The patch for PR 34671 created an optimization opportunity that
is currently missed.
The "any" and "all" functions with dim=1 now accept an argument that
can be any logical type. Only the lowest-order byte is used for
evaluation.
Now consider the following:
program main
logical(kind=1), dimension(2,2) :: a
logical, dimension(2) :: b
data a / .false., .true., .false., .false. /
b = any(a,dim=1)
end program main
This gets translated to
_gfortran_any_l1 (&atmp.2, &parm.1, &C.511);
{
integer(kind=4) S.4;
S.4 = 0;
while (1)
{
if (S.4 > 1) goto L.1;
b[S.4] = (logical(kind=4)) (*(logical(kind=1)[2] *)
atmp.2.data)[S.4];
S.4 = S.4 + 1;
}
L.1:;
}
}
where a simple call to _gfortran_any_l4 would have been enough.
--
Summary: logical kinds for any and all
Product: gcc
Version: 4.3.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: enhancement
Priority: P3
Component: fortran
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: tkoenig at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34816
