Two sets of performance test data for eval_javascript skills

A Weibo post by @老赵 "Is the code generated by eval really inefficient? http://t.cn/zWTUBEo Contains personal attacks, please don't enter if you don't like it."
triggered the recent heated discussion about eval During the discussion, the leader @Franky and Hui Da @otakustay also gave wonderful data analysis.
I happened to have done a similar test before, so I followed suit and joined in the fun, providing two sets of data for your reference.

Update 1: Thanks to @otakustay’s guidance, in order to eliminate the impact of the eval('') call itself on the results, a new set of data A3 and B3 has been added. And retest all old data.
Update 2: Thanks to Mo Da @貘吃荍香 for the powerful Paizhuan, added 1). Test data after eval coverage of A4, B4; A5, B5; 2). A6, B6 eval aliases; 3). A7,B7 eval.call.

Test environment:
a. Machine: Intel(R) Corei7-2720 2.2Ghz (4 cores 8 threads), memory 8Gb
b. OS: Windows 7 Enterprise SP1 64-bit
c. Browser:
b.1 Google Chrome 21.0.1180.79 m
b.2 Firefox 14.0.1
b.3 IE9.0.8112.16421
d. Test method
d .1 Each use case is tested 5 times, and the minimum time is taken.
d.2 Firebug or Chrome Console was not turned on during the test. Turning on these tools will double the time, making it difficult to get the results of this use case within the effective time

Use Case A1:
We inline Call empty eval("") in the function

Copy code The code is as follows:

!function () {
var a = 1,
b = 2,
c = true;
function func() {
var d = 2;
e = !c;
eval("");
}
for (var i = 0; i < 2999999; i ) {
func(i, i 1, i 2);
}
}();

Use case A2:
Comment out eval("") in the inline function

Copy code The code is as follows:

!function() {
var a = 1,
b = 2,
c = true;
function func() {
var d = 2;
e = !c;
//eval("");
}
for (var i = 0; i < 2999999 ; i ) {
func(i, i 1, i 2);
}
}();

Use case A3:
excludes eval("" ) The impact of the call itself, we call eval("") in the outer function

Copy the code The code is as follows:

!function() {
var a = 1,
b = 2,
c = true;
function func() {
var d = 2;
e = !c;
}
for (var i = 0; i < 2999999; i ) {
eval("");
func(i, i 1, i 2 ; >
Copy code

The code is as follows:


function eval(){}

!function() { var a = 1, b = 2, c = true; function func() { var d = 2; e = !c;

eval("");

}
for (var i = 0; i < 2999999; i ) {
func(i, i 1, i 2);
}
}();


Use case A5:

It is also a function call, not eval and another empty function f




Copy code

The code is as follows:


function f(){}

!function() { var a = 1, b = 2, c = true; function func() { var d = 2; e = !c;

f("");

}
for (var i = 0; i < 2999999; i ) {
func(i, i 1, i 2);
}
}();


Use Case A6:
Assign eval to another variable f, and then call f




Copy the code

The code is as follows:


var f = eval;

!function() { var a = 1, b = 2, c = true; function func() { var d = 2; e = ! c;

f("");

}
for (var i = 0; i < 2999999; i ) {
func(i, i 1, i 2);
}
}();


Use Case A7:

Use eval.call to call




Copy Code

The code is as follows:


!function() {

var a = 1, b = 2, c = true; function func() { var d = 2; e = !c; eval.call(null, '');

}

for (var i = 0; i < 2999999; i ) {
func(i, i 1, i 2);
}
}();

A组测试结果： 

	A1	A2	A3	A4	A5	A6	A7	A1 : A2	A1 : A3	A1 : A4	A4 : A5
Chrome	1612ms	8ms	1244ms	897ms	7ms	718ms	680ms	201.5	1.3	1.8	128.1
Firefox	2468ms	69ms	732ms	2928ms	134ms	5033ms	4984ms	35.8	3.4	0.8	21.9
IE	1207ms	23ms	233ms	1147ms	37ms	148ms	224ms	52.5	5.2	1.0	31.0

用例B1：

复制代码 代码如下:

for (var i = 0; i < 2999999; i++) {
!function() {
var a = 1,
b = 2,
c = true;
!function () {
var d = 2;
e = !c;
eval("");
}();
}();
}

用例B2：

复制代码 代码如下:

for (var i = 0; i < 2999999; i++) {
!function() {
var a = 1,
b = 2,
c = true;
!function () {
var d = 2;
e = !c;
//eval("");
}();
}();
}

用例B3：

复制代码 代码如下:

for (var i = 0; i < 2999999; i++) {
!function() {
var a = 1,
b = 2,
c = true;
!function () {
var d = 2;
e = !c;
}();
}();
eval("");
}

用例B4：

复制代码 代码如下:

var eval = function(){}
for (var i = 0; i < 2999999; i++) {
!function() {
var a = 1,
b = 2,
c = true;
!function () {
var d = 2;
e = !c;
eval("");
}();
}();
}

用例B5：

复制代码 代码如下:

var f = function(){}
for (var i = 0; i < 2999999; i++) {
!function() {
var a = 1,
b = 2,
c = true;
!function () {
var d = 2;
e = !c;
f("");
}();
}();
}

用例B6：

复制代码 代码如下:

var f = eval;
for (var i = 0; i < 2999999; i++) {
!function() {
var a = 1,
b = 2,
c = true;
!function () {
var d = 2;
e = !c;
f("");
}();
}();
}

用例B7:

复制代码 代码如下:

for (var i = 0; i < 2999999; i++) {
!function() {
var a = 1,
b = 2,
c = true;
!function () {
var d = 2;
e = !c;
eval.call(null, '');
}();
}();
}

B组测试结果：

	B1	B2	B3	B4	B5	B6	B7	B1 : B3	B1 : B2	B1 : B4	B4 : B5
Chrome	1569ms	134ms	1093ms	1022ms	173ms	830ms	916ms	11.7	1.4	1.5	5.9
Firefox	5334ms	1017ms	5503ms	5280ms	1171ms	6797ms	6883ms	5.2	1.0	1.0	4.5
IE	3933ms	560ms	680ms	4118ms	583ms	745ms	854ms	7.0	5.8	1.0	111.3

Conclusion ( is limited to CASE in the text):

1. Repeated calls to eval itself are very time-consuming, even empty eval("");

2. eval has an impact on the execution efficiency of inline functions, which varies depending on the specific environment and code;

3. We can see that no matter which browser, whether it is group A or group B, 2 and 5 have better speed. It shows that no matter how the eval of the inline function in the example is called (even if eval is overwritten by an empty function), it will still have a great impact on the operating efficiency. The inference is (black box inference, non-authoritative, probably just conjecture) that as long as eval is found in the inline function, even if the eval is an overridden empty function, all externally defined variables and other contents will be initialized in Scope Variables. to the current Scope. Similarly, eval will have a greater impact on the optimization function of the JS engine at runtime for inline functions and reduce execution efficiency.

4. To make a digression, although IE10 is not used, but IE9 is used, the performance of eval processing is very good. IE has always been criticized by developers, but its rapid growth is worthy of recognition. This example is a good proof.

A more detailed analysis of the reasons. The descriptions in the following articles are very detailed and will not be repeated again. Welcome:) Especially Mo Da...

@老赵’s 《Is the code generated by eval really inefficient? 》
@Franky’s "Popular Science on Eval"
@otakustay’s "A Brief Talk on the Influence of Eval"