标签：nolimits right 之求 sum wedge 代码优化 整型 align left
在 C/C++ 中, 直接利用 (x + y) >> 1 来计算 \(\left\lfloor {\left( {x + y} \right)/2} \right\rfloor\) (两个整数的平均值并向下取整)以及直接利用 (x + y + 1) >> 1 来计算 \(\left\lceil {\left( {x + y} \right)/2} \right\rceil\) (两个整数的平均值并向上取整)的结果可能有误, 因为 (x + y) >> 1 和 (x + y + 1) >> 1 中的 x + y 可能会发生数值溢出. 而 \(\left\lfloor {\left( {x + y} \right)/2} \right\rfloor\) 和 \(\left\lceil {\left( {x + y} \right)/2} \right\rceil\) 的结果是不可能数值溢出的, 这就引发我们思考可不可能通过某种方式来规避平均值计算中的数值溢出.
方式一
利用如下公式
\(\begin{align} \left\lfloor {\left( {x + y} \right)/2} \right\rfloor = \left\lfloor {x/2} \right\rfloor + \left\lfloor {y/2} \right\rfloor + \left\lfloor {\left( {x\bmod 2 + y\bmod 2} \right)/2} \right\rfloor \hfill \\ \left\lceil {\left( {x + y} \right)/2} \right\rceil = \left\lfloor {x/2} \right\rfloor + \left\lfloor {y/2} \right\rfloor + \left\lceil {\left( {x\bmod 2 + y\bmod 2} \right)/2} \right\rceil \hfill \\ \end{align}\)
下面是对上述两式的证明:
\(\begin{align}
\left\lfloor {\left( {x + y} \right)/2} \right\rfloor &= \left\{ {\begin{array}{*{20}{c}}
{m + n}&{x = 2m,y = 2n} \\
{m + n}&{x = 2m + 1,y = 2n} \\
{m + n}&{x = 2m,y = 2n + 1} \\
{m + n + 1}&{x = 2m + 1,y = 2n + 1}
\end{array}} \right. \\
&= \left\lfloor {x/2} \right\rfloor + \left\lfloor {y/2} \right\rfloor + \left\lfloor {\left( {x\bmod 2 + y\bmod 2} \right)/2} \right\rfloor \\
\end{align}\)
\(\begin{align} \left\lceil {\left( {x + y} \right)/2} \right\rceil &= \left\{ {\begin{array}{*{20}{c}} {m + n}&{x = 2m,y = 2n} \\ {m + n + 1}&{x = 2m + 1,y = 2n} \\ {m + n + 1}&{x = 2m,y = 2n + 1} \\ {m + n + 1}&{x = 2m + 1,y = 2n + 1} \end{array}} \right. \\ &= \left\lfloor {x/2} \right\rfloor + \left\lfloor {y/2} \right\rfloor + \left\lceil {\left( {x\bmod 2 + y\bmod 2} \right)/2} \right\rceil \\ \end{align}\)
其中 \(m,n\) 均为整数.
借用上面的公式可以 \(\left\lfloor {\left( {x + y} \right)/2} \right\rfloor\) 转化为如下的 C/C++ 代码 (据说这段代码还被申请了专利):
(x >> 1) + (y >> 1) + (x & y & 1);
可以将 \(\left\lceil {\left( {x + y} \right)/2} \right\rceil\) 转化为如下的 C/C++ 代码:
(x >> 1) + (y >> 1) + ((x | y) & 1);
这两段代码都不会发生数值溢出.
方式二
设 x 和 y 只能取 0 和 1 值, 则:
| x | y | x + y | x ^ y | x & y | x | y | 2*(x & y) + (x ^ y) | 2*(x | y) - (x ^ y) |
|---|---|---|---|---|---|---|---|
| 0 | 0 | 0 | 0 | 0 | 0 | 0 + 0 = 0 | 0 - 0 = 0 |
| 0 | 1 | 1 | 1 | 0 | 1 | 0 + 1 = 1 | 10 - 1 = 1 |
| 1 | 0 | 1 | 1 | 0 | 1 | 0 + 1 = 1 | 10 - 1 = 1 |
| 1 | 1 | 10 | 0 | 1 | 1 | 10 + 0 = 10 | 10 - 0 = 10 |
注意上表中的 10 是二进制下的 10, 即十进制下的 2, & 是逻辑与操作, | 是逻辑或运算, ^ 是逻辑异或操作.
由上表可见 x + y = 2*(x & y) + (x ^ y) = 2*(x | y) - (x ^ y).
无符号整型
对于无符号整型, 设 \(x = \sum\nolimits_{i = 0}^{n - 1} {{u_i}{2^i}}\) 和 \(y = \sum\nolimits_{i = 0}^{n - 1} {{v_i}{2^i}}\).
\(\begin{align} x + y &= \sum\nolimits_{i = 0}^{n - 1} {{u_i}{2^i}} {\text{ + }}\sum\nolimits_{i = 0}^{n - 1} {{v_i}{2^i}} \\ &= \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i} + {v_i}} \right){2^i}} \\ &= \sum\nolimits_{i = 0}^{n - 1} {\left( {2 \times \left( {{u_i}\& {v_i}} \right) + \left( {{u_i} \wedge {v_i}} \right)} \right){2^i}} \\ &= 2\sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}\& {v_i}} \right){2^i}} + \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i} \wedge {v_i}} \right){2^i}} \\ \end{align}\)
\(\begin{align} \left\lfloor {\left( {x + y} \right)/2} \right\rfloor &= \left\lfloor {\sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}\& {v_i}} \right){2^i}} + \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i} \wedge {v_i}} \right){2^{i - 1}}} } \right\rfloor \\ &= \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}\& {v_i}} \right){2^i}} + \sum\nolimits_{i = 1}^{n - 1} {\left( {{u_i} \wedge {v_i}} \right){2^{i - 1}}} \\ \end{align}\)
上式用 C/C++语言可以表示为:
(x & y) + ((x ^ y) >> 1);
\(\begin{align} x + y &= \sum\nolimits_{i = 0}^{n - 1} {{u_i}{2^i}} {\text{ + }}\sum\nolimits_{i = 0}^{n - 1} {{v_i}{2^i}} \\ &= \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i} + {v_i}} \right){2^i}} \\ &= \sum\nolimits_{i = 0}^{n - 1} {\left( {2 \times \left( {{u_i}|{v_i}} \right) - \left( {{u_i} \wedge {v_i}} \right)} \right){2^i}} \\ &= 2\sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}|{v_i}} \right){2^i}} - \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i} \wedge {v_i}} \right){2^i}} \\ \end{align}\)
\(\begin{align} \left\lceil {\left( {x + y} \right)/2} \right\rceil &= \left\lceil {\sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}|{v_i}} \right){2^i}} - \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i} \wedge {v_i}} \right){2^{i - 1}}} } \right\rceil \\ &= \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}|{v_i}} \right){2^i}} - \sum\nolimits_{i = 1}^{n - 1} {\left( {{u_i} \wedge {v_i}} \right){2^{i - 1}}} \\ \end{align}\)
上式用 C/C++ 语言可以表示为:
(x | y) - ((x ^ y) >> 1);
有符号整型
对于有符号整型, 设 \(x = - {u_{n - 1}}{2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {{u_i}{2^i}}\) 和 \(y = - {v_{n - 1}}{2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {{v_i}{2^i}}\).
\(\begin{align} x + y &= - {u_{n - 1}}{2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {{u_i}{2^i}} - {v_{n - 1}}{2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {{v_i}{2^i}} \\ &= - \left( {{u_{n - 1}} + {v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {\left( {{u_i} + {v_i}} \right){2^i}} \\ &= - \left( {2 \times \left( {{u_{n - 1}}\& {v_{n - 1}}} \right) + \left( {{u_{n - 1}} \wedge {v_{n - 1}}} \right)} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {\left( {2 \times \left( {{u_i}\& {v_i}} \right) + \left( {{u_i} \wedge {v_i}} \right)} \right){2^i}} \\ &= 2\left( { - \left( {{u_{n - 1}}\& {v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}\& {v_i}} \right){2^i}} } \right) + \left( { - \left( {{u_{n - 1}} \wedge {v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {\left( {{u_i} \wedge {v_i}} \right){2^i}} } \right) \\ \end{align}\)
\(\begin{align} \left\lfloor {\left( {x + y} \right)/2} \right\rfloor &= \left\lfloor {\left( { - \left( {{u_{n - 1}}\& {v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}\& {v_i}} \right){2^i}} } \right) + \left( { - \left( {{u_{n - 1}} \wedge {v_{n - 1}}} \right){2^{n - 2}} + \sum\nolimits_{i = 0}^{n - 2} {\left( {{u_i} \wedge {v_i}} \right){2^{i - 1}}} } \right)} \right\rfloor \\ &= \left( { - \left( {{u_{n - 1}}\& {v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}\& {v_i}} \right){2^i}} } \right) + \left( { - \left( {{u_{n - 1}} \wedge {v_{n - 1}}} \right){2^{n - 2}} + \sum\nolimits_{i = 1}^{n - 2} {\left( {{u_i} \wedge {v_i}} \right){2^{i - 1}}} } \right) \\ \end{align}\)
上式用 C/C++ 语言可以表示为:
(x & y) + ((x ^ y) >> 1);
\(\begin{align} x + y &= - {u_{n - 1}}{2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {{u_i}{2^i}} - {v_{n - 1}}{2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {{v_i}{2^i}} \\ &= - \left( {{u_{n - 1}} + {v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {\left( {{u_i} + {v_i}} \right){2^i}} \\ &= - \left( {2 \times \left( {{u_{n - 1}}|{v_{n - 1}}} \right) - \left( {{u_{n - 1}} \wedge {v_{n - 1}}} \right)} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {\left( {2 \times \left( {{u_i}|{v_i}} \right) - \left( {{u_i} \wedge {v_i}} \right)} \right){2^i}} \\ &= 2\left( { - \left( {{u_{n - 1}}|{v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}|{v_i}} \right){2^i}} } \right) - \left( { - \left( {{u_{n - 1}} \wedge {v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 2} {\left( {{u_i} \wedge {v_i}} \right){2^i}} } \right) \\ \end{align}\)
\(\begin{align} \left\lceil {\left( {x + y} \right)/2} \right\rceil &= \left\lceil {\left( { - \left( {{u_{n - 1}}|{v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}|{v_i}} \right){2^i}} } \right) - \left( { - \left( {{u_{n - 1}} \wedge {v_{n - 1}}} \right){2^{n - 2}} + \sum\nolimits_{i = 0}^{n - 2} {\left( {{u_i} \wedge {v_i}} \right){2^{i - 1}}} } \right)} \right\rceil \\ &= \left( { - \left( {{u_{n - 1}}|{v_{n - 1}}} \right){2^{n - 1}} + \sum\nolimits_{i = 0}^{n - 1} {\left( {{u_i}|{v_i}} \right){2^i}} } \right) - \left( { - \left( {{u_{n - 1}} \wedge {v_{n - 1}}} \right){2^{n - 2}} + \sum\nolimits_{i = 1}^{n - 2} {\left( {{u_i} \wedge {v_i}} \right){2^{i - 1}}} } \right) \\ \end{align}\)
上式用 C/C++ 语言可以表示为:
(x | y) - ((x ^ y) >> 1);
综合
综合上面的分析, 可见对于有符号整型和无符号整型,
\(\left\lceil {\left( {x + y} \right)/2} \right\rceil\) 都可以用 C/C++ 语言表示为:
(x & y) + ((x ^ y) >> 1);
\(\left\lceil {\left( {x + y} \right)/2} \right\rceil\) 都可以用 C/C++ 语言表示为:
(x | y) - ((x ^ y) >> 1);
参考:
版权声明
版权声明：自由分享，保持署名-非商业用途-非衍生，知识共享3.0协议。
如果你对本文有疑问或建议，欢迎留言！转载请保留版权声明！
如果你觉得本文不错, 也可以用微信赞赏一下哈.
标签：nolimits,right,之求,sum,wedge,代码优化,整型,align,left 来源： https://www.cnblogs.com/quarryman/p/12989553.html
本站声明： 1. iCode9 技术分享网（下文简称本站）提供的所有内容，仅供技术学习、探讨和分享； 2. 关于本站的所有留言、评论、转载及引用，纯属内容发起人的个人观点，与本站观点和立场无关； 3. 关于本站的所有言论和文字，纯属内容发起人的个人观点，与本站观点和立场无关； 4. 本站文章均是网友提供，不完全保证技术分享内容的完整性、准确性、时效性、风险性和版权归属；如您发现该文章侵犯了您的权益，可联系我们第一时间进行删除； 5. 本站为非盈利性的个人网站，所有内容不会用来进行牟利，也不会利用任何形式的广告来间接获益，纯粹是为了广大技术爱好者提供技术内容和技术思想的分享性交流网站。