var.par.nc: Change Parallel Access Mode

In RNetCDF: Interface to 'NetCDF' Datasets

Change Parallel Access Mode

Description

Change the parallel access mode of a NetCDF variable from independent to collective and vice versa.

Usage

var.par.nc(ncfile, variable, access="NC_COLLECTIVE")

Arguments

ncfile	Object of class NetCDF which points to the NetCDF dataset (as returned from open.nc).
variable	Numeric ID or name of the variable for which to change the parallel access mode. Use "NC_GLOBAL" to change the parallel access mode for all variables in the dataset.
access	Parallel access mode as one of the following strings: "NC_COLLECTIVE" or "NC_INDEPENDENT".

Details

Parallel file access is either collective (all processors must participate) or independent (any processor may access the data without waiting for others). Data reads and writes (i.e. calls to var.put.nc and var.get.nc) are independent by default. Use this function to change the parallel access mode for a variable from independent to collective mode or vice versa.

All netCDF metadata writing operations are collective - all creation of groups, types, variables, dimensions, or attributes.

Note that when the file format is "classic" or "offset64", the change always applies to all variables in the file, even if a single variable is specified in argument variable.

Author(s)

Pavel Michna, Milton Woods

References

https://www.unidata.ucar.edu/software/netcdf/

Examples

## Not run: 
# This example assumes that the NetCDF library was built with MPI support,
# and that both RNetCDF and pbdMPI are installed in R.
# If the example code is stored in a file myexample.R,
# run R under MPI using a command similar to:
# SHELL> mpiexec -np 2 Rscript --vanilla myexample.R

library(pbdMPI, quiet = TRUE)
library(RNetCDF, quiet = TRUE)

# Get MPI parameters
init()
rank <- comm.rank()
size <- comm.size()

# Define dimensions and data
nr <- 5
nc_local <- 4
nc <- nc_local * size
data_local <- matrix(rank, nrow=nr, ncol=nc_local)

# Open file for parallel access and define metadata
filename <- "myexample.nc"
info.create()
ncid <- create.nc(filename, format="netcdf4", mpi_comm=comm.c2f(), mpi_info=info.c2f())
rdim <- dim.def.nc(ncid, "rows", nr)
cdim <- dim.def.nc(ncid, "cols", nc)
varid <- var.def.nc(ncid, "data", "NC_INT", c(rdim, cdim))

# Use collective I/O
var.par.nc(ncid, "data", "NC_COLLECTIVE")

# Write data
var.put.nc(ncid, varid, data_local, start=c(1,rank*nc_local+1), count=c(nr,nc_local))

# Finish up
close.nc(ncid)
info.free()

finalize()

## End(Not run)

RNetCDF documentation built on May 29, 2024, 2:41 a.m.