[Bug fortran/31014] New: missed-optimization: unnecessary invokation of _gfortran_internal_pack

[burnus at gcc dot gnu dot org]

Calling _gfortran_internal_pack is not needed, if it is clear that the array is
contiguous. If it is further known that the called procedure uses 
dimension(*), the creation of the array struct is also unnessarily.

In the following, the  _gfortran_internal_pack call itself is unneeded:

  external foo

  ! At compile-time known bounds, could be a local or dummy variable
  ! Could also be the dummy argument "a(*)",
  ! if one specifies below the bounds "a(1:<some value>[:1])"
  real :: a(2:12)

  call foo( a(:) ) 
  call foo( a(2:) )
  call foo( a(:12) )
  call foo( a(2:12:1) )

The current tree looks like:
    struct array1_real4 parm.0;
    parm.0.dtype = 281;
    [...] // Initialize array struct, OK!

    D.1362 = _gfortran_internal_pack (&parm.0);
    foo (D.1362);
        _gfortran_internal_unpack (&parm.0, D.1362);
        _gfortran_internal_free (D.1362);

The last four lines can be replaced by a simple "foo(&parm.0)".


Note that for one dimension, even
  call foo (a(2:5) ) ! reduced ubound
or
  call foo ( a(4:)) ! increased lbound
are possible without needing _gfortran_internal_pack.


If one knows that the called subroutine takes dimension(*) then there is also
no need to fill the array struct:

 interface
   subroutine foo(x)
     real :: x(*)
   end subroutine foo
 end interface
 real :: a(*) ! Can also be, e.g. "a(1:12)"

 call foo(a(2:5:1))
end subroutine

Here, one can simply call

   foo(&a[1])

instead of filling the array struct and calling then _gfortran_internal_pack.


-- 
           Summary: missed-optimization: unnecessary invokation of
                    _gfortran_internal_pack
           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: burnus at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31014