Michael Shell wrote:
On Fri, 25 Sep 2015 03:07:58 +0100
Ken Moffat <[email protected]> wrote:
My own A4, when it was alive, showed no problems (apart from only
having two cores). My A10 (kaveri) is fine, and there I use make
-j5 because it has an SSD. So, from my experience only the Phenom
is a problem.
Ken,
If it is a problem in the kernel or with make, then I'd say that is
something "somebody outta fix".
This may be something more relevant to the LFS list, but have you
heard about or tried the make 4.1 vfork patch (March 2015)?
0001-Fix-paralellism-by-reverting-to-use-of-vfork.patch:
http://savannah.gnu.org/bugs/?44555
http://savannah.gnu.org/bugs/download.php?file_id=33382
In one of the comments it was stated:
"With make.git at 4.1-11-ga80a8b8, "make -j10" typically takes
about 3 minutes. With the attached patch applied, "make -j10"
typically takes about 11s."
I could not apply that patch to make-4.1 without modification. There
are no template files in the tarball. After removing sections for those
files, the patch applied with offsets, but the build was find and make
check passed with no failures.
patching file configure.ac
patching file job.c
Hunk #1 succeeded at 1154 (offset -6 lines).
Hunk #2 succeeded at 1520 (offset -27 lines).
Hunk #3 succeeded at 1546 (offset -27 lines).
patching file makeint.h
Hunk #1 succeeded at 535 (offset -17 lines).
patching file remote-cstms.c
I sometimes watch a long build with top and the current make seems to
use as many cores as specified.
As a test, I rebuilt mesa at -j8. With the standard version of make,
156.5 Elapsed Time - mesa-10.6.6
and with the patched version:
156.8 Elapsed Time - mesa-10.6.6
so the patch doesn't make any difference for timing, at least with my
i7-5820K CPU.
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/blfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
