File: MPE_Seq_begin.4

package info (click to toggle)
mpich 1.1.0-3
  • links: PTS
  • area: main
  • in suites: hamm
  • size: 22,116 kB
  • ctags: 27,349
  • sloc: ansic: 193,435; sh: 11,172; fortran: 6,545; makefile: 5,801; cpp: 5,020; tcl: 3,548; asm: 3,536; csh: 1,079; java: 614; perl: 183; awk: 168; sed: 70; f90: 62
file content (42 lines) | stat: -rw-r--r-- 1,063 bytes parent folder | download | duplicates (2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
.TH MPE_Seq_begin 4 "4/16/1997" " " "MPE"
.SH NAME
MPE_Seq_begin \-  Begins a sequential section of code.   
.SH SYNOPSIS
.nf
#include <stdio.h>
#include "mpe.h"
void MPE_Seq_begin( comm, ng )
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
mpe_seq.c