User-Agent: Microsoft-Entourage/9.0.2509
Date: Thu, 20 Sep 2001 11:27:05 -0500
Subject: FW: blocklength zero in MPI_Type_struct 
From: Steven Huss-Lederman <huss@XXXXXXXXXX>
To: William Gropp <gropp@XXXXXXXXXXX>
In-Reply-To: <3B5EAA9E.1743FEE3@XXXXXXXXXXXX>
Mime-version: 1.0
Content-type: text/plain; charset="US-ASCII"
Content-transfer-encoding: 7bit

Dear Bill,

This is an old message I never responded to.  It raises the question of 0
length blocks.  In MPI-1.1 p. 63 line 29 says they are nonnegative.  Thus 0
seems OK.  The formula on p. 64 lines 26-42 uses bl -1 so if the blocklength
is 0 it seems to give a funny value.  I'm hoping I'm missing something
obvious.  Help?

Steve

------ Forwarded Message
From: Peter Ganster <P.Ganster@XXXXXXXXXXXX>
Reply-To: mpi-21@mpi-forum.org
Date: Wed, 25 Jul 2001 13:16:46 +0200
To: mpi-comments@XXXXXXXXXXXXX
Subject: blocklength zero in MPI_Type_struct

Hello,

I have a question:

What happens if one use blocklenghts of zero in the call of
MPI_Type_struct: Is the behavior implementation dependent or is it
definded by the MPI-1 or MPI-2 standard?


Example:

 typedef struct {
   double a, b[2], c;
 } S;

 int count = 3;
 int blocklens[3] =  {0, 2, 0};
 MPI_Aint disps[3] = {0, offsetof(S, b), sizeof(S)};
 MPI_Datatype types[3] = {MPI_BYTE, MPI_DOUBLE, MPI_BYTE};
 MPI_Datatype newtype;

 MPI_Type_struct(count, blocklens, disps, types, &newtype);


And second, are there any technical standard papers (like the C language
standard paper of ANSI) that gives precise answers on questions like
above?



with kind regards
Peter Ganster



------ End of Forwarded Message