Cycles: Remove MetalRT experimental status, and add "auto enable" option

_(NOTE: This is a clone of [PR 114067](https://projects.blender.org/blender/blender/pulls/114067), but targeting `blender-v4.0-release` as originally intended)_

This PR removes the "experimental" disclaimer from the MetalRT control now that the unit tests all render correctly with it enabled. As well as "Off" and "On", this adds a third "Auto" setting - a new default which can be used to pick the best option.

Pull Request: https://projects.blender.org/blender/blender/pulls/114232

intern/cycles/blender/addon/properties.py

@@ -1472,11 +1472,16 @@ class CyclesPreferences(bpy.types.AddonPreferences):
         default=False,
     )
 
-    use_metalrt: BoolProperty(
-        name="MetalRT (Experimental)",
+    metalrt: EnumProperty(
+        name="MetalRT",
         description="MetalRT for ray tracing uses less memory for scenes which use curves extensively, and can give better "
-                    "performance in specific cases. However this support is experimental and some scenes may render incorrectly",
-        default=False,
+                    "performance in specific cases.",
+        default='AUTO',
+        items=(
+            ('OFF', "Off", "Disable MetalRT (uses BVH2 layout for intersection queries)"),
+            ('ON', "On", "Enable MetalRT for intersection queries"),
+            ('AUTO', "Auto", "Automatically pick the fastest intersection method"),
+        ),
     )
 
     use_hiprt: BoolProperty(
@@ -1711,7 +1716,7 @@ class CyclesPreferences(bpy.types.AddonPreferences):
                 if is_arm64:
                     col.prop(self, "kernel_optimization_level")
                 if has_rt_api_support:
-                    col.prop(self, "use_metalrt")
+                    col.prop(self, "metalrt")
 
         if compute_device_type == 'HIP':
             has_cuda, has_optix, has_hip, has_metal, has_oneapi, has_hiprt = _cycles.get_device_types()

intern/cycles/blender/device.cpp

@@ -39,8 +39,17 @@ void static adjust_device_info_from_preferences(DeviceInfo &info, PointerRNA cpr
     info.has_peer_memory = false;
   }
 
-  if (info.type == DEVICE_METAL && !get_boolean(cpreferences, "use_metalrt")) {
-    info.use_hardware_raytracing = false;
+  if (info.type == DEVICE_METAL) {
+    MetalRTSetting use_metalrt = (MetalRTSetting)get_enum(
+        cpreferences, "metalrt", METALRT_NUM_SETTINGS, METALRT_AUTO);
+
+    info.use_hardware_raytracing = info.use_metalrt_by_default;
+    if (use_metalrt == METALRT_OFF) {
+      info.use_hardware_raytracing = false;
+    }
+    else if (use_metalrt == METALRT_ON) {
+      info.use_hardware_raytracing = true;
+    }
   }
 
   if (info.type == DEVICE_ONEAPI && !get_boolean(cpreferences, "use_oneapirt")) {

intern/cycles/device/device.h

@@ -67,6 +67,14 @@ enum KernelOptimizationLevel {
   KERNEL_OPTIMIZATION_NUM_LEVELS
 };
 
+enum MetalRTSetting {
+  METALRT_OFF = 0,
+  METALRT_ON = 1,
+  METALRT_AUTO = 2,
+
+  METALRT_NUM_SETTINGS
+};
+
 class DeviceInfo {
  public:
   DeviceType type;
@@ -83,6 +91,7 @@ class DeviceInfo {
   bool has_peer_memory;         /* GPU has P2P access to memory of another GPU. */
   bool has_gpu_queue;           /* Device supports GPU queue. */
   bool use_hardware_raytracing; /* Use hardware instructions to accelerate ray tracing. */
+  bool use_metalrt_by_default;  /* Use MetalRT by default. */
   KernelOptimizationLevel kernel_optimization_level; /* Optimization level applied to path tracing
                                                       * kernels (Metal only). */
   DenoiserTypeMask denoisers;                        /* Supported denoiser types. */
@@ -106,6 +115,7 @@ class DeviceInfo {
     has_peer_memory = false;
     has_gpu_queue = false;
     use_hardware_raytracing = false;
+    use_metalrt_by_default = false;
     denoisers = DENOISER_NONE;
   }
 

intern/cycles/device/metal/device_impl.mm

@@ -164,9 +164,14 @@ MetalDevice::MetalDevice(const DeviceInfo &info, Stats &stats, Profiler &profile
     texture_bindings_3d = [mtlDevice newBufferWithLength:8192 options:default_storage_mode];
     stats.mem_alloc(buffer_bindings_1d.allocatedSize + texture_bindings_2d.allocatedSize +
                     texture_bindings_3d.allocatedSize);
-    /* command queue for path-tracing work on the GPU */
+
+    /* Command queue for path-tracing work on the GPU. In a situation where multiple
+     * MetalDeviceQueues are spawned from one MetalDevice, they share the same MTLCommandQueue.
+     * This is thread safe and just as performant as each having their own instance. It also
+     * adheres to best practices of maximizing the lifetime of each MTLCommandQueue. */
     mtlComputeCommandQueue = [mtlDevice newCommandQueue];
-    /* command queue for non-tracing work on the GPU */
+
+    /* Command queue for non-tracing work on the GPU. */
     mtlGeneralCommandQueue = [mtlDevice newCommandQueue];
 
     /* Acceleration structure arg encoder, if needed */