CUDA, Exception thrown: *.exe has triggered a breakpoint. occurred

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello,

I have this code, I don’t know what is going on, it is just a simple code for CUDA, but when I try to debug or just run it, it always gives this error.

aaa.exe has triggered a breakpoint. occurred

It says that it stopped at exceptions, “An exception was caught by the debugger, and user settings indicate that break should occur”

I have tried to mark and unmark several exceptions, but nothing seems to work.

At any time I try to access a CUDA function, in this case, it happens when it reaches line 54.

I am not sure if this is a bug or what, I have tried VS pro (2015, 2017, 2019), CUDA 10 and CUDA 10.1 with all its combinations. But I’m still getting this error.

I am working on Windows 10.0.17763 (Build 17763)

I actually worked on more complicated codes a few months ago with no problems, but now this thing is apparently nonsense.

Help, please.

#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include <cuda.h>

#include <stdio.h>
#include <iostream>
#include <fstream>




// Helper function for using CUDA to add vectors in parallel.
void cuda_code(float* a, float* x, int N, int dim);

__global__ void kernel_func(float* a, float* x, int N, int dim)
{
	int i = threadIdx.x;
	printf("Hello from thread %d \n", i);
	
	__syncthreads;
	//c[i] = a[i] + b[i];
}

void readm(float* m, int n);
void printm(float *m, float*x, int n);

int main() {
	float* m;
	float* x;
	int dim;
	int N = 10;
	dim = (N + 1) * N;

	m = (float*)malloc(dim);
	x = (float*)malloc(N);

	readm(m, N);

	cuda_code(m, x, N, dim);

	cudaDeviceReset();

	return 0;
}

void cuda_code(float* a, float* x, int N, int dim)
{
	size_t size1 = dim * sizeof(float);
	size_t size2 = N * sizeof(float);
	float* dev_a;
	float* dev_x;

	// Allocate GPU buffers for three vectors (two input, one output).
	cudaMalloc((void**)&dev_a, size1);
	cudaMalloc((void**)&dev_x, size2);

	// Copy input vectors from host memory to GPU buffers.
	cudaMemcpy(dev_a, a, dim * sizeof(float), cudaMemcpyHostToDevice);

	//cudaMemcpy(dev_x, x, N * sizeof(float), cudaMemcpyHostToDevice);

	// Launch a kernel on the GPU with one thread for each element.
	kernel_func << <1, 1 >> > (dev_a, dev_x, N, dim);
	cudaDeviceSynchronize();

	// Copy output vector from GPU buffer to host memory.
	cudaMemcpy(x, dev_x, N * sizeof(float), cudaMemcpyDeviceToHost);

	cudaFree(dev_a);
	cudaFree(dev_x);

}

void readm(float* m, int n) {
	using namespace std;
	ifstream myfile;
	myfile.open("matrix.txt");

	if (myfile.is_open()) {
		for (int j = 0; j < n; j++)
		{
			for (int i = 0; i < (n + 1); i++)
			{
				int ij = i + ((n + 1) * j);
				myfile >> m[ij];
			}
		}

	}


	myfile.close();
}

void printm(float* m, float* x, int n) {
	std::ofstream myfile;
	myfile.open("matrix2.txt");

	if (myfile.is_open()) {
		for (int j = 0; j < n; j++)
		{
			for (int i = 0; i < n + 1; i++)
			{
				myfile << m[i + (n + 1) * j] << "\t";
			}
			myfile << "\n";
		}
		for (int j = 0; j < n; j++)
		{
			myfile << x[j] << "\n";
		}

	}
	myfile.close();
}
2

add proper cuda error checking to your code

not sure what that is? google “proper CUDA error checking”, take the first hit, start reading, and apply it to your code.

Then report whatever error code is being emitted from the function call on line 54

it happens to be the first CUDA call in your program, so my guess would be that you don’t have a functional cuda environment in your machine

3

Hello,

Two things,

I fixed it already, thank you for your time. In line 34 I had

m = (float*)malloc(dim);

I needed to put.

m = (float*)malloc(dim * sizeof(float));

and it started to work perfectly, as usual.

I before fixing it, I tried the CUDA error checking, but even that was giving violation exception.

Thank you