Just remarks

jsmaby at virgo.umeche.maine.edu jsmaby at virgo.umeche.maine.edu
Mon May 27 10:40:17 PDT 2002

> Think about it - if gerhard used his 2.2GHz P4 to do the timings, and
> you had (say) a 733MHz celeron... would you be able to predict the build
> time based on his figure?

Yes, after compiling one medium-sized package (static bash would do nicely),
and doing a little arithmatic.

> Now, an average figure - does that make it /any/ easier to guess how
> long your build will take?

To an extent, yes.  The speed depends on lots of things, like memory
bandwith, amount of memory (less than 32mb _really_ hurts), hard drive
speed, etc.  By averaging over many configurations, one creates a mythical
machine that has all those properties averaged, and a given machine will
be more likely to match (using the speed ratio calculated from the first
package's build time).  Things do start to go non-linear in 486's, and
they would likely mess up the average (or rather, systems with < 32mb ram).

A while back, the idea of changing the build time units to ed's (the time
it takes ed to compile) was discused (although I think static bashes would
make more sence).  I think Gerard's answer at the time was that it had
to wait until after 3.0.  I guess it's after 3.0 now, so perhaps it
should be implemented.

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