vv_get_jid: Divide Job ID by Ranks

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

Divide Job ID by Ranks

Description

This function obtains job id which can be used to divide jobs.

Usage

get.jid(n, method = .pbd_env$SPMD.CT$divide.method[1], all = FALSE,
        comm = .pbd_env$SPMD.CT$comm, reduced = FALSE)

Arguments

n	total number of jobs.
method	a way to divide jobs.
all	indicate if return all id for each processor.
comm	a communicator number.
reduced	indicate if return should be a reduced representation.

Details

n is total number of jobs needed to be divided into all processors (comm.size(comm), i.e. 1:n will be split according to the rank of processor (comm.rank(comm)) and method. Job id will be returned. Currently, three possible methods are provided.

"block" will use return id's which are nearly equal size blocks. For example, 7 jobs in 4 processors will have jid=1 for rank 0, jid=2,3 for rank 1, jid=4,5 for rank 2, and jid=6,7 for rank 3.

"block0" will use return id's which are nearly equal size blocks, in the opposite direction of "block". For example, 7 jobs in 4 processors will have jid=1,2 for rank 0, jid=3,4 for rank 1, jid=5,6 for rank 2, and jid=7 for rank 3.

"cycle" will use return id's which are nearly equal size in cycle. For example, 7 jobs in 4 processors will have jid=1,5 for rank 0, jid=2,6 for rank 1, jid=3,7 for rank 2, and jid=4 for rank 3.

Value

get.jid() returns a vector containing job id for each individual processor if all = FALSE. While it returns a list containing all job id for all processor if all = TRUE. The list has length equal to comm.size.

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

task.pull() and comm.chunk().

Examples

### Save code in a file "demo.r" and run with 2 processors by
### SHELL> mpiexec -np 2 Rscript demo.r

spmd.code <- "
### Initialize
suppressMessages(library(pbdMPI, quietly = TRUE))

### Examples.
comm.cat(\">>> block\n\", quiet = TRUE)
jid <- get.jid(7, method = \"block\")
comm.print(jid, all.rank = TRUE)

comm.cat(\">>> cycle\n\", quiet = TRUE)
jid <- get.jid(7, method = \"cycle\")
comm.print(jid, all.rank = TRUE)

comm.cat(\">>> block (all)\n\", quiet = TRUE)
alljid <- get.jid(7, method = \"block\", all = TRUE)
comm.print(alljid)

comm.cat(\">>> cycle (all)\n\", quiet = TRUE)
alljid <- get.jid(7, method = \"cycle\", all = TRUE)
comm.print(alljid)

### Finish.
finalize()
"
pbdMPI::execmpi(spmd.code, nranks = 2L)

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