R語言中for循環的並行處理方式

Posted on by

前言

本文用於記錄筆者在將R語言中的for語句並行化處理中的一些問題。

實驗

這裡使用foreach和doParallel包提供的函數實現for語句的並行處理。

for語句腳本

func <- function(x, y, z) {
  return(x^y/z)
}
# >>> main <<<
x <- 2
y <- 3
z <- 1:100000
start <- (proc.time())[3][[1]]
a <- 0
for (i_z in z) {
  a <- a + func(x, y, i_z)
}
end <- (proc.time())[3][[1]]
print(paste('Result = ', round(a, 2), ', time = ', (end-start), 's', sep=''))

輸出：

[1] “Result = 96.72, time = 0.177s”

並行化版本

library(foreach)
library(doParallel)
func <- function(x, y, z) {
  return(x^y/z)
}
# >>> main <<<
x <- 2
y <- 3
z <- 1:100000
start <- (proc.time())[3][[1]]
cl <- makeCluster(12)
registerDoParallel(cl)
a <- foreach(z=z, .combine='rbind') %dopar% func(x, y, z)
a <- sum(a)
stopCluster(cl)
end <- (proc.time())[3][[1]]
print(paste('Result = ', round(a, 2), ', time = ', (end-start), 's', sep=''))

輸出：

[1] “Result = 96.72, time = 37.988s”

總結

1、這裡發現並行化所用時間大於非並行化所用過的時間，是因為需要執行的操作（func函數）過於簡單，而foreach處理時會有額外的資源消耗。此時foreach額外消耗的資源遠大於需要執行的操作所需的資源，因此會導致並行化後反而使用的時間增加瞭。所以對於一些復雜的操作比較適合使用並行化的策略。

2、foreach函數的.packages參數可以為並行化函數傳遞額外需要的包。

3、foreach中的參數為需要在func中循環的變量，其他固定的變量則在func中傳入。參數可以是data.frame類型。

補充：R語言–for循環語句的使用

R語言for循壞語句的使用（多個for)

對於多個for循還語句，R語言的執行順序（以3個for為例）：從外向內單個執行，裡邊循還完整，再往外一層，直到全部完成。話不多說，上例子：

代碼：

library(data.table)
mm<-data.table()
m<-c(1,2,3,4,5)
n<-c('a','b','c','d','e')
o<-c(6,7,8,9,10)
for (i1 in m){
  for ( i2 in n){
     for (i3 in o){
       print(c(i1,i2,i3))
       aa<-data.table(i1,i2,i3)
       bb<-rbind(mm,aa)
       
     }
  }
}

執行結果：

[1] "1" "a" "6"
[1] "1" "a" "7"
[1] "1" "a" "8"
[1] "1" "a" "9"
[1] "1"  "a"  "10"
[1] "1" "b" "6"
[1] "1" "b" "7"
[1] "1" "b" "8"
[1] "1" "b" "9"
[1] "1"  "b"  "10"
[1] "1" "c" "6"
[1] "1" "c" "7"
[1] "1" "c" "8"
[1] "1" "c" "9"
[1] "1"  "c"  "10"
[1] "1" "d" "6"
[1] "1" "d" "7"
[1] "1" "d" "8"
[1] "1" "d" "9"
[1] "1"  "d"  "10"
[1] "1" "e" "6"
[1] "1" "e" "7"
[1] "1" "e" "8"
[1] "1" "e" "9"
[1] "1"  "e"  "10"
[1] "2" "a" "6"
[1] "2" "a" "7"
[1] "2" "a" "8"
[1] "2" "a" "9"
[1] "2"  "a"  "10"
[1] "2" "b" "6"
[1] "2" "b" "7"
[1] "2" "b" "8"
[1] "2" "b" "9"
[1] "2"  "b"  "10"
[1] "2" "c" "6"
[1] "2" "c" "7"
[1] "2" "c" "8"
[1] "2" "c" "9"
[1] "2"  "c"  "10"
[1] "2" "d" "6"
[1] "2" "d" "7"
[1] "2" "d" "8"
[1] "2" "d" "9"
[1] "2"  "d"  "10"
[1] "2" "e" "6"
[1] "2" "e" "7"
[1] "2" "e" "8"
[1] "2" "e" "9"
[1] "2"  "e"  "10"
[1] "3" "a" "6"
[1] "3" "a" "7"
[1] "3" "a" "8"
[1] "3" "a" "9"
[1] "3"  "a"  "10"
[1] "3" "b" "6"
[1] "3" "b" "7"
[1] "3" "b" "8"
[1] "3" "b" "9"
[1] "3"  "b"  "10"
[1] "3" "c" "6"
[1] "3" "c" "7"
[1] "3" "c" "8"
[1] "3" "c" "9"
[1] "3"  "c"  "10"
[1] "3" "d" "6"
[1] "3" "d" "7"
[1] "3" "d" "8"
[1] "3" "d" "9"
[1] "3"  "d"  "10"
[1] "3" "e" "6"
[1] "3" "e" "7"
[1] "3" "e" "8"
[1] "3" "e" "9"
[1] "3"  "e"  "10"
[1] "4" "a" "6"
[1] "4" "a" "7"
[1] "4" "a" "8"
[1] "4" "a" "9"
[1] "4"  "a"  "10"
[1] "4" "b" "6"
[1] "4" "b" "7"
[1] "4" "b" "8"
[1] "4" "b" "9"
[1] "4"  "b"  "10"
[1] "4" "c" "6"
[1] "4" "c" "7"
[1] "4" "c" "8"
[1] "4" "c" "9"
[1] "4"  "c"  "10"
[1] "4" "d" "6"
[1] "4" "d" "7"
[1] "4" "d" "8"
[1] "4" "d" "9"
[1] "4"  "d"  "10"
[1] "4" "e" "6"
[1] "4" "e" "7"
[1] "4" "e" "8"
[1] "4" "e" "9"
[1] "4"  "e"  "10"
[1] "5" "a" "6"
[1] "5" "a" "7"
[1] "5" "a" "8"
[1] "5" "a" "9"
[1] "5"  "a"  "10"
[1] "5" "b" "6"
[1] "5" "b" "7"
[1] "5" "b" "8"
[1] "5" "b" "9"
[1] "5"  "b"  "10"
[1] "5" "c" "6"
[1] "5" "c" "7"
[1] "5" "c" "8"
[1] "5" "c" "9"
[1] "5"  "c"  "10"
[1] "5" "d" "6"
[1] "5" "d" "7"
[1] "5" "d" "8"
[1] "5" "d" "9"
[1] "5"  "d"  "10"
[1] "5" "e" "6"
[1] "5" "e" "7"
[1] "5" "e" "8"
[1] "5" "e" "9"
[1] "5"  "e"  "10"

以上為個人經驗，希望能給大傢一個參考，也希望大傢多多支持WalkonNet。如有錯誤或未考慮完全的地方，望不吝賜教。

推薦閱讀：

    None Found