From: Erez Haba <erezh@XXXXXXXXXXXXX>
Date: January 23, 2008 12:19:20 PM CST
To: "mpi-21@mpi-forum.org" <mpi-21@XXXXXXXXXXXXX>
Subject: RE: [mpi-21] Proposal EH2 & MPI_INPLACE
Reply-To: mpi-21@XXXXXXXXXXXXX

Thanks Hubert,

This is the same discussion we had at the MPI Forum meeting re const.

I looked up (again) the C99 standard (and the C89) and the standard  
does not preclude a function from changing the constant buffer as  
long as it not done through a pointer that is passed in as const.

That is, for the following function

         void foo(const void* buf1, void* buf2);

it is perfectly alright to change buf2 even if it overlaps with buf1.

I will bring up this point in the *const* proposal.


For your reference, the C99 standard words are:

------
         const int *ptr_to_constant;

The contents of any object pointed to by ptr_to_constant shall not be  
modified through that pointer, but ptr_to_constant itself may be  
changed to point to another object.

-------

Note the "...through that pointer..." text.

Thanks,
.Erez


-----Original Message-----
From: owner-mpi-21@XXXXXXXXXXXXX
On Behalf Of Hubert Ritzdorf
Sent: Wednesday, January 23, 2008 10:01 AM
To: mpi-21@XXXXXXXXXXXXX
Subject: Re: [mpi-21] Proposal EH2 & MPI_INPLACE



Erez Haba wrote:
> Using the send buffer pointer?
>
> Can you give an example?
>
>
If you define 2 derived datatypes send_type and recv_type, which don't
overlap,  you can execute for example MPI_Gather in the following way

{
     int buf[1000];

      MPI_Gather (buf, 1, send_type, buf, 1, recv_type, 0,  
MPI_COMM_WORLD);

    ....

}

In this case buf is send and recv buffer of MPI_Gather and
buf is IN and OUT argument of MPI_Gather in root process 0.


Hubert