Messages in this thread | | | Date | Sat, 19 Jul 2008 09:14:20 -0700 | From | Arjan van de Ven <> | Subject | Re: [patch 1/3] fastboot: Create a "asynchronous" initlevel |
| |
On Sat, 19 Jul 2008 09:08:05 -0700 Daniel Walker <dwalker@mvista.com> wrote:
> On Sat, 2008-07-19 at 08:35 -0700, Arjan van de Ven wrote: > > > > Ok .. I'm not trying to rush you .. With that said, the increased > > > number of threads seems like a natural direction to take.. I would > > > assume the returns would diminish depending on the number of > > > threads per the hardware.. For instance , your addition of one > > > thread might even harm the boot time on some single core embedded > > > systems .. > > > > well it still improves the cases where msleep() happens... even on > > single core. (fwiw the test machine I reported numbers from is > > single core, albeit with hyperthreading) > > > > How did you measure the boot time? I didn't notice any infrastructure > for that in the patches.. >
well.. we already have the printk-timestamps and the initcall_debug also timestamps individual initcalls...
the way I measured it was based on the printk timestamp of the last thing the kernel did before giving control to init. (which was a printk I added just for this purpose) (and the bootchart tool also confirmed the same data, based on system uptime)
fwiw I have more patches coming to help boottime once these are done, but these 3 were self contained and ready for posting, the others need some more cleanup first.
-- If you want to reach me at my work email, use arjan@linux.intel.com For development, discussion and tips for power savings, visit http://www.lesswatts.org
| |