NVIDIA Cuda error "all CUDA-capable devices are busy or unavailable" on OSX

Quite often, I get the CUDA library to completely fail and return with an error 46 ("all CUDA-capable devices are busy or unavailable") even for simple calls like cudaMalloc. The code runs successfully if I restart the computer, but this is far from ideal. This problem is apparently quite common.

My setup is the following:

• OSX 10.6.8
• NVIDIA CUDA drivers : CUDA Driver Version: 4.0.31 (latest)
• GPU Driver Version: 1.6.36.10 (256.00.35f11)

I tried many solutions from the Nvidia forum, but it didn't work. I don't want to reboot every time it happens. I also tried to unload and reload the driver with a procedure I assume to be correct (may not be)

kextunload -b com.nvidia.CUDA
kextload -b com.nvidia.CUDA

But still it does not work. How can I kick the GPU (or CUDA) back into sanity ?

This is the device querying result

 CUDA Device Query (Runtime API) version (CUDART static linking)

Found 1 CUDA Capable device(s)

Device 0: "GeForce 9400M"
  CUDA Driver Version / Runtime Version          4.0 / 4.0
  CUDA Capability Major/Minor version number:    1.1
  Total amount of global memory:                 254 MBytes (265945088 bytes)
  ( 2) Multiprocessors x ( 8) CUDA Cores/MP:     16 CUDA Cores
  GPU Clock Speed:                               1.10 GHz
  Memory Clock rate:                             1075.00 Mhz
  Memory Bus Width:                              128-bit
  Max Texture Dimension Size (x,y,z)             1D=(8192), 2D=(65536,32768), 3D=(2048,2048,2048)
  Max Layered Texture Size (dim) x layers        1D=(8192) x 512, 2D=(8192,8192) x 512
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       16384 bytes
  Total number of registers available per block: 8192
  Warp size:                                     32
  Maximum number of threads per block:           512
  Maximum sizes of each dimension of a block:    512 x 512 x 64
  Maximum sizes of each dimension of a grid:     65535 x 65535 x 1
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             256 bytes
  Concurrent copy and execution:                 No with 0 copy engine(s)
  Run time limit on kernels:                     Yes
  Integrated GPU sharing Host Memory:            Yes
  Support host page-locked memory mapping:       Yes
  Concurrent kernel execution:                   No
  Alignment requirement for Surfaces:            Yes
  Device has ECC support enabled:                No
  Device is using TCC driver mode:               No
  Device supports Unified Addressing (UVA):      No
  Device PCI Bus ID / PCI location ID:           2 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 4.0, CUDA Runtime Version = 4.0, NumDevs = 1, Device = GeForce 9400M
[deviceQuery] test results...
PASSED

This is an example of code that may fail (although in normal conditions it does not)

#include <stdio.h>

__global__ void add(int a, int b, int *c) {
    *c = a + b;
}

int main(void) {
    int c;
    int *dev_c;

    cudaMalloc( (void **) &dev_c, sizeof(int)); // fails here, returning 46

    add<<<1,1>>>(2,7,dev_c);
    cudaMemcpy(&c, dev_c, sizeof(int), cudaMemcpyDeviceToHost);
    printf("hello world, %d\n",c);
    cudaFree( dev_c);
    return 0;
}

I also found out that occasionally I get to revert back to a sane behavior without a reboot. I still don't know what triggers it.