Howdy folks,

   Figured I'd do this while I'm sober and tired!

I usually build in ram now that the machines I build with either have 1GB or
2GB of DDR in them. For a sample of how fast it can be:

P4 3.2GHz with 2GB using dual channel linear DDR400/PC3200
with 4 512MB modules I get this time with glibc:

10m47s, that's just building not installing or configuring.

On the P4 1.8GHz with 1GB at DDR200 I get something:
well it is under 30 mins and over 10m47s.

If your machine only has 1GB ram tmpfs should give you around 400 megs to work
with and around 1GB if you have 2GB of ram. This is how I usually build:

I put all the source in $LFS/sources and then mkdir $LFS/sources/build and mount
tmpfs to build, something like mount -t tmpfs shm $LFS/sources/build.
Then so I don't have to type more '../'s than I want I just link everything in 
sources
to build.

When building binutils, put it on a hard drive in case the system drops and you 
lose
what was in the tmpfs and then can't adjust the toolchain later.

If 1GB ram is being used and there is only around 400 megs for tmpfs, extract 
the source
on the drive and build gcc, glibc, and any other large packages to the drive 
and then just
build the source in ram using a separate configure and build directory if 
allowable. The kernel
tree can be extracted and built all in tmpfs if ya get enough ram for tmpfs.

If 2GB ram is being used then it can all be done in tmpfs and heed the warning 
on binutils.
I built a whole LFS in less than 2 hours last night. Of course I skipped a few 
things like
bootstrap since I knew the compiler would work just fine on the systems. Even 
with bootstrapping it would have taken less than 2 hours. No checks either.

If ya have ram, might as well use it!

William
-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-chat
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to