注意:下面所有的算法都是用C语言编写的，但是应该可以移植到你所选择的语言中(当它们没有那么快的时候不要看着我:)

选项

低内存(32位int, 32位机器)(从这里):

unsigned int
reverse(register unsigned int x)
{
    x = (((x & 0xaaaaaaaa) >> 1) | ((x & 0x55555555) << 1));
    x = (((x & 0xcccccccc) >> 2) | ((x & 0x33333333) << 2));
    x = (((x & 0xf0f0f0f0) >> 4) | ((x & 0x0f0f0f0f) << 4));
    x = (((x & 0xff00ff00) >> 8) | ((x & 0x00ff00ff) << 8));
    return((x >> 16) | (x << 16));

}

来自著名的Bit Twiddling Hacks页面:

最快(查找表):

static const unsigned char BitReverseTable256[] = 
{
  0x00, 0x80, 0x40, 0xC0, 0x20, 0xA0, 0x60, 0xE0, 0x10, 0x90, 0x50, 0xD0, 0x30, 0xB0, 0x70, 0xF0, 
  0x08, 0x88, 0x48, 0xC8, 0x28, 0xA8, 0x68, 0xE8, 0x18, 0x98, 0x58, 0xD8, 0x38, 0xB8, 0x78, 0xF8, 
  0x04, 0x84, 0x44, 0xC4, 0x24, 0xA4, 0x64, 0xE4, 0x14, 0x94, 0x54, 0xD4, 0x34, 0xB4, 0x74, 0xF4, 
  0x0C, 0x8C, 0x4C, 0xCC, 0x2C, 0xAC, 0x6C, 0xEC, 0x1C, 0x9C, 0x5C, 0xDC, 0x3C, 0xBC, 0x7C, 0xFC, 
  0x02, 0x82, 0x42, 0xC2, 0x22, 0xA2, 0x62, 0xE2, 0x12, 0x92, 0x52, 0xD2, 0x32, 0xB2, 0x72, 0xF2, 
  0x0A, 0x8A, 0x4A, 0xCA, 0x2A, 0xAA, 0x6A, 0xEA, 0x1A, 0x9A, 0x5A, 0xDA, 0x3A, 0xBA, 0x7A, 0xFA,
  0x06, 0x86, 0x46, 0xC6, 0x26, 0xA6, 0x66, 0xE6, 0x16, 0x96, 0x56, 0xD6, 0x36, 0xB6, 0x76, 0xF6, 
  0x0E, 0x8E, 0x4E, 0xCE, 0x2E, 0xAE, 0x6E, 0xEE, 0x1E, 0x9E, 0x5E, 0xDE, 0x3E, 0xBE, 0x7E, 0xFE,
  0x01, 0x81, 0x41, 0xC1, 0x21, 0xA1, 0x61, 0xE1, 0x11, 0x91, 0x51, 0xD1, 0x31, 0xB1, 0x71, 0xF1,
  0x09, 0x89, 0x49, 0xC9, 0x29, 0xA9, 0x69, 0xE9, 0x19, 0x99, 0x59, 0xD9, 0x39, 0xB9, 0x79, 0xF9, 
  0x05, 0x85, 0x45, 0xC5, 0x25, 0xA5, 0x65, 0xE5, 0x15, 0x95, 0x55, 0xD5, 0x35, 0xB5, 0x75, 0xF5,
  0x0D, 0x8D, 0x4D, 0xCD, 0x2D, 0xAD, 0x6D, 0xED, 0x1D, 0x9D, 0x5D, 0xDD, 0x3D, 0xBD, 0x7D, 0xFD,
  0x03, 0x83, 0x43, 0xC3, 0x23, 0xA3, 0x63, 0xE3, 0x13, 0x93, 0x53, 0xD3, 0x33, 0xB3, 0x73, 0xF3, 
  0x0B, 0x8B, 0x4B, 0xCB, 0x2B, 0xAB, 0x6B, 0xEB, 0x1B, 0x9B, 0x5B, 0xDB, 0x3B, 0xBB, 0x7B, 0xFB,
  0x07, 0x87, 0x47, 0xC7, 0x27, 0xA7, 0x67, 0xE7, 0x17, 0x97, 0x57, 0xD7, 0x37, 0xB7, 0x77, 0xF7, 
  0x0F, 0x8F, 0x4F, 0xCF, 0x2F, 0xAF, 0x6F, 0xEF, 0x1F, 0x9F, 0x5F, 0xDF, 0x3F, 0xBF, 0x7F, 0xFF
};

unsigned int v; // reverse 32-bit value, 8 bits at time
unsigned int c; // c will get v reversed

// Option 1:
c = (BitReverseTable256[v & 0xff] << 24) | 
    (BitReverseTable256[(v >> 8) & 0xff] << 16) | 
    (BitReverseTable256[(v >> 16) & 0xff] << 8) |
    (BitReverseTable256[(v >> 24) & 0xff]);

// Option 2:
unsigned char * p = (unsigned char *) &v;
unsigned char * q = (unsigned char *) &c;
q[3] = BitReverseTable256[p[0]]; 
q[2] = BitReverseTable256[p[1]]; 
q[1] = BitReverseTable256[p[2]]; 
q[0] = BitReverseTable256[p[3]];

您可以将此想法扩展到64位整数，或者为了速度而牺牲内存(假设L1数据缓存足够大)，并使用一个64k条目查找表一次反向16位。

---

其他人

简单的

unsigned int v;     // input bits to be reversed
unsigned int r = v & 1; // r will be reversed bits of v; first get LSB of v
int s = sizeof(v) * CHAR_BIT - 1; // extra shift needed at end

for (v >>= 1; v; v >>= 1)
{   
  r <<= 1;
  r |= v & 1;
  s--;
}
r <<= s; // shift when v's highest bits are zero

更快(32位处理器)

unsigned char b = x;
b = ((b * 0x0802LU & 0x22110LU) | (b * 0x8020LU & 0x88440LU)) * 0x10101LU >> 16; 

更快(64位处理器)

unsigned char b; // reverse this (8-bit) byte
b = (b * 0x0202020202ULL & 0x010884422010ULL) % 1023;

如果您想在32位整型上执行此操作，只需反转每个字节中的位，并反转字节的顺序。那就是:

unsigned int toReverse;
unsigned int reversed;
unsigned char inByte0 = (toReverse & 0xFF);
unsigned char inByte1 = (toReverse & 0xFF00) >> 8;
unsigned char inByte2 = (toReverse & 0xFF0000) >> 16;
unsigned char inByte3 = (toReverse & 0xFF000000) >> 24;
reversed = (reverseBits(inByte0) << 24) | (reverseBits(inByte1) << 16) | (reverseBits(inByte2) << 8) | (reverseBits(inByte3);

---

结果

我对两种最有希望的解决方案进行了基准测试，查找表和按位and(第一个)。测试机器是一台带有4GB DDR2-800和酷睿2 Duo T7500 @ 2.4GHz, 4MB L2缓存的笔记本电脑;YMMV。我在64位Linux上使用gcc 4.3.2。OpenMP(和GCC绑定)用于高分辨率计时器。

reverse.c

#include <stdlib.h>
#include <stdio.h>
#include <omp.h>

unsigned int
reverse(register unsigned int x)
{
    x = (((x & 0xaaaaaaaa) >> 1) | ((x & 0x55555555) << 1));
    x = (((x & 0xcccccccc) >> 2) | ((x & 0x33333333) << 2));
    x = (((x & 0xf0f0f0f0) >> 4) | ((x & 0x0f0f0f0f) << 4));
    x = (((x & 0xff00ff00) >> 8) | ((x & 0x00ff00ff) << 8));
    return((x >> 16) | (x << 16));

}

int main()
{
    unsigned int *ints = malloc(100000000*sizeof(unsigned int));
    unsigned int *ints2 = malloc(100000000*sizeof(unsigned int));
    for(unsigned int i = 0; i < 100000000; i++)
      ints[i] = rand();

    unsigned int *inptr = ints;
    unsigned int *outptr = ints2;
    unsigned int *endptr = ints + 100000000;
    // Starting the time measurement
    double start = omp_get_wtime();
    // Computations to be measured
    while(inptr != endptr)
    {
      (*outptr) = reverse(*inptr);
      inptr++;
      outptr++;
    }
    // Measuring the elapsed time
    double end = omp_get_wtime();
    // Time calculation (in seconds)
    printf("Time: %f seconds\n", end-start);

    free(ints);
    free(ints2);

    return 0;
}

reverse_lookup.c

#include <stdlib.h>
#include <stdio.h>
#include <omp.h>

static const unsigned char BitReverseTable256[] = 
{
  0x00, 0x80, 0x40, 0xC0, 0x20, 0xA0, 0x60, 0xE0, 0x10, 0x90, 0x50, 0xD0, 0x30, 0xB0, 0x70, 0xF0, 
  0x08, 0x88, 0x48, 0xC8, 0x28, 0xA8, 0x68, 0xE8, 0x18, 0x98, 0x58, 0xD8, 0x38, 0xB8, 0x78, 0xF8, 
  0x04, 0x84, 0x44, 0xC4, 0x24, 0xA4, 0x64, 0xE4, 0x14, 0x94, 0x54, 0xD4, 0x34, 0xB4, 0x74, 0xF4, 
  0x0C, 0x8C, 0x4C, 0xCC, 0x2C, 0xAC, 0x6C, 0xEC, 0x1C, 0x9C, 0x5C, 0xDC, 0x3C, 0xBC, 0x7C, 0xFC, 
  0x02, 0x82, 0x42, 0xC2, 0x22, 0xA2, 0x62, 0xE2, 0x12, 0x92, 0x52, 0xD2, 0x32, 0xB2, 0x72, 0xF2, 
  0x0A, 0x8A, 0x4A, 0xCA, 0x2A, 0xAA, 0x6A, 0xEA, 0x1A, 0x9A, 0x5A, 0xDA, 0x3A, 0xBA, 0x7A, 0xFA,
  0x06, 0x86, 0x46, 0xC6, 0x26, 0xA6, 0x66, 0xE6, 0x16, 0x96, 0x56, 0xD6, 0x36, 0xB6, 0x76, 0xF6, 
  0x0E, 0x8E, 0x4E, 0xCE, 0x2E, 0xAE, 0x6E, 0xEE, 0x1E, 0x9E, 0x5E, 0xDE, 0x3E, 0xBE, 0x7E, 0xFE,
  0x01, 0x81, 0x41, 0xC1, 0x21, 0xA1, 0x61, 0xE1, 0x11, 0x91, 0x51, 0xD1, 0x31, 0xB1, 0x71, 0xF1,
  0x09, 0x89, 0x49, 0xC9, 0x29, 0xA9, 0x69, 0xE9, 0x19, 0x99, 0x59, 0xD9, 0x39, 0xB9, 0x79, 0xF9, 
  0x05, 0x85, 0x45, 0xC5, 0x25, 0xA5, 0x65, 0xE5, 0x15, 0x95, 0x55, 0xD5, 0x35, 0xB5, 0x75, 0xF5,
  0x0D, 0x8D, 0x4D, 0xCD, 0x2D, 0xAD, 0x6D, 0xED, 0x1D, 0x9D, 0x5D, 0xDD, 0x3D, 0xBD, 0x7D, 0xFD,
  0x03, 0x83, 0x43, 0xC3, 0x23, 0xA3, 0x63, 0xE3, 0x13, 0x93, 0x53, 0xD3, 0x33, 0xB3, 0x73, 0xF3, 
  0x0B, 0x8B, 0x4B, 0xCB, 0x2B, 0xAB, 0x6B, 0xEB, 0x1B, 0x9B, 0x5B, 0xDB, 0x3B, 0xBB, 0x7B, 0xFB,
  0x07, 0x87, 0x47, 0xC7, 0x27, 0xA7, 0x67, 0xE7, 0x17, 0x97, 0x57, 0xD7, 0x37, 0xB7, 0x77, 0xF7, 
  0x0F, 0x8F, 0x4F, 0xCF, 0x2F, 0xAF, 0x6F, 0xEF, 0x1F, 0x9F, 0x5F, 0xDF, 0x3F, 0xBF, 0x7F, 0xFF
};

int main()
{
    unsigned int *ints = malloc(100000000*sizeof(unsigned int));
    unsigned int *ints2 = malloc(100000000*sizeof(unsigned int));
    for(unsigned int i = 0; i < 100000000; i++)
      ints[i] = rand();

    unsigned int *inptr = ints;
    unsigned int *outptr = ints2;
    unsigned int *endptr = ints + 100000000;
    // Starting the time measurement
    double start = omp_get_wtime();
    // Computations to be measured
    while(inptr != endptr)
    {
    unsigned int in = *inptr;  

    // Option 1:
    //*outptr = (BitReverseTable256[in & 0xff] << 24) | 
    //    (BitReverseTable256[(in >> 8) & 0xff] << 16) | 
    //    (BitReverseTable256[(in >> 16) & 0xff] << 8) |
    //    (BitReverseTable256[(in >> 24) & 0xff]);

    // Option 2:
    unsigned char * p = (unsigned char *) &(*inptr);
    unsigned char * q = (unsigned char *) &(*outptr);
    q[3] = BitReverseTable256[p[0]]; 
    q[2] = BitReverseTable256[p[1]]; 
    q[1] = BitReverseTable256[p[2]]; 
    q[0] = BitReverseTable256[p[3]];

      inptr++;
      outptr++;
    }
    // Measuring the elapsed time
    double end = omp_get_wtime();
    // Time calculation (in seconds)
    printf("Time: %f seconds\n", end-start);

    free(ints);
    free(ints2);

    return 0;
}

我在几个不同的优化中尝试了这两种方法，在每个级别上进行了3次试验，每次试验逆转了1亿个随机无符号整数。对于查找表选项，我尝试了按位hacks页面上给出的两种方案(选项1和2)。结果如下所示。

位和

mrj10@mjlap:~/code$ gcc -fopenmp -std=c99 -o reverse reverse.c
mrj10@mjlap:~/code$ ./reverse
Time: 2.000593 seconds
mrj10@mjlap:~/code$ ./reverse
Time: 1.938893 seconds
mrj10@mjlap:~/code$ ./reverse
Time: 1.936365 seconds
mrj10@mjlap:~/code$ gcc -fopenmp -std=c99 -O2 -o reverse reverse.c
mrj10@mjlap:~/code$ ./reverse
Time: 0.942709 seconds
mrj10@mjlap:~/code$ ./reverse
Time: 0.991104 seconds
mrj10@mjlap:~/code$ ./reverse
Time: 0.947203 seconds
mrj10@mjlap:~/code$ gcc -fopenmp -std=c99 -O3 -o reverse reverse.c
mrj10@mjlap:~/code$ ./reverse
Time: 0.922639 seconds
mrj10@mjlap:~/code$ ./reverse
Time: 0.892372 seconds
mrj10@mjlap:~/code$ ./reverse
Time: 0.891688 seconds

查阅表(选项1)

mrj10@mjlap:~/code$ gcc -fopenmp -std=c99 -o reverse_lookup reverse_lookup.c
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.201127 seconds              
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.196129 seconds              
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.235972 seconds              
mrj10@mjlap:~/code$ gcc -fopenmp -std=c99 -O2 -o reverse_lookup reverse_lookup.c
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 0.633042 seconds              
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 0.655880 seconds              
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 0.633390 seconds              
mrj10@mjlap:~/code$ gcc -fopenmp -std=c99 -O3 -o reverse_lookup reverse_lookup.c
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 0.652322 seconds              
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 0.631739 seconds              
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 0.652431 seconds  

查找表(选项2)

mrj10@mjlap:~/code$ gcc -fopenmp -std=c99 -o reverse_lookup reverse_lookup.c
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.671537 seconds
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.688173 seconds
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.664662 seconds
mrj10@mjlap:~/code$ gcc -fopenmp -std=c99 -O2 -o reverse_lookup reverse_lookup.c
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.049851 seconds
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.048403 seconds
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.085086 seconds
mrj10@mjlap:~/code$ gcc -fopenmp -std=c99 -O3 -o reverse_lookup reverse_lookup.c
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.082223 seconds
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.053431 seconds
mrj10@mjlap:~/code$ ./reverse_lookup
Time: 1.081224 seconds

结论

如果您关心性能，请使用选项1(字节寻址速度慢，这是意料中事)的查找表。如果您需要从系统中挤出最后一个字节的内存(如果您关心位反转的性能，您可能会这样做)，那么按位- and方法的优化版本也不会太糟糕。

警告

是的，我知道基准代码完全是一种hack。关于如何改进它的建议非常受欢迎。我知道的事情:

I don't have access to ICC. This may be faster (please respond in a comment if you can test this out). A 64K lookup table may do well on some modern microarchitectures with large L1D. -mtune=native didn't work for -O2/-O3 (ld blew up with some crazy symbol redefinition error), so I don't believe the generated code is tuned for my microarchitecture. There may be a way to do this slightly faster with SSE. I have no idea how, but with fast replication, packed bitwise AND, and swizzling instructions, there's got to be something there. I know only enough x86 assembly to be dangerous; here's the code GCC generated on -O3 for option 1, so somebody more knowledgable than myself can check it out:

32位

.L3:
movl    (%r12,%rsi), %ecx
movzbl  %cl, %eax
movzbl  BitReverseTable256(%rax), %edx
movl    %ecx, %eax
shrl    $24, %eax
mov     %eax, %eax
movzbl  BitReverseTable256(%rax), %eax
sall    $24, %edx
orl     %eax, %edx
movzbl  %ch, %eax
shrl    $16, %ecx
movzbl  BitReverseTable256(%rax), %eax
movzbl  %cl, %ecx
sall    $16, %eax
orl     %eax, %edx
movzbl  BitReverseTable256(%rcx), %eax
sall    $8, %eax
orl     %eax, %edx
movl    %edx, (%r13,%rsi)
addq    $4, %rsi
cmpq    $400000000, %rsi
jne     .L3

编辑:我还尝试在我的机器上使用uint64_t类型，看看是否有任何性能提升。性能比32位快10%左右，无论您是一次使用64位类型对两个32位整型反转位，还是实际上将64位值的一半反转位，性能都几乎相同。汇编代码如下所示(对于前一种情况，一次为两个32位整型反转位):

.L3:
movq    (%r12,%rsi), %rdx
movq    %rdx, %rax
shrq    $24, %rax
andl    $255, %eax
movzbl  BitReverseTable256(%rax), %ecx
movzbq  %dl,%rax
movzbl  BitReverseTable256(%rax), %eax
salq    $24, %rax
orq     %rax, %rcx
movq    %rdx, %rax
shrq    $56, %rax
movzbl  BitReverseTable256(%rax), %eax
salq    $32, %rax
orq     %rax, %rcx
movzbl  %dh, %eax
shrq    $16, %rdx
movzbl  BitReverseTable256(%rax), %eax
salq    $16, %rax
orq     %rax, %rcx
movzbq  %dl,%rax
shrq    $16, %rdx
movzbl  BitReverseTable256(%rax), %eax
salq    $8, %rax
orq     %rax, %rcx
movzbq  %dl,%rax
shrq    $8, %rdx
movzbl  BitReverseTable256(%rax), %eax
salq    $56, %rax
orq     %rax, %rcx
movzbq  %dl,%rax
shrq    $8, %rdx
movzbl  BitReverseTable256(%rax), %eax
andl    $255, %edx
salq    $48, %rax
orq     %rax, %rcx
movzbl  BitReverseTable256(%rdx), %eax
salq    $40, %rax
orq     %rax, %rcx
movq    %rcx, (%r13,%rsi)
addq    $8, %rsi
cmpq    $400000000, %rsi
jne     .L3

这个线程引起了我的注意，因为它处理了一个简单的问题，即使对于现代CPU也需要大量的工作(CPU周期)。有一天我也站在那里，有同样的¤#%“#”问题。我得翻几百万字节。然而，我知道我所有的目标系统都是基于现代英特尔的，所以让我们开始优化到极致!!

所以我使用了Matt J的查找代码作为基础。我正在基准测试的系统是i7 haswell 4700eq。

Matt J的查找位翻转400亿字节:大约0.272秒。

然后我继续尝试，看看英特尔的ISPC编译器是否可以向量化反向的算术。c。

我不打算在这里用我的发现来烦你，因为我尝试了很多来帮助编译器找到东西，无论如何，我最终得到了大约0.15秒的性能来bitflip 400亿字节。这是一个伟大的减少，但对于我的应用程序，这仍然是方式方式太慢。

所以人们让我展示世界上最快的基于英特尔的bitflipper。定时:

时间到bitflip 400000000字节:0.050082秒!!!!!

// Bitflip using AVX2 - The fastest Intel based bitflip in the world!!
// Made by Anders Cedronius 2014 (anders.cedronius (you know what) gmail.com)

#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <omp.h>

using namespace std;

#define DISPLAY_HEIGHT  4
#define DISPLAY_WIDTH   32
#define NUM_DATA_BYTES  400000000

// Constants (first we got the mask, then the high order nibble look up table and last we got the low order nibble lookup table)
__attribute__ ((aligned(32))) static unsigned char k1[32*3]={
        0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,0x0f,
        0x00,0x08,0x04,0x0c,0x02,0x0a,0x06,0x0e,0x01,0x09,0x05,0x0d,0x03,0x0b,0x07,0x0f,0x00,0x08,0x04,0x0c,0x02,0x0a,0x06,0x0e,0x01,0x09,0x05,0x0d,0x03,0x0b,0x07,0x0f,
        0x00,0x80,0x40,0xc0,0x20,0xa0,0x60,0xe0,0x10,0x90,0x50,0xd0,0x30,0xb0,0x70,0xf0,0x00,0x80,0x40,0xc0,0x20,0xa0,0x60,0xe0,0x10,0x90,0x50,0xd0,0x30,0xb0,0x70,0xf0
};

// The data to be bitflipped (+32 to avoid the quantization out of memory problem)
__attribute__ ((aligned(32))) static unsigned char data[NUM_DATA_BYTES+32]={};

extern "C" {
void bitflipbyte(unsigned char[],unsigned int,unsigned char[]);
}

int main()
{

    for(unsigned int i = 0; i < NUM_DATA_BYTES; i++)
    {
        data[i] = rand();
    }

    printf ("\r\nData in(start):\r\n");
    for (unsigned int j = 0; j < 4; j++)
    {
        for (unsigned int i = 0; i < DISPLAY_WIDTH; i++)
        {
            printf ("0x%02x,",data[i+(j*DISPLAY_WIDTH)]);
        }
        printf ("\r\n");
    }

    printf ("\r\nNumber of 32-byte chunks to convert: %d\r\n",(unsigned int)ceil(NUM_DATA_BYTES/32.0));

    double start_time = omp_get_wtime();
    bitflipbyte(data,(unsigned int)ceil(NUM_DATA_BYTES/32.0),k1);
    double end_time = omp_get_wtime();

    printf ("\r\nData out:\r\n");
    for (unsigned int j = 0; j < 4; j++)
    {
        for (unsigned int i = 0; i < DISPLAY_WIDTH; i++)
        {
            printf ("0x%02x,",data[i+(j*DISPLAY_WIDTH)]);
        }
        printf ("\r\n");
    }
    printf("\r\n\r\nTime to bitflip %d bytes: %f seconds\r\n\r\n",NUM_DATA_BYTES, end_time-start_time);

    // return with no errors
    return 0;
}

printf是用来调试的。

这是主要的工作:

bits 64
global bitflipbyte

bitflipbyte:    
        vmovdqa     ymm2, [rdx]
        add         rdx, 20h
        vmovdqa     ymm3, [rdx]
        add         rdx, 20h
        vmovdqa     ymm4, [rdx]
bitflipp_loop:
        vmovdqa     ymm0, [rdi] 
        vpand       ymm1, ymm2, ymm0 
        vpandn      ymm0, ymm2, ymm0 
        vpsrld      ymm0, ymm0, 4h 
        vpshufb     ymm1, ymm4, ymm1 
        vpshufb     ymm0, ymm3, ymm0         
        vpor        ymm0, ymm0, ymm1
        vmovdqa     [rdi], ymm0
        add     rdi, 20h
        dec     rsi
        jnz     bitflipp_loop
        ret

代码占用32个字节，然后屏蔽掉蚕食。高啃角右移了4。然后使用vpshufb和ymm4 / ymm3作为查找表。我可以使用一个单独的查找表，但我将不得不在ORing再次一起啃啃之前向左移动。

还有更快的翻转比特的方法。但我被绑定到单线程和CPU，所以这是我能实现的最快速度。你能做一个快一点的版本吗?

关于使用Intel C/ c++编译器内在等效命令，请不要发表任何评论…

通用的

C代码。以1字节输入数据num为例。

    unsigned char num = 0xaa;   // 1010 1010 (aa) -> 0101 0101 (55)
    int s = sizeof(num) * 8;    // get number of bits
    int i, x, y, p;
    int var = 0;                // make var data type to be equal or larger than num

    for (i = 0; i < (s / 2); i++) {
        // extract bit on the left, from MSB
        p = s - i - 1;
        x = num & (1 << p);
        x = x >> p;
        printf("x: %d\n", x);

        // extract bit on the right, from LSB
        y = num & (1 << i);
        y = y >> i;
        printf("y: %d\n", y);

        var = var | (x << i);       // apply x
        var = var | (y << p);       // apply y
    }

    printf("new: 0x%x\n", new);

另一个基于循环的解决方案，在数量较低时快速退出(在c++中用于多种类型)

template<class T>
T reverse_bits(T in) {
    T bit = static_cast<T>(1) << (sizeof(T) * 8 - 1);
    T out;

    for (out = 0; bit && in; bit >>= 1, in >>= 1) {
        if (in & 1) {
            out |= bit;
        }
    }
    return out;
}

或者C语言中unsigned int

unsigned int reverse_bits(unsigned int in) {
    unsigned int bit = 1u << (sizeof(T) * 8 - 1);
    unsigned int out;

    for (out = 0; bit && in; bit >>= 1, in >>= 1) {
        if (in & 1)
            out |= bit;
    }
    return out;
}

这不是人类能做的工作!．.． 但非常适合做机器

这是2015年，距离第一次提出这个问题已经过去了6年。编译器从此成为我们的主人，而我们作为人类的工作只是帮助它们。那么，把我们的意图传达给机器的最佳方式是什么呢?

位反转是如此普遍，以至于你不得不怀疑为什么x86不断增长的ISA没有包含一次性完成它的指令。

原因是:如果你给编译器一个真正简洁的意图，位反转应该只需要大约20个CPU周期。让我向你展示如何制作reverse()并使用它:

#include <inttypes.h>
#include <stdio.h>

uint64_t reverse(const uint64_t n,
                 const uint64_t k)
{
        uint64_t r, i;
        for (r = 0, i = 0; i < k; ++i)
                r |= ((n >> i) & 1) << (k - i - 1);
        return r;
}

int main()
{
        const uint64_t size = 64;
        uint64_t sum = 0;
        uint64_t a;
        for (a = 0; a < (uint64_t)1 << 30; ++a)
                sum += reverse(a, size);
        printf("%" PRIu64 "\n", sum);
        return 0;
}

使用Clang版本>= 3.6，-O3， -march=native(用Haswell测试)编译这个示例程序，使用新的AVX2指令提供美术质量代码，运行时为11秒处理~ 10亿reverse()秒。这是~10 ns每反向()，0.5 ns CPU周期假设2 GHz，我们将达到甜蜜的20个CPU周期。

对于单个大数组，您可以在访问RAM一次所需的时间内放入10个reverse() ! 你可以在访问L2缓存LUT两次的时间里放入1个reverse()。

注意:这个示例代码应该可以作为一个不错的基准运行几年，但是一旦编译器足够聪明，可以优化main()只输出最终结果，而不是真正计算任何东西，它最终就会开始显得过时了。但目前它只用于展示reverse()。

对于喜欢递归的人来说，这是另一个解决方案。

这个想法很简单。 将输入除以一半并交换两部分，继续直到达到单个位。

Illustrated in the example below.

Ex : If Input is 00101010   ==> Expected output is 01010100

1. Divide the input into 2 halves 
    0010 --- 1010

2. Swap the 2 Halves
    1010     0010

3. Repeat the same for each half.
    10 -- 10 ---  00 -- 10
    10    10      10    00

    1-0 -- 1-0 --- 1-0 -- 0-0
    0 1    0 1     0 1    0 0

Done! Output is 01010100

这里有一个递归函数来求解。(注意，我使用了unsigned int，所以它可以用于sizeof(unsigned int)*8位的输入。

递归函数有两个参数-需要位的值 要反转的值和值中的比特数。

int reverse_bits_recursive(unsigned int num, unsigned int numBits)
{
    unsigned int reversedNum;;
    unsigned int mask = 0;

    mask = (0x1 << (numBits/2)) - 1;

    if (numBits == 1) return num;
    reversedNum = reverse_bits_recursive(num >> numBits/2, numBits/2) |
                   reverse_bits_recursive((num & mask), numBits/2) << numBits/2;
    return reversedNum;
}

int main()
{
    unsigned int reversedNum;
    unsigned int num;

    num = 0x55;
    reversedNum = reverse_bits_recursive(num, 8);
    printf ("Bit Reversal Input = 0x%x Output = 0x%x\n", num, reversedNum);

    num = 0xabcd;
    reversedNum = reverse_bits_recursive(num, 16);
    printf ("Bit Reversal Input = 0x%x Output = 0x%x\n", num, reversedNum);

    num = 0x123456;
    reversedNum = reverse_bits_recursive(num, 24);
    printf ("Bit Reversal Input = 0x%x Output = 0x%x\n", num, reversedNum);

    num = 0x11223344;
    reversedNum = reverse_bits_recursive(num,32);
    printf ("Bit Reversal Input = 0x%x Output = 0x%x\n", num, reversedNum);
}

输出如下:

Bit Reversal Input = 0x55 Output = 0xaa
Bit Reversal Input = 0xabcd Output = 0xb3d5
Bit Reversal Input = 0x123456 Output = 0x651690
Bit Reversal Input = 0x11223344 Output = 0x22cc4488