On Mon, 31 Dec 2001 00:27:54 +0100, Peter Graf wrote: > [EMAIL PROTECTED] wrote: > > > I have measured the time it takes to boot Linux on Q40 (starting at the > > penguin): > > > > Kernel 2.2.17 : 62 sec > > Kernel 2.4.17 : 36 sec > > > > I used the same installation, of course. > > Thierry, your kernel seems to be mega fast. > > Indeed. I know I'm comparing apples and oranges, but I was really > puzzled, when I booted the same kernel version on a 1200 MHz x86 PC. > Took 45 sec, measured after LILO, just about the same amount of deamons > started. > > So all the probing and fiddling with the PC hardware has reached a point, > where a machine with 3000% higher clocked x86 CPU boots Linux slower than > a well-configured 68040 machine. A little bit crazy.
No, very logical on the contrary: keep in mind that most of the hardware must be awaited for while probing takes place (i.e. you have to include delay loops in your probing code so that the hardware has the time to setup and reply your queries). The delay loops involved are usually processor speed independant (else some weird things could happen, like in Windoze95 which is so dumb that with fast processor, e.g K6 >300MHz, the initialization fails resulting in a magnificent crash). This is why, beyond some point, whatever is your processor speed, initializing (or probing) a given hardware will take the same time... The hard drive speed is also a significant (mostly processor speed independant) factor for system such as Linux (which are, unlike SMSQ/E, heavily hard-disk based and must load all their daemons from disk). Thierry.
