Re: [PATCH] proposed scheduler enhancements and fixes

From: Dimitris Michailidis (dimitris@cthulhu.engr.sgi.com)
Date: Sun Feb 27 2000 - 22:07:07 EST

On 27-Feb-2000 Andrea Arcangeli wrote:
> On Sun, 27 Feb 2000, Linus Torvalds wrote:
>
>>It is not really single-threaded. In fact, in theory it could use 2 CPU*s
>
> Ok. I was thinking with a pipe buffer to 1 byte but it's PAGE_SIZE large
> instead of course ;).
>
> So my smp scheduler changes are probably the culprit for the lat_ctx
> numbers.

After extensive experimentation it turns out that the culprit is this test

        if (p->policy == SCHED_OTHER) {
                tsk = cpu_curr(best_cpu);
                if (p->avg_slice + tsk->avg_slice < cacheflush_time)
                        goto out_no_target;
        }

that was moved and changed recently. This made Linux more aggressive about
exploiting idle CPUs and apparently lat_ctx likes to execute on 1 CPU.
Putting the above code back makes 2.3.47 a tad faster than 2.3.40. dbench
remains miserable. 

-- 
Dimitris Michailidis                    dimitris@engr.sgi.com

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 21:00:18 EST