Potential Speedup - gcc-3.2.x compiling large c++ projects

Lee Hancock lee at nospamhancock1980.freeserve.co.uk
Tue Feb 18 13:58:46 PST 2003

Greg Schafer wrote:

> Hi
> Apologies for being slightly off-topic, but I believe this is worth
> mentioning to the general testing public.
> Did you know that RH have their own gcc-3.2 branch in the gcc CVS? Well,
> it pays to keep an eye on those sneaky devils :-)
> This patch caught my eye:-
> http://gcc.gnu.org/ml/gcc-cvs/2003-02/msg00155.html
> It seems to have stemmed from a discussion on the gcc list where it has
> come to light that a certain constant default relating to memory
> management is way too low for modern hardware.
> The start of the (long) thread is here:-
> http://gcc.gnu.org/ml/gcc/2003-01/msg01336.html
> Note that the thread is talking about gcc-3.3. 3.3 actually has command
> line switches to tweak the parameters under discussion while 3.2 does not.
> Also, this part of the system (garbage collection) was considerably
> re-written in gcc-3.3 so the gains in 3.2.2 might not be as significant.
> Intrigued by the patch on the RH branch, I tried an experiment here by
> compiling up a stock 3.2.1 but with the RH patch applied. Compiling c++
> projects was definitely faster, but only by a few seconds on the smallish
> projects I tried it on.
> Someone with more patience than me, and at least 256 megs of ram, may want
> to give it a whirl for some larger projects (QT, KDE, MOZ, oOO or
> whatever) and see if it makes a worthwhile difference to the overall
> compile time.
> e.g. install the tweaked gcc with "--prefix=/opt/gcc-test".
> Then compile the projects with:-
> "CC=/opt/gcc-test/bin/gcc CXX=/opt/gcc-test/bin/g++ ./configure blah blah"
> Greg

Ive got 1gig of memory in my machine, i wonder if i change 
+#define GGC_MIN_LAST_ALLOCATED (16 * 1024 * 1024) to
+#define GGC_MIN_LAST_ALLOCATED (128 * 1024 * 1024) i will see more of an
increase in speedup? Might try a complete rebuild over the week end.

Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe lfs-dev' in the subject header of the message

More information about the lfs-dev mailing list