[Impeller] Remove the unused Vulkan queue submit task runner (#50511)

see #50484

impeller/renderer/backend/vulkan/context_vk.cc

@@ -132,12 +132,6 @@ void ContextVK::Setup(Settings settings) {
     return;
   }
 
-  queue_submit_thread_ = std::make_unique<fml::Thread>("IplrVkQueueSub");
-  queue_submit_thread_->GetTaskRunner()->PostTask([]() {
-    // submitKHR is extremely cheap and mostly blocks on an internal fence.
-    fml::RequestAffinity(fml::CpuAffinity::kEfficiency);
-  });
-
   raster_message_loop_ = fml::ConcurrentMessageLoop::Create(
       std::min(4u, std::thread::hardware_concurrency()));
   raster_message_loop_->PostTaskToAllWorkers([]() {
@@ -507,10 +501,6 @@ const vk::Device& ContextVK::GetDevice() const {
   return device_holder_->device.get();
 }
 
-const fml::RefPtr<fml::TaskRunner> ContextVK::GetQueueSubmitRunner() const {
-  return queue_submit_thread_->GetTaskRunner();
-}
-
 const std::shared_ptr<fml::ConcurrentTaskRunner>
 ContextVK::GetConcurrentWorkerTaskRunner() const {
   return raster_message_loop_->GetTaskRunner();
@@ -525,7 +515,6 @@ void ContextVK::Shutdown() {
   fence_waiter_.reset();
   resource_manager_.reset();
 
-  queue_submit_thread_->Join();
   raster_message_loop_->Terminate();
 }
 

impeller/renderer/backend/vulkan/context_vk.h

@@ -133,18 +133,6 @@ class ContextVK final : public Context,
   const std::shared_ptr<fml::ConcurrentTaskRunner>
   GetConcurrentWorkerTaskRunner() const;
 
-  /// @brief A single-threaded task runner that should only be used for
-  ///        submitKHR.
-  ///
-  /// SubmitKHR will block until all previously submitted command buffers have
-  /// been scheduled. If there are no platform views in the scene (excluding
-  /// texture backed platform views). Then it is safe for SwapchainImpl::Present
-  /// to return before submit has completed. To do so, we offload the submit
-  /// command to a specialized single threaded task runner. The single thread
-  /// ensures that we do not queue up too much work and that the submissions
-  /// proceed in order.
-  const fml::RefPtr<fml::TaskRunner> GetQueueSubmitRunner() const;
-
   std::shared_ptr<SurfaceContextVK> CreateSurfaceContext();
 
   const std::shared_ptr<QueueVK>& GetGraphicsQueue() const;
@@ -192,7 +180,6 @@ class ContextVK final : public Context,
   std::shared_ptr<CommandPoolRecyclerVK> command_pool_recycler_;
   std::string device_name_;
   std::shared_ptr<fml::ConcurrentMessageLoop> raster_message_loop_;
-  std::unique_ptr<fml::Thread> queue_submit_thread_;
   std::shared_ptr<GPUTracerVK> gpu_tracer_;
   std::shared_ptr<DescriptorPoolRecyclerVK> descriptor_pool_recycler_;
   std::shared_ptr<CommandQueue> command_queue_vk_;