另一个汉明权重算法，如果你使用的是BMI2 CPU:

the_weight = __tzcnt_u64(~_pext_u64(data[i], data[i]));

我给出了两个算法来回答这个问题，

package countSetBitsInAnInteger;

import java.util.Scanner;

public class UsingLoop {

    public static void main(String[] args) {
        Scanner in = new Scanner(System.in);
        try {
            System.out.println("Enter a integer number to check for set bits in it");
            int n = in.nextInt();
            System.out.println("Using while loop, we get the number of set bits as: " + usingLoop(n));
            System.out.println("Using Brain Kernighan's Algorithm, we get the number of set bits as: " + usingBrainKernighan(n));
            System.out.println("Using ");
        }
        finally {
            in.close();
        }
    }

    private static int usingBrainKernighan(int n) {
        int count = 0;
        while(n > 0) {
            n& = (n-1);
            count++;
        }
        return count;
    }

    /*
        Analysis:
            Time complexity = O(lgn)
            Space complexity = O(1)
    */

    private static int usingLoop(int n) {
        int count = 0;
        for(int i=0; i<32; i++) {
            if((n&(1 << i)) != 0)
                count++;
        }
        return count;
    }

    /*
        Analysis:
            Time Complexity = O(32) // Maybe the complexity is O(lgn)
            Space Complexity = O(1)
    */
}

这不是最快或最好的解决方案，但我以自己的方式发现了同样的问题，我开始反复思考。最后我意识到它可以这样做，如果你从数学方面得到这个问题，画一个图，然后你发现它是一个有周期部分的函数，然后你意识到周期之间的差异……所以你看:

unsigned int f(unsigned int x)
{
    switch (x) {
        case 0:
            return 0;
        case 1:
            return 1;
        case 2:
            return 1;
        case 3:
            return 2;
        default:
            return f(x/4) + f(x%4);
    }
}

这是一个可移植的模块(ANSI-C)，它可以在任何架构上对每个算法进行基准测试。

你的CPU有9位字节?目前它实现了2个算法，K&R算法和一个字节查找表。查找表的平均速度比K&R算法快3倍。如果有人能想出办法使“黑客的喜悦”算法可移植，请随意添加它。

#ifndef _BITCOUNT_H_
#define _BITCOUNT_H_

/* Return the Hamming Wieght of val, i.e. the number of 'on' bits. */
int bitcount( unsigned int );

/* List of available bitcount algorithms.  
 * onTheFly:    Calculate the bitcount on demand.
 *
 * lookupTalbe: Uses a small lookup table to determine the bitcount.  This
 * method is on average 3 times as fast as onTheFly, but incurs a small
 * upfront cost to initialize the lookup table on the first call.
 *
 * strategyCount is just a placeholder. 
 */
enum strategy { onTheFly, lookupTable, strategyCount };

/* String represenations of the algorithm names */
extern const char *strategyNames[];

/* Choose which bitcount algorithm to use. */
void setStrategy( enum strategy );

#endif

.

#include <limits.h>

#include "bitcount.h"

/* The number of entries needed in the table is equal to the number of unique
 * values a char can represent which is always UCHAR_MAX + 1*/
static unsigned char _bitCountTable[UCHAR_MAX + 1];
static unsigned int _lookupTableInitialized = 0;

static int _defaultBitCount( unsigned int val ) {
    int count;

    /* Starting with:
     * 1100 - 1 == 1011,  1100 & 1011 == 1000
     * 1000 - 1 == 0111,  1000 & 0111 == 0000
     */
    for ( count = 0; val; ++count )
        val &= val - 1;

    return count;
}

/* Looks up each byte of the integer in a lookup table.
 *
 * The first time the function is called it initializes the lookup table.
 */
static int _tableBitCount( unsigned int val ) {
    int bCount = 0;

    if ( !_lookupTableInitialized ) {
        unsigned int i;
        for ( i = 0; i != UCHAR_MAX + 1; ++i )
            _bitCountTable[i] =
                ( unsigned char )_defaultBitCount( i );

        _lookupTableInitialized = 1;
    }

    for ( ; val; val >>= CHAR_BIT )
        bCount += _bitCountTable[val & UCHAR_MAX];

    return bCount;
}

static int ( *_bitcount ) ( unsigned int ) = _defaultBitCount;

const char *strategyNames[] = { "onTheFly", "lookupTable" };

void setStrategy( enum strategy s ) {
    switch ( s ) {
    case onTheFly:
        _bitcount = _defaultBitCount;
        break;
    case lookupTable:
        _bitcount = _tableBitCount;
        break;
    case strategyCount:
        break;
    }
}

/* Just a forwarding function which will call whichever version of the
 * algorithm has been selected by the client 
 */
int bitcount( unsigned int val ) {
    return _bitcount( val );
}

#ifdef _BITCOUNT_EXE_

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

/* Use the same sequence of pseudo random numbers to benmark each Hamming
 * Weight algorithm.
 */
void benchmark( int reps ) {
    clock_t start, stop;
    int i, j;
    static const int iterations = 1000000;

    for ( j = 0; j != strategyCount; ++j ) {
        setStrategy( j );

        srand( 257 );

        start = clock(  );

        for ( i = 0; i != reps * iterations; ++i )
            bitcount( rand(  ) );

        stop = clock(  );

        printf
            ( "\n\t%d psudoe-random integers using %s: %f seconds\n\n",
              reps * iterations, strategyNames[j],
              ( double )( stop - start ) / CLOCKS_PER_SEC );
    }
}

int main( void ) {
    int option;

    while ( 1 ) {
        printf( "Menu Options\n"
            "\t1.\tPrint the Hamming Weight of an Integer\n"
            "\t2.\tBenchmark Hamming Weight implementations\n"
            "\t3.\tExit ( or cntl-d )\n\n\t" );

        if ( scanf( "%d", &option ) == EOF )
            break;

        switch ( option ) {
        case 1:
            printf( "Please enter the integer: " );
            if ( scanf( "%d", &option ) != EOF )
                printf
                    ( "The Hamming Weight of %d ( 0x%X ) is %d\n\n",
                      option, option, bitcount( option ) );
            break;
        case 2:
            printf
                ( "Please select number of reps ( in millions ): " );
            if ( scanf( "%d", &option ) != EOF )
                benchmark( option );
            break;
        case 3:
            goto EXIT;
            break;
        default:
            printf( "Invalid option\n" );
        }

    }

 EXIT:
    printf( "\n" );

    return 0;
}

#endif

对于232查找表和逐个遍历每个位之间的折中方法:

int bitcount(unsigned int num){
    int count = 0;
    static int nibblebits[] =
        {0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
    for(; num != 0; num >>= 4)
        count += nibblebits[num & 0x0f];
    return count;
}

从http://ctips.pbwiki.com/CountBits

我觉得很无聊，于是对三种方法进行了十亿次迭代。编译器是gcc -O3。CPU就是第一代Macbook Pro里装的东西。

最快的是3.7秒:

static unsigned char wordbits[65536] = { bitcounts of ints between 0 and 65535 };
static int popcount( unsigned int i )
{
    return( wordbits[i&0xFFFF] + wordbits[i>>16] );
}

第二名是相同的代码，但查找的是4个字节而不是2个半字。这花了大约5.5秒。

第三名是“横向加法”法，用时8.6秒。

第四名是GCC的__builtin_popcount()，仅为11秒。

一次一个比特的计数方法要慢得多，我厌倦了等待它完成。

因此，如果您最关心的是性能，那么请使用第一种方法。如果您关心它，但又不想在上面花费64Kb的RAM，那么可以使用第二种方法。否则，请使用可读的(但速度较慢)一次一位的方法。

很难想象在什么情况下你会想要使用比特旋转方法。

编辑:这里也有类似的结果。