Daniele Bordes wrote:
Thank you for your quick and exhaustive reply, Dan. I try to answer
your questions:
cpu0 is probably active only to respond to your keypresses and to fork
your processes (ps -A, prstat, etc).
Perhaps it is so, but I stop Simics simulation (ctrl+c) and type
Ruby0.dump-stats when no commands are in execution.
dump-stats reports the stats for the entire execution history.
Also, did you issue instruction-fetch-mode instruction-fetch-trace,
istc-disable, dstc-disable to Simics?
Yes, I use those commands
How about cpu-switch-time 1?
Try running a mulitthreaded benchmark, and be sure to use
processor_bind() to get all the processors busy.
Well, the fact is that I first noticed the "strange" statistics just
when I tried to run some OCEAN multithreaded simulations with one
thread per CPU (using processor_bind() ), and I hoped that in an "idle
machine" case the situation was different, but it's the same.
Did you run the benchmark for long time? Did you start ruby after the
processor_bind() calls to avoid the single-cpu startup transient?
_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.
|
|