Almost nine months have passed since Nvidia released the first beta version of its Compute Unified Device Architecture (CUDA) software development kit. CUDA provides a C-like application programming interface through which developers can harness Nvidia GPUs to run general-purpose computing applications.
Today, Nvidia tells us it has unleashed version 1.1 of CUDA, bringing additions like support for 64-bit versions of Windows XP and extra examples of source code that takes advantage of multi-GPU systems. The Windows flavor of CUDA 1.1 will become available today from Nvidia's CUDA site, while Linux users will have to wait until next Monday, November 19, before they can get their hands on the new compiler and SDK.
Starting with the release of CUDA 1.1, Nvidia says it will begin implementing its CUDA driver in its standard graphics drivers. The move will remove the need to install special software to run CUDA applications, Nvidia says, thereby "enabling the seamless deployment of CUDA accelerated applications across multiple markets." So far, both Nvidia's GeForce 8-series graphics card and Tesla "GPU Computing Processors" can run CUDA apps.
In a separate announcement, Nvidia boasts that 20 universities worldwide are actively using CUDA to teach parallel programming to students. Those universities include Stanford, Berkeley, and Purdue. "Many more" are considering the addition of Nvidia tools to their curriculums, the company adds.