On Saturday 05 February 2005 09:10, Timothy Miller wrote: > Something noteworthy about the precision loss for the reciprocal is > that it will not result in CUMULATIVE error. Numbers are processed > and send down the pipeline to be used for whatever, and never get > used again. Unless you count the case of writing out and then > reading back in, but then you're down to 8 bits anyhow.
Right. The main problem for me in the past in trying to work with lower precision was in handling all possible ranges. Floating point in a few key places will really help that. The fleshed out render model is going to answer these questions perfectly, especially if hooked into Mesa. In that case, we could use my little game engine demo (among many other possible OpenGL programs) to run simulations. I'll hook in a replay mode so any defects somebody happens to stumble on interactively can be reproduced exactly. As far as C++ idioms go, operator overloading is not a big deal to me, it's just a pet peeve. However, hooking a render.cpp into Mesa might be tricky, have you considered the linking issues? Building Mesa 6.2.1 now, let's see how hard it is to weld these two bits together. Regards, Daniel _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
