I agree. You now have this nifty geometry processor sitting on your card. There has to be more you can do with it.
It would be cool to see some OEM's releasing some extension to allow bezier patch rendering.
As far as depth of field goes the new 3dfx part is supposed to be able to do it with their T-Buffers. Also full scene AA (blurring the edges).
I doubt a physics engine or hit detection is a possibilty though, as these are pretty implementation dependent on the game. With bezier patches, I could send the control points to the card, it could render them without further intervention. Things like physics and hit detection require return values (and a stall in the video card).
I understand where you ar coming from. But there are certain features that I think can be supported across the board, and emulated where it doesn't exist. For instance our product uses bones based animation. If your card supports HW multiple matrices we use them, else we use the software imp (which was written before HW multiple matrices existed in consumer space). I think that OEM's in most cases are doing a good job of selecting technologies that can be adapted to very quickly. For instance we all interpolate between key-frames at some point right? So now we get that HW accelerated by ATI cards. Prolly through an OGL extension. I think that OEM's are taking your concerns (or rather our concerns) into carefull consideration when designing these features. Except Matrox, I can't see a real good reason to learn all that bump mapping stuff expesially if I can't have a software fallback that runs decent. But HW geometry parts, and higher fillrate are always good things.
I think it is going to be interesting to see what kinds of interesting things these OEMs are going to be doing with those nifty geometry accellerators. I mean that is a lot of power to just be HWA the matrix transformations. Things like the keyframe interpolations sound very promising. It would be really cool to be able to submit the start mesh, then the tareget, then specify a value that is a linear interpretation. Also, lets not forget NVIDIA's extension to do primative skinning via 2 HW matrices (ATI says they are going to do 4). Just try and imagine what else those geometry proccessors can be made to do. And from a developers standpoint I am putting my money on OGL as the API that is best positioned to expose these new HW features.
Slashdot Mirror
It would be cool to see some OEM's releasing some extension to allow bezier patch rendering.
As far as depth of field goes the new 3dfx part is supposed to be able to do it with their T-Buffers. Also full scene AA (blurring the edges).
I doubt a physics engine or hit detection is a possibilty though, as these are pretty implementation dependent on the game. With bezier patches, I could send the control points to the card, it could render them without further intervention. Things like physics and hit detection require return values (and a stall in the video card).
Sorry about the bold part, I added b tags when I should have added p tags ;) But I think my comment was a good one none-the-less ;)
I understand where you ar coming from. But there are certain features that I think can be supported across the board, and emulated where it doesn't exist. For instance our product uses bones based animation. If your card supports HW multiple matrices we use them, else we use the software imp (which was written before HW multiple matrices existed in consumer space). I think that OEM's in most cases are doing a good job of selecting technologies that can be adapted to very quickly. For instance we all interpolate between key-frames at some point right? So now we get that HW accelerated by ATI cards. Prolly through an OGL extension. I think that OEM's are taking your concerns (or rather our concerns) into carefull consideration when designing these features. Except Matrox, I can't see a real good reason to learn all that bump mapping stuff expesially if I can't have a software fallback that runs decent. But HW geometry parts, and higher fillrate are always good things.
I think it is going to be interesting to see what kinds of interesting things these OEMs are going to be doing with those nifty geometry accellerators. I mean that is a lot of power to just be HWA the matrix transformations. Things like the keyframe interpolations sound very promising. It would be really cool to be able to submit the start mesh, then the tareget, then specify a value that is a linear interpretation. Also, lets not forget NVIDIA's extension to do primative skinning via 2 HW matrices (ATI says they are going to do 4). Just try and imagine what else those geometry proccessors can be made to do. And from a developers standpoint I am putting my money on OGL as the API that is best positioned to expose these new HW features.