[Bf-blender-cvs] [c31f4e8] cycles_hair_bvh: Cycles: Bring child visibility check

[Sergey Sharybin]

Commit: c31f4e835ac1d76854994660170696583ad26bc7
Author: Sergey Sharybin
Date:   Fri Apr 29 12:01:49 2016 +0200
Branches: cycles_hair_bvh
https://developer.blender.org/rBc31f4e835ac1d76854994660170696583ad26bc7

Cycles: Bring child visibility check

It is slower for simple cases, but could be faster in cases like complex
ray visibility settings.

Need some investigation about how to do visiblity check without time penalty.

===================================================================

M	intern/cycles/kernel/geom/geom_bvh_hair.h

===================================================================

diff --git a/intern/cycles/kernel/geom/geom_bvh_hair.h b/intern/cycles/kernel/geom/geom_bvh_hair.h
index f7c6283..1feb317 100644
--- a/intern/cycles/kernel/geom/geom_bvh_hair.h
+++ b/intern/cycles/kernel/geom/geom_bvh_hair.h
@@ -143,12 +143,21 @@ int ccl_device bvh_hair_intersect_node_unaligned(KernelGlobals *kg,
 	const ssef tNear = max4(tnear_x, tnear_y, tnear_z, tnear);
 	const ssef tFar = min4(tfar_x, tfar_y, tfar_z, tfar);
 	const sseb vmask = tNear <= tFar;
-	int mask = (int)movemask(vmask);
 
 	dist[0] = tNear.f[0];
 	dist[1] = tNear.f[1];
 
+	int mask = (int)movemask(vmask);
+
+#  ifdef __VISIBILITY_FLAG__
+	/* this visibility test gives a 5% performance hit, how to solve? */
+	float4 cnodes = kernel_tex_fetch(__bvh_curve_nodes, nodeAddr*BVH_UNALIGNED_NODE_SIZE+8);
+	int cmask = (((mask & 1) && (__float_as_uint(cnodes.z) & visibility))? 1: 0) |
+	            (((mask & 2) && (__float_as_uint(cnodes.w) & visibility))? 2: 0);
+	return cmask;
+#  else
 	return mask & 3;
+#  endif
 }
 
 int ccl_device_inline bvh_hair_intersect_node_aligned(KernelGlobals *kg,
@@ -180,12 +189,20 @@ int ccl_device_inline bvh_hair_intersect_node_aligned(KernelGlobals *kg,
 	const ssef tminmax = minmax ^ pn;
 	const sseb lrhit = tminmax <= shuffle<2, 3, 0, 1>(tminmax);
 
-	int mask = movemask(lrhit);
-
 	dist[0] = tminmax[0];
 	dist[1] = tminmax[1];
 
+	int mask = movemask(lrhit);
+
+#  ifdef __VISIBILITY_FLAG__
+	/* this visibility test gives a 5% performance hit, how to solve? */
+	float4 cnodes = kernel_tex_fetch(__bvh_curve_nodes, nodeAddr*BVH_UNALIGNED_NODE_SIZE+8);
+	int cmask = (((mask & 1) && (__float_as_uint(cnodes.z) & visibility))? 1: 0) |
+	            (((mask & 2) && (__float_as_uint(cnodes.w) & visibility))? 2: 0);
+	return cmask;
+#  else
 	return mask & 3;
+#  endif
 }
 
 int ccl_device_inline bvh_hair_intersect_node(KernelGlobals *kg,