为什么不迭代地除以2呢?

count = 0
while n > 0
  if (n % 2) == 1
    count += 1
  n /= 2  

我同意这不是最快的，但是“最好”这个词有点含糊不清。我认为“最好”应该有一个清晰的元素

下面是示例代码，可能很有用。

private static final int[] bitCountArr = new int[]{0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4, 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5, 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5, 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5, 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7, 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5, 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7, 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6, 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7, 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7, 4, 5, 5, 6, 5, 6, 6, 7, 5, 6, 6, 7, 6, 7, 7, 8};
private static final int firstByteFF = 255;
public static final int getCountOfSetBits(int value){
    int count = 0;
    for(int i=0;i<4;i++){
        if(value == 0) break;
        count += bitCountArr[value & firstByteFF];
        value >>>= 8;
    }
    return count;
}

对于那些想要在c++ 11中为任何无符号整数类型作为consexpr函数的人(tacklelib/include/tacklelib/utility/math.hpp):

#include <stdint.h>
#include <limits>
#include <type_traits>

const constexpr uint32_t uint32_max = (std::numeric_limits<uint32_t>::max)();

namespace detail
{
    template <typename T>
    inline constexpr T _count_bits_0(const T & v)
    {
        return v - ((v >> 1) & 0x55555555);
    }

    template <typename T>
    inline constexpr T _count_bits_1(const T & v)
    {
        return (v & 0x33333333) + ((v >> 2) & 0x33333333);
    }

    template <typename T>
    inline constexpr T _count_bits_2(const T & v)
    {
        return (v + (v >> 4)) & 0x0F0F0F0F;
    }

    template <typename T>
    inline constexpr T _count_bits_3(const T & v)
    {
        return v + (v >> 8);
    }

    template <typename T>
    inline constexpr T _count_bits_4(const T & v)
    {
        return v + (v >> 16);
    }

    template <typename T>
    inline constexpr T _count_bits_5(const T & v)
    {
        return v & 0x0000003F;
    }

    template <typename T, bool greater_than_uint32>
    struct _impl
    {
        static inline constexpr T _count_bits_with_shift(const T & v)
        {
            return
                detail::_count_bits_5(
                    detail::_count_bits_4(
                        detail::_count_bits_3(
                            detail::_count_bits_2(
                                detail::_count_bits_1(
                                    detail::_count_bits_0(v)))))) + count_bits(v >> 32);
        }
    };

    template <typename T>
    struct _impl<T, false>
    {
        static inline constexpr T _count_bits_with_shift(const T & v)
        {
            return 0;
        }
    };
}

template <typename T>
inline constexpr T count_bits(const T & v)
{
    static_assert(std::is_integral<T>::value, "type T must be an integer");
    static_assert(!std::is_signed<T>::value, "type T must be not signed");

    return uint32_max >= v ?
        detail::_count_bits_5(
            detail::_count_bits_4(
                detail::_count_bits_3(
                    detail::_count_bits_2(
                        detail::_count_bits_1(
                            detail::_count_bits_0(v)))))) :
        detail::_impl<T, sizeof(uint32_t) < sizeof(v)>::_count_bits_with_shift(v);
}

谷歌测试库中的附加测试:

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

namespace {
    template <typename T>
    inline uint32_t _test_count_bits(const T & v)
    {
        uint32_t count = 0;
        T n = v;
        while (n > 0) {
            if (n % 2) {
                count += 1;
            }
            n /= 2;
        }
        return count;
    }
}

TEST(FunctionsTest, random_count_bits_uint32_100K)
{
    srand(uint_t(time(NULL)));
    for (uint32_t i = 0; i < 100000; i++) {
        const uint32_t r = uint32_t(rand()) + (uint32_t(rand()) << 16);
        ASSERT_EQ(_test_count_bits(r), count_bits(r));
    }
}

TEST(FunctionsTest, random_count_bits_uint64_100K)
{
    srand(uint_t(time(NULL)));
    for (uint32_t i = 0; i < 100000; i++) {
        const uint64_t r = uint64_t(rand()) + (uint64_t(rand()) << 16) + (uint64_t(rand()) << 32) + (uint64_t(rand()) << 48);
        ASSERT_EQ(_test_count_bits(r), count_bits(r));
    }
}

对于JavaScript，你可以使用一个查找表来计算一个32位值的设置位的数量(这段代码可以很容易地翻译成C语言)。此外，添加了8位和16位版本，以供通过网络搜索查找的人使用。

const COUNT_BITS_TABLE = makeLookupTable() function makeLookupTable() { const table = new Uint8Array(256) for (let i = 0; i < 256; i++) { table[i] = (i & 1) + table[(i / 2) | 0]; } return table } function countOneBits32(n) { return COUNT_BITS_TABLE[n & 0xff] + COUNT_BITS_TABLE[(n >> 8) & 0xff] + COUNT_BITS_TABLE[(n >> 16) & 0xff] + COUNT_BITS_TABLE[(n >> 24) & 0xff]; } function countOneBits16(n) { return COUNT_BITS_TABLE[n & 0xff] + COUNT_BITS_TABLE[(n >> 8) & 0xff] } function countOneBits8(n) { return COUNT_BITS_TABLE[n & 0xff] } console.log('countOneBits32', countOneBits32(0b10101010000000001010101000000000)) console.log('countOneBits32', countOneBits32(0b10101011110000001010101000000000)) console.log('countOneBits16', countOneBits16(0b1010101000000000)) console.log('countOneBits8', countOneBits8(0b10000010))

一个快速的c#解决方案，使用预先计算的字节位计数表，并根据输入大小进行分支。

public static class BitCount
{
    public static uint GetSetBitsCount(uint n)
    {
        var counts = BYTE_BIT_COUNTS;
        return n <= 0xff ? counts[n]
             : n <= 0xffff ? counts[n & 0xff] + counts[n >> 8]
             : n <= 0xffffff ? counts[n & 0xff] + counts[(n >> 8) & 0xff] + counts[(n >> 16) & 0xff]
             : counts[n & 0xff] + counts[(n >> 8) & 0xff] + counts[(n >> 16) & 0xff] + counts[(n >> 24) & 0xff];
    }

    public static readonly uint[] BYTE_BIT_COUNTS =
    {
        0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4,
        1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
        1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
        2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
        1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
        2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
        2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
        3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
        1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
        2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
        2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
        3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
        2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
        3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
        3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
        4, 5, 5, 6, 5, 6, 6, 7, 5, 6, 6, 7, 6, 7, 7, 8
    };
}

我认为Brian Kernighan的方法也很有用… 它的迭代次数和设置位个数一样多。因此，如果我们有一个32位的单词，只设置了高位，那么它将只经过一次循环。

int countSetBits(unsigned int n) { 
    unsigned int n; // count the number of bits set in n
    unsigned int c; // c accumulates the total bits set in n
    for (c=0;n>0;n=n&(n-1)) c++; 
    return c; 
}

出版于1988年的C编程语言第二版(由Brian W. Kernighan和Dennis M. Ritchie编写)在练习2-9中提到了这一点。2006年4月19日，Don Knuth向我指出，这种方法“是由Peter Wegner在CACM 3(1960)， 322中首次发表的。(同样由德里克·莱默(Derrick Lehmer)独立发现，并于1964年在贝肯巴赫(Beckenbach)编辑的一本书中出版。)