rendering capabilites

[rooly]

Yeah, i was wondering, is it just my computer, or does the software renderer seem inherently bugged on everyone elses?
its almost as if the code showed opposite occlusion or something...
someone expound on this issue please?

[Guest]

try irrlicht 014 it has a new software renderer

[hybrid]

I had a quick look into the sources and found at least one obvious bug in the ordering of vertices (sorting such that 1->3 is ascending Y-axis, but taking v1.y as upper left coordinate), but this did not fix the bug. But shouldn't be too hard to find it - probably I'll have some spare time this weekend.

[rooly]

about that gfx, i did use the v14, and that was the bugged version. my v12 worked out just fine.

[hybrid]

GFXStyler meant the new Software2 Renderer which is definitely better than the old one. Much more features, slightly improved performance and correct culling. The old software renderer does not work correctly for several versions now, the clipping is too eagerly and removes visible triangles as soon as the first pixel moves out of view.
BTW: sorting is correct and all main culling parts are working correctly. Could not figure out the problem, yet.

[Eternl Knight]

The new software shader is pretty good all things considered. I haven't pushed it hard yet, but it can render the demo without problems and that is pretty much good enough for me (as I will only use it for the "intro" before getting the user to select a hardware driver).

If you want to really push it - you couold try integrating swShader. It's an LGPL (windows-only) software "shader" renderer. It compiles pixel shader files into assembly code for execution (so pretty much the best performance you can get without coding the assembler yourself).

--EK

[bitplane]

as for the original software renderer, it throws away all triangles that should be clipped. This is a one line change.