On 5 June 2011 04:51, Ralph Versteegen <[email protected]> wrote: > On 5 June 2011 04:22, Ralph Versteegen <[email protected]> wrote: >> On 5 June 2011 04:08, Jay Tennant <[email protected]> wrote: >>>> From: [email protected] >>>> Sent: Saturday, June 04, 2011 8:05 AM >>>> >>>> jay >>>> 2011-06-04 06:05:10 -0700 (Sat, 04 Jun 2011) >>>> 119 >>>> Fixed some graphical artifacts. There's one known left: a problem in the >>>> clipping interpolation of texture coordinates. >>>> --- >>>> U wip/rasterizer.cpp >>> >>> Eh. Turns out it's a problem elsewhere... but I have no idea where. >>> I think it'd be better if I converted to floating point everywhere >>> and slowly converted sections back until I see where it's failing. >>> For all I know, it could be completely unrelated to the precision >>> of the fixed point integers. >>> >>> Does that seem like a good approach? >> >> So it is a problem with the fixed point arithmetic? Have you tried >> modifying FPint to use a double internally? >> > > By the way, I'd like to make the code portable (at least compilable by > gcc) but I need to wait until you tell me you're not working on it. In > particular, I'll switch to stdint.h (eg. int32_t instead of __int32) > so it'll be a pretty big change. >
I spent some time playing around with this. Lots of fun. I added a wrapper function for drawing rotated+scaled Frames (wrappers are needed anyway since FB can't call the QuadRasterizer methods), and a test menu in Custom to try it out (via the testing menu accessed by typing 'spam' at the toplevel editing menu). However rasterizer.cpp, matrixMath.cpp, rasterizer_wrap.cpp are only linked if you compile with "scons raster=1". This way you don't have to worry about breaking builds if you break my wrapper or gcc support. I've committed to SVN everything except my patch to fix the Visual-C++-isms in the code. That patch is here for anyone who wants to take a look: https://gitorious.org/~tmc/ohrrpgce-svn/tmc/commit/3f4b6a3270bbeebb695433ce5cfc929127ba2563 At some point I'll add support for Visual-C++ to the SConscript. Until then you'll need mingw to build and try it out, so here's a screenshot: http://dl.dropbox.com/u/16291524/quad0.png Aside from the obvious graphic bugs (and I also notice that sometimes a line is drawn to the left edge of the screen from the top edge of the quad) I totally forgot about the problem of making triangles in a mesh line up perfectly! Good luck with that! The matrix code seems to work fine, and I didn't manage to make it crash :) _______________________________________________ Ohrrpgce mailing list [email protected] http://lists.motherhamster.org/listinfo.cgi/ohrrpgce-motherhamster.org
