To make it short: using any texture matrix != 0 doesn´t work. Multiplying texture coordinates with such a matrix results in a constant (nullvector?) vector, thus the formerly textured geometry now looks untextured
This bug was first introduced with the pre-4.6 betas and is still present in catalyst 4.7.
I checked my GLSL code on a FX, its working there. It worked on ATI up to catalyst 4.5.
Can anyone confirm this?
thx,
skynet
PS: the ATI_fragment_shader emulation on R300 class hardware is also broken since 4.6
Yeah, I noticed that too. The strange thing is that it works in some cases and not in others. In my main project I use gl_TextureMatrix[2] for spotlights (projection matrix), no problems here. But in a demo, gl_TextureMatrix[0] is the only one that works.
If you have a sample showing the problem, pass it on to me at epersson (at) ati.com and I’ll take a look at it.
Didn’t understand what you meant with “ATI_fragment_shader emulation”. It’s not emulated, it’s run in hardware, though with full floating point rather than fixed point as on the R200 series. That extension btw works fine with all my old demos, so if you have any problems I guess it’s more specific to what you’re doing. If you have a sample of that, pass that on to me too.
Ok, I´ll try to put some stuff together. I´ve already mailed to ATI´s devrel. But they didn´t respond yet. They usually respond very quickly, mostly within a day…
About the ATI_fs problem. Well, it certainly is some kind of emulation in hardware. Those fragment shader setups get translated into a “real” fragment programs, while on R200 hw they get directly mapped onto the hardware.
I know that (all?) ATI_fs demos from ATI still work… I have now idea why my app fails. It might be, because I´m using ARB_vp. None of the ATI_fs demos I´ve seen is using that combination. But what would you say, if one of your programs suddenly fails after a driver update, while it worked perfectly with older versions and is still working perfectly on R200 class hardware? I honestly doubt, that I am to blame.
The ATI driver guys should just check, what they changed from 4.5 to pre-4.6 beta, and examine how that could affect ATI_fs (on R300 hw) and GLSL´s texture matrix handling.
