mpe_seq_begin.4

fortran并行计算包

.TH MPE_Seq_begin 4 "11/2/2007" " " "MPE"
.SH NAME
MPE_Seq_begin \-  Begins a sequential section of code.   
.SH SYNOPSIS
.nf
void MPE_Seq_begin( MPI_Comm comm, int ng )
.fi
.SH INPUT PARAMETERS
.PD 0
.TP
.B comm 
- Communicator to sequentialize.  
.PD 1
.PD 0
.TP
.B ng   
- Number in group.  This many processes are allowed to execute
at the same time.  Usually one.  
.PD 1

.SH NOTES
'MPE_Seq_begin' and 'MPE_Seq_end' provide a way to force a section of code
to
be executed by the processes in rank order.  Typically, this is done
with
.nf
MPE_Seq_begin( comm, 1 );
<code to be executed sequentially>
MPE_Seq_end( comm, 1 );
.fi

Often, the sequential code contains output statements (e.g., 'printf') to
be executed.  Note that you may need to flush the I/O buffers before
calling 'MPE_Seq_end'; also note that some systems do not propagate I/O in
any
order to the controling terminal (in other words, even if you flush the
output, you may not get the data in the order that you want).
.SH LOCATION
../src/misc/src/mpe_seq.c