compile error using CUDA 2.0 'cudafe' died due to signal 11

Accelerated Computing CUDA CUDA Programming and Performance
1

I have just installed CUDA 2.0 driver, toolkit, and sdk on fedora 8, and then tried to compile my codes that worked on CUDA 1.1. Unfortunately, I have the following compile error:

nvcc error: ‘cudafe’ died due to signal 11 (Invalid memory reference)

Could anybody please tell me what is likely to cause that error?

Nicky

2

Its potentially a bug in cudafe. Can you post a test app which reproduces this behavior, along with the exact command that triggers it?

Also, please post the output from running “nvcc -V” and an nvidia-bug-report.log.

thanks,
Lonni

3

Hi

Here’s a small test app that produces the behaviour on my system:

template<unsigned int N>

__global__ void

moo( float[N] ) {

}

int main() {

  float boo[3] = {1.0, 2.0, 3.0};

  moo<3><<<dim3(2,1),dim3(16,1)>>>(boo);

  return 0;

}

output:

[font=“Courier”]modesty ~ $ nvcc -V

nvcc: NVIDIA ® Cuda compiler driver

Copyright © 2005-2007 NVIDIA Corporation

Built on Tue_Jun_10_05:42:45_PDT_2008

Cuda compilation tools, release 1.1, V0.2.1221

modesty ~ $ nvcc reproduce.cu

nvcc error : ‘cudafe’ died due to signal 11 (Invalid memory reference)

modesty ~ $[/font]

Best,

Chris

4

Thanks, I’m able to replicate this, and have opened bug 439534 to have it investigated further. I’ll let you know if there are any immediate solutions or workarounds.

5

This should be fixed in the next release.

6

Heya,

I have the same problem - i publicate it on that forum:

[url=“http://www.gpgpu.org/forums/viewtopic.php?t=5396”]http://www.gpgpu.org/forums/viewtopic.php?t=5396[/url]

if some one know what it’s wrong i be glad