Just a thought that popped into my head

Bruce Dubbs bdubbs at swbell.net
Tue May 7 10:25:43 PDT 2002


Ian Molton wrote:
> On Tue, 7 May 2002 07:43:21 -0400
> Gerard Beekmans <gerard at linuxfromscratch.org> wrote:

>>>and for the 'real evidence', read the gcc mailinglist. gcc 3 is up to 
 >>>20% slower than 2.95.x in some cases, and not significantly faster 
in any others.
>>Exactly, _some_ cases. What cases are they? I'm not going to read through a
>>year's worth of email archives at GCC. I don't have the time. You come up
>>with some URL's and I'll be more than happy to read them.

> why do I need an URL? dont you believe me?

Trust, but verify.

>I think we should stick to gcc 2.95.3 (maybe up that to 2.95.4), and have 
 >gcc *3* as the optional compiler, considering gcc *3* is the unstable 
one,
 >by its authors own admission.

I went to http://gcc.gnu.org/releases.html and they certainly don't 
indicate that 3.04 is "unstable".

I just did a benchmark that does indicate that 3.04 is slower in compile 
speed.  I did a chroot to a partition that is set up with gcc 3.04 and 
went to the jpeg source directory.

make clean
time make

real    0m52.247s
user    0m47.750s
sys     0m4.120s
root:/usr/src/jpeg-6b# gcc --version
3.0.4
root:/usr/src/jpeg-6b#

and from another window

real    0m45.068s
user    0m39.950s
sys     0m3.940s
[root at lfs3 /mnt/lfs2/usr/src/jpeg-6b]# gcc --version
2.95.2.1
[root at lfs3 /mnt/lfs2/usr/src/jpeg-6b]#

This does indicate a 16% increase in compile time, but unless you are 
compiling glibc, xfree86, or kde a lot, this probably is not a big issue.

I suppose someone needs to do some benchmarks on execution speed, but 
that is much more difficult because of the combinations of optimizations 
and the sensitivity of the benchmarks to the type of code being compiled.

   -- Bruce

-- 
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