516900 - Repeated call to document.getElementById is slower in firefox than Safari.

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Biju]

Attachment: getelement.html

Repeated call to document.getElementById is slower in firefox than Safari.

function $(id, refresh) {
  if (typeof id == "string") {
      id = !refresh && $[id] || ($[id] = document.getElementById(id));
  }
  return id;
}

function dd($) {
  var _$ = document;
  _$.$ = _$.getElementById;
  var i, d1, d2, d3, d4, d5, id = "alink", k = 1000000; //ie. 1,000,000

  function __(id) {return document.getElementById(id);}

  d1 = new Date;
  for (i = 0; i < k; i++) _$.$("alink");
  d2 = new Date;
  for (i = 0; i < k; i++) __("alink");
  d3 = new Date;
  for (i = 0; i < k; i++) $("alink");
  d4 = new Date;
  for (i = 0; i < k; i++)$[id]||$(id);
  d5 = new Date;
  for (i = 0; i < k; i++)$["alink"]||$("alink");
  d6 = new Date;
  for (i = 0; i < k; i++)$.alink ||$("alink");
  d7 = new Date;

  return [d2 - d1, d3 - d2, d4 - d3, d5 - d4, d6 - d5, d7 - d6];
}

dd($);


I ran above code on various PC and got following result.
And here is the summary of what I observed.
1. Safari is doing some kind of caching like 
   I do with my $() function and beat Firefox on speed

2. Namoroka/Trunk build executing shorter code ie _$.$() 
   slower than calling my javascript function __()
   which in turn call getElementById

3. On slower PC in some case Shiretoko is faster than Minefield


====== PC 1 =========
Loop Count : 1000000   //ie. 1,000,000
Safari 4
Result : 165,293,223,101,98,15

Namoroka/3.6a2pre
Result : 1792,1356,312,109,70,70

====== PC 2 =========
Loop Count : 1000000   //ie. 1,000,000

Gecko/20090915 Minefield/3.7a1pre
Result : 1354,559,335,71,10,9

Gecko/20090915 Namoroka/3.6a2pre
Result : 1312,583,361,67,13,9

Gecko/20090915 Shiretoko/3.5.4pre
Result : 1931,3866,346,108,10,11

====== PC 3 =========
Loop Count : 300000    //ie. 300,000 (slow PC so reduced loop count)
Gecko/20090914 Shiretoko/3.5.4pre
Result : 2401,4474,310,53,7,8

Gecko/20090915 Minefield/3.7a1pre
Result : 1690,1193,350,148,109,110

Comment 1

[Boris Zbarsky [:bzbarsky]]

This can also be seen in the dromaeo "getElementById (not in document)" test.

Should remeasure once we have the new DOM bindings.

Comment 2

[Guilherme Lima]

Great news: Firefox is generally faster than Chrome! It's slower only on the first one.

Nightly 26: 241,99,15,8,9,4
Chrome 29: 159,360,186,12,11,12

Comment 3

[(mostly gone) XtC4UaLL [:xtc4uall]]

FWIW, attachment 400951 [details] somewhat regressed recently for the second result value (d3 - d2):
Firefox 31 (Aurora,  CSet: b982a8afab52): 209,  95,  12, 7,  6, 6
Firefox 32 (Nightly, CSet: 00c2ba04f8d4): 231, 650,  12, 9,  7, 7
Chrome 37                               : 134, 528, 208, 9, 10, 9

I tracked this down to

Last good revision: cbe4f69c2e9c (2014-05-27)
First bad revision: e017c15325ae (2014-05-28)
Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=cbe4f69c2e9c&tochange=e017c15325ae

and further

https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=a3f2754fcfb8&tochange=79520ed7bc4e

=> by Bug 789261 (defaulting dom.window_experimental_bindings;true)

Comment 4

[Olli Pettay [:smaug][[e-mail]]]

This bug is ancient. Please file new bugs for recent regressions.

Comment 5

[(mostly gone) XtC4UaLL [:xtc4uall]]

(In reply to Olli Pettay [:smaug] from comment #4)
> This bug is ancient. Please file new bugs for recent regressions.

Done per Bug 1022736.

Comment 6

[Guilherme Lima]

Nightly 48: 137,150,15,6,4,3
Chrome 49: 169,965,268,4,4,5

Would be nice to compare to Safari.

Comment 7

[Biju]

Safari
Loop Count : 1000,000
Mozilla/5.0 (Macintosh; Intel Mac OS X 10_10_5) AppleWebKit/601.5.17 (KHTML, like Gecko) Version/9.1 Safari/601.5.17
27,119,154,37,8,9
26,120,150,37,7,8
26,117,150,36,8,8
26,119,149,39,9,7
27,116,151,37,7,8
26,119,149,36,8,8
26,116,153,37,7,8
26,119,151,37,7,8
26,116,149,37,8,8
26,115,154,38,7,8


Firefox
Loop Count : 1000,000
Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:45.0) Gecko/20100101 Firefox/45.0

49,71,7,4,2,1
49,71,4,4,1,2
49,72,6,3,2,2
49,70,5,3,2,1
49,70,5,3,2,2
49,71,5,3,2,1
50,71,5,3,2,1
52,69,5,3,2,2
49,70,7,3,2,2
49,70,7,3,2,1

Comment 8

[Mayank Bansal]

Chrome: https://share.firefox.dev/3XbkyJG (222ms)
Nightly: https://share.firefox.dev/3SV2u3W (100ms)

Calling this fixed.