試圖找出執行時間差異的原因

我撰寫了下面的代碼來計算 Collat??z 序列。基本演算法步驟與 CRAN 提供的庫“ collat??z ”中使用的代碼幾乎相同。（源 tarball 在這里） 但是，我的 hack 代碼運行速度比庫的hailstone_sequence函式慢 10 到 50 倍。
我不知道為什么，除非通過存盤重用的常量lockBinding會有所不同。任何人都可以看看這兩組代碼并提出導致差異的原因嗎？我查看了運行的結果，Rprof并沒有什么明顯的地方。

{{我的代碼如下}}

collatz<-function(x, div=2, mul=3, add= 1, maxKnown=1, maxiter = 1000) {
y <- as.bigz(x[1])  #silent dumping
maxKnown <- as.bigz(floor(maxKnown))
cyclic = FALSE

for (jj in 2:maxiter) {
    
    y[[jj]] <- collatz_fun(y[[jj-1]],div,mul,add)
   # bigz/bigz returns a bigq even if that bigq has denominator 1
    # so we do a divq, "%/%", instead of div, to just get the bigz.
    if (y[[jj]] <= maxKnown   ) break 
    if(any(y[[jj]] == y[[1:(jj-1)]]))   {
        cyclic = TRUE
        break
    }
}
    if (jj >= maxiter)  warning('not converged (yet)')
    return(invisible(list(y = y, div = div, mul=mul, add= add, cyclic = cyclic) ) )
}

collatz_fun <- function(n, d, m, a){
    if (n%%d ==0 ) ( n %/% d ) else (m*n)  a
}

uj5u.com熱心網友回復：

原來我是在自欺欺人！對于一些特定的起始值，例如 737 ， hailstone_sequence速度更快。但是，總的來說，我的“collat??z”代碼更快。這是一個表格，顯示了大小增加的亂數的 tictoc 時間（以毫秒為單位）。

   Collatz, ms hailstone, ms start                            
 [1,] "2"         "2"           "89"                             
 [2,] "7"         "8"           "505"                            
 [3,] "10"        "13"          "4760"                           
 [4,] "5"         "8"           "46125"                          
 [5,] "57"        "89"          "706047"                         
 [6,] "52"        "77"          "7426598"                        
 [7,] "114"       "179"         "71693193"                       
 [8,] "76"        "128"         "195808257"                      
 [9,] "75"        "111"         "4082229980"                     
[10,] "89"        "133"         "29816905510"                    
[11,] "56"        "86"          "850451975896"                   
[12,] "240"       "390"         "5056177278457"                  
[13,] "179"       "288"         "23497279632257"                 
[14,] "133"       "217"         "667081939494479"                
[15,] "183"       "286"         "8724848384524922"               
[16,] "126"       "199"         "20376642404589490"              
[17,] "175"       "271"         "195397921609459067"             
[18,] "96"        "146"         "9068592584569983840"            
[19,] "203"       "317"         "98435723276783667011"           
[20,] "189"       "288"         "906945269554050299597"          
[21,] "586"       "934"         "1942663519120549652842"         
[22,] "288"       "414"         "32708032219025335530871"        
[23,] "331"       "459"         "815908776331672807666938"       
[24,] "379"       "581"         "6894029370205376937364345"      
[25,] "811"       "1182"        "49809117889125237920557852"     
[26,] "415"       "605"         "583415680232489269692462895"    
[27,] "559"       "793"         "9220767227146002168685568453"   
[28,] "733"       "1082"        "60679473842196546847954821109"  
[29,] "894"       "1228"        "698657027947326650828429401951" 
[30,] "1320"      "1932"        "8336688931823299843178876023479"

標籤：r表现

上一篇：使用具有99.7%快取率的Cloudflare時，服務器位置有多重要？

下一篇：優化去除名字中敬語的功能