电脑：这才是世界上最绝望的死法

userId_undefined

‮༺ཌༀ猫条一ༀད༻

秩序白银

2025-02-19 19:32:58

发布于：江苏

150阅读
13回复
2点赞

注：所有代码均有AI书写，本人一概不负责（AI预估CPU会烧到95摄氏度，慎用!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!


#include <iostream>
#include <vector>
#include <thread>
#include <atomic>
#include <Windows.h>
#include <immintrin.h> // SIMD指令

using namespace std;

constexpr int W = 120;   // 控制台宽度
constexpr int H = 40;    // 控制台高度
constexpr int NUM_PARTICLES = 500000; // 粒子数量
constexpr float GRAVITY = 0.0005f;    // 重力加速度

// 内存对齐结构体 (32字节对齐)
struct alignas(32) Particle {
    __m256 pos_x;   // 8个粒子的X坐标
    __m256 pos_y;   // 8个粒子的Y坐标
    __m256 vel_x;
    __m256 vel_y;
    __m256 color;
};

vector<Particle*> particles; // 粒子数组
atomic<bool> running{true};

// SIMD重力计算
void apply_gravity(__m256& vel_y) {
    vel_y = _mm256_add_ps(vel_y, _mm256_set1_ps(GRAVITY));
}

// 多线程更新函数
void update_thread(int start, int end) {
    const __m256 bounds_x = _mm256_set1_ps(W-1);
    const __m256 bounds_y = _mm256_set1_ps(H-1);
    const __m256 damp = _mm256_set1_ps(-0.7f);
    
    for (int i = start; i < end; ++i) {
        Particle& p = *particles[i];
        
        // 应用重力
        apply_gravity(p.vel_y);
        
        // 更新位置
        p.pos_x = _mm256_add_ps(p.pos_x, p.vel_x);
        p.pos_y = _mm256_add_ps(p.pos_y, p.vel_y);
        
        // 边界碰撞检测
        __m256 mask_x = _mm256_cmp_ps(p.pos_x, bounds_x, _CMP_GT_OQ);
        p.pos_x = _mm256_blendv_ps(p.pos_x, bounds_x, mask_x);
        p.vel_x = _mm256_blendv_ps(p.vel_x, _mm256_mul_ps(p.vel_x, damp), mask_x);
        
        __m256 mask_y = _mm256_cmp_ps(p.pos_y, bounds_y, _CMP_GT_OQ);
        p.pos_y = _mm256_blendv_ps(p.pos_y, bounds_y, mask_y);
        p.vel_y = _mm256_blendv_ps(p.vel_y, _mm256_mul_ps(p.vel_y, damp), mask_y);
    }
}

void render(HANDLE hConsole) {
    // 使用内存映射直接操作控制台缓冲区
    CHAR_INFO buffer[H][W] = {0};
    
    for (auto& p : particles) {
        float* px = (float*)&p->pos_x;
        float* py = (float*)&p->pos_y;
        float* clr = (float*)&p->color;
        
        for (int i = 0; i < 8; ++i) { // 处理每个SIMD包中的8个粒子
            int x = min(max((int)px[i], 0), W-1);
            int y = min(max((int)py[i], 0), H-1);
            
            buffer[y][x].Char.UnicodeChar = 0x2588;
            buffer[y][x].Attributes = (int)(clr[i] * 255) % 0xFF;
        }
    }
    
    // 批量写入控制台
    COORD size = {W, H};
    COORD coord = {0, 0};
    SMALL_RECT rect = {0, 0, W-1, H-1};
    WriteConsoleOutput(hConsole, (CHAR_INFO*)buffer, size, coord, &rect);
}

int main() {
    HANDLE hConsole = CreateConsoleScreenBuffer(
        GENERIC_READ | GENERIC_WRITE, 0, NULL, CONSOLE_TEXTMODE_BUFFER, NULL);
    SetConsoleActiveScreenBuffer(hConsole);
    
    // 预分配内存（对齐内存）
    const int num_blocks = (NUM_PARTICLES + 7) / 8;
    particles.resize(num_blocks);
    for (auto& p : particles) {
        p = (Particle*)_aligned_malloc(sizeof(Particle), 32);
        // 初始化SIMD数据...
    }
    
    // 启动多个工作线程
    const int num_threads = thread::hardware_concurrency();
    vector<thread> threads;
    int per_thread = num_blocks / num_threads;
    
    while (running) {
        auto start = chrono::high_resolution_clock::now();
        
        // 并行更新
        for (int i = 0; i < num_threads; ++i) {
            int start = i * per_thread;
            int end = (i == num_threads-1) ? num_blocks : start + per_thread;
            threads.emplace_back(update_thread, start, end);
        }
        
        // 等待所有线程完成
        for (auto& t : threads) t.join();
        threads.clear();
        
        // 渲染
        render(hConsole);
        
        // 限帧
        auto end = chrono::high_resolution_clock::now();
        auto duration = chrono::duration_cast<chrono::milliseconds>(end - start);
        if (duration.count() < 33) {
            this_thread::sleep_for(chrono::milliseconds(33 - duration.count()));
        }
    }
    
    // 清理内存
    for (auto p : particles) _aligned_free(p);
    return 0;
}
userId_undefined
去预览
0/2000

全部评论 1

热门讨论