[Mesa-dev] [PATCH] nir: Remove sRGB colorspace conversion round-trip.
mattst88 at gmail.com
Mon May 18 15:15:32 PDT 2015
Some shaders in Civilization V and Beyond Earth do
pow(pow(x, 2.2), 0.454545)
which is converting to and from sRGB colorspace.
A more general rule that replaces pow(pow(a, b), c) with pow(a, b * c)
actually regresses two shaders in Sun Temple in which the result of the
inner pow is used twice, once by another pow and once by another
instruction. Also, since 2.2 * 0.454545 isn't exactly one, the more
general pattern would have still left us with a pow, and I'm 2.2 *
0.454545 percent sure that's not what they want.
instructions in affected programs: 934 -> 886 (-5.14%)
I also see other uses of similar constants, like
but none that seem to be used in a round-trip like this.
Maybe we'd rather just have the general rule and another rule that replaces
pow(x, y) with x where y is ~around~ 1.0?
src/glsl/nir/nir_opt_algebraic.py | 2 ++
1 file changed, 2 insertions(+)
diff --git a/src/glsl/nir/nir_opt_algebraic.py b/src/glsl/nir/nir_opt_algebraic.py
index fa03922..eace791 100644
@@ -156,6 +156,8 @@ optimizations = [
(('fpow', a, 2.0), ('fmul', a, a)),
(('fpow', a, 4.0), ('fmul', ('fmul', a, a), ('fmul', a, a))),
(('fpow', 2.0, a), ('fexp2', a)),
+ (('fpow', ('fpow', a, 2.2), 0.454545), a),
+ (('fpow', ('fabs', ('fpow', a, 2.2)), 0.454545), ('fabs', a)),
(('fsqrt', ('fexp2', a)), ('fexp2', ('fmul', 0.5, a))),
(('frcp', ('fexp2', a)), ('fexp2', ('fneg', a))),
(('frsq', ('fexp2', a)), ('fexp2', ('fmul', -0.5, a))),
More information about the mesa-dev