This is the mail archive of the fortran@gcc.gnu.org mailing list for the GNU Fortran project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Module functions are not inlined


> This is a head's up: I looked at PR 32817
> (http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32817 ), and discovered that
> functions form used modules are not inlined by gfortran.  I always
> thought they were, so just in case other people think the same as I did,
> I send this mail :)

This is slightly more complicated: contained functions can be inlined
within a module (see the polyhedron test suite for examples), for instance 
you can get

call	___free_input_MOD_next
[ibook-dhum] lin/test% gfc -S -O3 fatigue.f90
[ibook-dhum] lin/test% egrep "call[^L]*_.*MOD" fatigue.s
call	___perdida_m_MOD_generalized_hookes_law
call	___perdida_m_MOD_generalized_hookes_law
call	___free_input_MOD_next
[ibook-dhum] lin/test% gfc -S -O3 -finline-limit=500 fatigue.f90
[ibook-dhum] lin/test% egrep "call[^L]*_.*MOD" fatigue.s
[ibook-dhum] lin/test% 

where (with the right inlining limit) the functions
perdida_m_MOD_generalized_hookes_law and ___free_input_MOD_next are
inlined.  I have seen an example (but cannot find it again) where a
function was called within the module containing it and in the main program
and was inlined in the first case but not in the second.

Cheers,

Dominique


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]