foreach：％do％需要比1-worker％dopar％更多的处理时间

Question

我准备了一个小基准来比较多个工人用于处理时的速度增益。

library(foreach)
library(doParallel)

f <- function(n){ #just a function that needs some time
  sample(n, 1000000, replace = T)
  sqrt(n)
}

before = Sys.time()
y = foreach(i=1:10000, .combine = c) %do% {
  f(i)
}
after = Sys.time()
after - before # Time difference of 1.764439 mins (105.86634s)


benchmark <- function(n){
  f <- function(n){
    sample(n, 1000000, replace = T)
    sqrt(n)
  }

  cl <- makeCluster(n)
  registerDoParallel(cl)

  before = Sys.time()
  y = foreach(i=1:10000, .combine = c) %dopar% {
    f(i)
  }
  after = Sys.time()

  stopCluster(cl)

  after - before
}

benchmark(1) #Time difference of 1.49561 mins (89.7366s)
benchmark(2) #Time difference of 44.76028 secs
benchmark(3) #Time difference of 32.92803 secs
benchmark(4) #Time difference of 28.984 secs
benchmark(5) #Time difference of 27.46466 secs
benchmark(6) #Time difference of 26.48457 secs
benchmark(7) #Time difference of 26.48769 secs
benchmark(8) #Time difference of 29.07004 secs

我预计%dopar%处理时间（仅使用1名工作人员）与%do%处理时间相似。但是%do%需要大约16秒。这是正常的吗？