GPGPU DEVELOPERS can now get their hands on Nvidia's Cuda Toolkit 3.0, which is aimed at the scientific community and related professions with big numbers to crunch.
With the Green Goblin's Fermi architecture finally due to land in the coming weeks, developers can start developing or adjusting code to cater for the company's native 64-bit GPU support, multiple copy engines, ECC reporting, concurrent kernel execution, optimised double precision and support for linear algebra libraries such as Blas and Lapack.
Coders will get C++ support and better GPGPU and graphics interoperability, which will allow them to use Direct3D 9, 10 and 11 and OpenGL for both Cuda processing and OpenCL applications, amongst others.
Support for device emulation mode has been packaged in a separate version of the Cuda C runtime library, which promises more sophisticated hardware debugging tools. However the company warned that it will be focusing on supporting these tools instead of the legacy device emulators.
Nvidia claims that there are also improved developer tools for Linux and the Tesla compute cluster system.
A full list of Nvidia's latest Cuda Toolkit 3.0 features along with links to download it can be found here. µ