[GPU] Display swap output in the trace viewer

Resolve output is unreliable because resolving may be done to a subregion of a texture and even to 3D textures, and to any color format

src/xenia/app/emulator_window.cc

@@ -683,6 +683,7 @@ EmulatorWindow::GetGuestOutputPaintEffectForCvarValue(
 ui::Presenter::GuestOutputPaintConfig
 EmulatorWindow::GetGuestOutputPaintConfigForCvars() {
   ui::Presenter::GuestOutputPaintConfig paint_config;
+  paint_config.SetAllowOverscanCutoff(true);
   paint_config.SetEffect(GetGuestOutputPaintEffectForCvarValue(
       cvars::postprocess_scaling_and_sharpening));
   paint_config.SetCasAdditionalSharpness(

src/xenia/gpu/command_processor.cc

@@ -916,9 +916,7 @@ bool CommandProcessor::ExecutePacketType3_XE_SWAP(RingBuffer* reader,
   uint32_t frontbuffer_height = reader->ReadAndSwap<uint32_t>();
   reader->AdvanceRead((count - 4) * sizeof(uint32_t));
 
-  if (!ignore_swap_) {
+  IssueSwap(frontbuffer_ptr, frontbuffer_width, frontbuffer_height);
-    IssueSwap(frontbuffer_ptr, frontbuffer_width, frontbuffer_height);
-  }
 
   ++counter_;
   return true;

src/xenia/gpu/command_processor.h

@@ -90,7 +90,6 @@ class CommandProcessor {
 
   virtual void ClearCaches();
 
-  void SetIgnoreSwap(bool ignore_swap) { ignore_swap_ = ignore_swap; }
   // "Desired" is for the external thread managing the post-processing effect.
   SwapPostEffect GetDesiredSwapPostEffect() const {
     return swap_post_effect_desired_;
@@ -265,8 +264,6 @@ class CommandProcessor {
   std::atomic<bool> worker_running_;
   kernel::object_ref<kernel::XHostThread> worker_thread_;
 
-  bool ignore_swap_ = false;
-
   std::queue<std::function<void()>> pending_fns_;
 
   // MicroEngine binary from PM4_ME_INIT

src/xenia/gpu/d3d12/d3d12_command_processor.cc

@@ -1718,10 +1718,6 @@ void D3D12CommandProcessor::OnGammaRampPWLValueWritten() {
 void D3D12CommandProcessor::IssueSwap(uint32_t frontbuffer_ptr,
                                       uint32_t frontbuffer_width,
                                       uint32_t frontbuffer_height) {
-  // FIXME(Triang3l): frontbuffer_ptr is currently unreliable, in the trace
-  // player it's set to 0, but it's not needed anyway since the fetch constant
-  // contains the address.
-
   SCOPE_profile_cpu_f("gpu");
 
   ui::Presenter* presenter = graphics_system_->presenter();

src/xenia/gpu/trace_player.cc

@@ -89,19 +89,15 @@ void TracePlayer::PlayTrace(const uint8_t* trace_data, size_t trace_size,
                             TracePlaybackMode playback_mode,
                             bool clear_caches) {
   playing_trace_ = true;
-  // Pass a copy of present_last_copy_ to the thread so it's not accessible by
-  // multiple threads at once.
-  bool present_last_copy = present_last_copy_;
   graphics_system_->command_processor()->CallInThread([=]() {
-    PlayTraceOnThread(trace_data, trace_size, playback_mode, clear_caches,
+    PlayTraceOnThread(trace_data, trace_size, playback_mode, clear_caches);
-                      present_last_copy);
   });
 }
 
 void TracePlayer::PlayTraceOnThread(const uint8_t* trace_data,
                                     size_t trace_size,
                                     TracePlaybackMode playback_mode,
-                                    bool clear_caches, bool present_last_copy) {
+                                    bool clear_caches) {
   auto memory = graphics_system_->memory();
   auto command_processor = graphics_system_->command_processor();
 
@@ -109,10 +105,6 @@ void TracePlayer::PlayTraceOnThread(const uint8_t* trace_data,
     command_processor->ClearCaches();
   }
 
-  if (present_last_copy) {
-    command_processor->SetIgnoreSwap(true);
-  }
-
   playback_percent_ = 0;
   auto trace_end = trace_data + trace_size;
 
@@ -252,11 +244,6 @@ void TracePlayer::PlayTraceOnThread(const uint8_t* trace_data,
 
   playing_trace_ = false;
 
-  if (present_last_copy) {
-    command_processor->SetIgnoreSwap(false);
-    command_processor->IssueSwap(0, 1280, 720);
-  }
-
   playback_event_->Set();
 }
 

src/xenia/gpu/trace_player.h

@@ -32,9 +32,6 @@ class TracePlayer : public TraceReader {
   TracePlayer(GraphicsSystem* graphics_system);
 
   GraphicsSystem* graphics_system() const { return graphics_system_; }
-  void SetPresentLastCopy(bool present_last_copy) {
-    present_last_copy_ = present_last_copy;
-  }
   int current_frame_index() const { return current_frame_index_; }
   int current_command_index() const { return current_command_index_; }
   bool is_playing_trace() const { return playing_trace_; }
@@ -53,13 +50,9 @@ class TracePlayer : public TraceReader {
   void PlayTrace(const uint8_t* trace_data, size_t trace_size,
                  TracePlaybackMode playback_mode, bool clear_caches);
   void PlayTraceOnThread(const uint8_t* trace_data, size_t trace_size,
-                         TracePlaybackMode playback_mode, bool clear_caches,
+                         TracePlaybackMode playback_mode, bool clear_caches);
-                         bool present_last_copy);
 
   GraphicsSystem* graphics_system_;
-  // Whether to present the results of the latest resolve instead of displaying
-  // the front buffer from the trace.
-  bool present_last_copy_ = false;
   int current_frame_index_;
   int current_command_index_;
   bool playing_trace_ = false;

src/xenia/gpu/trace_viewer.cc

@@ -114,7 +114,7 @@ bool TraceViewer::Setup() {
   // Main display window.
   assert_true(app_context().IsInUIThread());
   window_ = xe::ui::Window::Create(app_context(), "xenia-gpu-trace-viewer",
-                                   1920, 1200);
+                                   1920, 1080);
   window_->AddListener(&window_listener_);
   window_->AddInputListener(&window_listener_, kZOrderTraceViewerInput);
   if (!window_->Open()) {
@@ -135,28 +135,27 @@ bool TraceViewer::Setup() {
   graphics_system_ = emulator_->graphics_system();
 
   player_ = std::make_unique<TracePlayer>(graphics_system_);
-  player_->SetPresentLastCopy(true);
 
   // Setup drawing to the window.
-  xe::ui::GraphicsProvider& graphics_provider = *graphics_system_->provider();
+  ui::Presenter* presenter = graphics_system_->presenter();
-  presenter_ = graphics_provider.CreatePresenter();
+  if (!presenter) {
-  if (!presenter_) {
     XELOGE("Failed to initialize the presenter");
     return false;
   }
+  xe::ui::GraphicsProvider& graphics_provider = *graphics_system_->provider();
   immediate_drawer_ = graphics_provider.CreateImmediateDrawer();
   if (!immediate_drawer_) {
     XELOGE("Failed to initialize the immediate drawer");
     return false;
   }
-  immediate_drawer_->SetPresenter(presenter_.get());
+  immediate_drawer_->SetPresenter(presenter);
   imgui_drawer_ =
       std::make_unique<xe::ui::ImGuiDrawer>(window_.get(), kZOrderImGui);
-  imgui_drawer_->SetPresenterAndImmediateDrawer(presenter_.get(),
+  imgui_drawer_->SetPresenterAndImmediateDrawer(presenter,
                                                 immediate_drawer_.get());
   trace_viewer_dialog_ = std::unique_ptr<TraceViewerDialog>(
       new TraceViewerDialog(imgui_drawer_.get(), *this));
-  window_->SetPresenter(presenter_.get());
+  window_->SetPresenter(presenter);
 
   return true;
 }

src/xenia/gpu/trace_viewer.h

@@ -21,7 +21,6 @@
 #include "xenia/ui/imgui_dialog.h"
 #include "xenia/ui/imgui_drawer.h"
 #include "xenia/ui/immediate_drawer.h"
-#include "xenia/ui/presenter.h"
 #include "xenia/ui/window.h"
 #include "xenia/ui/window_listener.h"
 #include "xenia/ui/windowed_app.h"
@@ -129,7 +128,6 @@ class TraceViewer : public xe::ui::WindowedApp {
   GraphicsSystem* graphics_system_ = nullptr;
   std::unique_ptr<TracePlayer> player_;
 
-  std::unique_ptr<xe::ui::Presenter> presenter_;
   std::unique_ptr<xe::ui::ImmediateDrawer> immediate_drawer_;
   std::unique_ptr<xe::ui::ImGuiDrawer> imgui_drawer_;
   std::unique_ptr<TraceViewerDialog> trace_viewer_dialog_;

src/xenia/gpu/vulkan/vulkan_command_processor.cc

@@ -269,11 +269,6 @@ void VulkanCommandProcessor::IssueSwap(uint32_t frontbuffer_ptr,
     return;
   }
 
-  if (!frontbuffer_ptr) {
-    // Trace viewer does this.
-    frontbuffer_ptr = last_copy_base_;
-  }
-
   std::vector<VkCommandBuffer> submit_buffers;
   if (frame_open_) {
     // TODO(DrChat): If the setup buffer is empty, don't bother queueing it up.
@@ -1108,9 +1103,6 @@ bool VulkanCommandProcessor::IssueCopy() {
 
   texture->in_flight_fence = current_batch_fence_;
 
-  // For debugging purposes only (trace viewer)
-  last_copy_base_ = texture->texture_info.memory.base_address;
-
   if (!frame_open_) {
     BeginFrame();
   } else if (current_render_state_) {

src/xenia/gpu/vulkan/vulkan_command_processor.h

@@ -106,9 +106,6 @@ class VulkanCommandProcessor : public CommandProcessor {
   uint32_t coher_base_vc_ = 0;
   uint32_t coher_size_vc_ = 0;
 
-  // Last copy base address, for debugging only.
-  uint32_t last_copy_base_ = 0;
-
   bool capturing_ = false;
   bool trace_requested_ = false;
   bool cache_clear_requested_ = false;

src/xenia/ui/presenter.cc

@@ -714,7 +714,8 @@ Presenter::GuestOutputPaintFlow Presenter::GetGuestOutputPaintFlow(
     // ratio while stretching throughout the entire surface's width, then limit
     // the Y cropping via letterboxing or stretching along X.
     uint32_t present_safe_area;
-    if (cvars::present_safe_area_y > 0 && cvars::present_safe_area_y < 100) {
+    if (config.GetAllowOverscanCutoff() && cvars::present_safe_area_y > 0 &&
+        cvars::present_safe_area_y < 100) {
       present_safe_area = uint32_t(cvars::present_safe_area_y);
     } else {
       present_safe_area = 100;
@@ -756,7 +757,8 @@ Presenter::GuestOutputPaintFlow Presenter::GetGuestOutputPaintFlow(
     // aspect ratio while stretching throughout the entire surface's height,
     // then limit the X cropping via letterboxing or stretching along Y.
     uint32_t present_safe_area;
-    if (cvars::present_safe_area_x > 0 && cvars::present_safe_area_x < 100) {
+    if (config.GetAllowOverscanCutoff() && cvars::present_safe_area_x > 0 &&
+        cvars::present_safe_area_x < 100) {
       present_safe_area = uint32_t(cvars::present_safe_area_x);
     } else {
       present_safe_area = 100;

src/xenia/ui/presenter.h

@@ -228,6 +228,11 @@ class Presenter {
     // In the sharpness setters, min / max with a constant as the first argument
     // also drops NaNs.
 
+    bool GetAllowOverscanCutoff() const { return allow_overscan_cutoff_; }
+    void SetAllowOverscanCutoff(bool new_allow_overscan_cutoff) {
+      allow_overscan_cutoff_ = new_allow_overscan_cutoff;
+    }
+
     Effect GetEffect() const { return effect_; }
     void SetEffect(Effect new_effect) { effect_ = new_effect; }
 
@@ -265,8 +270,10 @@ class Presenter {
     void SetDither(bool new_dither) { dither_ = new_dither; }
 
    private:
-    // Tools, rather than the emulator itself, must use kBilinear as the image
+    // Tools, rather than the emulator itself, must not allow overscan cutoff
-    // must be as close to the original front buffer as possible.
+    // and must use the kBilinear effect as the image must be as close to the
+    // original front buffer as possible.
+    bool allow_overscan_cutoff_ = false;
     Effect effect_ = Effect::kBilinear;
     float cas_additional_sharpness_ = kCasAdditionalSharpnessDefault;
     uint32_t fsr_max_upsampling_passes_ = kFsrMaxUpscalingPassesMax;