Example change from real-world code:

     1.7 -  if saVolume in attr then fragment+='vec3 norm2 = max(pow((abs(normalize(norm))-0.2)*7.0, vec3(3.0, 3.0, 3.0)), 0.0);norm2 /= dot(norm2, vec3(1.0, 1.0, 1.0));';
     1.8 +  if saVolume in attr then fragment+='vec3 tmpn = (abs(normalize(norm))-0.2)*7.0; vec3 norm2 = max(tmpn*tmpn*tmpn, 0.0);norm2 /= dot(norm2, vec3(1.0, 1.0, 1.0));';

(In reply to Samuel Rødal from comment #0)
> Changing this code to x * x * x gives a noticable performance increase on an
> r600.
> 
> Is this something mesa could automatically optimize?

Samuel, your bug is really old, from 2012, is this still an issue?

The bug may well exist, or it might be fixed by SB. The bug is 6 years old and the original reporting isn't responding. Closing.