Profiling Tools Interfaces for GPU (PTI for GPU) is a set of Getting Started Documentation and Tools Library to start performance analysis on Intel(R) Processor Graphics easily
It would be great to have median metric alongside with average or have a choice. Many benchmarks run warm-up trials and then a lot of main trials of the same workload. Having median is useful to exclude spikes which usually happen during warm-up runs.
I refer to the Average metric from an example report below:
=== API Timing Results: ===
Total Execution Time (ns): 418056422
Total API Time (ns): 407283268
Function, Calls, Time (ns), Time (%), Average (ns), Min (ns), Max (ns)
zeCommandQueueSynchronize, 4, 182529847, 44.82, 45632461, 45271728, 46364532
zeModuleCreate, 1, 111687828, 27.42, 111687828, 111687828, 111687828
zeCommandQueueExecuteCommandLists, 4, 108593458, 26.66, 27148364, 1756304, 102803947
zeCommandListAppendMemoryCopy, 12, 2493748, 0.61, 207812, 62061, 1037087
Hi,
It would be great to have median metric alongside with average or have a choice. Many benchmarks run warm-up trials and then a lot of main trials of the same workload. Having median is useful to exclude spikes which usually happen during warm-up runs.
I refer to the Average metric from an example report below: