I just checked the code and to be honest there is no difference,
except for the initialization part. The code is clearly "copy&paste"
from one version to the other. The FORTRAN specific code is related
to the number of elements in the array, in the ompi pointer array we
limit the max size of the array to the largest Fortran integer. I
don't think this is a real limitation, as I doubt anyone will
allocate an array of 2^32-1 pointers. Actually, we can even keep this
max size, as the orte pointer array is not supposed to grow that big
At the light of these, it seems there is a 4th option. Merge
orte_pointer_array and ompi_pointer_array in one opal_pointer_array.
I volunteer to do the merge, but not before next week.
On Oct 17, 2007, at 10:43 PM, Tim Prins wrote:
> The openib and udapl btls currently use the orte_pointer_array class.
> This is a problem for me as I am trying to implement the RSL. So,
> as far
> as I can tell, there are 3 options:
> 1. Move the orte_pointer_array class to opal. This would be quite
> to do and makes sense in that there is nothing in the
> specific to orte.
> 2. Change the udapl and openib btls to use a simple C array. There is
> currently a ticket filed (https://svn.open-mpi.org/trac/ompi/ticket/
> to do this in the openib btl.
> 3. Change the btls to use the ompi_pointer_array. This might not be a
> good idea since the above ticket says that the ompi_pointer array was
> intentionally not used.
> Any of these options are fine with me, although if #2 is picked
> else would probably need to do most of the work.
> devel mailing list
- application/pkcs7-signature attachment: smime.p7s