0

我正在尝试在 CBMC 中获取数组的所有排列。对于小情况,例如 [1,2,3],我想我可以写

i1 = nondet()
i2 = nondet()
i3 = nondet()
assume (i > 0 && i < 4); ...
assume (i1 != i2 && i2 != i3 && i1 != i3);
// do stuffs with i1,i2,i3

但是对于较大的元素,代码会非常混乱。所以我的问题是有没有更好/通用的方式来表达这一点?

4

1 回答 1

1

基于 Craig 的使用数组的建议,您可以循环遍历要置换的值,并以节点方式选择尚未被占用的位置。例如,像这样的循环(其中所有值的序列都预初始化为 -1)。

for(int i = 1; i <= count; ++i) {
  int nondet;
  assume(nondet >= 0 && nondet < count);
  // ensure we don't pick a spot already picked
  assume(sequence[nondet] == -1); 
  sequence[nondet] = i;
}

所以一个完整的程序看起来像这样:

#include <assert.h>
#include <memory.h>

int sum(int *array, int count) {
    int total = 0;
    for(int i = 0; i < count; ++i) {
        total += array[i];
    }
    return total;
}

int main(){

    int count = 5; // 1, ..., 6
    int *sequence = malloc(sizeof(int) * count);

    // this isn't working - not sure why, but constant propagator should
    // unroll the loop anyway
    // memset(sequence, -1, count);
    for(int i = 0; i < count; ++i) {
        sequence[i] = -1;
    }

    assert(sum(sequence, count) == -1 * count);

    for(int i = 1; i <= count; ++i) {
      int nondet;
      __CPROVER_assume(nondet >= 0);
      __CPROVER_assume(nondet < count);
      __CPROVER_assume(sequence[nondet] == -1); // ensure we don't pick a spot already picked
      sequence[nondet] = i;
    }

    int total = (count * (count + 1)) / 2;
    // verify this is a permuation
    assert(sum(sequence, count) == total);
}

但是,对于大于 6 的值,这非常慢(尽管我没有将它与您的方法进行比较 - 它不会被卡住展开,它会卡住解决)。

于 2020-05-07T12:02:46.803 回答