I did some testing tonight using the environment variable MAKEFLAGS=-j2.
The results for Chapter 5 are below. I used the followign to enter
chroot for Chapter 6:
/usr/sbin/chroot $(MOUNT_PT) /tools/bin/env -i HOME=/root TERM="$$TERM"
PS1='\u:\w\$$ ' PATH=/bin:/usr/bin:/sbin:/usr/sbin:/tools/bin
MAKEFLAGS=-j2 /tools/bin/bash --login +h
But that didn't have much affect on the build times other than making
gawk crash during the tests. I'll have to investigate Chapter 6 some more.
After we get 6.6 released, I think we might want to consider adding to
the General Compilation Instructions:
export JOBS=-j2
for systems with multiple cores and add to gcc, glibc, and perhaps binutils
make $JOBS
For systems where JOBS is not defined, it would do nothing, but we can
get a good speedup for the large builds. I don't see any real advantage
to adding $JOBS to the other packages because the real time advantage is
minimal.
If we added it to binutils-pass1, it would mess up SBU times a lot.
We would need to discuss that.
-- Bruce
----------------------
LFS 6.6-rc1 build times in seconds
Chapter 5
Package j1 j2 % diff
[binutils-pass1] 91 56 61.5 35
[gcc-pass1] 442 276 62.4 166
[linux-headers] 6 5 83.3 1
[glibc] 500 334 66.8 166
[binutils-pass2] 94 58 61.7 36
[gcc-pass2] 606 380 62.7 226
[tcl] 33 21 63.6 12
[expect] 4 4 100.0 0
[dejagnu] 5 4 80.0 1
[ncurses] 50 35 70.0 15
[bash] 39 29 74.4 10
[bzip2] 3 1 33.3 2
[coreutils] 60 49 81.7 11
[diffutils] 8 8 100.0 0
[findutils] 22 19 86.4 3
[gawk] 20 16 80.0 4
[gettext] 59 48 81.4 11
[grep] 9 7 77.8 2
[gzip] 14 13 92.9 1
[m4] 17 15 88.2 2
[make] 10 8 80.0 2
[patch] 6 5 83.3 1
[perl] 58 43 74.1 15
[sed] 11 10 90.9 1
[tar] 24 21 87.5 3
[texinfo] 18 14 77.8 4
2209 1479 67.0
Chapter 5 (minutes) 36.8 24.7
Chapter 5 SBU 24.3 26.4
Total (minutes) 198.84 185.92
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
