Hi,
I’m using Gentoo on a 64 bit system and my GPU is a GTX260.
# nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2007 NVIDIA Corporation
Built on Thu_Jun_19_03:38:28_PDT_2008
Cuda compilation tools, release 2.0, V0.2.1221
# emerge --search nvidia-cuda
* dev-util/nvidia-cuda-sdk
Latest version available: 2.1.1215.2015
Latest version installed: 2.02.0807.1535
Size of files: 45,525 kB
Homepage: http://developer.nvidia.com/cuda
Description: NVIDIA CUDA Software Development Kit
License: CUDPP
* dev-util/nvidia-cuda-toolkit
Latest version available: 2.1
Latest version installed: 2.0
Size of files: 42,527 kB
Homepage: http://developer.nvidia.com/cuda
Description: NVIDIA CUDA Toolkit
License: NVIDIA
I can’t upgrage to 2.1 because it’s a gentoo bug.
# nvcc -m64 --ptxas-options=-v -I /opt/cuda/include -I /opt/cuda/sdk/common/inc -L /opt/cuda/lib -lcuda -lcudart -o project-64 project-64.cu
ptxas info : Compiling entry function '__globfunc__Z8cuda_desPciS_iS_'
ptxas info : Used 23 registers, 248+246 bytes smem, 76 bytes cmem[1], 60 bytes cmem[14]
ptxas info : Compiling entry function '__globfunc__Z13cuda_init_descc'
nvcc error : 'ptxas' died due to signal 11 (Invalid memory reference)
Maybe the cause of my problem :
__device__ static long long myarray[4][4096]; // global GPU var
__device__ void function()
{
//some code here
// and modification of array (global var)
myarray[1][654] = 1; // Just an example
}
I also tested without the static statement but it’s still failling on the same error.
It seems to be a known issue in CUDA 1.1 but I’m using CUDA 2.0.
Does anyone have anything to help me to identify the problem ? any clue ?
Thanks