cmake modifications break initRandom

dnouri / cuda-convnet

My fork of Alex Krizhevsky's cuda-convnet from 2013 where I added dropout, among other features.

http://code.google.com/p/cuda-convnet/

254 stars 147 forks source link

cmake modifications break initRandom #12

Closed invisibleroads closed 10 years ago

invisibleroads commented 10 years ago

@kashif

The following error appears under the following case when running convnet.py.

ab2c1fd5af1f647330de336396f706e0f19ab6d9 Fedora 20 64-bit GeForce 610M

src/nvmatrix/nvmatrix.cu(279) : getLastCudaError() CUDA error : initRandom: Kernel execution failed : (8) invalid device function .

But the error does not appear in the following cases:

ab2c1fd5af1f647330de336396f706e0f19ab6d9 Amazon ec2 g2.2xlarge
0d9e86bc2ed6401249481a8da61bcbe1a6cdeec8 Fedora 20 64-bit GeForce 610M

kashif commented 10 years ago

@invisibleroads for that you might want to uncomment the CMakeList.txt line and add the compute capability of your card. As far as I remember for 610M it is 2.0 or 2.1?

list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_20,code=sm_21)

invisibleroads commented 10 years ago

Aha, I confirm that using the following configuration in CMakeLists.txt results in convnet.py running successfully on GeForce 610M. Thank you so much. You know the answers!

list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_20,code=sm_20)
#list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_30,code=sm_30)
#list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_32,code=sm_32)
#list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_35,code=sm_35)
#list(APPEND CUDA_NVCC_FLAGS -gencode arch=compute_50,code=sm_50)