[ILUG] Thoughts on a compile box

[Kenn Humborg]

> Quoting <NBBBIGEGHIGMPCNKHCECGEHHDKAA.kenn at bluetree.ie>
> by Kenn Humborg <kenn at bluetree.ie>:
> 
> > > Come up with a .config and send it round. Then everyone wold 
> be making the
> > > same kernel. Pick an exact kernel too. Maybe leave make dep 
> and clean out
> > > in case disk factors come in..
> > 
> > Build it twice.  The first time will pull pretty much
> > everything into the buffer cache (assuming you have loads
> > of RAM).  The second time should then give a fairly
> > good indication of the speed of your CPU/chipset/RAM combo.
> 
> Of course, unless you normally compile your kernel twice, it's the
> first number you should be looking at, and so you come to realise that
> the speed of your disk subsystem is probably more of a factor than the
> speed of your CPUs.

I _do_ normally compile it multiple times.  I'm doing kernel
development.  And I'll be compiling GCC/binutils toolchains
as well, which take about three times as long to compile as
a kernel.

I expect that 128MB of RAM should have no trouble caching most of
the kernel tree (certainly all the source files that actually
get read and the object files that get created).

Later,
Kenn