GPU Branch Divergence

Performance Profiler: Measuring Throughput & Utilization

Throughput
0%
Utilization
0%

Warp Setup

Toggle threads to see how "Wasteful Cycles" increase with divergence.

ℹ️ Analysis

Ready. Configure threads and step to start profiling.
Current PC:
IDLE

Metrics Guide

Throughput: Useful work done per hardware cycle. Divergence drops this because "waiting" threads do zero work.

Utilization: Ratio of active execution lanes vs total available lanes in the warp.

Path A
Path B