On 15 Jan 2002, at 22:00, Robin Stevens wrote:

> On an otherwise idle Linux system, I've been noticing that the
> per-iteration speed has been varying during the course of a single
> primality test.  Until Saturday afternoon I'd been getting a fairly
> consistent 0.188/0.189s time when the system was idle.   It then increased
> to around 0.206s.  It fell again last night to around 0.199s (see logs
> below - percentages trimmed for 80-character neatness).

With stop/start or system reboots between?

Some systems appear prone to variability of a few percent from 
program start to program start. Starting mprime immediately the 
system boots is usually very good unless the boot included actual 
partition checks (either forced, or because the partition check 
interval has expired). When the assignment finishes, the new 
assignment almost invariably starts at the fastest speed the 
system will ever run at. 

In my experience, the speed doesn't change much however long 
mprime is left running the same assignment, but stopping and 
restarting can make a significant difference. It might be worth trying 
that if you find that it's running slow for some unknown reason. 

Probably the reason for the variability has something to do with the 
way the program code and data get assigned to physical memory 
locations. BTW Prime95 behaves similarly under Windows.

Regards
Brian Beesley
_________________________________________________________________________
Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

Reply via email to