compile speed on lfs 5.0

mrnobby no at spam.please
Wed May 5 15:18:14 PDT 2004


"John Gay" <johngay at eircom.net> wrote in message
news:200405041544.12204.johngay at eircom.net...
> On Mon 03 May 2004 19:49, mrnobby wrote:
> > "shelton" <shelton at onr.com> wrote in message
> > news:40957D37.4070108 at onr.com...
> >
> > > mrnobby wrote:
> > > > I've built LFS 5.0 on a P233 MMX.
> > > >
> > > > I'm rebuilding LFS 5.0 using the above LFS 5.0 build as the host
on
> >
> > the
> >
> > > > same machine.
> > > >
> > > > Why are my compile times now twice as long?
> > > >
> > > > Mr Nobby.
> > >
> > > has the hardware changed at all?
> > > what services and programs are you running while compiling?
> > > if this isn't a live production must have online all the
> > > time system
> > > (which is doubtful, why rebuild a live mission critical system)
> > > then try dropping down to runlevel 1 when building.
> >
> > Nothing has changed. I built LFS on RH7, now I'm trying to build LFS
on
> > LFS, nothing has changed in between. Changing to init 1 won't halve
the
> > time will it?
> >
> I'm not sure if this is applicable, but on Chris's Bleeding Edge LFS
disk, the
> instructions mention that newer versions of GCC are slower than older
ones.
>
> I would guess that the RedHat version of GCC is significantly faster
than the
> LFS version. This would explain the slower compile times
>
> But, that does not mean the box will run slow. GCC is notorious for
creating
> extremely large hash tables and other memory structures in memory, and
if
> memory is limited, swapping slows everything down. But the resulting
binaries
> show run as fast or faster.
>
> Just the thought.
Thanks, unfortunatley, it was not only gcc, it was *all* packages
created for stage 1. They were all 2x longer to within a few minutes.
I'm not bothering with stage 2 until I fathom this out.

I'm wondering if RH has some sort of special optimizer (caching???) but
if they have they would have to make it available under the gnu license
wouldn't they?

Mr Nobby.





More information about the lfs-support mailing list