+The profiler is based on the following idea: The client-side testing application is executed. The target on-card operation is executed multiple times, but with a different constant (called performance trap ID) specifying what to measure during current call. Depending on the current value of performance trap ID, on-card operation is prematurely interrupted once the the specified performance trap matching provided ID is reached. The corresponding client-side measurements are collected and processed to compute time difference between consecutive traps.
0 commit comments