LUMINANCE32F and ...16F FBOs on ATI?

[Hampel]

Does ATI support LUMINANCE32F and ...16F color attachments for FBOs? After reading back the content after rendering into the FBO, all values are clamped to [0, 1]. The same code works as expected (without clamping) on NVidia systems. If I change the type to RGBA_32F it also works for ATI systems. But I don't need the extra 3 components and they are a huge waste of memory and bandwidth...

I also tried the GL_clamp_color_ARB extension, disabling the color clamps, but without success...

[scratt]

I noticed noone else replied yet.. FWIW, and I am not sure what you are doing, but I run HDR in an FBO on ATI using GL_RGBA8.

Like I said I am not sure if that's even helpful, but I had problems with some of the float formats on ATI, and LUMINANCE in general.

I didn't investigate it much further after I found a solution that works for me.

I was just thinking if you are reading it back, then you might be able to fudge something with shaders and a different format.

[Nick H]

LUMINANCE 3F2 and 16F should work fine. Do you have a sample with the issue?

[Nicolas Lelong]

Did you try with a GL_R32F or GL_R16F attachment ?

[dletozeun]

Are you sure you will save bandwidth and memory usage with LUMINANCE32F or LUMINANCE16F formats? It is not the case with the fixed point LUMINANCE format. The luminance value is duplicated in all RGB and A channels, but maybe it is not the case with floating point format. Need to be checked.

[martinsm]

ATI_OpenGL_Programming_and_Optimization_Guide.pdf says that Luminance formats 32f and 16f and also fixed point formats like lum8, lum16. nv_ogl_texture_formats.pdf says the same thing for nvidia except lum16f format which uses lum_alpha_16f internally.