On Wed, 2016-09-21 at 17:39 +0200, Patrick Ohly wrote:
> new-rmwork-new-scheduler
> elapsed: 42:58.54
> final disk usage: 12873MiB
> max disk usage: 14230MiB
> A bit better in terms of max disk usage than with the completion
> scheduler, but not by much. My observation is that in practice there
> aren't that many ready tasks, so often the choice of priorities among
> them doesn't have such a big influence. The lower bound of the build
> time is determined by the length of the do_configure->do_compile
> critical path and it does not matter much when the other tasks run, as
> long as they run at some point in parallel to that. Perhaps that is
> different for larger distros. I might redo this benchmark with Ostro OS
> XT...

When building the ostro-xt-image-noswupd in the different
configurations, I get these numbers:

elapsed: 1:52:59
final disk usage: 135963MiB
max disk usage: 136521MiB

elapsed: 1:55:55
final disk usage: 33753MiB
max disk usage: 66832MiB

elapsed: 1:53:33
final disk usage: 33755MiB
max disk usage: 69757MiB

elapsed: 1:53:13
final disk usage: 33754MiB
max disk usage: 65933MiB

elapsed: 1:53:57
final disk usage: 33754MiB
max disk usage: 69173MiB

Here the "rmwork" scheduler again shows some advantages over the
"completion" scheduler (faster build, lower disk usage) as long as the
"additional tasks" are turned off. Looks like doing additional work
(like do_configure) is hurting more than it helps.

Best Regards, Patrick Ohly

The content of this message is my personal opinion only and although
I am an employee of Intel, the statements I make here in no way
represent Intel's position on the issue, nor am I authorized to speak
on behalf of Intel on this matter.

Openembedded-core mailing list

Reply via email to