Cycles: Use native saturate function for CUDA

This more a workaround for CUDA optimizer which can't optimize clamp(x, 0, 1)
into a single instruction and uses 4 instructions instead.

Original patch by @lockal with own modification:

  Don't make changes outside of the kernel. They don't make any difference
  anyway and term saturate() has a bit different meaning outside of kernel.

This gives around 2% of speedup in Barcelona file, but in more complex shader
setups with lots of math nodes with clamping speedup could be much nicer.

Subscribers: dingto

Projects: #cycles

Differential Revision: https://developer.blender.org/D1224

1 files changed, 5 insertions, 5 deletions

diff --git a/intern/cycles/kernel/kernel_film.h b/intern/cycles/kernel/kernel_film.h
index 4668b40b86d..f9e9b413898 100644
--- a/intern/cycles/kernel/kernel_film.h
+++ b/intern/cycles/kernel/kernel_film.h
@@ -27,7 +27,7 @@ ccl_device float4 film_map(KernelGlobals *kg, float4 irradiance, float scale)
 	result.z = color_scene_linear_to_srgb(result.z*exposure);
 
 	/* clamp since alpha might be > 1.0 due to russian roulette */
-	result.w = clamp(result.w, 0.0f, 1.0f);
+	result.w = saturate(result.w);
 
 	return result;
 }
@@ -37,10 +37,10 @@ ccl_device uchar4 film_float_to_byte(float4 color)
 	uchar4 result;
 
 	/* simple float to byte conversion */
-	result.x = (uchar)clamp(color.x*255.0f, 0.0f, 255.0f);
-	result.y = (uchar)clamp(color.y*255.0f, 0.0f, 255.0f);
-	result.z = (uchar)clamp(color.z*255.0f, 0.0f, 255.0f);
-	result.w = (uchar)clamp(color.w*255.0f, 0.0f, 255.0f);
+	result.x = (uchar)(saturate(color.x)*255.0f);
+	result.y = (uchar)(saturate(color.y)*255.0f);
+	result.z = (uchar)(saturate(color.z)*255.0f);
+	result.w = (uchar)(saturate(color.w)*255.0f);
 
 	return result;
 }