Things like getting monitor information through DDC take a long time. Also, because clocks are being set there are PLL settling times that need to be observed least you mess things up by touching the hardware before the PLLs lock. It may be that particular drivers have unnecessary waits here and there, but you really have to address them one by one, profiling to find the wait and then investigating whether or not it's really necessary. I have worked on systems with boot times times on the order of a few seconds but these weren't trying to access VESA bios services or probing monitors through DDC. If you remove stuff like that, the boot times will probably be dominated by IDE probing, but then you end up having to hardcode things like monitor EDIDs that the driver would usually try to probe. Mark. On Mon, 29 May 2006, Jun OKAJIMA wrote: > I am working on making a very fast booting technology for Linux. > Check this: > http://www.machboot.com/ > This boots within 10sec with x48 CD-ROM. > I dont intend to blow my own horn, but not so bad result, dont you? > > > One of the bottle neck at the moment is, XFree86 driver. > For example, VESA driver seems to spend several seconds for > hardware initialization. Other drivers (S3 or such) also > needs several seconds. > > My questions are: > > 1. This delay really comes from H/W init (or detect)? > 2. If so, is there any way to reduce this delay? > For example, if I hard-code all parameters about H/W to a driver, > It starts faster? > > Especially, I have much interest in tweaking VESA driver. > > --- Okajima, Jun. Tokyo, Japan. > _______________________________________________ > XFree86 mailing list > XFree86@xxxxxxxxxxx > http://XFree86.Org/mailman/listinfo/xfree86 > _______________________________________________ XFree86 mailing list XFree86@xxxxxxxxxxx http://XFree86.Org/mailman/listinfo/xfree86
