OpenCL's implementation sucks !

I was fighting against segfault in my program when i realized that it was coming from nvcuda.dll.
So, apparently you can’t execute the same kernel twice with different parameters each time or you could get a SEGFAULT (access violation) (i say ‘could’ because it doesn’t come everytime…).
To resolve this, i needed to put the work for the 2 buffers (image2d) in the same kernel.

I don’t want to spend my time on this kind of problem so please nvidia, make your driver more robust !

Code?

I never had these problems