This week at SIGGRAPH, the Khronos Group has released a new update to their cross-platform, open standard 2D and 3D graphics API. I sat down with Barthold Lichtenbelt, the OpenGL ARB Working Group Chair at Khronos and Sr. Manager of the Core OpenGL driver at NVIDIA to talk about the new release.
Barthold emphasized the OpenGL-based ecosystem as an important focus for the 4.1 update to OpenGL, particularly with the feature convergence between OpenGL, OpenGL ES (embedded systems) and WebGL. OpenGL 4.1 now has full compatibility between OpenGL ES 2.0 to make it easier to develop for both desktop and mobile platforms.
OpenGL 4.1 has also improved the interoperability between OpenGL and OpenCL, the Khronos Group standard for doing parallel processing on the GPU by letting OpenGL sync objects linked with OpenCL event objects.
As the OpenGL standard matures and the adoption continues to rise, Barthold told me that one of the things they’ve been working on is streamlining the API. Version 4.0 removed some depreciated features that no longer make sense or have been replaced by better methods. Developers using these elements aren’t out of luck, though, and can continue to build code compatible with new versions by including a compatibility extension.
Khronos' vision for the OpenGL ecosystem
With WebGL and the growing number of other OpenGL implementations, security is becoming a larger concern and the new version has new features for preventing buffer overflow issues—as well as preventing one instance of OpenGL from crashing other instances whenever possible.
The most interesting part of the conversation for me was discussing their thoughts on OpenGL and DirectX, as well as OpenCL and competing CUDA. Barthold’s perspective is that there are strengths to all four of these different technologies and they want users to pick the best tool for the job. For example, if an a user wants to write low-level parallel processing on NVIDIA GPU’s their best choice will probably be to use CUDA. However, if they want to write an application that is portable across multiple platforms then OpenCL is the tool for the job.
Articles RSS