Cycles: optimize device updates

This optimizes device updates (during user edits or frame changes in
the viewport) by avoiding unnecessary computations. To achieve this,
we use a combination of the sockets' update flags as well as some new
flags passed to the various managers when tagging for an update to tell
exactly what the tagging is for (e.g. shader was modified, object was
removed, etc.).

Besides avoiding recomputations, we also avoid resending to the devices
unmodified data arrays, thus reducing bandwidth usage. For OptiX and
Embree, BVH packing was also multithreaded.

The performance improvements may vary depending on the used device (CPU
or GPU), and the content of the scene. Simple scenes (e.g. with no adaptive
subdivision or volumes) rendered using OptiX will benefit from this work
the most.

On average, for a variety of animated scenes, this gives a 3x speedup.

Reviewed By: #cycles, brecht

Maniphest Tasks: T79174

Differential Revision: https://developer.blender.org/D9555

1 files changed, 19 insertions, 14 deletions

diff --git a/intern/cycles/render/hair.cpp b/intern/cycles/render/hair.cpp
index 896e798b6f9..e94cad6b32e 100644
--- a/intern/cycles/render/hair.cpp
+++ b/intern/cycles/render/hair.cpp
@@ -494,33 +494,38 @@ void Hair::pack_curves(Scene *scene,
   }
 }
 
-void Hair::pack_primitives(PackedBVH &pack, int object, uint visibility)
+void Hair::pack_primitives(PackedBVH *pack, int object, uint visibility, bool pack_all)
 {
   if (curve_first_key.empty())
     return;
 
-  const size_t num_prims = num_segments();
-  pack.prim_tri_index.reserve(pack.prim_tri_index.size() + num_prims);
-  pack.prim_type.reserve(pack.prim_type.size() + num_prims);
-  pack.prim_visibility.reserve(pack.prim_visibility.size() + num_prims);
-  pack.prim_index.reserve(pack.prim_index.size() + num_prims);
-  pack.prim_object.reserve(pack.prim_object.size() + num_prims);
-  // 'pack.prim_time' is unused by Embree and OptiX
+  /* If the BVH does not have to be recreated, we can bail out. */
+  if (!pack_all) {
+    return;
+  }
+
+  unsigned int *prim_tri_index = &pack->prim_tri_index[optix_prim_offset];
+  int *prim_type = &pack->prim_type[optix_prim_offset];
+  unsigned int *prim_visibility = &pack->prim_visibility[optix_prim_offset];
+  int *prim_index = &pack->prim_index[optix_prim_offset];
+  int *prim_object = &pack->prim_object[optix_prim_offset];
+  // 'pack->prim_time' is unused by Embree and OptiX
 
   uint type = has_motion_blur() ?
                   ((curve_shape == CURVE_RIBBON) ? PRIMITIVE_MOTION_CURVE_RIBBON :
                                                    PRIMITIVE_MOTION_CURVE_THICK) :
                   ((curve_shape == CURVE_RIBBON) ? PRIMITIVE_CURVE_RIBBON : PRIMITIVE_CURVE_THICK);
 
+  size_t index = 0;
   for (size_t j = 0; j < num_curves(); ++j) {
     Curve curve = get_curve(j);
-    for (size_t k = 0; k < curve.num_segments(); ++k) {
-      pack.prim_tri_index.push_back_reserved(-1);
-      pack.prim_type.push_back_reserved(PRIMITIVE_PACK_SEGMENT(type, k));
-      pack.prim_visibility.push_back_reserved(visibility);
+    for (size_t k = 0; k < curve.num_segments(); ++k, ++index) {
+      prim_tri_index[index] = -1;
+      prim_type[index] = PRIMITIVE_PACK_SEGMENT(type, k);
+      prim_visibility[index] = visibility;
       // Each curve segment points back to its curve index
-      pack.prim_index.push_back_reserved(j + prim_offset);
-      pack.prim_object.push_back_reserved(object);
+      prim_index[index] = j + prim_offset;
+      prim_object[index] = object;
     }
   }
 }