Please explain me about differences in memory allocation and freeing between CUDA 2.0 and CUDA 3.0
I used functions cudaMalloc and cudaFree when my programm used CUDA 2.0, programm works fine.
Now when i moved to new version of CUDA i have device memory leaks.
I noticed in CUDA 3.0 Programming guide descriptions of new functions cuMemAlloc, cuMemFree, but cudaMalloc and cudaFree also used in test examples in this guide. Nevertheless old functions compiles as well.