an accounting bug; the anomaly flag fired on nearly every short phase in the live verification run. Only flag (and record non_cpu_raw_ms) when the deficit exceeds both 100 us and 1% of the phase wall. Larger real anomalies keep the previous behaviour.
This commit is contained in:
@@ -2477,8 +2477,11 @@ def _emit_loop_detail(lt: LoopTimings, attrs: dict, child, dp_rank: int) -> None
|
||||
if cpu_count == count:
|
||||
non_cpu_ms = max(0.0, (total_ns - cpu_ns) / 1e6)
|
||||
attrs[prefix + "_non_cpu_ms"] = non_cpu_ms
|
||||
if total_ns < cpu_ns:
|
||||
attrs[prefix + "_non_cpu_raw_ms"] = (total_ns - cpu_ns) / 1e6
|
||||
# Only a deficit beyond clock granularity is an accounting
|
||||
# anomaly; a few hundred ns of skew on a 2 ms phase is noise.
|
||||
deficit_ns = cpu_ns - total_ns
|
||||
if deficit_ns > max(100_000, total_ns // 100):
|
||||
attrs[prefix + "_non_cpu_raw_ms"] = -deficit_ns / 1e6
|
||||
attrs[prefix + "_accounting_anomaly"] = True
|
||||
accounting_anomalies += 1
|
||||
if count > 1:
|
||||
|
||||
Reference in New Issue
Block a user