On Tuesday 01 February 2005 16:13, Timothy Miller wrote:
That's the thing... we've discussed coverage, full-screen antialiasing, subpixel coordinates and the like, and the experts here have almost unanimously said that they're unnecessary. (Some can be emulated anyhow.)
If we decide to support it now or in the future, it'll matter, but for now, it seems to be something we can just drop.
If you have a compelling argument contrary to that, feel free to discuss it with the others. Note that the design, as it is, has probably seriously blown our transistor budget, so don't get your hopes up, even about some things that are ALREADY in there.
Yes, it's reality. I have been making some effort to understand exactly what this FPGA is capable of, and how that maps to the 3D pipeline. Did you say the part would be the XC3S2000?
That is, you might say, the point of reference. If we have to use a chip from another vendor, we will, but in quantity, that one's a reasonable price and fast enough. It's also quite large in transistor count.
A key point here is that the design will fall back to iterating when it runs out of dedicated components, which means that instead of blowing the transistor budget, it just slows down. Hopefully, not to a crawl.
Yes, and I'm going to have to put some effort into figuring out how to do that in some cases.
Another point is, the first rev doesn't have to implement everything, it only absolutely needs to implement enough to run xscreensaver-gl and we will win.
Well, many people will be disappointed if we leave out any of the features we have in there. It's been whittled down. People have offered their opinions on what's necessary and not. Some even disagree with me on the priorites of some features I DO have, but they were trivial to implement.
BTW, font-rendering from xft is pre-antialiased. It's just a grayscale bitmap that we use as the alpha channel while texturing.
Right, and subpixel positioning can be expressed through U and V. I'll go back and sort through the archives and see if I can find the arguments. My intuition is still that it's a better and more regular design to send X and Y in the same format. It's hardly enough of an overhead to worry about. This is probably a good time to segue into exactly what the DMA command stream is going to look like. I know you talked about it a little, and worried that the bandwidth could be excessive without some clever tricks.
I just sent another email on that... _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
