Oh man, I remember my first games industry job - having to write a billion different combinations of pentium hand-optimized polygon renderers. (Convex/concave,

Oh man, I remember my first games industry job - having to write a billion different combinations of pentium hand-optimized polygon renderers. (Convex/concave, textured, gouraud, textured+gouraud, etc...). We did some similar occlusion stuff in our (admittedly terrible) game ("Assault Rigs", Psygnosis). We sliced foreground object geometry along a plane where necessary to resolve sort issues with the background - and also to allow the scenery to cast "fake" shadows on parts of the object. I also came up with bitfield-represented PVS independently of the Quake guys :) If I remember correctly, I did a special-case constant-z renderer for big scenery polys, but the in-game geometry used self-modifying code to set the constant U/V gradients for the spanline. Ugh. Of course on the PS1, there was no software rendering anyway, and we just had to subdivide foreground polys to avoid affine-mapping artifacts. I'm still not sure how we dealt with all the seaming issues that that can introduce. Ah, happy days :)