Le 16/04/2013 13:47, LM a écrit : > > Ken Moffat wrote: > > LFS itself took about 3 hours. > > That is very impressive. Was experimenting with a build of just gcc > over the weekend and that took 4 hours by itself. Didn't try the -j > option. Some of the systems I've tried it on just locked up with it. > I try to maintain a list of packages who do not accept the -j option, for jhalfs (actually, not tested for a while). All I have is: BLACK_LIST="attr autoconf coreutils dejagnu gettext gperf groff man-db vim"
attr and gperf are for the systemd branch. Only the tests actually need -j1, as well as for coreutils and vim. For the others, I have not tested for a while, but they do not take long anyway. All the LFS packages not listed in $BLACK_LIST build fine on linux with -jX (X=3..15, tried quite a few distributions on virtual machines, both 32 et 64 bit, on two 64 bit hosts, one with a core i5 at 2,6 GHz, and the other one with two quadcore CPU, Xeon E5620 at 1,6 Ghz with hyperthreading). FWIW, a full 7.3 build on a bi-processor Xeon E5620 (quad core) at 1,6 GHz with -j5 took ~2 h, running only the toolchain tests (chapter 6 glibc, binutils, gmp, mpfr, mpc and gcc). With -j5, gcc took 31 mn, build, test, install. I have almost the same timings with a core i5 at 2,6 GHz (slightly faster actually). Pierre -- http://linuxfromscratch.org/mailman/listinfo/blfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
