If you're the kind of person that gets off on future specification targets for application programming interfaces, you can check out the full presentation here.
The XNA team shared some of their goals for the next iteration of the API, DirectX 10.1, as well as a few of the longer-term aims for the progression of gaming on the PC.
10.1 will aim to eliminate some programmability limitations when it comes to the unified shader architecture established in 10, as well as laying down a minimum capability for anti-aliasing in hardware. 4x Multi-Sampling AA will be minimum for 10.1 - although we suspect that there will be many cards that can theoretically tick the box, yet are totally useless at this in practice. The new iteration will also add better support for arrays of textures using cube maps, with the aim of acceleration texture processing.
Looking forward to DirectX 11, which is surely a few years away yet, the Vole's stated high-end goals are to bring down the cost of authoring art for games, integrating with new PC hardware better, and enabling more interaction between the user and the game in the form of better physics, AI and other such gameplay functions.
To help with art costs - which are swiftly spiralling upwards, as games get bigger, longer and more detailed, thanks to next generation consoles - Microsoft will be implementing procedural content generation to help create larger textures. The API could also adjust image quality on the fly, within an engine - so if a game started to slow down as a level got more and more complex and busy, DirectX could scale back the detail level on screen to up the frame rate.
In terms of hardware integration, the XNA team pointed out that a number of hardware trends will need to be looked at. CPUs are adopting more and more parallel cores, and DX11+ will sport better support for parallel resource creation on threads, and a more friendly driver interface for working with parallel threading.
Stream processing and GPGPU are big movers and shakers in the graphics world, and DX11+ will bring in support for this functionality. Since GPUs are massively parallel processors, they could be potentially used for super-high precision arithmetic, which is much slower than the floating point single precision currently used. µ
Sign up for INQbot – a weekly roundup of the best from the INQ