Hi Gao, Many CPUs will support multiple "hardware threads" per core. Hardware threads on the same core share the floating-point arithmetic unit of the core. The operating system will see each hardware thread (or "logical core") as a separate "core". So the question is whether the 120 cores are logical cores or real hardware cores. The easiest way to do this is to just look up the specification for your processor. Running one MPI process per logical core will result in multiple processes competing for the same floating-point arithmetic unit, so for compute-intensive tasks, this is detrimental to performance. I suspect this is what is happening here. Kind regards, Chris On 09/11/2024 13:31, Gao wrote:
This email from feigao2024@126.com originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list <https://spam.ic.ac.uk/SpamConsole/Senders.aspx> to disable email stamping for this address.
Hello everyone
I run the Compressible solver in parallel, when I used 120 cores, it was slower than 68 cores. I used the commond that "mpirun -n 64 CmpressibleSolver Cyl.xml ". Could someone give me some advice? The computer has 128 cores, 2 NUMA.
With best regards
_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
-- Chris Cantwell Reader in Computational Engineering Department of Aeronautics Imperial College London South Kensington Campus London SW7 2AZ Tel: +44 (0)20 759 45050 Email:c.cantwell@imperial.ac.uk www.imperial.ac.uk/people/c.cantwell