利用R的Rmpi來做平行化運算..

筆電Toshiba Satellite C640, Linux Mint 13

速度, 本來需要578.18 secs, 加上平行化後只需209.99 secs.


R 程式碼

rm(list=ls())
library(snow)
myfun <- function(ii) {
    set.seed(ii*100)
    ret = mean(rnorm(10000))
    ret
}
cluster = makeCluster(4,type="MPI")

### Normal
tt0 = Sys.time()
aa = sapply(1:100000,myfun)
tt1 = Sys.time()
bb = parSapply(cluster,1:100000,myfun)
tt2 = Sys.time()

stopCluster(cluster)

cat(sprintf(" - %.4f, %.4f\n",difftime(tt1,tt0,units="secs"),difftime(tt2,tt1,units="secs")))