Last chance! 50% off unlimited learning
Sale ends in
This method lets a rank send an object to the other rank and
receive an object from another rank in the same communicator.
The default return is x
.
sendrecv(x, x.buffer = NULL,
rank.dest = (comm.rank(.pbd_env$SPMD.CT$comm) + 1) %%
comm.size(.pbd_env$SPMD.CT$comm),
send.tag = .pbd_env$SPMD.CT$tag,
rank.source = (comm.rank(.pbd_env$SPMD.CT$comm) - 1) %%
comm.size(.pbd_env$SPMD.CT$comm),
recv.tag = .pbd_env$SPMD.CT$tag,
comm = .pbd_env$SPMD.CT$comm, status = .pbd_env$SPMD.CT$status)
an object to be sent from a rank.
a buffer to store x
sent from the other rank.
a rank of destination where x
send to.
a send tag number.
a source rank where x
sent from.
a receive tag number.
a communicator number.
a status number.
A x
is returned by default.
For calling spmd.sendrecv.*()
:
signature(x = "ANY", x.buffer = "ANY")
signature(x = "integer", x.buffer = "integer")
signature(x = "numeric", x.buffer = "numeric")
signature(x = "raw", x.buffer = "raw")
A corresponding sendrecv()
should be evoked at the corresponding ranks
rank.dest
and rank.source
.
rank.dest
and rank.source
can be as.integer(NULL)
to
create a silent sendrecv operation which is more efficient than setting
rank.dest
and rank.source
to be equal.
Programming with Big Data in R Website: http://r-pbd.org/
# NOT RUN {
### Save code in a file "demo.r" and run with 2 processors by
### SHELL> mpiexec -np 2 Rscript demo.r
spmd.code <- "
### Initial.
suppressMessages(library(pbdMPI, quietly = TRUE))
init()
.comm.size <- comm.size()
.comm.rank <- comm.rank()
### Examples.
N <- 5
x <- (1:N) + N * .comm.size
y <- sendrecv(matrix(x, nrow = 1))
comm.print(y, rank.print = 1)
### Finish.
finalize()
"
# execmpi(spmd.code, nranks = 2L)
# }
Run the code above in your browser using DataLab