VB.NET位运算之交换两个数字的效率分析

一、问题描述

近日，同事Kings给我看了一段奇怪的C#代码，鄙人对位运算了解甚少，初看之下懵了，还真不知道其功能，代码如下：

Int32 a=10,b=20;
a=a^b;
b=a^b;
a=a^b;

不过，我相信各位看客肯定都遇到过类似代码，对了，就是不用中间变量完成交换两个数字的功能。

出于个人喜好，我习惯性的想到：那么这种通过位运算的方式进行数字交换的效率和普通交换方式的效率相比，孰高孰低呢？

还是测试。

二、问题剖析

先用C#语言来测试，代码如下：

private void ExchangeMethodTest()
{
Int32 a = 10, b = 20;
Console.WriteLine(op_format, "Cycles", count);

Stopwatch sw = Stopwatch.StartNew();
for (Int32 i = 0; i < count; i++)
{
Int32 temp = 0;
temp = b;
b = a;
a = temp;
}
Console.WriteLine(op_format, "Common Exchange Method", sw.Elapsed);

for (Int32 i = 0; i < count; i++)
{
a = a ^ b;
b = a ^ b;
a = a ^ b;
}
Console.WriteLine(op_format, "Use Bit Exchange Method", sw.Elapsed);
}

测试结果截图如下：





哎，C#位运算的效率怎么会这么低呐？

再看看VB的，不过我想效率应该也不怎么样，测试一下再说吧，代码如下：

Private Sub ExchangeMethodTest()
Dim intA As Int32 = 10, intB As Int32 = 20
Dim sw As Stopwatch = Stopwatch.StartNew()
For i As Int32 = 0 To count
Dim j As Int32
j = intB
intB = intA
intA = j
Next
Console.WriteLine(op_format, "Common Exchange Method", sw.Elapsed)

sw = Stopwatch.StartNew()
For i As Int32 = 0 To count
intA = intA Xor intB
intB = intA Xor intB
intA = intA Xor intB
Next
Console.WriteLine(op_format, "Use Bit Exchange Method", sw.Elapsed)
End Sub

运行结果如下截图：





哦？VB中位运算的效率是还高于C#的！

但是不论是C#还是VB，在位运算的效率都要低于普通的交换方法。看来，这种位运算不仅难懂，而且效率不高(这里仅只.net平台的C#和VB)。

反汇编(以VB为例)显示结果无法复制文本，只好截图了：





一下是IL代码：

.method private static void ExchangeMethodTest() cil managed
{
.maxstack 3
.locals init (
[0] int32 intA,
[1] int32 intB,
[2] class [System]System.Diagnostics.Stopwatch sw,
[3] int32 i,
[4] int32 j,
[5] int32 V_5,
[6] int32 VB$CG$t_i4$S0)
L_0000: nop
L_0001: ldc.i4.s 10
L_0003: stloc.0
L_0004: ldc.i4.s 20
L_0006: stloc.1
L_0007: ldstr "{0,-27}:{1}"
L_000c: ldstr "Cycles"
L_0011: ldc.i4 0x3b9aca00
L_0016: box int32
L_001b: call void [mscorlib]System.Console::WriteLine(string, object, object)
L_0020: nop
L_0021: call class [System]System.Diagnostics.Stopwatch [System]System.Diagnostics.Stopwatch::StartNew()
L_0026: stloc.2
L_0027: ldc.i4.0
L_0028: stloc.3
L_0029: ldloc.1
L_002a: stloc.s j
L_002c: ldloc.0
L_002d: stloc.1
L_002e: ldloc.s j
L_0030: stloc.0
L_0031: nop
L_0032: ldloc.3
L_0033: ldc.i4.1
L_0034: add.ovf
L_0035: stloc.3
L_0036: ldloc.3
L_0037: ldc.i4 0x3b9aca00
L_003c: stloc.s VB$CG$t_i4$S0
L_003e: ldloc.s VB$CG$t_i4$S0
L_0040: ble.s L_0029
L_0042: ldstr "{0,-27}:{1}"
L_0047: ldstr "Common Exchange Method"
L_004c: ldloc.2
L_004d: callvirt instance valuetype [mscorlib]System.TimeSpan [System]System.Diagnostics.Stopwatch::get_Elapsed()
L_0052: box [mscorlib]System.TimeSpan
L_0057: call void [mscorlib]System.Console::WriteLine(string, object, object)
L_005c: nop
L_005d: call class [System]System.Diagnostics.Stopwatch [System]System.Diagnostics.Stopwatch::StartNew()
L_0062: stloc.2
L_0063: ldc.i4.0
L_0064: stloc.s V_5
L_0066: ldloc.0
L_0067: ldloc.1
L_0068: xor       <---异或操作
L_0069: stloc.0
L_006a: ldloc.0
L_006b: ldloc.1
L_006c: xor       <---异或操作
L_006d: stloc.1
L_006e: ldloc.0
L_006f: ldloc.1
L_0070: xor       <---异或操作
L_0071: stloc.0
L_0072: nop
L_0073: ldloc.s V_5
L_0075: ldc.i4.1
L_0076: add.ovf
L_0077: stloc.s V_5
L_0079: ldloc.s V_5
L_007b: ldc.i4 0x3b9aca00
L_0080: stloc.s VB$CG$t_i4$S0
L_0082: ldloc.s VB$CG$t_i4$S0
L_0084: ble.s L_0066
L_0086: ldstr "{0,-27}:{1}"
L_008b: ldstr "Use Bit Exchange Method"
L_0090: ldloc.2
L_0091: callvirt instance valuetype [mscorlib]System.TimeSpan [System]System.Diagnostics.Stopwatch::get_Elapsed()
L_0096: box [mscorlib]System.TimeSpan
L_009b: call void [mscorlib]System.Console::WriteLine(string, object, object)
L_00a0: nop
L_00a1: nop
L_00a2: ret
}

三、总结

综上所述，交换两个数字还是使用中间变量吧。

另外，截图中的反汇编和IL代码还望有高手指点一二！

不知C/C++在位运算交换两个数字时的效率如何，还望高手指点一二！