GPU: Avoid using garbage size for non-cb0 storage buffers (#4999)

* GPU: Avoid using garbage size for non-cb0 storage buffers

In the depths area, Tears of the Kingdom uses a global memory access with address on constant buffer slot 6. This isn't standard and thus doesn't actually have a size 8 bytes after it, so we were reading back a garbage size that ended up very large (at least in version 1.1.0), and would synchronize a lot of data per frame.

This PR makes storage buffers created from addresses outside constant buffer slot 0 get their size as the number of bytes remaining in the GPU mapping starting at the given virtual address. This should bound the buffer to a reasonable size, and ideally stop it crossing into other memory.

* Limit max size

* Add TODO

* Feedback

.github/workflows/build.yml

@@ -27,6 +27,7 @@ jobs:
   build:
     name: ${{ matrix.OS_NAME }} (${{ matrix.configuration }})
     runs-on: ${{ matrix.os }}
+    timeout-minutes: 35
     strategy:
       matrix:
         os: [ubuntu-latest, macOS-latest, windows-latest]
@@ -109,6 +110,7 @@ jobs:
   build_macos:
     name: macOS Universal (${{ matrix.configuration }})
     runs-on: ubuntu-latest
+    timeout-minutes: 35
     strategy:
       matrix:
         configuration: [ Debug, Release ]
@@ -150,4 +152,4 @@ jobs:
         with:
           name: ava-ryujinx-${{ matrix.configuration }}-${{ env.RYUJINX_BASE_VERSION }}+${{ steps.git_short_hash.outputs.result }}-macos_universal
           path: "publish_ava/*.tar.gz"
-        if: github.event_name == 'pull_request'
+        if: github.event_name == 'pull_request'

.github/workflows/flatpak.yml

@@ -12,6 +12,7 @@ concurrency: flatpak-release
 
 jobs:
   release:
+    timeout-minutes: 35
     runs-on: ubuntu-latest
 
     env:

.github/workflows/nightly_pr_comment.yml

@@ -7,6 +7,7 @@ jobs:
   pr_comment:
     if: github.event.workflow_run.event == 'pull_request' && github.event.workflow_run.conclusion == 'success'
     runs-on: ubuntu-latest
+    timeout-minutes: 35
     steps:
       - uses: actions/github-script@v6
         with:
@@ -65,4 +66,4 @@ jobs:
             } else {
               core.info(`Creating a comment`);
               await github.rest.issues.createComment({repo, owner, issue_number, body});
-            }
+            }

.github/workflows/release.yml

@@ -46,6 +46,7 @@ jobs:
   release:
     name: Release ${{ matrix.OS_NAME }}
     runs-on: ${{ matrix.os }}
+    timeout-minutes: 35
     strategy:
       matrix:
         os: [ ubuntu-latest, windows-latest ]
@@ -143,13 +144,14 @@ jobs:
   macos_release:
     name: Release MacOS universal
     runs-on: ubuntu-latest
+    timeout-minutes: 35
     steps:
       - uses: actions/checkout@v3
 
       - uses: actions/setup-dotnet@v3
         with:
           global-json-file: global.json
-          
+
       - name: Setup LLVM 14
         run: |
           wget https://apt.llvm.org/llvm.sh
@@ -205,4 +207,4 @@ jobs:
     needs: release
     with:
       ryujinx_version: "1.1.${{ github.run_number }}"
-    secrets: inherit
+    secrets: inherit

distribution/macos/updater.sh

@@ -36,4 +36,9 @@ sleep 1
 # Now replace and reopen.
 rm -rf "$INSTALL_DIRECTORY"
 mv "$NEW_APP_DIRECTORY" "$INSTALL_DIRECTORY"
-open -a "$INSTALL_DIRECTORY" --args "$APP_ARGUMENTS"
+
+if [ "$#" -le 3 ]; then
+    open -a "$INSTALL_DIRECTORY"
+else
+    open -a "$INSTALL_DIRECTORY" --args "$APP_ARGUMENTS"
+fi

src/Ryujinx.Ava/AppHost.cs

@@ -86,7 +86,7 @@ namespace Ryujinx.Ava
         private KeyboardHotkeyState         _prevHotkeyState;
 
         private long _lastCursorMoveTime;
-        private bool _isCursorInRenderer;
+        private bool _isCursorInRenderer = true;
 
         private bool _isStopped;
         private bool _isActive;
@@ -160,7 +160,9 @@ namespace Ryujinx.Ava
 
             ConfigurationState.Instance.HideCursor.Event += HideCursorState_Changed;
 
-            _topLevel.PointerMoved += TopLevel_PointerMoved;
+            _topLevel.PointerMoved += TopLevel_PointerEnterOrMoved;
+            _topLevel.PointerEnter += TopLevel_PointerEnterOrMoved;
+            _topLevel.PointerLeave += TopLevel_PointerLeave;
 
             if (OperatingSystem.IsWindows())
             {
@@ -183,7 +185,7 @@ namespace Ryujinx.Ava
             _gpuCancellationTokenSource = new CancellationTokenSource();
         }
 
-        private void TopLevel_PointerMoved(object sender, PointerEventArgs e)
+        private void TopLevel_PointerEnterOrMoved(object sender, PointerEventArgs e)
         {
             if (sender is MainWindow window)
             {
@@ -201,6 +203,12 @@ namespace Ryujinx.Ava
                 }
             }
         }
+
+        private void TopLevel_PointerLeave(object sender, PointerEventArgs e)
+        {
+            _isCursorInRenderer = false;
+        }
+
         private void UpdateScalingFilterLevel(object sender, ReactiveEventArgs<int> e)
         {
             _renderer.Window?.SetScalingFilter((Graphics.GAL.ScalingFilter)ConfigurationState.Instance.Graphics.ScalingFilter.Value);
@@ -446,7 +454,9 @@ namespace Ryujinx.Ava
             ConfigurationState.Instance.Graphics.ScalingFilterLevel.Event  -= UpdateScalingFilterLevel;
             ConfigurationState.Instance.Graphics.AntiAliasing.Event        -= UpdateAntiAliasing;
 
-            _topLevel.PointerMoved -= TopLevel_PointerMoved;
+            _topLevel.PointerMoved -= TopLevel_PointerEnterOrMoved;
+            _topLevel.PointerEnter -= TopLevel_PointerEnterOrMoved;
+            _topLevel.PointerLeave -= TopLevel_PointerLeave;
 
             _gpuCancellationTokenSource.Cancel();
             _gpuCancellationTokenSource.Dispose();

src/Ryujinx.Ava/Common/ApplicationHelper.cs

@@ -193,7 +193,7 @@ namespace Ryujinx.Ava.Common
                             if (nca.Header.ContentType == NcaContentType.Program)
                             {
                                 int dataIndex = Nca.GetSectionIndexFromType(NcaSectionType.Data, NcaContentType.Program);
-                                if (nca.Header.GetFsHeader(dataIndex).IsPatchSection())
+                                if (nca.SectionExists(NcaSectionType.Data) && nca.Header.GetFsHeader(dataIndex).IsPatchSection())
                                 {
                                     patchNca = nca;
                                 }

src/Ryujinx.Graphics.Gpu/Engine/Threed/StateUpdater.cs

@@ -23,6 +23,8 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
         public const int PrimitiveRestartStateIndex = 12;
         public const int RenderTargetStateIndex = 27;
 
+        private const ulong MaxUnknownStorageSize = 0x100000;
+
         private readonly GpuContext _context;
         private readonly GpuChannel _channel;
         private readonly DeviceStateWithShadow<ThreedClassState> _state;
@@ -356,7 +358,19 @@ namespace Ryujinx.Graphics.Gpu.Engine.Threed
 
                     SbDescriptor sbDescriptor = _channel.MemoryManager.Physical.Read<SbDescriptor>(sbDescAddress);
 
-                    _channel.BufferManager.SetGraphicsStorageBuffer(stage, sb.Slot, sbDescriptor.PackAddress(), (uint)sbDescriptor.Size, sb.Flags);
+                    uint size;
+                    if (sb.SbCbSlot == 0)
+                    {
+                        // Only trust the SbDescriptor size if it comes from slot 0.
+                        size = (uint)sbDescriptor.Size;
+                    }
+                    else
+                    {
+                        // TODO: Use full mapped size and somehow speed up buffer sync.
+                        size = (uint)_channel.MemoryManager.GetMappedSize(sbDescriptor.PackAddress(), MaxUnknownStorageSize);
+                    }
+
+                    _channel.BufferManager.SetGraphicsStorageBuffer(stage, sb.Slot, sbDescriptor.PackAddress(), size, sb.Flags);
                 }
             }
         }

src/Ryujinx.Graphics.Gpu/GpuContext.cs

@@ -99,6 +99,7 @@ namespace Ryujinx.Graphics.Gpu
         private bool _pendingSync;
 
         private long _modifiedSequence;
+        private ulong _firstTimestamp;
 
         /// <summary>
         /// Creates a new instance of the GPU emulation context.
@@ -123,6 +124,8 @@ namespace Ryujinx.Graphics.Gpu
             DeferredActions = new Queue<Action>();
 
             PhysicalMemoryRegistry = new ConcurrentDictionary<ulong, PhysicalMemory>();
+
+            _firstTimestamp = ConvertNanosecondsToTicks((ulong)PerformanceCounter.ElapsedNanoseconds);
         }
 
         /// <summary>
@@ -217,7 +220,8 @@ namespace Ryujinx.Graphics.Gpu
         /// <returns>The current GPU timestamp</returns>
         public ulong GetTimestamp()
         {
-            ulong ticks = ConvertNanosecondsToTicks((ulong)PerformanceCounter.ElapsedNanoseconds);
+            // Guest timestamp will start at 0, instead of host value. 
+            ulong ticks = ConvertNanosecondsToTicks((ulong)PerformanceCounter.ElapsedNanoseconds) - _firstTimestamp;
 
             if (GraphicsConfig.FastGpuTime)
             {

src/Ryujinx.Graphics.Gpu/Memory/MemoryManager.cs

@@ -637,6 +637,33 @@ namespace Ryujinx.Graphics.Gpu.Memory
             return UnpackPaFromPte(pte) + (va & PageMask);
         }
 
+        /// <summary>
+        /// Translates a GPU virtual address and returns the number of bytes that are mapped after it.
+        /// </summary>
+        /// <param name="va">GPU virtual address to be translated</param>
+        /// <param name="maxSize">Maximum size in bytes to scan</param>
+        /// <returns>Number of bytes, 0 if unmapped</returns>
+        public ulong GetMappedSize(ulong va, ulong maxSize)
+        {
+            if (!ValidateAddress(va))
+            {
+                return 0;
+            }
+
+            ulong startVa = va;
+            ulong endVa = va + maxSize;
+
+            ulong pte = GetPte(va);
+
+            while (pte != PteUnmapped && va < endVa)
+            {
+                va += PageSize - (va & PageMask);
+                pte = GetPte(va);
+            }
+
+            return Math.Min(maxSize, va - startVa);
+        }
+
         /// <summary>
         /// Gets the kind of a given memory page.
         /// This might indicate the type of resource that can be allocated on the page, and also texture tiling.

src/Ryujinx.Graphics.Vulkan/BufferHolder.cs

@@ -56,6 +56,7 @@ namespace Ryujinx.Graphics.Vulkan
         private int _writeCount;
         private int _flushCount;
         private int _flushTemp;
+        private int _lastFlushWrite = -1;
 
         private ReaderWriterLock _flushLock;
         private FenceHolder _flushFence;
@@ -200,14 +201,21 @@ namespace Ryujinx.Graphics.Vulkan
         {
             if (_baseType == BufferAllocationType.Auto)
             {
-                if (_writeCount >= WriteCountThreshold || _setCount >= SetCountThreshold || _flushCount >= FlushCountThreshold)
+                // When flushed, wait for a bit more info to make a decision.
+                bool wasFlushed = _flushTemp > 0;
+                int multiplier = wasFlushed ? 2 : 0;
+                if (_writeCount >= (WriteCountThreshold << multiplier) || _setCount >= (SetCountThreshold << multiplier) || _flushCount >= (FlushCountThreshold << multiplier))
                 {
                     if (_flushCount > 0 || _flushTemp-- > 0)
                     {
                         // Buffers that flush should ideally be mapped in host address space for easy copies.
                         // If the buffer is large it will do better on GPU memory, as there will be more writes than data flushes (typically individual pages).
                         // If it is small, then it's likely most of the buffer will be flushed so we want it on host memory, as access is cached.
-                        DesiredType = Size > DeviceLocalSizeThreshold ? BufferAllocationType.DeviceLocalMapped : BufferAllocationType.HostMapped;
+
+                        bool hostMappingSensitive = _gd.Vendor == Vendor.Nvidia;
+                        bool deviceLocalMapped = Size > DeviceLocalSizeThreshold || (wasFlushed && _writeCount > _flushCount * 10 && hostMappingSensitive) || _currentType == BufferAllocationType.DeviceLocalMapped;
+
+                        DesiredType = deviceLocalMapped ? BufferAllocationType.DeviceLocalMapped : BufferAllocationType.HostMapped;
 
                         // It's harder for a buffer that is flushed to revert to another type of mapping.
                         if (_flushCount > 0)
@@ -215,17 +223,18 @@ namespace Ryujinx.Graphics.Vulkan
                             _flushTemp = 1000;
                         }
                     }
-                    else if (_writeCount >= WriteCountThreshold)
+                    else if (_writeCount >= (WriteCountThreshold << multiplier))
                     {
                         // Buffers that are written often should ideally be in the device local heap. (Storage buffers)
                         DesiredType = BufferAllocationType.DeviceLocal;
                     }
-                    else if (_setCount > SetCountThreshold)
+                    else if (_setCount > (SetCountThreshold << multiplier))
                     {
                         // Buffers that have their data set often should ideally be host mapped. (Constant buffers)
                         DesiredType = BufferAllocationType.HostMapped;
                     }
 
+                    _lastFlushWrite = -1;
                     _flushCount = 0;
                     _writeCount = 0;
                     _setCount = 0;
@@ -418,7 +427,12 @@ namespace Ryujinx.Graphics.Vulkan
 
             WaitForFlushFence();
 
-            _flushCount++;
+            if (_lastFlushWrite != _writeCount)
+            {
+                // If it's on the same page as the last flush, ignore it.
+                _lastFlushWrite = _writeCount;
+                _flushCount++;
+            }
 
             Span<byte> result;
 

src/Ryujinx/Ui/RendererWidgetBase.cs

@@ -321,27 +321,27 @@ namespace Ryujinx.Ui
 
             _toggleDockedMode = toggleDockedMode;
 
-            if (ConfigurationState.Instance.Hid.EnableMouse.Value)
+            if (_isMouseInClient)
             {
-                if (_isMouseInClient)
+                if (ConfigurationState.Instance.Hid.EnableMouse.Value)
                 {
                     Window.Cursor = _invisibleCursor;
                 }
-            }
-            else
-            {
-                switch (_hideCursorMode)
+                else
                 {
-                    case HideCursorMode.OnIdle:
-                        long cursorMoveDelta = Stopwatch.GetTimestamp() - _lastCursorMoveTime;
-                        Window.Cursor = (cursorMoveDelta >= CursorHideIdleTime * Stopwatch.Frequency) ? _invisibleCursor : null;
-                        break;
-                    case HideCursorMode.Always:
-                        Window.Cursor = _invisibleCursor;
-                        break;
-                    case HideCursorMode.Never:
-                        Window.Cursor = null;
-                        break;
+                    switch (_hideCursorMode)
+                    {
+                        case HideCursorMode.OnIdle:
+                            long cursorMoveDelta = Stopwatch.GetTimestamp() - _lastCursorMoveTime;
+                            Window.Cursor = (cursorMoveDelta >= CursorHideIdleTime * Stopwatch.Frequency) ? _invisibleCursor : null;
+                            break;
+                        case HideCursorMode.Always:
+                            Window.Cursor = _invisibleCursor;
+                            break;
+                        case HideCursorMode.Never:
+                            Window.Cursor = null;
+                            break;
+                    }
                 }
             }
         }