以uint为例,当计算过程中(比如9999^6)产生大于UINT_MAX(2^32 - 1)的值的时候,编译时会产生integer overflow,即数值溢出,最后的结果也被截断.
1.如何检测 :https://www.quora.com/How-do-I-prevent-integer-overflow-in-C++(有墙)
贴上里面提供的示例代码:
- #include <iostream>
- #include <cstdlib>
- #include <climits>
- int main() {
- int x = 0x70000000;
- int y = 0x70000000;
- bool underflow = false;
- bool overflow = false;
- if ((x > 0) && (y > INT_MAX - x)) overflow = true;
- if ((x < 0) && (y < INT_MIN - x)) underflow = true;
-
- if(overflow) {
- std::cerr << "overflow will occur\n";
- exit(-1);
- }
-
- if(underflow) {
- std::cerr << "underflow will occur\n";
- exit(-1);
- }
- int z = x + y; // UH OH!
- std::cout << z << "\n";
- }
2.如何解决:可以使用数组或者字符串模拟,例如使用std::string模拟乘法计算.只要内存够,多大的数都存的下(这里类似pure python的处理方式:使用pure python做运算,不会产生overflow,但是使用Numpy则会有overflow的问题,具体可参考:https://mortada.net/can-integer-operations-overflow-in-python.html)
- string bigmul(string a, string b) {
- int size = a.size() + b.size();
- char* res = new char[size];
- memset(res, a.size() + b.size(), 0);
- for (int i = b.size() - 1; i >= 0; i--) {
- for (int j = a.size() - 1; j >= 0; j--) {
- res[i+j+1] += (b[i] - '0')* (a[j] - '0');
- res[i+j] += res[i+j+1]/10;
- res[i+j+1] = res[i+j+1]%10;
- }
- }
- for (int i = 0; i < size; i++) {
- res[i] += '0';
- }
- return res[0] == '0' ? string(res+1) : string(res);
- }