Cudaa kernel launch failure

Hi, I have made cuda c mex file after I run the mex file outputs are obtain but when gpu enable commands are run . I get an error cuda kernel launch failure. Further investigating I came to know that error is "gpu device busy or currantly unavailable "

[Joss Knight]

Probably your MEX kernel is still running when you try to access the result. Did you call cudaDeviceSynchronize() or cudaStreamSynchronize() after your MEX computation?

[ARSALAN RIAZ]

Using mex file in loop I think some memory is leaked. Output argument are not overwritten some how!