The "9999/sec" metric is misleading because seems to be an average over the entire time that the fuzzer has spent running, which includes the half a minute that was spent initialising, so even though the actual rate is about 131k execs/sec, the reported rate keeps on rising towards that figure for a long time (it's about right after 15m of running).
I'd suggest that either we start the overall timer for execs/sec when the execs actually start completing, or that it reflect only the rate since the last time the figure was reported.
On balance, I think I'd prefer to see the latter, because that means it's easy to see if things are slowing down (for example by something else interfering with the amount of available CPU time), and that was my intuitive understanding of what the metric was showing me until I realised otherwise.
If you do need to find out the overall rate, it's not hard to do the calculation yourself by dividing total execs by elapsed time.
The text was updated successfully, but these errors were encountered: