Concurrent Fluent calculations in Linux
I have 3 blades with 2 AMD chips each (each chip with 4 cores) all running Linux. How do I get concurrent instances of Fluent 6.3 (parallel version) to use different cores? When I start two instances of Fluent (say each using 4 cores) on the same machine, they use the same cores while the others remain inactive. Any suggestions? Thanks.
I presume you are using version 6.3.26 where this was a known problem. If possible, request a post-release version from your supplier: the problem was fixed in version 6.3.33, and I have used 6.3.35 successfully as well.
In the meantime, there is a manual work around that was provided on the Online Technical Support site of Fluent (Solution 1440).
For example, after launching a 4-core run, type the following in the command window:
(set-affinity "0 4 1 5 2 6 3 7")
This will move the node processes 0,1,2,3 to the processor cores 4,5,6,7.
The strings contains a sequence of pairs of numbers, each pair consisting of a process id and a cpu/core id. Process ids are 0, 1, 2,... for the node processes, and the core IDs can be found in /proc/cpuinfo on Linux.
Hope this helps.
|All times are GMT -4. The time now is 11:28.|