Function count_lanes

advanced_source/cuda_graph_annotations_tutorial.py:512–519

Count kernels per lane (tid).

(trace)

def compare_traces(raw_trace, annotated_trace):
    """Compare kernel distribution before and after annotation."""
    def count_lanes(trace):
        """Count kernels per lane (tid)."""
        counter = Counter(
            event["tid"]
            for event in trace["traceEvents"]
            if event.get("cat") == "kernel"
        )
        return dict(sorted(counter.items()))

    raw_lanes = count_lanes(raw_trace)
    annotated_lanes = count_lanes(annotated_trace)