Re: OT? hypertread on or off for SMP kernel
On Thu, 2004-12-09 at 16:53, Miquel van Smoorenburg wrote:
> In article <[🔎] 1102603975.27189.503.camel@localhost.localdomain>,
> michael <linux@networkingnewsletter.org.uk> wrote:
> >However, for one simple test (v.low memory, v.low comms, high compute)
> >case which is Fortran90 compiled using mpich, I actually found turning
> >HT off helped -- when it was on then for -np 2 there was no speed up
> >(top implied they ran on same physical processor). I presume this was
> >because the CPU was fully busy anyhow so the 2nd thread had to wait (you
> >seem to imply the opposite but I don't follow why!).
>
> Turning on the "SMT scheduler support" in the kernel is supposed
> to prevent exactly this. If two jobs are runnable, they will be
> run on 2 different CPUs (if you have them), not on 1 CPU with 2 threads.
>
> Mike.
>
Thanks that's useful.
The question thus becomes whether I can compile 'mpich' such that it can
take advantage of SMT and how well it can, or not.
Michael
Reply to: