Thought about speeding up startup...

[Matt Zimmerman]

On Fri, Oct 22, 2004 at 12:26:21AM +0100, Colin Watson wrote:

> On Thu, Oct 21, 2004 at 04:11:27PM -0700, Matt Zimmerman wrote:
> > In Oxford, I did some primitive profiling of this, and a huge part of the
> > delay is actually parsing out the (large) pcimap file using only the shell.
> > I believe it was about 300ms per PCI device, which adds up to a significant
> > amount of time.  Most of the modules are actually fairly quick to load.
> > The whole pipeline is started up and stopped again for each device; I have
> > 17 PCI devices in my laptop, and 11 in my desktop.
> > 
> > The trouble is that hotplug runs so early, that it can't rely on having much
> > available in terms of useful programming languages or other
> > infrastructure...
> 
> Would it be very hard to write that in C?

No, it wouldn't.  It would move hotplug from arch: all to arch: any, though.
Probably not an entirely unreasonable tradeoff, if we choose that road.

-- 
 - mdz