[FORTRAN PATCH] Reduce gfc_match_strings usage (part 1)
Tobias Schlüter
tobias.schlueter@physik.uni-muenchen.de
Sun Mar 11 17:27:00 GMT 2007
Bud Davis wrote:
> --- FX Coudert <fxcoudert@gmail.com> wrote:
>> A few days ago, I was wondering how much we could
>> improve gfortran's
>> compilation speed,
>
>
> with the results shown, the answer seems to be 'not
> much'. my usual response when see in a gprof output
> like this is to give up.... there will be a lot of
> work for a couple of percent.
>
> when i optimise, i like to see gains in the X 5 or X
> 30 range, not in + 0.03 . :):)
There's one well-known quadratic bottleneck, which is only relevant for
legacy codes: the resoultion of labels and GOTOs, see PR 18937. I've
never felt bothered enough to actually pull this through, but the gains
could be quite significant for some users.
- Tobi
More information about the Fortran
mailing list