On 9/6/05, Martin Jeppesen <[EMAIL PROTECTED]> wrote: > > > Generally no. Desktop usage involves lots of large rectangles, which > > we handle well. And I won't tolerate the kind of laziness I see in > > the DDX modules for some graphics chips. We damn well will accelerate > > zero-lines that are long enough to benefit from it. There are > > multiple ways to accelerate text. We'll have a real hardware cursor. > > There's no excuse for accelerating only solid fills and bitblts. > > Accelerated text? That would be so cool!!
Would it? For every chip I have ever written X11 drivers for, I have always accelerated AT LEAST the following: - bitblt (copyarea) - copyplane (1 to N and N to N) - solid fill - lines and segments - pixmaps (caching in off-screen memory, with sophisticated replacement algorithms) - planemasks and raster-ops - dashed (1D stipple) lines - stippled (2D stipple) lines (which is odd, I know, but someone wanted it fast) - character glyphs of all sizes - special case for terminal fonts to get even more speed - spans (where mi decomposes things like polygons and circles into horizontal spans) - color tiling of spans and rectangles - stippling of spans and rectangles - large ellipses (solid and dashed) Each one has different sets of special cases. For instance, for the i128, I accelerated 8xN, 8xN, 10xN, 6xN, 7xN, and 12xN characters differently from how I did all other sizes just so I could get more performance. I have a long list of chips I've supported, which I'm not sure I should mention, some of which did not have nice ways of supporting some of these functions, but I came up with ways of doing them anyhow. As a result of this, we tended to be able to beat competitors in xmarks scores, even those who were using faster hardware, and we usually beat those who used the same hardware. When the hardware is available, and we can talk some people into helping with the drivers, I'm going to be giving plenty of suggestions, and I'll be sending lots of patches, because the situation we have here is that to compensate for slow hardware, we must produce a fantastic driver that gets squeezes every last bit of capability and speed out of what we have. > > If it could speed up/smooth the scrolling in Gnome Terminal that would be > so nice! Those Anti Aliased fonts really kills it. And antialiased fonts is something that we can probably accelerate well with OGA. > > > Doom 3 uses the GPU in ways that are completely different from how the > > desktop (regular 2D apps) does. But you might get good performance on > > a AGP bus. > > Compared to my current card, that will be an improvement=) Diamond Viper > 550 TNT 16MB AGP 2X. Had it since 1998 =) > > > > > Will the 3D preformance in OGP be comparable to some old 3D cards? > > > > Better in some ways, comparable in others. Generally, we've > > implemented the most important parts of the OpenGL spec. > > Wouldn't it be fun if e.g. Toms hardware took it in one of their tests?=) > Toms is known to choose resolusions and games that makes the hardware they > would like to win look better. So if they doesn't read this post, we might > be able to win =) I'm sure that a reputable news source like Tom's Hardware never biases their tests. > > > > > Would it be likely that the OGC could be sold from an EU contry? I am > > > thinking about that it is rather expensive to buy from the US if you > live in > > > Europe. > > > > I hope some places will distribute it. > > What would it require? Could I e.g. becom a distributor? Yeah, sure. Why not. If you don't have any money to get an inventory, you could just start trying to get people to place orders. When you get 100 orders, you order 100, and you get a discount that you get to pocket when you sell. Then when you have built enough capital, you can just buy an inventory of them. > > > > Well, the basic stuff isn't a problem. It's just certain kinds of > > priveleged bus accesses. We'll let the hackers tackle that after > > we've gotten the usual ones finished. :) > > > > Sounds like a plan =) > > > > _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
