(j3.2006) MPI usage problems in Fortran

[Craig Rasmussen]

On Mar 18, 2008, at 4:17 PM, Aleksandar Donev wrote:

> On Tuesday 18 March 2008 15:04, Craig Rasmussen wrote:
>> Plus if I declare a dummy to have the target
>> attribute, the actual argument need not have the target attribute so
>> the user simply won't add it.  I want these errors to be caught at
>> compiler time.
> That's exactly my "beef" with asynchronous. There are no rules or even
> specifics about the matching of the attribute across argument  
> association.
> Why don't you write the code (including an explicit interface for  
> MPI_Isend!)
> with ASYNCHRONOUS and then I can comment on what the standard (in  
> my reading)
> actually says/guarantees about it.
>


Fair enough.

interface
     function MPI_Isend(buf, count, datatype, &
                        dest, tag, comm, request) BIND(C,  
name="MPI_Isend") result(err)
       use, intrinsic :: ISO_C_BINDING
       import :: MPI_HANDLE
       implicit none
       type(C_PTR), value, asynchronous, intent(in) :: buf     ! LEAD  
to compile error about volatile and C527
       integer(C_INT), value, intent(in) :: count
       type(MPI_HANDLE), value, intent(in) :: datatype
       integer(C_INT), value, intent(in) :: dest
       integer(C_INT), value, intent(in) :: tag
       type(MPI_HANDLE), value, intent(in) :: comm
       type(MPI_HANDLE), intent(out) :: request
       integer(C_INT) :: err
     end function MPI_Isend
end interface

use, intrinsic :: ISO_C_BINDING
use MPI3
type(MPI_HANDLE) :: request
integer, asynchronous :: buf(10)
integer(C_INT) :: err

err = MPI_Isend(C_LOC(buf), 10, MPI_INTEGER, 1, 1, MPI_COMM_WORLD,  
request)

end

There are three problems with this as I see it:
1. VALUE and SYNCHRONOUS lead to a compile error (is this correct?)
2. buf doesn't have the TARGET attribute --- so change C_LOC to allow  
ASYNCHRONOUS attributes as well
3. How long is buf tainted as Bill suggests?  Let's say within the  
scope of the variable and it is the users responsibility to add  
ASYNCHRONOUS if tainted beyond the scope of the variable (argument  
association).

Regards,
Craig