http://bugs.winehq.org/show_bug.cgi?id=7284
--- Comment #36 from Tobias Jakobi liquid.acid@gmx.net 2009-03-30 07:27:21 --- We should probably do some of these checks:
vs.1.1 mov oPos, v0 rcp r3.x, c0.x mov oD0, r3.xxxx;
I assume rcp produces +INFTY in r3.x, which should result in 0xffffffff (full white) after oD0 is converted to RGBA (it should be clamped then, right?)
Then insert some scaling after rcp, for example with a 0.5 factor.
The current hack assumes that data is already clamped right after rcp, but I think this is wrong. Adding the scaling shouldn't make a difference, however it does if we clamp right after rcp.