On 4/18/06, Menno Lageman [EMAIL PROTECTED] wrote:
Holger Berger wrote:
I think one part of this jigsaw is the disk bottleneck. If you build
ON on a tmpfs volume you should have a far better CPU utilisation on
Niagara.
Nothing beats real data, so I ran a nightly on a tmpfs file system with
Menno Lageman [EMAIL PROTECTED] wrote:
Indeed, setting it to 2 increases build time to almost 4 hours. I need
to take back my statement that increasing max jobs to 32 doesn't
decrease build time though (must have loked at the wrong logfile):
jobstime
4 2:26
8
Joerg Schilling wrote:
Menno Lageman [EMAIL PROTECTED] wrote:
Indeed, setting it to 2 increases build time to almost 4 hours. I need
to take back my statement that increasing max jobs to 32 doesn't
decrease build time though (must have loked at the wrong logfile):
jobstime
On 4/14/06, Menno Lageman [EMAIL PROTECTED] wrote:
I happen to have access to a T2000 (1 GHz, 32 strands) for a couple of days,
so I ran a nightly of on20050327:
Nightly distributed build completed: Thu Apr 13 22:17:16 CEST 2006
Total build time
real2:26:51
This
Holger Berger wrote:
I think one part of this jigsaw is the disk bottleneck. If you build
ON on a tmpfs volume you should have a far better CPU utilisation on
Niagara.
Nothing beats real data, so I ran a nightly on a tmpfs file system with
max jobs = 32. The build time decreases from 1:53
I happen to have access to a T2000 (1 GHz, 32 strands) for a couple of days, so
I ran a nightly of on20050327:
Nightly distributed build completed: Thu Apr 13 22:17:16 CEST 2006
Total build time
real2:26:51
This was with the default max concurrent jobs = 4. Increasing
On Fri, Apr 14, 2006 at 01:06:21AM -0700, Menno Lageman wrote:
This was with the default max concurrent jobs = 4. Increasing it to 32
through $HOME/.make.machines did not decrease build time, because the
build process is mostly serial as Bart noted earlier. Apart from short
periods where 32
Do you think it's possible to change the build, such that more can go on in
parallel?
Or, are there some limiting steps that are hard to eliminate (if so, where do
you think they are?)?
Thanks,
Mike
Danek Duvall wrote:
On Fri, Apr 14, 2006 at 01:06:21AM -0700, Menno Lageman wrote:
This