lkml.org 
[lkml]   [2007]   [Dec]   [14]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Major regression on hackbench with SLUB (more numbers)

Christoph,

Welcome back from your vacation! :-)

On Thu, 13 Dec 2007, Christoph Lameter wrote:
>
> In an extreme case (boot with slub_min_order=9 to get huge page sized
> slabs) SLUB can win against SLAB:
>
> N=10 Time: 0.338 Minimally faster
> N=20 Time: 0.560 10% faster
> N=50 Time: 1.353 15% faster
>

I booted back to slub with slub_min_order=9 on the 64way and you are
indeed right about this.

# tests/hackbench 90
Time: 2.109
# chrt -f 10 tests/hackbench 90
Time: 3.775
# tests/hackbench 90
Time: 1.583
# chrt -f 10 tests/hackbench 90
Time: 1.833
# tests/hackbench 90
Time: 1.601
# chrt -f 10 tests/hackbench 90
Time: 3.078
# tests/hackbench 90
Time: 1.321
# chrt -f 10 tests/hackbench 90
Time: 1.933



A flux between 1.321 and 3.775 (*)(the rt runs were higher). But still,
good job at narrowing this down. Now the hard part. Figuring out a way
that will be nice to all users.

-- Steve

(*) The previous runs (without the command line option) were 13 seconds or
so.


\
 
 \ /
  Last update: 2007-12-14 07:19    [W:0.105 / U:1.308 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site