This has gone though a couple of iterations as I've considered what I wanted to do;
The first was a multi-threaded game engine and various related things, this mostly came about as I purchased a dual core processor last year. However this project was veto'd by my course leader on account of a lack of sane deliverable... curses!
The second idea was to produce a prototype for a game focused scripting language, not really orignal however good scope for research and a deliverable at the end of it all. I was all set todo this until a better choice was presented to me;
Transmission Line Matrix Theory.
I admit that, right now, I don't completely understand it, as far as I can see it's related to forces being transmitted between points, the class example I've seen is for producing water like ripples.
One of the lectures at my college is doing a PhD todo with this and last year a student took his work and produced a project based on projecting the ripples onto a mesh via normal mapping/changing.
As my lecture, Andrew, was going on about this an idea struck me; using the GPU todo all this work!
It's an extension on the project of last year, some decent scope (I can witter on about GPU arch a bit I guess) and a damned cool looking delivable at the end of it all.
I still need to formalise the propsal and email the guy doing the PhD with my ideas for research, but Andrew seems to think it's a good idea and is pretty sure he'll run with it (infact he said something along the lines of including that he recommends I do this in the email).
There are a couple of things which I need to sort out;
- Which API
- Will D3D10 hardware be around soon enuff to take advantage of it
Right now, I can do most of the work via R2VB on my current X1900 card but if I could get hold of a D3D10 card the possiblity of doing some of this via the GS and producing a comparsion would be pretty cool.
Ofcourse, this also effects the API choice somewhat; for example I'll need hardware which can do it and an API which will support it; R2VB can be done via OGL however unless an extension to GLSL appears GS support is out until late next year when OGL3.0 is ment to come about.. I'm hoping ATI get that extension out quickly, infact I'm gonna try and bother them about this as it's kinda critical.
If they can't, then my GS support relies on D3D10, Vista and the hardware being about. Infact, I need to see if GS emutlation is support in the D3D10 Beta SDK which is lurking around, if it is then it's not such a huge problem, but real harware would be nice [smile]
So, that looks to be the direction I'll be taking this year... could mean learning D3D's way of doing things, but I'm not that bothered about that, infact it gives me something else to include in my research [grin]
One thing I learnt from my project is that its better to pick something less ambitious but do a awesome job of it than to over-stretch yourself and not finish it to such a high quality. I think I tended more towards the latter [headshake]
btw, the D3D10 reference rasterizer implents the GS so you can do prototyping work already. I'm not expecting much (if any) hardware before the launch of Vista next Jan/Feb [headshake]
Bare in mind that relying on specialist software/hardware might give you headaches later on. I had to jump through numerous hoops to get my university to set up a demonstration machine with the right hardware/software for my project...
Keep us informed of progress [smile]