linux - Outliers during Performance Evaluation -
I'm trying to measure some performance using Intels RDTSC, and it's variations that occur during different testers In most cases, my benchmark in C requires 3000000 Mio cycles, however, in some cases the same execution takes 5000000, it is almost two times more. I tried not to have any intensive workload running in parallel so that I get good performance estimates. Any ideas where this huge time difference can come? I know that there can be interconnection and stuff, but I did not expect such a great variety in time!
PS: I'm running it on a Pentium processor on which Linux is running.
Thanks for the response, John
I think the answer is:
<
You have insufficient control over a modern OS.
/ html>
Comments
Post a Comment