When working with raster datasets I often encounter performance issues caused by the large filesizes. I thus wrote up a little R function that invokes gdal_translate which would split the raster into parts which makes subsequent processing more CPU friendly. I didn't use built-in R functions simply because performance is much better when using gdal from the command line...

The screenshot to the left shows a raster in QGIS that was split into four parts with the below script.

## get filesnames (assuming the datasets were downloaded already.
## please see http://thebiobucket.blogspot.co.at/2013/06/use-r-to-bulk-download-digital.html
## on how to download high-resolution DEMs)
setwd("D:/GIS_DataBase/DEM")
files <- dir(pattern = ".hgt")
 
## function for single file processing mind to replace the PATH to gdalinfo.exe!
## s = division applied to each side of raster, i.e. s = 2 gives 4 tiles, 3 gives 9, etc.
split_raster <- function(file, s = 2) {
     
    filename <- gsub(".hgt", "", file)
    gdalinfo_str <- paste0("\"C:/OSGeo4W64/bin/gdalinfo.exe\" ", file)
       
    # pick size of each side
    x <- as.numeric(gsub("[^0-9]", "", unlist(strsplit(system(gdalinfo_str, intern = T)[3], ", "))))[1]
    y <- as.numeric(gsub("[^0-9]", "", unlist(strsplit(system(gdalinfo_str, intern = T)[3], ", "))))[2]
     
    # t is nr. of iterations per side
    t <- s - 1
    for (i in 0:t) {
        for (j in 0:t) {
            # [-srcwin xoff yoff xsize ysize] src_dataset dst_dataset
            srcwin_str <- paste("-srcwin ", i * x/s, j * y/s, x/s, y/s)
            gdal_str <- paste0("\"C:/OSGeo4W64/bin/gdal_translate.exe\" ", srcwin_str, " ", "\"", file, "\" ", "\"", filename, "_", i, "_", j, ".tif\"")
            system(gdal_str)
        }
    }
}
 

## process all files and save to same directory
mapply(split_raster, files, 2)