NVIDIA Details CUDA 9 Features, Allows C++14 In Device Code

NVIDIA at their annual GPU Technology Conference (GTC'17) have provided more public details about the forthcoming CUDA 9 compute update.

Among the work slated for CUDA 9 is Volta GPU architecture, a new programming model of Cooperative Groups, a new API for programming Tensor Core, faster library routines, various new algorithms, a new visual profiler, improved compiler performance, support for C++14 in device code, and an expanded developer platform. The expanded compiler support includes Clang 3.9 and GCC 6.

CUDA/GPGPU developers wishing to learn more about the plans for CUDA 9.0 can read an extensive overview of the prominent changes via this NVIDIA developer blog post.

CUDA 9 release candidates are expected to begin appearing soon.