7:00 - 8:00 UTC please

[Eric Wong]

testing online reduce

[Eric Wong]

On Jun 16, 11:38 pm, Eric Wong <normalper...@gmail.com> wrote:
> testing online reduce

NJOBS=32 NPARTS=128 (read/write):
real 7:29.7
user 2:17:05.2
sys 22:16.3

Trying pthreads-io now (not looking good)...

[Eric Wong]

On Jun 17, 12:19 am, Eric Wong <normalper...@gmail.com> wrote:
> On Jun 16, 11:38 pm, Eric Wong <normalper...@gmail.com> wrote:
> Trying pthreads-io now (not looking good)...

(pthreads-io + online reduce NJOBs=32 NPARTS=128)
real 8:27.8
user 2:03:29.6
sys 24:02.8

[Eric Wong]

Again with fewer threads:
pthreads-io + online reduce NJOBS=16 NPARTS=128
real 8:17.2
user 1:53:01.2
sys 23:18.1

[Eric Wong]

All a balancing act (NJOBS=24 NPARTS=144)
real 7:46.1
user 2:10:04.6
sys 25:21.2

[Eric Wong]

On Jun 17, 12:19 am, Eric Wong <normalper...@gmail.com> wrote:

> On Jun 16, 11:38 pm, Eric Wong <normalper...@gmail.com> wrote:
>
> > testing online reduce
>
> NJOBS=32 NPARTS=128 (read/write):
> real 7:29.7
> user 2:17:05.2
> sys 22:16.3

I guess this is the best I could do tonight. I'm pretty definitely I/
O-bound
at this point but the extra context-switch from using pthreads I/O
does't work so well. Maybe I'll look into using mmap properly, to
do I/O, since whatever I'm doing with mmap doesn't seem to be
working...

This was 0cfa5afe23f08f2077700df7ca408d5c4051faea from my git repo

[Eric Wong]

Also updated http://bogomips.org/wf2/README and
http://wikis.sun.com/display/WideFinder/Results

[Eric Wong]

Actually, this was e80624c7513c15a4d6154fd4c0c18927cd5b1b5c

I had to rewrite the git commit because I was on a different machine
and used the wrong email address. I didn't want that one to be public