From e527544b766bb01fb95bff1d98fe5a246aaf7398 Mon Sep 17 00:00:00 2001
From: Jeroen Bakker <jbakker>
Date: Wed, 13 Nov 2019 08:03:44 +0100
Subject: Cycles: OpenCL Performance

When using OpenCL with Cycles the rendering time increased substantial.
After doing some tests the bottleneck was found in 4d voronoi and 2d and
3d smooth voronoi.

This change will hide these behind a specific compile directive so the
speed will improve.

AMD RX480 + BMW scene

    2.80 (3:10)
    2.81 (5:48)
    2.81 excluding 4d voronoi+2d/3d smooth (3:50)

Reviewed By: sergey

Differential Revision: https://developer.blender.org/D6231
---
 intern/cycles/kernel/svm/svm.h         |  6 ------
 intern/cycles/kernel/svm/svm_types.h   |  7 ++++++-
 intern/cycles/kernel/svm/svm_voronoi.h |  7 +++++++
 intern/cycles/render/nodes.h           | 12 ++++++++++++
 4 files changed, 25 insertions(+), 7 deletions(-)

(limited to 'intern/cycles')

diff --git a/intern/cycles/kernel/svm/svm.h b/intern/cycles/kernel/svm/svm.h
index 4fc5e633e3b..18f086d6726 100644
--- a/intern/cycles/kernel/svm/svm.h
+++ b/intern/cycles/kernel/svm/svm.h
@@ -214,9 +214,6 @@ CCL_NAMESPACE_END
 
 CCL_NAMESPACE_BEGIN
 
-#define NODES_GROUP(group) ((group) <= __NODES_MAX_GROUP__)
-#define NODES_FEATURE(feature) ((__NODES_FEATURES__ & (feature)) != 0)
-
 /* Main Interpreter Loop */
 ccl_device_noinline void svm_eval_nodes(KernelGlobals *kg,
                                         ShaderData *sd,
@@ -545,9 +542,6 @@ ccl_device_noinline void svm_eval_nodes(KernelGlobals *kg,
   }
 }
 
-#undef NODES_GROUP
-#undef NODES_FEATURE
-
 CCL_NAMESPACE_END
 
 #endif /* __SVM_H__ */
diff --git a/intern/cycles/kernel/svm/svm_types.h b/intern/cycles/kernel/svm/svm_types.h
index a6f2327dada..01710f713ac 100644
--- a/intern/cycles/kernel/svm/svm_types.h
+++ b/intern/cycles/kernel/svm/svm_types.h
@@ -48,12 +48,17 @@ CCL_NAMESPACE_BEGIN
 #define NODE_FEATURE_HAIR (1 << 1)
 #define NODE_FEATURE_BUMP (1 << 2)
 #define NODE_FEATURE_BUMP_STATE (1 << 3)
+#define NODE_FEATURE_VORONOI_EXTRA (1 << 4)
 /* TODO(sergey): Consider using something like ((uint)(-1)).
  * Need to check carefully operand types around usage of this
  * define first.
  */
 #define NODE_FEATURE_ALL \
-  (NODE_FEATURE_VOLUME | NODE_FEATURE_HAIR | NODE_FEATURE_BUMP | NODE_FEATURE_BUMP_STATE)
+  (NODE_FEATURE_VOLUME | NODE_FEATURE_HAIR | NODE_FEATURE_BUMP | NODE_FEATURE_BUMP_STATE | \
+   NODE_FEATURE_VORONOI_EXTRA)
+
+#define NODES_GROUP(group) ((group) <= __NODES_MAX_GROUP__)
+#define NODES_FEATURE(feature) ((__NODES_FEATURES__ & (feature)) != 0)
 
 typedef enum ShaderNodeType {
   NODE_END = 0,
diff --git a/intern/cycles/kernel/svm/svm_voronoi.h b/intern/cycles/kernel/svm/svm_voronoi.h
index bdfaed8845f..2ad22592eef 100644
--- a/intern/cycles/kernel/svm/svm_voronoi.h
+++ b/intern/cycles/kernel/svm/svm_voronoi.h
@@ -986,6 +986,7 @@ ccl_device void svm_node_tex_voronoi(KernelGlobals *kg,
                         &color_out,
                         &position_out_2d);
           break;
+#if NODES_FEATURE(NODE_FEATURE_VORONOI_EXTRA)
         case NODE_VORONOI_SMOOTH_F1:
           voronoi_smooth_f1_2d(coord_2d,
                                smoothness,
@@ -996,6 +997,7 @@ ccl_device void svm_node_tex_voronoi(KernelGlobals *kg,
                                &color_out,
                                &position_out_2d);
           break;
+#endif
         case NODE_VORONOI_F2:
           voronoi_f2_2d(coord_2d,
                         exponent,
@@ -1029,6 +1031,7 @@ ccl_device void svm_node_tex_voronoi(KernelGlobals *kg,
                         &color_out,
                         &position_out);
           break;
+#if NODES_FEATURE(NODE_FEATURE_VORONOI_EXTRA)
         case NODE_VORONOI_SMOOTH_F1:
           voronoi_smooth_f1_3d(coord,
                                smoothness,
@@ -1039,6 +1042,7 @@ ccl_device void svm_node_tex_voronoi(KernelGlobals *kg,
                                &color_out,
                                &position_out);
           break;
+#endif
         case NODE_VORONOI_F2:
           voronoi_f2_3d(coord,
                         exponent,
@@ -1060,6 +1064,8 @@ ccl_device void svm_node_tex_voronoi(KernelGlobals *kg,
       position_out = safe_divide_float3_float(position_out, scale);
       break;
     }
+
+#if NODES_FEATURE(NODE_FEATURE_VORONOI_EXTRA)
     case 4: {
       float4 coord_4d = make_float4(coord.x, coord.y, coord.z, w);
       float4 position_out_4d;
@@ -1106,6 +1112,7 @@ ccl_device void svm_node_tex_voronoi(KernelGlobals *kg,
       w_out = position_out_4d.w;
       break;
     }
+#endif
     default:
       kernel_assert(0);
   }
diff --git a/intern/cycles/render/nodes.h b/intern/cycles/render/nodes.h
index 13e9b65fa07..62037b0d381 100644
--- a/intern/cycles/render/nodes.h
+++ b/intern/cycles/render/nodes.h
@@ -220,6 +220,18 @@ class VoronoiTextureNode : public TextureNode {
     return NODE_GROUP_LEVEL_2;
   }
 
+  virtual int get_feature()
+  {
+    int result = ShaderNode::get_feature();
+    if (dimensions == 4) {
+      result |= NODE_FEATURE_VORONOI_EXTRA;
+    }
+    else if (dimensions >= 2 && feature == NODE_VORONOI_SMOOTH_F1) {
+      result |= NODE_FEATURE_VORONOI_EXTRA;
+    }
+    return result;
+  }
+
   int dimensions;
   NodeVoronoiDistanceMetric metric;
   NodeVoronoiFeature feature;
-- 
cgit v1.2.3