Cycles: Speedup by avoiding extra calculations in noise texture when unneeded

Noise texture is now faster when the color socket is unused. Potential for
speedup spotted by @nutel.

Some performance results:

                     Render Time Before    After    Difference
Gooseberry benchmark         47:51.34    45:55.57       -4%
Koro                         12:24.92    12:18.46     -0.8%
Simple cube (Color socket)      48.53       48.72     +0.3%
Simple cube (Fac socket)        48.74       32.78    -32.7%
Goethe displacement           1:21.18     1:08.47    -15.6%
Cycles brick displacement     3:02.38     2:16.76    -25.0%
Large displacement scene     23:54.12    20:09.62    -15.6%

Reviewed By: sergey

Differential Revision: https://developer.blender.org/D2513

1 files changed, 19 insertions, 27 deletions

diff --git a/intern/cycles/kernel/svm/svm_noisetex.h b/intern/cycles/kernel/svm/svm_noisetex.h
index 62ff38cf1c5..0347ab7b193 100644
--- a/intern/cycles/kernel/svm/svm_noisetex.h
+++ b/intern/cycles/kernel/svm/svm_noisetex.h
@@ -18,50 +18,42 @@ CCL_NAMESPACE_BEGIN
 
 /* Noise */
 
-ccl_device_inline void svm_noise(float3 p, float detail, float distortion, float *fac, float3 *color)
-{
-	int hard = 0;
-
-	if(distortion != 0.0f) {
-		float3 r, offset = make_float3(13.5f, 13.5f, 13.5f);
-
-		r.x = noise(p + offset) * distortion;
-		r.y = noise(p) * distortion;
-		r.z = noise(p - offset) * distortion;
-
-		p += r;
-	}
-
-	*fac = noise_turbulence(p, detail, hard);
-	*color = make_float3(*fac,
-		noise_turbulence(make_float3(p.y, p.x, p.z), detail, hard),
-		noise_turbulence(make_float3(p.y, p.z, p.x), detail, hard));
-}
-
 ccl_device void svm_node_tex_noise(KernelGlobals *kg, ShaderData *sd, float *stack, uint4 node, int *offset)
 {
 	uint co_offset, scale_offset, detail_offset, distortion_offset, fac_offset, color_offset;
 
 	decode_node_uchar4(node.y, &co_offset, &scale_offset, &detail_offset, &distortion_offset);
+	decode_node_uchar4(node.z, &color_offset, &fac_offset, NULL, NULL);
 
 	uint4 node2 = read_node(kg, offset);
 
 	float scale = stack_load_float_default(stack, scale_offset, node2.x);
 	float detail = stack_load_float_default(stack, detail_offset, node2.y);
 	float distortion = stack_load_float_default(stack, distortion_offset, node2.z);
-	float3 co = stack_load_float3(stack, co_offset);
+	float3 p = stack_load_float3(stack, co_offset) * scale;
+	int hard = 0;
 
-	float3 color;
-	float f;
+	if(distortion != 0.0f) {
+		float3 r, offset = make_float3(13.5f, 13.5f, 13.5f);
+
+		r.x = noise(p + offset) * distortion;
+		r.y = noise(p) * distortion;
+		r.z = noise(p - offset) * distortion;
 
-	svm_noise(co*scale, detail, distortion, &f, &color);
+		p += r;
+	}
 
-	decode_node_uchar4(node.z, &color_offset, &fac_offset, NULL, NULL);
+	float f = noise_turbulence(p, detail, hard);
 
-	if(stack_valid(fac_offset))
+	if(stack_valid(fac_offset)) {
 		stack_store_float(stack, fac_offset, f);
-	if(stack_valid(color_offset))
+	}
+	if(stack_valid(color_offset)) {
+		float3 color = make_float3(f,
+			noise_turbulence(make_float3(p.y, p.x, p.z), detail, hard),
+			noise_turbulence(make_float3(p.y, p.z, p.x), detail, hard));
 		stack_store_float3(stack, color_offset, color);
+	}
 }
 
 CCL_NAMESPACE_END