mrq/GLdc
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Hacky fix for ortho projections

Browse Source
This commit is contained in:
Luke Benstead 2020-04-07 20:44:29 +01:00
parent 8add817086
commit 5407e9643e
1 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
GL/draw.c
View File

@ -1173,10 +1173,14 @@ GL_FORCE_INLINE void divide(SubmissionTarget* target) {
vertex->xyz[0] *= f; vertex->xyz[0] *= f;
vertex->xyz[1] *= f; vertex->xyz[1] *= f;
/* We have to use 1/z instead of reusing 1/w from above because /* FIXME: OK, so. This is using 1/w as Z. Which is great for perspective projections
* orthographic projections always product a W == 1 and so would not * but absolutely broken for orthographic ones (because all Z values end up with 1.0).
* correctly render ordered transparent surfaces */ * To fix this, we need to switch to 1/z *BUT* that breaks clipping as it's currently
vertex->xyz[2] = MATH_Fast_Invert(vertex->xyz[2]); * written... I'm not sure why.
* Also, switching to 1/z results in coordinates between -1 and 1 (well, inf) and Z
* coordinates *MUST* be > 0.0f to render on the PVR. So we'll need to make use
* of glDepthRange when that happens */
vertex->xyz[2] = f;
/* FIXME: Consider taking glDepthRange into account. PVR is designed to use 1/w /* FIXME: Consider taking glDepthRange into account. PVR is designed to use 1/w
* which is unlike most GPUs - this apparently provides advantages. * which is unlike most GPUs - this apparently provides advantages.