I came across a code in which an instance of mtgp32_kernel_params is allocated via cudaMalloc but not freed. When I run that code with cuda-memcheck, it does not give an error message. Does this have to do with the fact that curandMakeMTGP32Constants and curandMakeMTGP32KernelState is run on the mtgp32_kernel_params instance? Or is this just faulty code?
The code can be found on https://codeload.github.com/gist/7466985/zip/1e0c21e9003cdef2df2130f920fb4789248e9cc1
Best and thank you in advance for your answers,