ROCm / rocprofiler

ROC profiler library. Profiling with perf-counters and derived metrics.
https://rocm.docs.amd.com/projects/rocprofiler/en/latest/
Other
116 stars 44 forks source link

Timing Issues with MI50 and performance metrics #83

Open arfio opened 2 years ago

arfio commented 2 years ago

When tracing with the performance counters enabled on MI50, the kernel timings are not correct. When testing against MI100, the problem does not appear.

OS: Ubuntu 20.04 Kernel: 5.4.0-104 ROCm version: 5.0.0 GPU: tested on 2 configurations one with MI50 and one with MI100

kikimych commented 2 years ago

Could you please provide small synthetic test, build instructions and profiler options used to run?