列举两个序列的交集

huangapple go评论76阅读模式
英文:

Enumerating intersections of two sequences

问题

I have two sequences:

  1. 一个无符号64位数 k 的倍数,例如 k=5{0, 5, 10, 15, 20, 25, ...}
  2. 64位掩码 m 的子集,加上64位常数 C,例如 m=0b1011{C+0, C+1, C+2, C+3, C+8, C+9, C+10, C+11}。注意:Cm 不重叠,即 (m & C) = 0

除了逐个检查 k 的每个倍数并检查是否符合第二个序列的条件,或者检查掩码 m 的每个子集并与 C 进行 OR 运算,然后检查是否能被 k 整除之外,还有其他方法可以列举这两个序列的交集吗?

64位数 P 在这两个序列的交集中,如果 P % k == 0 并且 (P & C) == C 并且 (P & ~(C | m)) = 0

注意:实际上很容易枚举第二个序列:

uint64_t m = ...; uint64_t C = ...;
uint64_t subset = 0;
do {
    uint64_t value = subset + C;
    print(value);
    // if (value % 5)   => in the intersection of the sequences
    subset = (subset - m) & m;
} while (subset != 0);

编辑:枚举的顺序无关紧要。

此外:我不一定希望枚举它们全部,但我想要一种迭代器,可以始终推进并获取下一个交集。

英文:

I have two sequences:

  1. Multiples of an unsigned 64-bit number k, example for k=5: {0, 5, 10, 15, 20, 25, ...}
  2. Subsets of the 64-bit mask m plus 64-bit constant C, example for m=0b1011: {C+0, C+1, C+2, C+3, C+8, C+9, C+10, C+11}. Note: C is disjoint from m, aka (m & C) = 0.

Aside from checking each and every multiple of k and checking if it matches the second sequence criteria, or checking every subset of the mask m and ORing with C and checking for division with k, are there any other ways I can enumerate all the intersection of those two sequences?

A 64-bit number P is in the intersection of those two sequences if P % k == 0 and (P & C) == C and (P & ~(C | m)) = 0.

Note: Enumerating over the second sequence is actually really easy:

uint64_t m = ..., C = ...;
uint64_t subset = 0;
do {
    uint64_t value = subset + C;
    print(value);
    // if (value % 5)   => in the intersection of the sequences
    subset = (subset - m) & m;
} while (subset != 0);

Edit: The order of enumeration doesn't matter

Also: I don't exactly hope to enumerate them all, but I want to have some sort of iterator which I can always advance and get the next intersection

答案1

得分: 1

C mod k 是一个常数,表示我们正在寻找满足对于模 k 而言其结果为特定值的 m 的子集 — 即当其与 C mod k 相加后结果为零模 k 的子集。

x = (k - (C mod k)) mod k

m 中的每个比特作为 2 的幂对于模 k 也是常数。因此,任务似乎是枚举那些总和为 xk 的组合。

例如,对于已见过的组合,可以通过将当前在 m 中的幂对 k 取模来生成新的组合。一个即时的优化可能是将在 m 中与 k 对模相等的幂进行分组。除此之外,我不知道有什么技巧可以加速这种搜索 —— 对于模 k 的组合状态数量可能取决于 km 两者。

英文:

C mod k is a constant that means we are looking for subsets of m that achieve one and only one value mod k — the one that when added to C mod k is zero mod k.

x = (k - (C mod k)) mod k

Each of the bits of m as a power of 2 is also constant mod k. So the task seems to be enumerating those combinations that sum to x mod k.

For example, for each of seen combinations, generate a new one by adding the current power of 2 in m mod k. One immediate optimisation could be to group the powers of 2 in m that are equal mod k. Other than that, I don't know of a trick to speed that kind of search up -- the number of combination states mod k could depend on both k and m.

huangapple
  • 本文由 发表于 2023年6月18日 21:07:18
  • 转载请务必保留本文链接:https://go.coder-hub.com/76500702.html
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定