Refine CopyBytes macro instruction in IA32/X64

Refine CopyBytes macro instruction in IA32/X64

Use DWORD/QWORD copy for more short cases

BUG=
R=svenpanne@chromium.org

Committed: http://code.google.com/p/v8/source/detail?r=17632

[Weiliang, 2013-11-08 08:10:52]

It improves the array.join by 30% for short cases and less than 1% regression
for long cases.


Micros

function foo() {
    var fruits = ["Banana", "Orange", "Apple", "Mango"];
    var energy = fruits.join();
    return energy;
}

function fooLong() {
    var fruits = ["BananaBananaBananaBananaBanana",
"OrangeOrangeOrangeOrangeOrange", "AppleAppleAppleAppleAppleAppleApple",
"MangoMangoMangoMangoMangoMangoMango"];
    var energy = fruits.join();
    return energy;
}

[Weiliang, 2013-11-09 14:33:20]

port to x64

[Sven Panne, 2013-11-12 08:07:13]

LGTM, I'll land it for you. Duff's device reloaded in assembler, well, almost...
;-)

[Sven Panne, 2013-11-12 09:08:59]

Committed patchset #2 manually as r17632 (presubmit successful).