Bruce Dubbs wrote: > Bryan Kadzban wrote: >> Bruce Dubbs wrote: >>> I did some testing tonight using the environment variable >>> MAKEFLAGS=-j2. >> I assume you have a dual-core CPU -- have you tried -j3? Any higher? > > Chapter 6 binutils: > > time make -j2 tooldir=/usr 68 seconds > time make -j3 tooldir=/usr 60 seconds > > time make -j2 check 55 seconds > time make -j3 check 53 seconds > > From jhalfs -j1 > Chapter 5 binutils 91 seconds > Chapter 6 170 seconds
So a *much* bigger improvement going from 1 to 2, than from 2 to 3. Very likely not worth going higher than your CPU-count then. > Comparing Chapter 6 > -j1 170 > -j2 133 + make install and other very short ops > -j3 123 + make install and other very short ops > > The only significant speedup seems to be for the 'make' itself. > There appears to be less than 10% gain going from -j2 to -j3 Yeah, seems not-worthwhile to me. Since most of this speedup is in the few giant core packages anyway, and the j2-to-j3 speedup is probably down in the noise... > Since most packages in LFS are less than 1 SBU, I don't see any real > advantage for those. Using -j2 for binutils, gcc, glibc, and maybe > gettext would speed things up IF you have a multi-processor or > multi-core system AND it's not your first LFS build. Yeah. > Right now I'm leaning against any changes to the book in this area. At > most, a mention in General Compilation Instructions is all that is > warranted. Especially for 6.6.
signature.asc
Description: OpenPGP digital signature
-- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
