Vulkan: Don't create preload command buffer outside a render pass (#3864)

* Vulkan: Don't create preload buffer outside a render pass

The preload command buffer is used to avoid render pass splits and barriers when updating buffer data. However, when a render pass is not active (for example, at the start of a pass, or during compute invocations) buffer uploads can be performed at any time, so the optimization isn't as useful.

This PR makes it so that the preload command buffer is only used for buffer updates outside of a render pass. It's still used for textures as I don't want to shake things up right now regarding how the preload buffer is obtained before some other changes, and texture updates are a lot rarer anyways.

Improves performance slightly in Pokemon Scarlet/Violet (43 -> 48), as it was switching to compute, writing a bunch of buffers inline, then dispatching, then flushing commands... It uses 1 command buffer instead of 2 every time it does this now. Maybe it would be nice to find a faster way to sync without creating so many command buffers in a short period of time.

* Address feedback

Ryujinx.Graphics.Gpu/GpuChannel.cs

@@ -67,6 +67,7 @@ namespace Ryujinx.Graphics.Gpu
 
             // Since the memory manager changed, make sure we will get pools from addresses of the new memory manager.
             TextureManager.ReloadPools();
+            MemoryManager.Physical.BufferCache.QueuePrune();
         }
 
         /// <summary>
@@ -77,6 +78,7 @@ namespace Ryujinx.Graphics.Gpu
         private void MemoryUnmappedHandler(object sender, UnmapEventArgs e)
         {
             TextureManager.ReloadPools();
+            MemoryManager.Physical.BufferCache.QueuePrune();
         }
 
         /// <summary>

Ryujinx.Graphics.Gpu/Memory/BufferCache.cs

@@ -28,6 +28,7 @@ namespace Ryujinx.Graphics.Gpu.Memory
 
         private readonly Dictionary<ulong, BufferCacheEntry> _dirtyCache;
         private readonly Dictionary<ulong, BufferCacheEntry> _modifiedCache;
+        private bool _pruneCaches;
 
         public event Action NotifyBuffersModified;
 
@@ -136,6 +137,11 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// <param name="size">Size in bytes of the buffer</param>
         public void ForceDirty(MemoryManager memoryManager, ulong gpuVa, ulong size)
         {
+            if (_pruneCaches)
+            {
+                Prune();
+            }
+
             if (!_dirtyCache.TryGetValue(gpuVa, out BufferCacheEntry result) ||
                 result.EndGpuAddress < gpuVa + size ||
                 result.UnmappedSequence != result.Buffer.UnmappedSequence)
@@ -158,17 +164,29 @@ namespace Ryujinx.Graphics.Gpu.Memory
         /// <returns>True if modified, false otherwise</returns>
         public bool CheckModified(MemoryManager memoryManager, ulong gpuVa, ulong size, out ulong outAddr)
         {
-            if (!_modifiedCache.TryGetValue(gpuVa, out BufferCacheEntry result) ||
-                result.EndGpuAddress < gpuVa + size ||
-                result.UnmappedSequence != result.Buffer.UnmappedSequence)
+            if (_pruneCaches)
             {
-                ulong address = TranslateAndCreateBuffer(memoryManager, gpuVa, size);
-                result = new BufferCacheEntry(address, gpuVa, GetBuffer(address, size));
-
-                _modifiedCache[gpuVa] = result;
+                Prune();
             }
 
-            outAddr = result.Address;
+            // Align the address to avoid creating too many entries on the quick lookup dictionary.
+            ulong mask = BufferAlignmentMask;
+            ulong alignedGpuVa = gpuVa & (~mask);
+            ulong alignedEndGpuVa = (gpuVa + size + mask) & (~mask);
+
+            size = alignedEndGpuVa - alignedGpuVa;
+
+            if (!_modifiedCache.TryGetValue(alignedGpuVa, out BufferCacheEntry result) ||
+                result.EndGpuAddress < alignedEndGpuVa ||
+                result.UnmappedSequence != result.Buffer.UnmappedSequence)
+            {
+                ulong address = TranslateAndCreateBuffer(memoryManager, alignedGpuVa, size);
+                result = new BufferCacheEntry(address, alignedGpuVa, GetBuffer(address, size));
+
+                _modifiedCache[alignedGpuVa] = result;
+            }
+
+            outAddr = result.Address | (gpuVa & mask);
 
             return result.Buffer.IsModified(result.Address, size);
         }
@@ -435,6 +453,54 @@ namespace Ryujinx.Graphics.Gpu.Memory
             }
         }
 
+        /// <summary>
+        /// Prune any invalid entries from a quick access dictionary.
+        /// </summary>
+        /// <param name="dictionary">Dictionary to prune</param>
+        /// <param name="toDelete">List used to track entries to delete</param>
+        private void Prune(Dictionary<ulong, BufferCacheEntry> dictionary, ref List<ulong> toDelete)
+        {
+            foreach (var entry in dictionary)
+            {
+                if (entry.Value.UnmappedSequence != entry.Value.Buffer.UnmappedSequence)
+                {
+                    (toDelete ??= new()).Add(entry.Key);
+                }
+            }
+
+            if (toDelete != null)
+            {
+                foreach (ulong entry in toDelete)
+                {
+                    dictionary.Remove(entry);
+                }
+            }
+        }
+
+        /// <summary>
+        /// Prune any invalid entries from the quick access dictionaries.
+        /// </summary>
+        private void Prune()
+        {
+            List<ulong> toDelete = null;
+
+            Prune(_dirtyCache, ref toDelete);
+
+            toDelete?.Clear();
+
+            Prune(_modifiedCache, ref toDelete);
+
+            _pruneCaches = false;
+        }
+
+        /// <summary>
+        /// Queues a prune of invalid entries the next time a dictionary cache is accessed.
+        /// </summary>
+        public void QueuePrune()
+        {
+            _pruneCaches = true;
+        }
+
         /// <summary>
         /// Disposes all buffers in the cache.
         /// It's an error to use the buffer manager after disposal.

Ryujinx.Graphics.Vulkan/BufferHolder.cs

@@ -210,7 +210,10 @@ namespace Ryujinx.Graphics.Vulkan
                 }
             }
 
-            if (cbs != null && !(_buffer.HasCommandBufferDependency(cbs.Value) && _waitable.IsBufferRangeInUse(cbs.Value.CommandBufferIndex, offset, dataSize)))
+            if (cbs != null &&
+                _gd.PipelineInternal.RenderPassActive &&
+                !(_buffer.HasCommandBufferDependency(cbs.Value) &&
+                _waitable.IsBufferRangeInUse(cbs.Value.CommandBufferIndex, offset, dataSize)))
             {
                 // If the buffer hasn't been used on the command buffer yet, try to preload the data.
                 // This avoids ending and beginning render passes on each buffer data upload.

Ryujinx.Graphics.Vulkan/PipelineBase.cs

@@ -49,7 +49,6 @@ namespace Ryujinx.Graphics.Vulkan
         private Auto<DisposableFramebuffer> _framebuffer;
         private Auto<DisposableRenderPass> _renderPass;
         private int _writtenAttachmentCount;
-        private bool _renderPassActive;
 
         private readonly DescriptorSetUpdater _descriptorSetUpdater;
 
@@ -73,6 +72,7 @@ namespace Ryujinx.Graphics.Vulkan
         private PipelineColorBlendAttachmentState[] _storedBlend;
 
         public ulong DrawCount { get; private set; }
+        public bool RenderPassActive { get; private set; }
 
         public unsafe PipelineBase(VulkanRenderer gd, Device device)
         {
@@ -838,6 +838,11 @@ namespace Ryujinx.Graphics.Vulkan
             stages.CopyTo(_newState.Stages.AsSpan().Slice(0, stages.Length));
 
             SignalStateChange();
+
+            if (_program.IsCompute)
+            {
+                EndRenderPass();
+            }
         }
 
         public void Specialize<T>(in T data) where T : unmanaged
@@ -1451,7 +1456,7 @@ namespace Ryujinx.Graphics.Vulkan
 
         private unsafe void BeginRenderPass()
         {
-            if (!_renderPassActive)
+            if (!RenderPassActive)
             {
                 var renderArea = new Rect2D(null, new Extent2D(FramebufferParams.Width, FramebufferParams.Height));
                 var clearValue = new ClearValue();
@@ -1467,18 +1472,18 @@ namespace Ryujinx.Graphics.Vulkan
                 };
 
                 Gd.Api.CmdBeginRenderPass(CommandBuffer, renderPassBeginInfo, SubpassContents.Inline);
-                _renderPassActive = true;
+                RenderPassActive = true;
             }
         }
 
         public void EndRenderPass()
         {
-            if (_renderPassActive)
+            if (RenderPassActive)
             {
                 PauseTransformFeedbackInternal();
                 Gd.Api.CmdEndRenderPass(CommandBuffer);
                 SignalRenderPassEnd();
-                _renderPassActive = false;
+                RenderPassActive = false;
             }
         }
 

Ryujinx.Graphics.Vulkan/ShaderCollection.cs

@@ -19,6 +19,7 @@ namespace Ryujinx.Graphics.Vulkan
 
         public bool HasMinimalLayout { get; }
         public bool UsePushDescriptors { get; }
+        public bool IsCompute { get; }
 
         public uint Stages { get; }
 
@@ -47,7 +48,6 @@ namespace Ryujinx.Graphics.Vulkan
         private VulkanRenderer _gd;
         private Device _device;
         private bool _initialized;
-        private bool _isCompute;
 
         private ProgramPipelineState _state;
         private DisposableRenderPass _dummyRenderPass;
@@ -91,7 +91,7 @@ namespace Ryujinx.Graphics.Vulkan
 
                 if (shader.StageFlags == ShaderStageFlags.ShaderStageComputeBit)
                 {
-                    _isCompute = true;
+                    IsCompute = true;
                 }
 
                 internalShaders[i] = shader;
@@ -163,7 +163,7 @@ namespace Ryujinx.Graphics.Vulkan
 
             try
             {
-                if (_isCompute)
+                if (IsCompute)
                 {
                     CreateBackgroundComputePipeline();
                 }