Page 1 of 1

Remote cores not fully used in error studies

Posted: Fri 19 Jul 2024 14:55
by Emmanuel
Hello,

When using Tracewin for error studies, tracelx64 processes run on the number of cores set in TraceWin, but with a very low efficiency. Each tracelx64 process uses a core at less than 15% (often between 5 and 12 %).

Is there any setting to make in TraWin to prevent such a behaviour ?

Than you for any help.

Emmanuel

Re: Remote cores not fully used in error studies

Posted: Fri 19 Jul 2024 21:00
by Didier
Dear Emmanuel,

It really depends on the duration of the calculation. If it's very short, the core won't be much used.
Also note that Tracelx64 is not parallelized.

Regards,

Didier

Re: Remote cores not fully used in error studies

Posted: Mon 29 Jul 2024 09:55
by Emmanuel
Dear Didier,

It should calculate a whole linac with hundreds thousands particles, what takes about thirty minutes for a single simulation.

So, a question comes to me : in error studies, how TraceWin spreads the calculation over the selected servers and cores ?

Thank you for any advice.

Emmanuel

Re: Remote cores not fully used in error studies

Posted: Sat 3 Aug 2024 11:34
by Didier
Dear Emmanuel,

TracelX64 is a simple single processor process that is just launched by the server or by TraceWin. The OS then manages the distribution of tasks to the different cores, so I have no way of controlling this.

Regards,

Didier