Question

这是我的代码

using namespace std;
#include <iostream>
#include <stdio.h>
#include <stdlib.h>
#define N 8000

void fillArray(int *data, int count){
    for(int i =0; i < count; i++)
        data[i] = (int) rand() / ((int) RAND_MAX);
}

__global__ void add(int* a, int *b){
    int add = 0;

    int tid = threadIdx.x + blockIdx.x * blockDim.x;
    if(tid < N){
        add = a[tid] + b[tid];
    }
}

__global__ void subtract(int* a, int *b){
    int subtract = 0;

    int tid = threadIdx.x + blockIdx.x * blockDim.x;
    if(tid < N){
        subtract = a[tid] - b[tid];
    }
}

float duration(int *devA, int *devB, int blocksPerGrid, int threadsPerBlock){

    cudaEvent_t start, stop;
    cudaEventCreate(&start);
    cudaEventCreate(&stop);
    cudaEventRecord(start,0);
    cudaEventRecord(stop,0);
    cudaEventSynchronize(stop);

    cudaMalloc((void**) &devA, N * sizeof(int));
    cudaMalloc((void**) &devB, N * sizeof(int));

    add<<<blocksPerGrid, threadsPerBlock>>>(devA,devB);

    float elapsedTime;
    cudaEventElapsedTime(&elapsedTime,start,stop);
    cudaEventDestroy(start);
    cudaEventDestroy(stop);

    return elapsedTime;
}



int main(void) {

    int *a = new int(N);
    int *b = new int(N);
    float dur = 0 ;



    fillArray(a, N);
    fillArray(b, N);

    dur =  duration(a,b,N,1);

    cout << "Global memory version:\n";
    cout << "Process completed in " << dur;
    cout << "for a data set of " << N << " integers.";


    return 0;
}

正如您所看到的，我在CPU端填充了fillArray函数。但填充数组函数给出错误：

malloc.c 3906 : sYSMalloc: Assertion bla bla

我在这里缺少什么？我只是尝试填充数组。我可能会遇到什么问题？事件，如果我删除持续时间功能中的添加功能，我得到此错误。这有什么问题？

Answer 1

错误在于创建a和b数组。正如@QWR和@talonmies所说，使用valgrind（或任何Windows substitute）可以帮助您找到此类错误的来源：

==8288== Invalid write of size 4
==8288==    at 0x400DD2: fillArray(int*, int) (kernel.cu:11)
==8288==    by 0x400F79: main (kernel.cu:63)
==8288==  Address 0x62783e4 is 0 bytes after a block of size 4 alloc'd
==8288==    at 0x4C2BA77: operator new(unsigned long) (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==8288==    by 0x400F41: main (kernel.cu:57)
==8288== 
==8288== Invalid write of size 4
==8288==    at 0x400DD2: fillArray(int*, int) (kernel.cu:11)
==8288==    by 0x400F8A: main (kernel.cu:64)
==8288==  Address 0x6278434 is 0 bytes after a block of size 4 alloc'd
==8288==    at 0x4C2BA77: operator new(unsigned long) (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==8288==    by 0x400F55: main (kernel.cu:58)

如果你改变：

int *a = new int(N);
int *b = new int(N);

要：

int *a = new int[N];
int *b = new int[N];

错误消失。实际上，你没有为数组分配内存，只是为了一个整数。

调试CUDA代码时，需要使用GPU /设备调试工具（cuda-memcheck，cuda-gdb）和CPU /主机工具（valgrind），因为错误可能发生在GPU和CPU都有。不要忘记使用nvcc：-G的两个调试标志进行编译以获取设备代码，并使用-g进行编译以获取主机代码。

对于good practice，您还应该在主页末尾delete使用数组：

delete [] a;
delete [] b;

当我尝试填充我的数组时，我收到malloc 3906错误

1 个答案: