I was wondering whether I can get some performace speedup for my application which uses A-Buffer with atomics. So, I tried to avoid using the atomics by employing the INTEL_fragment_shader_ordering extension which seems to be supported on my R9 270X (Catalyst 15.7, Windows 7 64-bit). After the implementation I found that my app causes the driver to crash after 10-15 seconds when run.
I did little debugging and found that the driver crashes even if the function beginFragmentShaderOrderingINTEL (from the extension) is only included in the fragment shader but not called. I included a call to the ...INTEL function into my own function named xyz which was not referenced from anywhere in the shader and the driver still crashed. When I commented out the ...INTEL call (using //) the app run without crash. The extension was declared in the fragment shader with require behavoiur and the shader compilation was successful.
Do you have any idea?
Thank You in advance,