我正在编写一个帮助程序类,打算用于从数据块中反向读取位。

我尝试做一个优化,使用“rol”指令掩盖数据。但是,令我惊讶的是,这实际上比每次访问期间创建新的位掩码要慢。

class reverse_bit_reader
{
public:
    static const size_t bits_per_block = sizeof(unsigned long)*8;
    static const size_t high_bit = 1 << (bits_per_block-1);

    reverse_bit_reader(const void* data, size_t size)
        : data_(reinterpret_cast<const unsigned long*>(data))
        , index_(size-1)
    {
            // Bits are stored in left to right order, potentially ignore the last bits
        size_t last_bit_index = index_ % bits_per_block;
        bit_mask_ = high_bit >> (last_bit_index+1);
        if(bit_mask_ == 0)
            bit_mask_ = high_bit;
    }

    bool next_bit1()
    {
        return get_bit(index_--);
    }

    bool next_bit2() // Why is next_bit1 faster?
    {
        __asm // Rotate bit_mask.
        {
            mov eax, [ecx+0];
            rol eax, 1;
            mov [ecx+0], eax;
        }
        return data_[index_-- / bits_per_block] & bit_mask_;
    }

    bool eof() const{return index_ < 0;}
private:

    bool get_bit(size_t index) const
    {
        const size_t block_index = index / bits_per_block;
        const size_t bit_index = index % bits_per_block;
        const size_t bit_mask = high_bit >> bit_index;
        return data_[block_index] & bit_mask;
    }

    unsigned long bit_mask_;
    int index_;
    const unsigned long* data_;
};

谁能解释为什么next_bit1比next_bit2快?

最佳答案

如果您要从最高有效位开始从长整数开始顺序读取位,并且您希望它尽可能快,那么您可以按照这些思路做些什么吗?

#define GETBIT ((theBit = (theLong < 0)), (theLong <<= 1), theBit)

关于c++ - 读位性能,我们在Stack Overflow上找到一个类似的问题:https://stackoverflow.com/questions/5779133/

10-09 05:58