The Linux Kernel: a decade of wasted cores -- a new paper that challenges the scheduler's effectiveness.
Thanks for the links.
Another problem is the Simultaneous Multithreading (SMT, what
Intel calls hyperthreading). Two or more threads are sharing the
same CPU core. If a high priority thread and a low priority thread
are running in the same core then they are both getting half the
resources. The Intel Knights Landing processor can even run four
threads per core. I have seen a high priority thread running at a
quarter of the normal speed because it was sharing a core with low
priority threads. The CPUID instruction on x86 processors can tell
how many threads can run in each core, but not which threads. This
is a severe weakness of the SMT principle. Apparently, Linux can't
avoid this problem. I don't know if any other operating system
can.
The Linux Kernel: a decade of wasted cores -- a new paper that challenges the scheduler's effectiveness.
https://www.ece.ubc.ca/~sasha/papers/eurosys16-final29.pdf
--
You received this message because you are subscribed to the Google Groups "Computing Performance" group.
To unsubscribe from this group and stop receiving emails from it, send an email to computing-perfor...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Best-F--
You received this message because you are subscribed to the Google Groups "Computing Performance" group.
To unsubscribe from this group and stop receiving emails from it, send an email to computing-perfor...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.