On 2009/10/03 06:00:41, Christian Plesner Hansen wrote:
> I think waiting to flatten until everything is done sounds reasonable.
> Just
> note that when using StringAdd the string won't be flattened until the
> first
> time it is used, whereas with the string builder the string is flattened
> when
it
> is returned from concat. If your benchmarks doesn't use the string after
> creating it (say by reading the last character) the results are misleading
since
> the cost won't include the flattening.
> On the other hand if concat'ing 20 strings together and reading the last
> character is faster with a StringAdd based concat than a StringBuilder
> then I
> agree it's better to use StringAdd.
well, here's the benchmark I used:
var t1 = new Date().getTime();
lala = "hello";
for(var i = 0; i < 50000; i++) {
lala = lala + "a" + "a" + "a" + "a" + "a" + "a" + "a" + "a" + "a" + "a"
+ "a" +
"a" + "a" + "a" + "a" + "a" + "a" + "a" + "a" + "a";
lala.charAt(lala.length-1);
}
print((new Date().getTime() - t1) + 'ms');
//------------
var t1 = new Date().getTime();
lala = "hello";
for(var i = 0; i < 50000; i++) {
lala =
lala.concat("a", "a", "a", "a", "a", "a", "a", "a", "a", "a", "a", "a",
"a", "a", "a", "a", "a", "a", "a", "a");
lala.charAt(lala.length-1);
}
print((new Date().getTime() - t1) + 'ms');
this reveals:
(1 string - not shown)
134ms (baseline)
120ms (concat using StringAdd)
807ms (concat using StringBuilder)
(20 strings - shown)
74ms (baseline - is actually faster, because it's affected by my other
pending
patch)
877ms (concat using StringAdd)
35201ms (concat using StringBuilder)
http://codereview.chromium.org/243053
--~--~---------~--~----~------------~-------~--~----~
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
-~----------~----~----~----~------~----~------~--~---
