NVIDIA announced the latest version, V.4.0, of the NVIDIA CUDA Toolkit to develop parallel applications for NVIDIA GPUs.
The NVIDIA CUDA 4.0 Toolkit was born to help making parallel programming easier, spreading developer number within application porting to NVIDIA’s GPUs.
New main features:
- NVIDIA GPUDirect 2.0 Technology – Offering support for peer-to-peer communication among GPUs within a single server or workstation.
- Unified Virtual Addressing (UVA) – Providing a single merged-memory address space for the main system memory and the GPU memories, enabling quicker and easier parallel programming.
- Thrust C++ Template Performance Primitives Libraries – Provideing a collection of powerful open source C++ parallel algorithms and data structures that ease programming for C++ developers.