Is there any other option except MEX to run a CUDA-written code in Matlab?

Hi,

I have written a CUDA function and made a MEX gateway to use it in Matlab. While it works fine in a for loop when i do not push the MEX function hard (inputs are large and i give it enough time with pause command between each iteration), it does not work fine if i push hard with small imput variables (which makes the processing faster).

My first question is that why this is happening. In the web, i have seen that other people had this issue, but i found no answer.

Second question: Is there any other way that I could use my CUDA-written code in Matlab?

Thanks in advance,

Moein.