Dick Balaska <dic### [at] buckosoftcom> wrote:
> Am 12/9/19 8:20 PM, also sprach jr:
> > alias nicepov='nice -n 19 povray'
> just fyi, "nice" really doesn't do anything, at least not since the
> invention of multicore CPUs. The scheduler just picks stuff and goes.
> This is an improvement, over where for several years cores could remain
> idle with tasks pending. Say you've got a normal priority task running
> one thread. It gets dispatched and since there are no other normal
> priority tasks pending, the scheduler is done, even though there are 3
> idle cores and several "low" priority tasks pending.
> There is a compile in replacement for the scheduler that fixes the
> current behavior, but the author recommends not using it and waiting for
> his new scheduler, which hopefully Linus will accept into the kernel.
> I got no answers to this question ...
> but further research led me to the statements above.
interesting, thanks, and new to me. I suspect I will simply continue to use it
in aliases and scripts though.
wrt the problem in the question, have you read 'man 7 cpuset'? no personal
experience, but the reading suggests it should be possible to sort of partition
the system resources and then set up to get the benefit of the "Scheduler load
btw, re 'qtpovmodeler', what's with the baseinsertrules.xml file not found?
(same for the docmap xml). also, and that may be my window manager to blame
(WindowMaker 0.95.7), main menu accelerator key bindings do not work for me.
Post a reply to this message