Re: next ballot

[William Gropp <gropp@mcs.anl.gov>]

At 01:30 PM 11/5/2001 +0100, Jesper Larsson Traeff wrote:

>What about the following replacement for p.120, lines 6-13. It clearly
>says that predefined operations including MPI_REPLACE behave uniformly,
>and that the extension to derived datatype is an extension that only
>applies to MPI_ACCUMULATE:

I like your proposed replacement.  I propose to substitute this text as 
item 7 in the ballot.  Are there any objections?
Bill


>---
>
>A new predefined operation, MPI_REPLACE, is defined. It corresponds
>to the associative function f(a,b)=b, i.e., the current value in the
>target memory is replaced by the value supplied by the origin. MPI_REPLACE
>can be used the same way as the other predifined operations.
>
>The datatype arguments can be any of the predefined datatypes as explained
>in Section 4.9, and in that case both origin and target datatype must be
>the same predefined datatype. For MPI_ACCUMULATE the predefined operations
>are extended to work for derived datatypes, where all basic components
>are of the {\em same} predefined datatype. Both datatype arguments must be
>constructed from the same predefined datatype. The operation {\sf op} applies
>to elements of that predefined datatype. The target datatype must not
>specify overlapping entries, and the target buffer must fit in the
>target window.
>
>---
>
>Jesper