Just remarks

Don Smith don_smith at att.net
Mon May 27 09:42:16 PDT 2002


Ian Molton wrote:
> 
> On Mon, 27 May 2002 08:00:47 -0400
> Gerard Beekmans <gerard at linuxfromscratch.org> wrote:
> 
> > It's all of just giving a ballpark figure of how long it is going to
> > take, along with approximately how much disk space you're going to
> > need. If you can't use the info then ignore it. Simple as that ;)
> 
> Well, a 500MHz PC takes **roughly** half as long as a 1GHz box (I know,
> I tried it)
> 
> below 500MHz, things like disc subsystem (PIO, UDMA, 33, 66), memory
> type / speed, all conspire to make build time estimates vary wildly.
> 
> and once build times for packages go over 45 mins I dont think most
> people care.

I dunno, I'd like to know if a package takes 1 hour or 4 hours. I could
eat dinner in 1 hour or mow the lawn in 4.

> 
> If we do add multiple speeds, I *dont* want to see 5 wasted lines at the
> top of each page in the book.

I agree with that. Just one time will do.
> 
> I think it'd be better to give RELATIVE compile times for thge packages.
> 
> that is, build all the packages, and group them into packages that take
> similar times to build.
> 
> so we would have:
> 
> glibc: class A (takes forever)
> gcc: class B (takes a while)
> kernel: class C (takes a while but not too bad)
> lilo: class G (nanoseconds)
> 
> this way, people get the best estimates - once they have built any
> package in a given category, they can expect packages in a similar
> category to take a similar time.

No, I'd rather have a number. I can always convert it to the speed of my
machine.
> 
> AND you dont have to get stats from dozens of machines - a machine that
> compiles the kernel in 2 hours will probably do gcc in about 2.5 hours.
> One that does the kernel in 2 mins will probably do gcc in about 3 or 4.
> 

But here it is easy to get stats from dozens of machine. Look at the
last time times were requested in this list. I saw at least a dozen
replies.

But, personally I don't care whether the times are an average or the
results from just one machine. What matters is the relative time between
different packages. Given those times and the results of building just
one package on my machine, I can estimate fairly acurately how long it
will take to build any other package or the entire load.

Don
-- 
Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe lfs-dev' in the subject header of the message



More information about the lfs-dev mailing list