int tid = pvm_mytid( void ) call pvmfmytid( tid )
The routine pvm_mytid() returns the TID of this process and can be called multiple times. It enrolls this process into PVM if this is the first PVM call. Any PVM system call (not just pvm_mytid) will enroll a task in PVM if the task is not enrolled before the call, but it is common practice to call pvm_mytid first to perform the enrolling.
int info = pvm_exit( void ) call pvmfexit( info )
The routine pvm_exit() tells the local pvmd that this process is leaving PVM. This routine does not kill the process, which can continue to perform tasks just like any other UNIX process. Users typically call pvm_exit right before exiting their C programs and right before STOP in their Fortran programs.
int numt = pvm_spawn(char *task, char **argv, int flag,
char *where, int ntask, int *tids )
call pvmfspawn( task, flag, where, ntask, tids, numt )
The routine pvm_spawn()
starts up ntask copies of an executable file task
on the virtual machine.
argv is a pointer to an array of arguments to task
with the end of the array specified by NULL.
If task takes no arguments, then argv is NULL.
The flag argument is used to specify options, and is a sum of:
Value Option Meaning
--------------------------------------------------------------------------
0 PvmTaskDefault PVM chooses where to spawn processes.
1 PvmTaskHost where argument is a particular host to spawn on.
2 PvmTaskArch where argument is a PVM_ARCH to spawn on.
4 PvmTaskDebug starts tasks under a debugger.
8 PvmTaskTrace trace data is generated.
16 PvmMppFront starts tasks on MPP front-end.
32 PvmHostCompl complements host set in where.
--------------------------------------------------------------------------
These names are predefined in pvm3/include/pvm3.h. In Fortran all the names are predefined in parameter statements which can be found in the include file pvm3/include/fpvm3.h.
PvmTaskTrace is a new feature in PVM 3.3. It causes spawned tasks to generate trace events . PvmTasktrace is used by XPVM (see Chapter 8). Otherwise, the user must specify where the trace events are sent in pvm_setopt().
On return, numt is set to the number of tasks successfully spawned or an error code if no tasks could be started. If tasks were started, then pvm_spawn() returns a vector of the spawned tasks' tids; and if some tasks could not be started, the corresponding error codes are placed in the last ntask - numt positions of the vector.
The pvm_spawn() call can also start tasks on multiprocessors. In the case of the Intel iPSC/860 the following restrictions apply. Each spawn call gets a subcube of size ntask and loads the program task on all of these nodes. The iPSC/860 OS has an allocation limit of 10 subcubes across all users, so it is better to start a block of tasks on an iPSC/860 with a single pvm_spawn() call rather than several calls. Two different blocks of tasks spawned separately on the iPSC/860 can still communicate with each other as well as any other PVM tasks even though they are in separate subcubes. The iPSC/860 OS has a restriction that messages going from the nodes to the outside world be less than 256 Kbytes.
int info = pvm_kill( int tid ) call pvmfkill( tid, info )
The routine pvm_kill() kills some other PVM task identified by TID. This routine is not designed to kill the calling task, which should be accomplished by calling pvm_exit() followed by exit().
int info = pvm_catchout( FILE *ff ) call pvmfcatchout( onoff )
The default is to have PVM write the stderr and stdout of spawned tasks to the log file /tmp/pvml.<uid>. The routine pvm_catchout causes the calling task to catch output from tasks subsequently spawned. Characters printed on stdout or stderr in children tasks are collected by the pvmds and sent in control messages to the parent task, which tags each line and appends it to the specified file (in C) or standard output (in Fortran). Each of the prints is prepended with information about which task generated the print, and the end of the print is marked to help separate outputs coming from several tasks at once.
If pvm_exit is called by the parent while output collection is in effect, it will block until all tasks sending it output have exited, in order to print all their output. To avoid this, one can turn off the output collection by calling pvm_catchout(0) before calling pvm_exit.
New capabilities in PVM 3.3 include the ability to register special PVM tasks to handle the jobs of adding new hosts, mapping tasks to hosts, and starting new tasks. This creates an interface for advanced batch schedulers (examples include Condor [7], DQS [6], and LSF [4]) to plug into PVM and run PVM jobs in batch mode. These register routines also create an interface for debugger writers to develop sophisticated debuggers for PVM.
The routine names are pvm_reg_rm(), pvm_reg_hoster(), and pvm_reg_tasker(). These are advanced functions not meant for the average PVM user and thus are not presented in detail here. Specifics can be found in Appendix B.