Using Multiple CPU Cores to Accelerate Boris Simulations on Linux

[Zhiyu Zhang]

Hi, Serban,

I would like to know how to utilize multiple CPU cores to speed up calculations when running Boris on a Linux system. In this system, GPU acceleration is not available, but the CPU has more than 16 cores. However, when running Python scripts, the performance still feels slow. Could you provide some guidance on how to optimize the use of CPU cores to improve the performance?

Thank you for your help!

Best regards,
Zhiyu

[Serban Lepadatu]

Hi Zhiyu,

When GPU computations are not enabled/available, all available CPU cores are used by default. You can adjust the number of cores (threads) used with the threads command, but there shouldn't be a need to. CPU computations will be much slower than GPU computations depending on the problem size, except for very small problem sizes due to GPU latencies (see e.g. https://www.boris-spintronics.uk/benchmarking-for-v3-0/).

Regards,

Serban