Re: some profiling numbers

[graydon hoare <graydon at redhat dot com>]

Zack Weinberg <zack@codesourcery.com> writes:

> > it can produce many more profiles, and annotated sources, yes. no
> > call graphs in the current release, but that will be
> > forthcoming. part of the idea with producing these reports was to
> > show the general technique and encourage those people interested in
> > performance tuning to give the profiler a try for
> > themselves. there's an endless list of experiments and results one
> > might want to run, and I can only run so many of them.
> 
> I'm interested in trying some experiments myself, but I have never
> managed to get oprofile to work, and you already have the setup.

phew! ok, so I've made the following changes to my experiment setup:

- configured with --disable-checking
- removed gcc/c-parse.c and gcc/cp/parse.[ch] before building
- selected a different set of compilers (3.0, 3.1, 3.2, 3.3, HEAD)
- rebuilt all the compilers with "make bootstrap"

and I've re-collected the linux kernel build sample set for each
compiler, dumped all the sample output, and all the annotated sources
for files which saw any samples at all. the results are rather large
(about 300mb of annotated source) but that means there's lots of stuff
to look through. who knokws, maybe optimization can be
parallelized. I've posted it all on:

http://www.off.net/~graydon/gcc

the data is obviously in a more "raw" form, so should be more amenable
to answering the myriad comparative questions you may have, without
requiring my interpretation.

-graydon