I’ve been working on running the simpleCUBLAS demo that comes with the CUDA SDK. The demo runs correctly under emulation, but the program fails to run when compiled to run on a GPU. Visual C++ reports an error 0xC0000005 when loading the libraries for the executable. This happens whether the code is built with the Debug or Release configuration. A similar issue is happening with another project that uses CUBLAS. Non-CUBLAS projects are running correctly, although only a few have been run (ie. bandwidthTest, radixSort).
Does anyone know what could be the cause?
Here is the Visual C++ output:
[codebox]‘simpleCUBLAS.exe’: Loaded ‘C:\Documents and Settings\All Users\Application Data\NVIDIA Corporation\NVIDIA GPU Computing SDK\C\bin\win32\Debug\simpleCUBLAS.exe’, Symbols loaded.
‘simpleCUBLAS.exe’: Loaded ‘C:\WINDOWS\system32\ntdll.dll’
‘simpleCUBLAS.exe’: Loaded ‘C:\WINDOWS\system32\kernel32.dll’
‘simpleCUBLAS.exe’: Loaded ‘C:\CUDA\bin\cublas.dll’, Binary was not built with debug information.
Debugger:: An unhandled non-continuable exception was thrown during process load
The program ‘ simpleCUBLAS.exe: Native’ has exited with code -1073741819 (0xc0000005).[/codebox]
Here is some information about the system and software involved:
Windows XP Service Pack 3, 32 bits
CUDA Driver version 190.38
CUDA Toolkit 2.3
CUDA SDK 2.3
Visual C++ 2008 Express Edition SP1 with Microsoft C/C++ Compiler 15.0.30729.1
Intel Core 2 CPU 6300 @ 1.86 Ghz
2.75 GB of RAM
Two graphics cards: GeForce 8500 GT, GeForce 8800 GTX