high precision model rendering

[Seth G. <sga...@li...>]

I've got the high precision model rendering working in software, but only
by not using the special case functions for non-clipped models.  These are
supposed to be more optimal because the transformation matrix is scaled
down by "1/2**32":

// do the scaling up of x and y to screen coordinates as part of the transform
// for the unclipped case (it would mess up clipping in the clipped case).
// Also scale down z, so 1/z is scaled 31 bits for free, and scale down x and y
// correspondingly so the projected x and y come out right
// FIXME: make this work for clipped case too?

	if (trivial_accept) {
		for (i = 0; i < 4; i++) {
			aliastransform[0][i] *= aliasxscale *
				(1.0 / ((float) 0x8000 * 0x10000));
			aliastransform[1][i] *= aliasyscale *
				(1.0 / ((float) 0x8000 * 0x10000));
			aliastransform[2][i] *= 1.0 / ((float) 0x8000 * 0x10000);
		}
	}

I see three options:
 * Drop this optimization
 * Implement this optimization in the other functions
 * Implement this optimization for both cases (as the FIXME suggests)
I couldn't figure out how to do either of the second two options
  __    __    _                                    _    __    __
_/  \__/  \__/  Seth Galbraith  "The Serpent Lord"  \__/  \__/  \_
 \__/  \__/  \_ sga...@kr...  #2244199 on ICQ  _/  \__/  \__/
_/  \__/  \__/    http://www.planetquake.com/gg     \__/  \__/  \_