On Wed, Sep 28, 2016 at 2:13 AM, <jf...@ms4.hinet.net> wrote: > If the load was distributed by the OS schedules across all cores, > does it means I can't make one core solely running a piece of codes > for me and so I have no contol on its performance?
In Unix, Python's os module may have sched_setaffinity() to set the CPU affinity for all threads in a given process. In Windows, you can use ctypes to call SetProcessAffinityMask, SetThreadAffinityMask, or SetThreadIdealProcessor (a hint for the scheduler). On a NUMA system you can call GetNumaNodeProcessorMask(Ex) to get the mask of CPUs that are on a given NUMA node. The cmd shell's "start" command supports "/numa" and "/affinity" options, which can be combined. -- https://mail.python.org/mailman/listinfo/python-list
