October 5, 2012 at 11:27

How much faster is Node.js than Windows Script Host? And what does this tell us?

    [quote]A reader who visits NodeJS.org , in the center of the page, sees four quotes from the leaders of site building, expressing the pleasure of Node. Quotes change every minute. With some luck (or patience when reloading the page), the reader will see praise from Claudio Caldato of Microsoft Open Technologies, Inc. - for your convenience, I quote this quote in the illustration on the right.

    What brought such praise to life? After all, it would seem that long before the Node engine even came to light, Microsoft had its own (built-in Windows, starting with Windows 98, as well as installed with Internet Explorer 5) tool for running scripts - Windows Script Host .

    The answer to this question is obvious: Node.js is much faster. (By the way, he also has an easier API. But speed is the main thing.)

    But how much faster is Node than WSH?

    It is not difficult to find out empirically. Take the script with which evgeniyup yesterday compared the performance of WSH with the performance of the virtual machine of its ObjectScript language. At the beginning of the script, add a dozen lines - a “crutch” for Node.js, which implements the WSH-function WScript.Echo in its absence.

    show (or hide again) the source code of the script
    // CScript to Node.js:
if (typeof WScript == "undefined") WScript = {};
if (typeof WScript.Echo == "undefined") WScript.Echo = function(){
  var i;
  var log = '';
  for (i=0; i < arguments.length; i++){
     log += arguments[i];
  }
  console.log(log);
}
var fannkuch = function(n)
{
  var p = [], q = [], s = [], sign = 1, maxflips = 0, sum = 0;
  var i;
  for(i=1; i<=n; i++) p[i] = q[i] = s[i] = i;
  for(;;){
    // Copy and flip.
    var q1 = p[1];				// Cache 1st element.
    if(q1 != 1){
      for(i=2; i<=n; i++) q[i] = p[i];		// Work on a copy.
      var flips = 1;
      for(;;){
        var qq = q[q1];
        if(qq == 1){				// ... until 1st element is 1.
          sum = sum + sign*flips;
          if(flips > maxflips){
            maxflips = flips;
          } // New maximum?
          break;
        }
        q[q1] = q1;
        if(q1 >= 4){
          var i = 2, j = q1 - 1
          for(;;){ var tmp = q[i]; q[i] = q[j]; q[j] = tmp; if(++i >= --j) break; }
        }
        q1 = qq; flips++;
      }
    }
    // Permute.
    if(sign == 1){
      var tmp = p[2]; p[2] = p[1]; p[1] = tmp; sign = -1;	// Rotate 1<-2.
    }else{
      var tmp = p[2]; p[2] = p[3]; p[3] = tmp; sign = 1;	// Rotate 1<-2 and 1<-2<-3.
      for(i = 3;; i++){
        // print "mark 4"
        var sx = s[i];
        if(sx != 1){ s[i] = sx-1; break; }
        if(i == n) return [sum, maxflips];	// Out of permutations.
        s[i] = i;
        // Rotate 1<-...<-i+1.
        var t = p[1]; for(var j = 1; j <= i; j++){ p[j] = p[j+1]; } p[i+1] = t;
      }
    }
  }
}
function getTimeSec(){
 	var d = new Date();
    return (d.getTime() + d.getMilliseconds() / 1000.0) / 1000.0;
}
var n = 10;
var start_time = getTimeSec();
var r = fannkuch(n);
var sum = r[0], flips = r[1];
WScript.Echo(
    sum,"\n",
    "Pfannkuchen(",n,") = ",flips,"\n",
    "time = ",(getTimeSec() - start_time),"\n"
)

    After that, it is enough to run this script twice (first in Node, then in WSH) - and we get the following result in the console (and in the screenshot ):

    [screenshot]

    The difference is two orders of magnitude! Computing that Node.js can do in a second, the Windows Script Host has been grinding for more than two minutes.

    I measured in Windows XP SP3 on a Pentium 4 processor (2.2 GHz). Each of you can independently repeat it at home according to the method described above.

    The conclusion is also quite simple and accessible to everyone: with the advent of the V8 engineJavaScript, we can say, survived a rebirth. Its computing capabilities have increased a hundredfold, although the rules of the language have not changed one iota (except for the appearance of some optional new products from the draft ECMAScript Harmony standard).

    In the real world, this only happens to people in superhero comics. And in programming, an awkward thing easily happens: the developer programmed in javascript, then V8 broads appeared, and all scripts began to work a hundred times faster.

    And this should give rise to further deep reflection on the prospects. For example: evgeniyup last measured , that PHP 5.3.3 is not much of a faster than WSH.Somewhere 1.2 times. Therefore, in this area there should be some potential for a sharp increase in speed by JIT compilation and other V8-like tricks, while until now all the so-called PHP accelerators have been engaged, at best, bytecode caching and attempts to optimize it , but not more. Imagine how the WWW world around us will change if PHP starts to execute a hundred times faster.

    Or at least an order of magnitude faster.
    Tags:

    Also popular now: