Re: GRR!! SMP=1 sucks

Alan Cox (alan@lxorguk.ukuu.org.uk)
Thu, 9 Apr 1998 23:42:31 +0100 (BST)

> FPU) and it should be marked as only runable on CPU A.. Then the kernel
> should do it's thing and decide if it want's to run the task again.. If so
> then it does.. If not then it saves the FPU state and marks it runnable on
> either cpu..

The problem is CPU's make their own indepedant scheduling decisions so CPU
B decides 'I need to run this' and has to signal CPU A to save the FPU
context and wait for it. I tried this in 2.0.x and it was much slower for
most things I tried. Partly perhaps because I should have played with
the scheduler weightings in hindsight

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu