Hello All,
Does this "inst_replay_overhead" mean the overhead caused by instruction cache misses?
Thanks,
Hello All,
Does this "inst_replay_overhead" mean the overhead caused by instruction cache misses?
Thanks,
Hi,smartvoice
It means “Average number of replays for each instructions executed”.
You can nvprof --query-metrics to get all description about metrics.
Veraj,
Is it due to the profiling overhead or instruction cache misses?
Hi,
Please refer the explain
https://stackoverflow.com/questions/35566178/how-to-explain-instruction-replay-in-cuda