代码

//written by r1
#include <immintrin.h>
#include <string.h>

// 启用AVX2指令集支持
#pragma GCC target("avx2")

// 每个step处理n个64字节块（n*64字节）
static inline void process_blocks(int blocks, const char** p1, const char** p2, unsigned* sum) {
    __m256i total = _mm256_setzero_si256();
    const __m256i ones = _mm256_set1_epi8(1);
    
    for (int i = 0; i < blocks; ++i) {
        // 加载64字节数据（分两个32字节块）
        __m256i a1 = _mm256_loadu_si256((const __m256i*)(*p1));
        __m256i a2 = _mm256_loadu_si256((const __m256i*)(*p2));
        __m256i b1 = _mm256_loadu_si256((const __m256i*)(*p1 + 32));
        __m256i b2 = _mm256_loadu_si256((const __m256i*)(*p2 + 32));

        // 生成比较掩码
        __m256i cmp1 = _mm256_cmpeq_epi8(a1, a2);
        __m256i cmp2 = _mm256_cmpeq_epi8(b1, b2);
        
        // 转换为位掩码并统计
        uint32_t mask = _mm256_movemask_epi8(cmp1) | 
                       ((uint64_t)_mm256_movemask_epi8(cmp2) << 32);
        *sum += __builtin_popcountll(mask);

        *p1 += 64;
        *p2 += 64;
    }
}

void solve(int n, int q, char* s1, char* s2, int* q_x, int* q_y, int* q_len, unsigned* ans) {
    // 预处理s1：0->1, 1->2, 2->0
    for (int i = 0; i < n; ++i) {
        s1[i] = "120"[s1[i] - '0'];
    }

    // 处理每个查询
    for (int k = 0; k < q; ++k) {
        const char* p1 = s1 + q_x[k];
        const char* p2 = s2 + q_y[k];
        unsigned len = q_len[k];
        unsigned sum = 0;

        // 处理完整64字节块（每次最多处理255块防止溢出）
        while (len >= 64) {
            int blocks = len / 64;
            if (blocks > 255) blocks = 255;
            
            process_blocks(blocks, &p1, &p2, &sum);
            len -= blocks * 64;
        }

        // 处理剩余字节（<64）
        for (unsigned i = 0; i < len; ++i) {
            sum += (p1[i] == p2[i]);
        }

        ans[k] = sum;
    }
}

评测结果