This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: huge object files. Expectable?
- From: Paul Richard Thomas <paul dot richard dot thomas at gmail dot com>
- To: "Andrew F. Nelson" <andy dot nelson at lanl dot gov>
- Cc: fortran at gcc dot gnu dot org
- Date: Sun, 21 Mar 2010 09:44:24 +0100
- Subject: Re: huge object files. Expectable?
- References: <4000E6E8-1D27-40F4-B6A7-4222ACF623C6@lanl.gov>
Dear Andy,
> Hopefully, I'm asking my questions in the right place. Looks like mostly
> developer traffic here rather than user traffic, of which this message
> is the latter. Please redirect as necessary.
Whilst this is a developer list, we are happy to take messages of this
kind; we are not particularly tolerant of requests for help with the
language itself or problem reports that are more general than gfortran
or are so discombobulated as to be incomprehensible :-)
....snip....
> Thread model: posix
> gcc version 4.5.0 20090604 (experimental) [trunk revision 148180] (GCC)
>
This version, since it goes back to last June, will be fairly close to
the 4.4 release but it might have some bugs that have been fixed for
the latter. That said, it should be pretty good.
>
> After getting it going and trying to compile a few things, I notice that
> occasionally I will get an extremely large object file: hundreds of MB in size.
> The specific files that seem to create these large .o files declare a number
> of arrays to rather large sizes and have no other functionality (i.e. no
> procedures). Based on that fact, it may be understandable that it makes
> a large .o file, but since I have never really observed this with other
> compilers over the years, I'm wondering:
>
> 1) is this something that is expectable with gfortran? (why? and why not with others?)
We have a known problem with initializers for large arrays.
> 2) is it some sort of pathological case that you haven't hit and therefore
> might be called 'regression' or some such?
A regression is a bug that was not present in earlier versions of the
compiler; typically this arise because a fix for one problem
introduces another. Usually, these are corner cases because our
testsuite is getting to be quite substantial.
>
> I can provide an example file that triggers this, if anyone wants to look
> at it.
If it is fairly well reduced, we would be happy to look at it.
Best regards
Paul