Skip to content

COB-121: Add tSubbandProcPerformance

Bram Veenboer requested to merge cob-121 into cob-master

This benchmark uses (a copy of) the parsets from RTCP/Cobalt/GPUProc/src/PerformanceTest/parsets/ to run gpu_load and compares the total runtime by the reference runtime. The reference runtime is measured by running the same parsets with 1000 iterations, while the test uses only 100 iterations. To allow for a little variation, a tolerance is provided. This benchmark can be extended by adding more parsets or adding reference files for different GPUs.

Merge request reports