Fix Vulkan blit-like operations swizzle (#5003)

* 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

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/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/DescriptorSetUpdater.cs

@@ -228,7 +228,12 @@ namespace Ryujinx.Graphics.Vulkan
             SignalDirty(DirtyFlags.Storage);
         }
 
-        public void SetTextureAndSampler(CommandBufferScoped cbs, ShaderStage stage, int binding, ITexture texture, ISampler sampler)
+        public void SetTextureAndSampler(
+            CommandBufferScoped cbs,
+            ShaderStage stage,
+            int binding,
+            ITexture texture,
+            ISampler sampler)
         {
             if (texture is TextureBuffer textureBuffer)
             {
@@ -251,6 +256,28 @@ namespace Ryujinx.Graphics.Vulkan
             SignalDirty(DirtyFlags.Texture);
         }
 
+        public void SetTextureAndSamplerIdentitySwizzle(
+            CommandBufferScoped cbs,
+            ShaderStage stage,
+            int binding,
+            ITexture texture,
+            ISampler sampler)
+        {
+            if (texture is TextureView view)
+            {
+                view.Storage.InsertWriteToReadBarrier(cbs, AccessFlags.ShaderReadBit, stage.ConvertToPipelineStageFlags());
+
+                _textureRefs[binding] = view.GetIdentityImageView();
+                _samplerRefs[binding] = ((SamplerHolder)sampler)?.GetSampler();
+
+                SignalDirty(DirtyFlags.Texture);
+            }
+            else
+            {
+                SetTextureAndSampler(cbs, stage, binding, texture, sampler);
+            }
+        }
+
         public void SetUniformBuffers(CommandBuffer commandBuffer, ReadOnlySpan<BufferAssignment> buffers)
         {
             for (int i = 0; i < buffers.Length; i++)

src/Ryujinx.Graphics.Vulkan/HelperShader.cs

@@ -415,7 +415,7 @@ namespace Ryujinx.Graphics.Vulkan
 
             var sampler = linearFilter ? _samplerLinear : _samplerNearest;
 
-            _pipeline.SetTextureAndSampler(ShaderStage.Fragment, 0, src, sampler);
+            _pipeline.SetTextureAndSamplerIdentitySwizzle(ShaderStage.Fragment, 0, src, sampler);
 
             Span<float> region = stackalloc float[RegionBufferSize / sizeof(float)];
 
@@ -625,7 +625,7 @@ namespace Ryujinx.Graphics.Vulkan
 
         private void BlitDepthStencilDraw(TextureView src, bool isDepth)
         {
-            _pipeline.SetTextureAndSampler(ShaderStage.Fragment, 0, src, _samplerNearest);
+            _pipeline.SetTextureAndSamplerIdentitySwizzle(ShaderStage.Fragment, 0, src, _samplerNearest);
 
             if (isDepth)
             {
@@ -1037,7 +1037,7 @@ namespace Ryujinx.Graphics.Vulkan
                     var srcView = Create2DLayerView(src, srcLayer + z, srcLevel + l, srcFormat);
                     var dstView = Create2DLayerView(dst, dstLayer + z, dstLevel + l);
 
-                    _pipeline.SetTextureAndSampler(ShaderStage.Compute, 0, srcView, null);
+                    _pipeline.SetTextureAndSamplerIdentitySwizzle(ShaderStage.Compute, 0, srcView, null);
                     _pipeline.SetImage(0, dstView, dstFormat);
 
                     int dispatchX = (Math.Min(srcView.Info.Width, dstView.Info.Width) + 31) / 32;
@@ -1177,7 +1177,7 @@ namespace Ryujinx.Graphics.Vulkan
                     var srcView = Create2DLayerView(src, srcLayer + z, 0, format);
                     var dstView = Create2DLayerView(dst, dstLayer + z, 0);
 
-                    _pipeline.SetTextureAndSampler(ShaderStage.Compute, 0, srcView, null);
+                    _pipeline.SetTextureAndSamplerIdentitySwizzle(ShaderStage.Compute, 0, srcView, null);
                     _pipeline.SetImage(0, dstView, format);
 
                     _pipeline.DispatchCompute(dispatchX, dispatchY, 1);
@@ -1313,7 +1313,7 @@ namespace Ryujinx.Graphics.Vulkan
                     var srcView = Create2DLayerView(src, srcLayer + z, 0, format);
                     var dstView = Create2DLayerView(dst, dstLayer + z, 0);
 
-                    _pipeline.SetTextureAndSampler(ShaderStage.Fragment, 0, srcView, null);
+                    _pipeline.SetTextureAndSamplerIdentitySwizzle(ShaderStage.Fragment, 0, srcView, null);
                     _pipeline.SetRenderTarget(
                         ((TextureView)dstView).GetView(format).GetImageViewForAttachment(),
                         (uint)dst.Width,
@@ -1384,7 +1384,7 @@ namespace Ryujinx.Graphics.Vulkan
 
         private void CopyMSAspectDraw(TextureView src, bool fromMS, bool isDepth)
         {
-            _pipeline.SetTextureAndSampler(ShaderStage.Fragment, 0, src, _samplerNearest);
+            _pipeline.SetTextureAndSamplerIdentitySwizzle(ShaderStage.Fragment, 0, src, _samplerNearest);
 
             if (isDepth)
             {

src/Ryujinx.Graphics.Vulkan/PipelineBase.cs

@@ -1098,6 +1098,11 @@ namespace Ryujinx.Graphics.Vulkan
             _descriptorSetUpdater.SetTextureAndSampler(Cbs, stage, binding, texture, sampler);
         }
 
+        public void SetTextureAndSamplerIdentitySwizzle(ShaderStage stage, int binding, ITexture texture, ISampler sampler)
+        {
+            _descriptorSetUpdater.SetTextureAndSamplerIdentitySwizzle(Cbs, stage, binding, texture, sampler);
+        }
+
         public void SetTransformFeedbackBuffers(ReadOnlySpan<BufferRange> buffers)
         {
             PauseTransformFeedbackInternal();