Why CUDA kernel does not launch in the VS 2013 with CUDA 9.0

Question

I have written a parallel program based on CUDA in Windows (GeForce GT 720M). I have installed the CUDA 9.0 Toolkit and the Visual Studio 2013. Everything is OK but when I compile the code and run it the output is wrong.

The program is:

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

__global__ void square(float * d_out, float * d_in)
{
    int idx = threadIdx.x;
    float f = d_in[idx];
    d_out[idx] = 50;
}

int main(int argc, char ** argv)
{
    const int ARRAY_SIZE = 64;
    const int ARRAY_BYTES = ARRAY_SIZE * sizeof(float);

    // generate the input array on the host
    float h_in[ARRAY_SIZE];
    for (int i = 0; i < ARRAY_SIZE; i++)
    {
        h_in[i] = float(i);
    }
    float h_out[ARRAY_SIZE];

    // declare GPU memory pointers
    float * d_in;
    float * d_out;

    // allocate GPU memory
    cudaMalloc((void **) &d_in, ARRAY_BYTES);
    cudaMalloc((void **) &d_out, ARRAY_BYTES);

    // transfer the array to the GPU
   cudaMemcpy(d_in, h_in, ARRAY_BYTES, cudaMemcpyHostToDevice);

    // launch the Kernel
    square << <1, ARRAY_SIZE >> >(d_out, d_in);

    // copy back the result array to the GPU
    cudaMemcpy(h_out, d_out, ARRAY_BYTES, cudaMemcpyDeviceToHost);

    // print out the resulting array
    for (int i = 0; i < ARRAY_SIZE; i++)
    {
        printf("%f", h_out[i]);
        printf(((i % 4) != 3) ? "\t" : "\n");
    }

    // free GPU memory allocation
    cudaFree(d_in);
    cudaFree(d_out);

    getchar();
    return 0;
}

When I run it the output is:

Also, I compiled it with the nvcc square.cu but the output is the same. I have the kernel launch syntax error in the VS but I think it is not related to the output (but the image is related to another program):

start by adding [proper cuda error checking](https://stackoverflow.com/questions/14038589/what-is-the-canonical-way-to-check-for-errors-using-the-cuda-runtime-api) to your code. Then recompile/run and report any errors that are indicated. — Robert Crovella, Dec 15 '17 at 17:29
@RobertCrovella CUDA error: no kernel image is available for execution on the device — Saeed Rahmani, Dec 15 '17 at 17:40
so are not compiling your for an architecture which will run on your GPU — talonmies, Dec 15 '17 at 18:18
The GeForce GT 720M is a [GF117 (Fermi-based) GPU](https://www.notebookcheck.net/NVIDIA-GeForce-GT-720M.90247.0.html). CUDA 9 dropped support for Fermi devices. CUDA 8 is the latest CUDA toolkit that still supports Fermi GPUs. — Robert Crovella, Dec 15 '17 at 18:26
@RobertCrovella Thanks. I saw it on the Wikipedia page. I will install the CUDA 8. — Saeed Rahmani, Dec 15 '17 at 19:01

score 1 · Accepted Answer · answered Dec 15 '17 at 19:55

1

The problem was the CUDA Toolkit version. For the GeForce GT 720M, the Compute Capability is 2.1 and it can be used by the CUDA 8.0.

answered Dec 15 '17 at 19:55

Saeed Rahmani

650
1
8
29

score 1 · Answer 2 · answered Mar 02 '18 at 13:54

1

Here's a table of CUDA Toolkit Version with their Compute Capabilities.

See section GPUs supported

answered Mar 02 '18 at 13:54

CarpeDiemKopi

316
3
13

Cannot believe that in 2020 the installation does not check your hardware to give you a warning that it does not support your hardware. This is bananas. I spend hours banging my head to the wall because I could copy data from host to device and vice versa but could not call a __global__ function. – Ali Nov 24 '20 at 04:43

Why CUDA kernel does not launch in the VS 2013 with CUDA 9.0

2 Answers2