ARB_fragment_program bugs (?)

I’ve noticed a few things that seem odd having to do with ARB_fragment_program on OS X 10.3:

1 - Texture matrix access in ARB_fragment_program seems to always return the transpose of the matrix. I wrote a shader at work in linux on an nv quadro card that made use of that matrix, and when I compiled it at home on my mac (exact same C program code as well as shader code) everything worked fine except that I had to use state.matrix.texture[0].transpose instead of state.matrix.texture[0] to prevent “backwards” texture transformations. To test this more thoroughly I wrote vertex and fragment programs that do the same thing using cube maps, and sure enough the texture matrix always pops up transposed in ARB_fp, but not ARB_vp. Has anyone else bumped into this?

2 - Passing parameters to programs seems to cause kernel panic some of the time. Seems when I make a call to glProgramLocalParameter4fARB or glProgramEnvParameter4fARB the entire os will screech to a halt when I swap buffers afterword. I can’t even force quit my app, and I have to do a hard restart. Again, anyone encountered this before?

For what its worth I’m using a dual 2.5ghz G5 with a radeon 9600xt on 10.3.5 - sort of bleeding edge mac stuff I know so I guess maybe some bugs are to be expected. The thing that scares me is that I had been developing for years on my old mac and never bumped into any problems like this, but since I got this G5 about two weeks ago I have bumped into what appear to be two opengl bugs already in that time.

it’s maybe the ATI Radeon 9600 ARB_fragment_program implementation that doesn’t work well.

Did you tried the OpenGL Shader Builder ? Because it’s using ARB_vertex_program and ARB_fragment_program.

This topic was automatically closed 183 days after the last reply. New replies are no longer allowed.