Cocos2d-x 3.0 Performance Benchamrk

This document is a brief introduction of testing sample and performance standard. I take this benchmark from two aspects-language compare and resource utilization compare.

Language Compare

This test is simply comparing the language performance by excuting Loop, Callback and MAT4. Take time consuming as the standard.

Loop:

N = 1000;
i = 0;
while(i < N)
{
	i++;
}

Callback:

N = 1000000;
i = 0;
sum = 0;
while(i < N)
{
	sum += add(i, 1);
}

N = 100000;
i = 0;
a = [...];
b = [...];
c = [...];
while(i < N)
{
	multiplayMatrix(a, b, c);
	i++;
}

####Device Overview

	Android	iOS
CPU	600MHz	1331MHz
RAM	256M	1G
OS	Android 2.2	iOS 7.03

####Testing Data

Android	Loop	Callback	MAT4
JSB(JIT)	3.7 ms	8268 ms	37899 ms
Lua(JIT)	0.8 ms	304 ms	899 ms
C++	0.1 ms	31.5 ms	295.4 ms

iOS	Loop	Callback	MAT4
JSB	1 ms	609.1 ms	9223 ms
Lua	0.3 ms	109.9 ms	581.7 ms
C++	0 ms	0 ms	29.6 ms

More detials:

C++ project

JSB project

Lua project

Resource Utilization Comapre

This test is computing the memory consuming when adding 1000, 10000 and 100000 nodes in the scene.

N = 1000

i = 0

while(i < N)
{
	auto sprite = Sprite::create("HelloWorld.png");
	addChile(sprite);
	i++;
}

####Device Overview

	iOS
CPU	1331MHz
RAM	1G
OS	iOS 7.03

####Testing Data

iOS	1000	10000	100000
JSB	2.1 M	11.8 M	128.6 M
Lua	0.7 M	13.1 M	118.6 M
C++	1.2 M	11.2 M	91.1 M