This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Some benchmarks with Apple's -fast
- From: Bradley Lucier <lucier at math dot purdue dot edu>
- To: gcc at gcc dot gnu dot org
- Cc: Bradley Lucier <lucier at math dot purdue dot edu>, nlucier at math dot purdue dot edu, feeley at iro dot umontreal dot ca
- Date: Wed, 24 Sep 2003 23:29:19 -0500
- Subject: Some benchmarks with Apple's -fast
I tried Gambit-C's Scheme benchmark suite compiled on my Cube with
Apple's new gcc with my default options (-O1 -fschedule-insns2 are the
biggies) and with Apple's -fast (plus disabling some things I needed
for correctness) option. The results (and all options) are at
http://www.math.purdue.edu/~lucier/bench2.html
I wanted to do this since bignum code compiled with the new options ran
about 10% faster than the same code compiled with IBM's xlc -O4.
On the whole, alas, my old set of options yielded faster code than with
the -fast options.
There was one source file that the -fast option took 10 times as long
to compile as with my default options. If someone's interested in
checking it out, it's at
http://www.math.purdue.edu/~lucier/pi.i.gz
Most of the time was spent in GCSE.
Brad