Cycles: MetalRT support (kernel side)

This patch adds MetalRT support to Cycles kernel code. It is mostly additive in nature or confined to Metal-specific code, however there are a few areas where this interacts with other code:

- MetalRT closely follows the Optix implementation, and in some cases (notably handling of transforms) it makes sense to extend Optix special-casing to MetalRT. For these generalisations we now have `__KERNEL_GPU_RAYTRACING__` instead of `__KERNEL_OPTIX__`.
- MetalRT doesn't support primitive offsetting (as with `primitiveIndexOffset` in Optix), so we define and populate a new kernel texture, `__object_prim_offset`, containing per-object primitive / curve-segment offsets. This is referenced and applied in MetalRT intersection handlers.
- Two new BVH layout enum values have been added: `BVH_LAYOUT_METAL` and `BVH_LAYOUT_MULTI_METAL_EMBREE` for XPU mode). Some host-side enum case handling has been updated where it is trivial to do so.

Ref T92212

Reviewed By: brecht

Maniphest Tasks: T92212

Differential Revision: https://developer.blender.org/D13353

5 files changed, 42 insertions, 2 deletions

diff --git a/intern/cycles/scene/geometry.cpp b/intern/cycles/scene/geometry.cpp
index bf426fc49f6..346b030817f 100644
--- a/intern/cycles/scene/geometry.cpp
+++ b/intern/cycles/scene/geometry.cpp
@@ -165,7 +165,8 @@ int Geometry::motion_step(float time) const
 bool Geometry::need_build_bvh(BVHLayout layout) const
 {
   return is_instanced() || layout == BVH_LAYOUT_OPTIX || layout == BVH_LAYOUT_MULTI_OPTIX ||
-         layout == BVH_LAYOUT_MULTI_OPTIX_EMBREE;
+         layout == BVH_LAYOUT_METAL || layout == BVH_LAYOUT_MULTI_OPTIX_EMBREE ||
+         layout == BVH_LAYOUT_MULTI_METAL_EMBREE;
 }
 
 bool Geometry::is_instanced() const
@@ -1247,7 +1248,8 @@ void GeometryManager::device_update_bvh(Device *device,
   VLOG(1) << "Using " << bvh_layout_name(bparams.bvh_layout) << " layout.";
 
   const bool can_refit = scene->bvh != nullptr &&
-                         (bparams.bvh_layout == BVHLayout::BVH_LAYOUT_OPTIX);
+                         (bparams.bvh_layout == BVHLayout::BVH_LAYOUT_OPTIX ||
+                          bparams.bvh_layout == BVHLayout::BVH_LAYOUT_METAL);
 
   BVH *bvh = scene->bvh;
   if (!scene->bvh) {
diff --git a/intern/cycles/scene/object.cpp b/intern/cycles/scene/object.cpp
index 69a2365f17c..bf224a81af5 100644
--- a/intern/cycles/scene/object.cpp
+++ b/intern/cycles/scene/object.cpp
@@ -530,6 +530,34 @@ void ObjectManager::device_update_object_transform(UpdateObjectTransformState *s
   }
 }
 
+void ObjectManager::device_update_prim_offsets(Device *device, DeviceScene *dscene, Scene *scene)
+{
+  BVHLayoutMask layout_mask = device->get_bvh_layout_mask();
+  if (layout_mask != BVH_LAYOUT_METAL && layout_mask != BVH_LAYOUT_MULTI_METAL_EMBREE) {
+    return;
+  }
+
+  /* On MetalRT, primitive / curve segment offsets can't be baked at BVH build time. Intersection
+   * handlers need to apply the offset manually. */
+  uint *object_prim_offset = dscene->object_prim_offset.alloc(scene->objects.size());
+  foreach (Object *ob, scene->objects) {
+    uint32_t prim_offset = 0;
+    if (Geometry *const geom = ob->geometry) {
+      if (geom->geometry_type == Geometry::HAIR) {
+        prim_offset = ((Hair *const)geom)->curve_segment_offset;
+      }
+      else {
+        prim_offset = geom->prim_offset;
+      }
+    }
+    uint obj_index = ob->get_device_index();
+    object_prim_offset[obj_index] = prim_offset;
+  }
+
+  dscene->object_prim_offset.copy_to_device();
+  dscene->object_prim_offset.clear_modified();
+}
+
 void ObjectManager::device_update_transforms(DeviceScene *dscene, Scene *scene, Progress &progress)
 {
   UpdateObjectTransformState state;
@@ -840,6 +868,7 @@ void ObjectManager::device_free(Device *, DeviceScene *dscene, bool force_free)
   dscene->object_motion.free_if_need_realloc(force_free);
   dscene->object_flag.free_if_need_realloc(force_free);
   dscene->object_volume_step.free_if_need_realloc(force_free);
+  dscene->object_prim_offset.free_if_need_realloc(force_free);
 }
 
 void ObjectManager::apply_static_transforms(DeviceScene *dscene, Scene *scene, Progress &progress)
diff --git a/intern/cycles/scene/object.h b/intern/cycles/scene/object.h
index f6dc57ee8b9..f983b58b59c 100644
--- a/intern/cycles/scene/object.h
+++ b/intern/cycles/scene/object.h
@@ -155,6 +155,7 @@ class ObjectManager {
 
   void device_update(Device *device, DeviceScene *dscene, Scene *scene, Progress &progress);
   void device_update_transforms(DeviceScene *dscene, Scene *scene, Progress &progress);
+  void device_update_prim_offsets(Device *device, DeviceScene *dscene, Scene *scene);
 
   void device_update_flags(Device *device,
                            DeviceScene *dscene,
diff --git a/intern/cycles/scene/scene.cpp b/intern/cycles/scene/scene.cpp
index 4230abe9a1b..452b5215836 100644
--- a/intern/cycles/scene/scene.cpp
+++ b/intern/cycles/scene/scene.cpp
@@ -69,6 +69,7 @@ DeviceScene::DeviceScene(Device *device)
       object_motion(device, "__object_motion", MEM_GLOBAL),
       object_flag(device, "__object_flag", MEM_GLOBAL),
       object_volume_step(device, "__object_volume_step", MEM_GLOBAL),
+      object_prim_offset(device, "__object_prim_offset", MEM_GLOBAL),
       camera_motion(device, "__camera_motion", MEM_GLOBAL),
       attributes_map(device, "__attributes_map", MEM_GLOBAL),
       attributes_float(device, "__attributes_float", MEM_GLOBAL),
@@ -312,6 +313,12 @@ void Scene::device_update(Device *device_, Progress &progress)
   if (progress.get_cancel() || device->have_error())
     return;
 
+  progress.set_status("Updating Primitive Offsets");
+  object_manager->device_update_prim_offsets(device, &dscene, this);
+
+  if (progress.get_cancel() || device->have_error())
+    return;
+
   progress.set_status("Updating Images");
   image_manager->device_update(device, this, progress);
 
diff --git a/intern/cycles/scene/scene.h b/intern/cycles/scene/scene.h
index 4af05349dd3..f8f672a079a 100644
--- a/intern/cycles/scene/scene.h
+++ b/intern/cycles/scene/scene.h
@@ -100,6 +100,7 @@ class DeviceScene {
   device_vector<DecomposedTransform> object_motion;
   device_vector<uint> object_flag;
   device_vector<float> object_volume_step;
+  device_vector<uint> object_prim_offset;
 
   /* cameras */
   device_vector<DecomposedTransform> camera_motion;