pycuda

I'm new to CUDA programming and I was wondering how the performance of pyCUDA is compared to programs implemented in plain C. Will the performance be roughly the same? Are there any bottle necks that I should be aware of? EDIT: I obviously tried to google this issue first, and was surprised to not find any information. i.e. I would have excepted that the pyCUDA people have this question answered in their FAQ. If you're using CUDA -- whether directly through C or with pyCUDA -- all the heavy numerical work you're doing is done in kernels that execute on the gpu and are written in CUDA C

I've got a problem that I want to split across multiple CUDA devices, but I suspect my current system architecture is holding me back; What I've set up is a GPU class, with functions that perform operations on the GPU (strange that). These operations are of the style for iteration in range(maxval): result[iteration]=gpuinstance.gpufunction(arguments,iteration) I'd imagined that there would be N gpuinstances for N devices, but I don't know enough about multiprocessing to see the simplest way of applying this so that each device is asynchronously assigned, and strangely few of the examples that

I am using 63 registers/thread ,so (32768 is maximum) i can use about 520 threads.I am using now 512 threads in this example. (The parallelism is in the function "computeEvec" inside global computeEHfields function function.) The problems are: 1) The mem check error below. 2) When i use numPointsRp>2000 it show me "out of memory" ,but (if i am not doing wrong) i compute the global memory and it's ok. -------------------------------UPDATED--------------------------- i run the program with cuda-memcheck and it gives me (only when numPointsRs>numPointsRp): ========= Invalid global read of size 4

I'm trying to run pycuda introductory tutorial after installing Visual C++ Express 2010 and all kinds of Nvidia drivers, SDK, etc. I get to mod = SourceModule(""" __global__ void doublify(float *a) { int idx = threadIdx.x + threadIdx.y*4; a[idx] *= 2; } """) without errors. But this call in IPython yields CompileError: nvcc compilation of c:\users\koj\appdata\local\temp\tmpbbhsca\kernel.cu failed [command: nvcc --cubin -arch sm_21 -m64 -IC:\Python27\lib\site-packages\pycuda\..\..\..\include\pycuda kernel.cu] [stderr: nvcc fatal : Visual Studio configuration file '(null)' could not be found for

I'm trying to make a pycuda wrapper inspired by scikits-cuda library for some operations provided in the new cuSolver library of Nvidia. I want to solve a linear system of the form AX=B by LU factorization, to perform that first use the cublasSgetrfBatched method from scikits-cuda, that give me the factorization LU; then with that factorization I want to solve the system using cusolverDnSgetrs from cuSolve that I want to wrap, when I perform the computation return status 3, the matrices that supose to give me the answer don't change, BUT the *devInfo is zero, looking in the cusolver's