|
|
|
|
@ -7,50 +7,44 @@
|
|
|
|
|
%
|
|
|
|
|
\subroutine{psb\_cdall}{Allocates a communication descriptor}
|
|
|
|
|
|
|
|
|
|
\syntax{call psb\_cdall}{m, n, parts, icontxt, desc\_a, info}
|
|
|
|
|
\syntax*{call psb\_cdall}{m, v, icontxt, desc\_a, info, flag}
|
|
|
|
|
\syntax{call psb\_cdall}{icontxt, desc\_a, info,mg=mg,parts=parts}
|
|
|
|
|
\syntax{call psb\_cdall}{icontxt, desc\_a, info,vg=vg,flag=flag}
|
|
|
|
|
\syntax{call psb\_cdall}{icontxt, desc\_a, info,vl=vl}
|
|
|
|
|
|
|
|
|
|
This subroutine initializes the communication descriptor associated
|
|
|
|
|
with an index space. It takes two forms depending on whether the user
|
|
|
|
|
specifies the domain partitioning through a subroutine or through a vector
|
|
|
|
|
with an index space. Exactly one of the optional arguments
|
|
|
|
|
\verb|parts|, \verb|vg|, \verb|vl| must be specified, thereby choosing
|
|
|
|
|
the specific initialization strategy:
|
|
|
|
|
\begin{description}
|
|
|
|
|
\item[\bf First Form: On Entry ]
|
|
|
|
|
\item[m] the number of rows of the problem.\\
|
|
|
|
|
Scope:{\bf global}.\\
|
|
|
|
|
Type:{\bf required}.\\
|
|
|
|
|
Specified as: an integer value.
|
|
|
|
|
\item[n] the number of columns of the problem.\\
|
|
|
|
|
Scope:{\bf global}.\\
|
|
|
|
|
Type:{\bf required}.\\
|
|
|
|
|
Specified as: an integer value. Currently constrained to be $m=n$.
|
|
|
|
|
\item[parts] the subroutine that defines the partitioning scheme.\\
|
|
|
|
|
Scope:{\bf global}.\\
|
|
|
|
|
Type:{\bf required}.\\
|
|
|
|
|
Specified as: a subroutine.
|
|
|
|
|
\item[\bf On Entry ]
|
|
|
|
|
\item[icontxt] the communication context.\\
|
|
|
|
|
Scope:{\bf global}.\\
|
|
|
|
|
Type:{\bf required}.\\
|
|
|
|
|
Specified as: an integer value.
|
|
|
|
|
\end{description}
|
|
|
|
|
\begin{description}
|
|
|
|
|
\item[\bf Second Form: On Entry ]
|
|
|
|
|
\item[m] the size of the index space.\\
|
|
|
|
|
\item[vg] Data allocation: each index $i\in \{1\dots mg\}$ is allocated
|
|
|
|
|
to process $vg(i)$.
|
|
|
|
|
Scope:{\bf global}.\\
|
|
|
|
|
Type:{\bf required}.\\
|
|
|
|
|
Specified as: an integer value $m>0$.
|
|
|
|
|
\item[v] Data allocation: each index $i\in \{1\dots m\}$ is allocated
|
|
|
|
|
to process $v(i)$.
|
|
|
|
|
Type:{\bf optional}.\\
|
|
|
|
|
Specified as: an integer array.
|
|
|
|
|
\item[flag] Specifies whether entries in $vg$ are zero- or one-based.
|
|
|
|
|
Scope:{\bf global}.\\
|
|
|
|
|
Type:{\bf required}.\\
|
|
|
|
|
Specified as: an integer array of size $m$.
|
|
|
|
|
\item[icontxt] the communication context.\\
|
|
|
|
|
Type:{\bf optional}.\\
|
|
|
|
|
Specified as: an integer value $0,1$, default $0$.
|
|
|
|
|
|
|
|
|
|
\item[mg] the (global) number of rows of the problem.\\
|
|
|
|
|
Scope:{\bf global}.\\
|
|
|
|
|
Type:{\bf required}.\\
|
|
|
|
|
Specified as: an integer value.
|
|
|
|
|
\item[flag] Specifies whether entries in $v$ are zero- or one-based.
|
|
|
|
|
Type:{\bf optional}.\\
|
|
|
|
|
Specified as: an integer value. It is required if \verb|parts| is
|
|
|
|
|
specified.
|
|
|
|
|
\item[parts] the subroutine that defines the partitioning scheme.\\
|
|
|
|
|
Scope:{\bf global}.\\
|
|
|
|
|
Type:{\bf required}.\\
|
|
|
|
|
Specified as: a subroutine.
|
|
|
|
|
\item[vl] Data allocation: the set of global indices belonging to the
|
|
|
|
|
calling process.
|
|
|
|
|
Scope:{\bf local}.\\
|
|
|
|
|
Type:{\bf optional}.\\
|
|
|
|
|
Specified as: an integer value $0,1$, default $0$.
|
|
|
|
|
Specified as: an integer array.
|
|
|
|
|
\end{description}
|
|
|
|
|
|
|
|
|
|
\begin{description}
|
|
|
|
|
@ -67,6 +61,28 @@ Specified as: an integer variable.\\
|
|
|
|
|
|
|
|
|
|
\section*{Notes}
|
|
|
|
|
\begin{enumerate}
|
|
|
|
|
\item Exactly one of the optional arguments \verb|parts|, \verb|vg|,
|
|
|
|
|
\verb|vl| must be specified, thereby choosing the initialization
|
|
|
|
|
strategy as follows:
|
|
|
|
|
\begin{description}
|
|
|
|
|
\item[parts] In this case we have a subroutine that takes as input a
|
|
|
|
|
index and the total number of indices in the space, and produces in
|
|
|
|
|
output a vector containing the set of processes (usually with just
|
|
|
|
|
one entry) to which the index should be assigned. If this argument
|
|
|
|
|
is specified, then it is mandatory to also specify the argument
|
|
|
|
|
\verb|mg|.
|
|
|
|
|
\item[vg] In this case the association between an index and a process
|
|
|
|
|
is specified via an integer vector; the size of the index space is
|
|
|
|
|
equal to the size of \verb|vg|, and each index $i$ is assigned to
|
|
|
|
|
the process $vg(i)$. The vector \verb|vg| must be identical on all
|
|
|
|
|
calling processes; its entries may have the ranges $(0\dots np-1)$
|
|
|
|
|
or $(1\dots np)$ according to the value of \verb|flag|.
|
|
|
|
|
\item[vl] In this case we are specifying the list of indices assigned
|
|
|
|
|
to the current process; thus, the global problem size $mg$ is given by
|
|
|
|
|
the sum of the sizes of the individual vectors \verb|vl| specified
|
|
|
|
|
on the calling processes. The subroutine will check that each entry
|
|
|
|
|
in the global index space $(1\dots mg)$ is specified exactly once.
|
|
|
|
|
\end{description}
|
|
|
|
|
\item On exit from this routine the descriptor is in the build state
|
|
|
|
|
\end{enumerate}
|
|
|
|
|
|
|
|
|
|
|
Salvatore Filippone
18 years ago