[Impeller] iOS/macOS: Only wait for command scheduling prior to present (redux)

impeller/renderer/backend/metal/command_buffer_mtl.mm

@@ -167,27 +167,6 @@ static bool LogMTLCommandBufferErrorIfPresent(id<MTLCommandBuffer> buffer) {
 
   [buffer_ commit];
 
-#if (FML_OS_MACOSX || FML_OS_IOS_SIMULATOR)
-  // We're using waitUntilScheduled on macOS and iOS simulator to force a hard
-  // barrier between the execution of different command buffers. This forces all
-  // renderable texture access to be synchronous (i.e. a write from a previous
-  // command buffer will not get scheduled to happen at the same time as a read
-  // in a future command buffer).
-  //
-  // Metal hazard tracks shared memory resources by default, and we don't need
-  // to do any additional work to synchronize access to MTLTextures and
-  // MTLBuffers on iOS devices with UMA. However, shared textures are disallowed
-  // on macOS according to the documentation:
-  // https://developer.apple.com/documentation/metal/mtlstoragemode/shared
-  // And so this is a stopgap solution that has been present in Impeller since
-  // multi-pass rendering/SaveLayer support was first set up.
-  //
-  // TODO(bdero): Remove this for all targets once a solution for resource
-  //              tracking that works everywhere is established:
-  //              https://github.com/flutter/flutter/issues/120406
-  [buffer_ waitUntilScheduled];
-#endif
-
   buffer_ = nil;
   return true;
 }

impeller/renderer/backend/metal/context_mtl.h

@@ -63,6 +63,8 @@ class ContextMTL final : public Context,
   // |Context|
   bool UpdateOffscreenLayerPixelFormat(PixelFormat format) override;
 
+  id<MTLCommandBuffer> CreateMTLCommandBuffer() const;
+
  private:
   id<MTLDevice> device_ = nullptr;
   id<MTLCommandQueue> command_queue_ = nullptr;

impeller/renderer/backend/metal/context_mtl.mm

@@ -289,4 +289,8 @@ static bool DeviceSupportsComputeSubgroups(id<MTLDevice> device) {
   return true;
 }
 
+id<MTLCommandBuffer> ContextMTL::CreateMTLCommandBuffer() const {
+  return [command_queue_ commandBuffer];
+}
+
 }  // namespace impeller

impeller/renderer/backend/metal/surface_mtl.h

@@ -43,9 +43,12 @@ class SurfaceMTL final : public Surface {
   id<MTLDrawable> drawable() const { return drawable_; }
 
  private:
+  std::weak_ptr<Context> context_;
   id<MTLDrawable> drawable_ = nil;
 
-  SurfaceMTL(const RenderTarget& target, id<MTLDrawable> drawable);
+  SurfaceMTL(const std::weak_ptr<Context>& context,
+             const RenderTarget& target,
+             id<MTLDrawable> drawable);
 
   // |Surface|
   bool Present() const override;

impeller/renderer/backend/metal/surface_mtl.mm

@@ -6,6 +6,7 @@
 
 #include "flutter/fml/trace_event.h"
 #include "impeller/base/validation.h"
+#include "impeller/renderer/backend/metal/context_mtl.h"
 #include "impeller/renderer/backend/metal/formats_mtl.h"
 #include "impeller/renderer/backend/metal/texture_mtl.h"
 #include "impeller/renderer/render_target.h"
@@ -111,12 +112,14 @@
   render_target_desc.SetStencilAttachment(stencil0);
 
   // The constructor is private. So make_unique may not be used.
-  return std::unique_ptr<SurfaceMTL>(
-      new SurfaceMTL(render_target_desc, current_drawable));
+  return std::unique_ptr<SurfaceMTL>(new SurfaceMTL(
+      context->weak_from_this(), render_target_desc, current_drawable));
 }
 
-SurfaceMTL::SurfaceMTL(const RenderTarget& target, id<MTLDrawable> drawable)
-    : Surface(target), drawable_(drawable) {}
+SurfaceMTL::SurfaceMTL(const std::weak_ptr<Context>& context,
+                       const RenderTarget& target,
+                       id<MTLDrawable> drawable)
+    : Surface(target), context_(context), drawable_(drawable) {}
 
 // |Surface|
 SurfaceMTL::~SurfaceMTL() = default;
@@ -127,7 +130,21 @@
     return false;
   }
 
+  auto context = context_.lock();
+  if (!context) {
+    return false;
+  }
+
+  // If a transaction is present, `presentDrawable` will present too early. And
+  // so we wait on an empty command buffer to get scheduled instead, which
+  // forces us to also wait for all of the previous command buffers in the queue
+  // to get scheduled.
+  id<MTLCommandBuffer> command_buffer =
+      ContextMTL::Cast(context.get())->CreateMTLCommandBuffer();
+  [command_buffer commit];
+  [command_buffer waitUntilScheduled];
   [drawable_ present];
+
   return true;
 }
 #pragma GCC diagnostic pop

shell/platform/darwin/ios/ios_surface_metal_impeller.mm

@@ -59,7 +59,7 @@
   // When there are platform views in the scene, the drawable needs to be presented in the same
   // transaction as the one created for platform views. When the drawable are being presented from
   // the raster thread, there is no such transaction.
-  layer.presentsWithTransaction = [[NSThread currentThread] isMainThread];
+  layer.presentsWithTransaction = YES;
 
   return layer;
 }