mpid_tbsend.tex

刚才是说明 现在是安装程序在 LINUX环境下进行编程的MPICH安装文件

\startmanpage
\mantitle{MPID{\tt \char`\_}tBsend}{tex}{10/7/2002}
\manname{MPID{\tt \char`\_}tBsend}
--- Attempt a send and return if it would block 
\subhead{Synopsis}
\startvb\begin{verbatim}
int MPID_tBsend( const void *buf, int count, MPI_Datatype datatype,
                int dest, int tag, MPID_Comm *comm, int context_offset )

\end{verbatim}
\endvb

\subhead{Notes}
This has the semantics of {\tt MPI{\tt \char`\_}Bsend}, except that it returns the internal
error code {\tt MPID{\tt \char`\_}WOULD{\tt \char`\_}BLOCK} if the message can't be sent immediately
(t is for "try").
\par
The reason that this interface is chosen over a query to check whether
a message {\em can} be sent is that the query approach is not
thread-safe.  Since the decision on whether a message can be sent
without blocking depends (among other things) on the state of flow
control managed by the device, this approach also gives the device
the greatest freedom in implementing flow control.  In particular,
if another MPI process can change the flow control parameters, then
even in a single-threaded implementation, it would not be safe to
return, for example, a message size that could be sent with {\tt MPI{\tt \char`\_}Bsend}.
\par
This routine allows an MPI implementation to optimize {\tt MPI{\tt \char`\_}Bsend
}for the case when the message can be delivered without blocking the
calling process.  An ADI implementation is free to have this routine
always return {\tt MPID{\tt \char`\_}WOULD{\tt \char`\_}BLOCK}, but is encouraged not to.
\par
To allow the MPI implementation to avoid trying this routine when it
is not implemented by the ADI, the C preprocessor constant {\tt MPID{\tt \char`\_}HAS{\tt \char`\_}TBSEND
}should be defined if this routine has a nontrivial implementation.
\par
\subhead{Module}
Communication
\endmanpage