This is the mail archive of the
mailing list for the GCC project.
Re: gcc compile-time performance take 2
> Am Son, 2002-05-19 um 03.06 schrieb Robert Dewar:
> > I don't think timing the spec builds is that useful except for gcc, and that
> > information is more usefully obtained by measuring the time for a bootstrap
> > which is a figure that is itself useful. I am going to see if we can extract
> > useful information automatically from the ACT nightly builds.
> I'm not exactly sure that measuring the time to build a changing
> codebase is all that useful. Having a codebase that is fixed seems
> more useful to me .
I believe both is usefull. SPEC build time should give us imager of the raw
compilation speed on C,C++ and fortran, while bootstrap should point out if we
add something particulary expensive for example into libstdc++ headers,
thats why I asked for both graphs. Hope Andreas's scripts will post the results
soon so we will see how much noisy the lines are.