Here's what I get, when I'm rendering 16bits (medical) images, with 2 differents cards : NVidia Quadro FX 580 / Intel Q45 Express Chipset, using the same (brightness/contrast) shader.
Posted Friday, 25 March, 2011 - 11:50 by the Fiddler
Looks like a precision issue. Does Intel support (and obey) precision modifiers in shaders? If yes, try setting everything to highp.
Additionally, Intel drivers have an option to downsample all 32bpp textures to 16bpp, make sure this is disabled (or try using floating-point textures which shouldn't be affected by this).
The Intel card has GLSL 1.20 and precision modifiers appears in GLSL 1.30.
The driver is correctly set to 32bpp textures, and I've also try to set all parameters to "quality".
Unsuccessfully.
Posted Thursday, 31 March, 2011 - 08:40 by Jonathan
Hi again,
I've tried using floating-point textures, but problem is still here.
Could it be possible that the Intel card don't support 16bits textures ?
I just noticed that isn't happening on 8bits textures..
Is there an extension to check this support (higher than 8 bits) ?
Comments
Re: [GLSL] Same shader and different cards, not same ...
Looks like a precision issue. Does Intel support (and obey) precision modifiers in shaders? If yes, try setting everything to
highp.Additionally, Intel drivers have an option to downsample all 32bpp textures to 16bpp, make sure this is disabled (or try using floating-point textures which shouldn't be affected by this).
Re: [GLSL] Same shader and different cards, not same ...
Thanks for the answer.
The Intel card has GLSL 1.20 and precision modifiers appears in GLSL 1.30.
The driver is correctly set to 32bpp textures, and I've also try to set all parameters to "quality".
Unsuccessfully.
I'll try using floating point texture !
Re: [GLSL] Same shader and different cards, not same ...
Hi again,
I've tried using floating-point textures, but problem is still here.
Could it be possible that the Intel card don't support 16bits textures ?
I just noticed that isn't happening on 8bits textures..
Is there an extension to check this support (higher than 8 bits) ?
Thanks .