search

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

rocprofiler accumulates results from previous runs with --stats option or without option #89

Open kikimych opened 2 years ago

kikimych commented 2 years ago

way to reproduce:

cat input.xml <metric name=SQ_WAVES ></metric>

sudo rm /tmp/*.txt

cd /opt/rocm-4.5.2/hip/samples/2_Cookbook/0_MatrixTranspose sudo make sudo rocprof -i input.xml ./MatrixTranspose

cd /opt/rocm-4.5.2/hip/samples/2_Cookbook/13_occupancy sudo make sudo rocprof -i input.xml ./occupancy

cat input.csv Index,KernelName,gpu-id,queue-id,queue-index,pid,tid,grd,wgr,lds,scr,vgpr,sgpr,fbar,sig,obj,SQ_WAVES 0,"matrixTranspose(float, float, int) [clone .kd]",0,0,0,16313,16313,1048576,16,0,0,8,24,0,0x0,0x7fd140c0a6c0,31251 1,"multiply(float, float, float*, int) [clone .kd]",0,0,4,16425,16425,1000448,1024,0,0,8,24,0,0x0,0x7fd40ac0a800,15632

occupancy result contains runs from matrixTranspose test. input.xml is copied to test directories