foreach/doParallel/PSOCK benchmark V2

suppressMessages(library(doParallel))
library(itertools)

args <- commandArgs(trailingOnly=TRUE)
n <- if (length(args) > 0) as.integer(args[1]) else 1000000
m <- if (length(args) > 1) as.integer(args[2]) else ceiling(n / 100)
set.seed(107)
td <- data.frame(val=rnorm(n), id=sample(m, n, replace=TRUE))

cl <- makePSOCKcluster(4)
registerDoParallel(cl)
workers <- getDoParWorkers()

vadd <- function(a, ...) {
  for (v in list(...))
    a <- a + v
  a
}

start <- proc.time()[3]
res <- foreach(ids=isplitVector(unique(td$id), chunks=workers),
               .combine='vadd',
               .multicombine=TRUE,
               .inorder=FALSE) %dopar% {
  r <- rep(0, NROW(td))
  for (i in ids)
    r[td$id == i] <- mean(td$val[td$id != i])
  r
}
elapsed <- proc.time()[3] - start

cat(sprintf("foreach/doParallel/PSOCK with %d workers:\n", workers))
cat(sprintf("Rows: %d, Unique IDs: %d\n", length(td$id), length(unique(td$id))))
cat(sprintf("Elapsed time: %f\n", elapsed))

library(digest)
cat(sprintf("MD5 hash: %s\n", digest(res)))

stopCluster(cl)