New CUDA 4.0 Release Makes Parallel Programming Easier

Unified Virtual Addressing, GPU-to-GPU Communication and Enhanced C++ Template Libraries Enable More Developers to Take Advantage of GPU Computing