panpvm - Implementation of PVM on top of Panda
SYNOPSIS
C:
#include <pvm3.h>
Fortran:
include ''fpmv3.h''
EXTENSIONS
C:
int panpvm_init(int *argc,
char*argv[]);
fortran:
subroutine panpvmfinit(ntids)
DESCRIPTION
This is a complete implementation of PVM, except for dynamic
process creation and group persistence.
All processes in the parallel run are started at the same
time, in the vein of Panda, and all execute the same code.
In short, the PANPVM parallelism is of the SPMD style.
Initialisation in C is done by calling
panpvm_init(&argc, argv),
where
argc
and
argv
are the options passed to
main().
panpvm_init()
returns the number of
parallel processes. In Fortran, function
panpvmfinit(ntids)
must be called
to initialise, and it returns the number of processors in
ntids.
After that, processes can investigate their identity by calling
pvm_mytid()/pvmfmytid(), which in
PANPVM returns the process's instance number, ranging from 0 to
ntids
- 1.
Typically, PANPVM programs are compiled with the
panpvmc(1)
script and run with the
prun(1)
utility.
SEE ALSO
panpvmc(1),
oc_panda(1),
prun(1)
FILES
- /usr/local/VU/panda_pvm
-
the root of the PANPVM directory tree.
- /usr/local/VU/panda_pvm/include
-
the PANPVM include directory, which contains
pvm3.h
and
fpvm3.h
- /usr/local/VU/panda_pvm/lib/$ARCH/libpanpvm.a
-
the PANPVM libraries for architecture $ARCH, see
panpvmc(1).
- /usr/local/VU/panda_pvm/fpvm
-
the root of the fortran PANPVM directory tree.
- /usr/local/VU/panda_pvm/fpvm/lib/$ARCH/libpanpvm.a
-
the fortran PANPVM libraries for architecture $ARCH, see
panpvmc(1).
Markup created by unroff 1.0, December 01, 1998.