yy_api_task_pull: Functions for Task Pull Parallelism

In pbdMPI: R Interface to MPI for HPC Clusters (Programming with Big Data Project)

Functions for Task Pull Parallelism

Description

These functions are designed for SPMD but assume that rank 0 is a manager and the rest are workers.

Usage

task.pull(jids, FUN, ..., rank.manager = .pbd_env$SPMD.CT$rank.root,
          comm = .pbd_env$SPMD.CT$comm, bcast = .pbd_env$SPMD.TP$bcast,
          barrier = .pbd_env$SPMD.TP$barrier,
          try = .pbd_env$SPMD.TP$try,
          try.silent = .pbd_env$SPMD.TP$try.silent)

task.pull.workers(FUN = function(jid, ...){ return(jid) }, ...,
                  rank.manager = .pbd_env$SPMD.CT$rank.root,
                  comm = .pbd_env$SPMD.CT$comm,
                  try = .pbd_env$SPMD.TP$try,
                  try.silent = .pbd_env$SPMD.TP$try.silent)
task.pull.manager(jids, rank.manager = .pbd_env$SPMD.CT$rank.root,
                 comm = .pbd_env$SPMD.CT$comm)

Arguments

jids	all job ids (a vector of positive integers).
FUN	a function to be evaluated by workers.
...	extra parameters for FUN.
rank.manager	rank of the manager from where jid is sent.
comm	a communicator number.
bcast	if bcast to all ranks.
barrier	if barrier for all ranks.
try	wheter to use try() to avoid crashes. CAUTION: try = FALSE is not safe and can crash all MPI/R jobs.
try.silent	turn off error messages from try().

Details

All of these functions are designed to emulate a manager/workers paradigm in an SPMD environment. If your chunk workloads are known and similar, consider a direct SPMD solution.

FUN is a user defined function which has jid as its first argument and other variables are given in ....

The manager will be queried by workers whenever a worker finishes a job to see if more jobs are available.

Value

A list with length comm.size() - 1 will be returned to the manager and NULL to the workers. Each element of the list contains the returns ret of their FUN results.

Author(s)

Wei-Chen Chen [e-mail], George Ostrouchov, Drew Schmidt, Pragneshkumar Patel, and Hao Yu.

References

Programming with Big Data in R Website: https://pbdr.org/

See Also

get.jid().

Examples

## Not run: 
### Under command mode, run the demo with 2 processors by
### (Use Rscript.exe for windows system)
# mpiexec -np 2 Rscript -e "demo(task_pull,'pbdMPI',ask=F,echo=F)"
### Or
# execmpi("demo(task_pull,'pbdMPI',ask=F,echo=F)", nranks = 2L)

## End(Not run)

pbdMPI documentation built on Sept. 18, 2024, 9:07 a.m.