#define NUM_THREADS 4

typedef struct {
    uint8_t *dst;
    const uint8_t *src;
    size_t size;
} ThreadData;

void *thread_copy(void *arg) {
    ThreadData *data = (ThreadData *)arg;
    memcpy(data->dst, data->src, data->size);
    return NULL;
}

void parallel_memcpy(uint8_t *dst, const uint8_t *src, size_t size) {
    pthread_t threads[NUM_THREADS];
    ThreadData thread_data[NUM_THREADS];
    size_t chunk_size = size / NUM_THREADS;
    size_t remaining = size % NUM_THREADS;

    for (int i = 0; i < NUM_THREADS; i++) {
        thread_data[i].dst = dst + i * chunk_size;
        thread_data[i].src = src + i * chunk_size;
        thread_data[i].size = chunk_size;
        if (i == NUM_THREADS - 1) {
            thread_data[i].size += remaining;
        }
        pthread_create(&threads[i], NULL, thread_copy, &thread_data[i]);
    }

    for (int i = 0; i < NUM_THREADS; i++) {
        pthread_join(threads[i], NULL);
    }
}

parallel_memcpy((uint8_t *)g_if_v_ctl._big_yuv_frame_buf.frame_vir_addr, (const uint8_t *)YuvBigImage, _FRAME_BIG_BUFF_SIZE);

利用多线程加速大块数据的拷贝过程。

最好在多核，大数据块的情况下使用该操作，否则可能会起到反作用。

可以探索使用指令集拷贝的方式，有不错的效果。