警告：'dot_prod' 可能在此函数中使用未初始化 [-Wmaybe-uninitialized]

Question

我一直在修改这段代码，但我似乎无法让它工作。当我运行它生成的错误文件时，我收到警告

dotp.c:39:11: warning: ‘dot_prod’ may be used uninitialized in this function [-Wmaybe-uninitialized]
   39 |   #pragma omp parallel for reduction(+: dot_prod) num_threads(num_td)
      |           ^~~

我不明白如何初始化 dot_prod 或者如果这样可以修复错误，谁能帮我修复它？非常感谢。

#include <omp.h>
#include <stdio.h>
#include <time.h>
#include <stdlib.h>
#include <math.h>


int main(int argc, char **argv){
        
    int exe_mode, num_td, vec_size;
    
    
    sscanf(argv[0], "%d", &exe_mode);
    sscanf(argv[1], "%d", &num_td);
    sscanf(argv[2], "%d", &vec_size);

    srand(1);

    int a[vec_size];
    int b[vec_size];
    int dot_prod,i;
    
    double start = omp_get_wtime();
    
    //initializing the vectors
    for( i = 0; i < vec_size; i++){ 
        a[i] = (int) (rand() % vec_size - vec_size/2);
        b[i] = (int) (rand() % vec_size - vec_size/2);
    }
    
    //Sequential execution
    if (exe_mode == 1){
        for( i = 0; i<vec_size;i++){
            dot_prod += a[i] * b[i];
        }
    }   
    //Parallel execution
    if (exe_mode == 2){
        #pragma omp parallel for reduction(+: dot_prod) num_threads(num_td)
        for(i=0; i<vec_size;i++){
            dot_prod += a[i] * b[i];
        }
    }
    //Parallel execution with vectorization
    if (exe_mode == 3){
        #pragma omp simd reduction(+: dot_prod)
        for( i=0; i<vec_size;i++){
            dot_prod += a[i] * b[i];
        }
    double runtime = omp_get_wtime()-start;
    printf("%.4e\t%i\n", runtime ,dot_prod);    
    }
    
    
    
    return 0;
}

/**void dotproduct(int exe_mode, int num_td, int vec_size){
    
    srand(1);
    
    int a[vec_size];
    int b[vec_size];
    int dot_prod;
    
    //initializing the vectors
    for(int i = 0; i < vec_size; i++){ 
        a[i] = (int) (rand() % vec_size - vec_size/2);
        b[i] = (int) (rand() % vec_size - vec_size/2);
    }
    
    //Sequential execution
    if (exe_mode == 1){
        for(int i=0; i<vec_size;i++){
            dot_prod += a[i] * b[i];
        }
    }   
    //Parallel execution
    if (exe_mode == 2){
        #pragma omp parallel for num_threads(num_td)
        for(int i=0; i<vec_size;i++){
            dot_prod += a[i] * b[i];
        }
    }
    //Parallel execution with vectorization
    if (exe_mode == 3){
        #pragma omp parallel for simd num_threads(num_td)
        for(int i=0; i<vec_size;i++){
            dot_prod += a[i] * b[i];
        }
    double runtime = omp_get_wtime();   
    printf("%.4e\t%i\n", runtime ,dot_prod);    
    }
}**/**strong text**

Answer 1

将 dot_prod 的声明更改为：

int dot_prod = 0;
int i;

Answer 2

dot_prod += (numeric expression) 是未定义的行为，如果你还没有初始化 dot_prod。

您初始化 dot_prod 的方式与初始化任何其他 int 变量的方式相同；通过将其设置为某个 pre-determined 值，例如零。

int dot_prod = 0;

如果没有 初始赋值， dot_prod 可以在声明时包含任何 int 值。您的 C 编译器只需要为您的变量分配一些内存；不需要在那里设置一些合理的初始值。你必须自己做。

警告：'dot_prod' 可能在此函数中使用未初始化 [-Wmaybe-uninitialized]

warning: ‘dot_prod’ may be used uninitialized in this function [-Wmaybe-uninitialized]

c

linux

openmp