lkml.org 
[lkml]   [2008]   [Jul]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [git pull] cpus4096 fixes

* Linus Torvalds <torvalds@linux-foundation.org> wrote:

> On Sun, 27 Jul 2008, Ingo Molnar wrote:
> >
> > Please pull the latest cpus4096-fixes git tree from:
>
> No. Not without explanations.
>
> Quite frankly, this "fix" looks like a huge stinking pile of sh*t.

Agreed :-/

NUMA-locality might have been a valid argument in favor of the massive
array of constant cpumasks (common usage is to use it for the current
cpu), if it wasnt all stupidly allocated on the boot node:

cpumask_of_cpu_map = alloc_bootmem_low(sizeof(cpumask_t) * nr_cpu_ids);
for (i = 0; i < nr_cpu_ids; i++)
cpu_set(i, cpumask_of_cpu_map[i]);

> And I further right, and you are so STUPID that you cannot see that
> you can share all the zero words?

That's fair to say :-/ I cannot talk for the others but it certainly
didnt occur to me and i should have caught it.

( To me it didnt occur because we rarely have the opportunity for such
rather clever optimizations, in most cases there's really just two
choices in practice: either maximally global, or maximally per cpu.
Here we can do something inbetween and overlap. I dont think we do any
comparable compression of constants in the kernel. That boxed in my
imagination i guess. )

Sorry about this - please ignore these patches, we'll rework them along
the lines you suggest. Your suggested data structure will indeed be both
simpler, smaller and more cache-efficient.

Ingo


\
 
 \ /
  Last update: 2008-07-27 23:07    [W:0.274 / U:0.052 seconds]
©2003-2014 Jasper Spaans. Advertise on this site