Ken Moffat wrote:
In the past, I have been planning to compare build times (and
perhaps sizes) using ld.bfd and ld.gold, but never got around to it.
For the moment I am mostly doing photo edits on my main build
machine, and it occurs to me that I can spare at least 4 cores to
run some tests. But what ? Firefox looks like a likely target, as
does libreoffice - but that needs to download a load of stuff on
each run, and past experience suggests that leaving it to run
overnight often fails (site being backed up). I will also note that
I don't use gold for QT5 (there have been problems), so by the same
token I doubt that building the panoply of weird and wonderful deps
for any of kde will be worth the time.
So, I suppose that leaves webkitgtk (I'll need to build a number of
deps, but I think I can adapt my main scripts).
The plan is to build (as root, if possible) and DESTDIR install each
package 6 times, 3 with each of the linkers, and dropping caches
before each run. Does that sound worthwhile ?
If so, any other interesting packages ?
I can't think of any that you haven't mentioned. My question is why use a
new linker at all? My understanding is that it may produce files faster,
but to me the link phase of a package is generally quite short compared to
the time it takes ot actually compile the packages. If the compile time
is 10x the link time and you get a 50% speedup (unlikely), you get a 5%
overall reduction in build time. To me that's not significant.
What am I missing?
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
