Implement HLE macro for DrawElementsIndirect (#3748)

* Implement HLE macro for DrawElementsIndirect

* Shader cache version bump

* Use GL_ARB_shader_draw_parameters extension on OpenGL

* Fix DrawIndexedIndirectCount on Vulkan when extension is not supported

* Implement DrawIndex

* Alignment

* Fix some validation errors

* Rename BaseIds to DrawParameters

* Fix incorrect index buffer and vertex buffer size in some cases

* Add HLE macros for DrawArraysInstanced and DrawElementsInstanced

* Perform a regular draw when indirect data is not modified

* Use non-indirect draw methods if indirect buffer was not GPU modified

* Only check if draw parameters match if the shader actually uses them

* Expose Macro HLE setting on GUI

* Reset FirstVertex and FirstInstance after draw

* Update shader cache version again since some people already tested this

* PR feedback

Co-authored-by: riperiperi <rhy3756547@hotmail.com>

7 files changed, 47 insertions, 5 deletions

diff --git a/Ryujinx.Graphics.Gpu/Shader/DiskCache/DiskCacheGpuAccessor.cs b/Ryujinx.Graphics.Gpu/Shader/DiskCache/DiskCacheGpuAccessor.cs
index 68ff4f2a..98748bf6 100644
--- a/Ryujinx.Graphics.Gpu/Shader/DiskCache/DiskCacheGpuAccessor.cs
+++ b/Ryujinx.Graphics.Gpu/Shader/DiskCache/DiskCacheGpuAccessor.cs
@@ -130,6 +130,12 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
         }
 
         /// <inheritdoc/>
+        public bool QueryHasConstantBufferDrawParameters()
+        {
+            return _oldSpecState.GraphicsState.HasConstantBufferDrawParameters;
+        }
+
+        /// <inheritdoc/>
         public InputTopology QueryPrimitiveTopology()
         {
             _newSpecState.RecordPrimitiveTopology();
diff --git a/Ryujinx.Graphics.Gpu/Shader/DiskCache/DiskCacheHostStorage.cs b/Ryujinx.Graphics.Gpu/Shader/DiskCache/DiskCacheHostStorage.cs
index e728c48c..0bdf4949 100644
--- a/Ryujinx.Graphics.Gpu/Shader/DiskCache/DiskCacheHostStorage.cs
+++ b/Ryujinx.Graphics.Gpu/Shader/DiskCache/DiskCacheHostStorage.cs
@@ -22,7 +22,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
         private const ushort FileFormatVersionMajor = 1;
         private const ushort FileFormatVersionMinor = 2;
         private const uint FileFormatVersionPacked = ((uint)FileFormatVersionMajor << 16) | FileFormatVersionMinor;
-        private const uint CodeGenVersion = 3831;
+        private const uint CodeGenVersion = 3747;
 
         private const string SharedTocFileName = "shared.toc";
         private const string SharedDataFileName = "shared.data";
@@ -159,6 +159,11 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
             /// Bit mask of the render target components written by the fragment stage.
             /// </summary>
             public int FragmentOutputMap;
+
+            /// <summary>
+            /// Indicates if the vertex shader accesses draw parameters.
+            /// </summary>
+            public bool UsesDrawParameters;
         }
 
         private readonly DiskCacheGuestStorage _guestStorage;
@@ -771,6 +776,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
                 images,
                 dataInfo.Stage,
                 dataInfo.UsesInstanceId,
+                dataInfo.UsesDrawParameters,
                 dataInfo.UsesRtLayer,
                 dataInfo.ClipDistancesWritten,
                 dataInfo.FragmentOutputMap);
@@ -796,6 +802,7 @@ namespace Ryujinx.Graphics.Gpu.Shader.DiskCache
             dataInfo.ImagesCount = (ushort)info.Images.Count;
             dataInfo.Stage = info.Stage;
             dataInfo.UsesInstanceId = info.UsesInstanceId;
+            dataInfo.UsesDrawParameters = info.UsesDrawParameters;
             dataInfo.UsesRtLayer = info.UsesRtLayer;
             dataInfo.ClipDistancesWritten = info.ClipDistancesWritten;
             dataInfo.FragmentOutputMap = info.FragmentOutputMap;
diff --git a/Ryujinx.Graphics.Gpu/Shader/GpuAccessor.cs b/Ryujinx.Graphics.Gpu/Shader/GpuAccessor.cs
index 44c26efb..b8cb1107 100644
--- a/Ryujinx.Graphics.Gpu/Shader/GpuAccessor.cs
+++ b/Ryujinx.Graphics.Gpu/Shader/GpuAccessor.cs
@@ -140,6 +140,12 @@ namespace Ryujinx.Graphics.Gpu.Shader
         }
 
         /// <inheritdoc/>
+        public bool QueryHasConstantBufferDrawParameters()
+        {
+            return _state.GraphicsState.HasConstantBufferDrawParameters;
+        }
+
+        /// <inheritdoc/>
         public InputTopology QueryPrimitiveTopology()
         {
             _state.SpecializationState?.RecordPrimitiveTopology();
diff --git a/Ryujinx.Graphics.Gpu/Shader/GpuChannelGraphicsState.cs b/Ryujinx.Graphics.Gpu/Shader/GpuChannelGraphicsState.cs
index 82252ced..b0727677 100644
--- a/Ryujinx.Graphics.Gpu/Shader/GpuChannelGraphicsState.cs
+++ b/Ryujinx.Graphics.Gpu/Shader/GpuChannelGraphicsState.cs
@@ -78,6 +78,11 @@ namespace Ryujinx.Graphics.Gpu.Shader
         public Array32<AttributeType> AttributeTypes;
 
         /// <summary>
+        /// Indicates that the draw is writing the base vertex, base instance and draw index to Constant Buffer 0.
+        /// </summary>
+        public readonly bool HasConstantBufferDrawParameters;
+
+        /// <summary>
         /// Creates a new GPU graphics state.
         /// </summary>
         /// <param name="earlyZForce">Early Z force enable</param>
@@ -93,6 +98,7 @@ namespace Ryujinx.Graphics.Gpu.Shader
         /// <param name="alphaTestCompare">When alpha test is enabled, indicates the comparison that decides if the fragment should be discarded</param>
         /// <param name="alphaTestReference">When alpha test is enabled, indicates the value to compare with the fragment output alpha</param>
         /// <param name="attributeTypes">Type of the vertex attributes consumed by the shader</param>
+        /// <param name="hasConstantBufferDrawParameters">Indicates that the draw is writing the base vertex, base instance and draw index to Constant Buffer 0</param>
         public GpuChannelGraphicsState(
             bool earlyZForce,
             PrimitiveTopology topology,
@@ -106,7 +112,8 @@ namespace Ryujinx.Graphics.Gpu.Shader
             bool alphaTestEnable,
             CompareOp alphaTestCompare,
             float alphaTestReference,
-            ref Array32<AttributeType> attributeTypes)
+            ref Array32<AttributeType> attributeTypes,
+            bool hasConstantBufferDrawParameters)
         {
             EarlyZForce = earlyZForce;
             Topology = topology;
@@ -121,6 +128,7 @@ namespace Ryujinx.Graphics.Gpu.Shader
             AlphaTestCompare = alphaTestCompare;
             AlphaTestReference = alphaTestReference;
             AttributeTypes = attributeTypes;
+            HasConstantBufferDrawParameters = hasConstantBufferDrawParameters;
         }
     }
 }
\ No newline at end of file
diff --git a/Ryujinx.Graphics.Gpu/Shader/ShaderCache.cs b/Ryujinx.Graphics.Gpu/Shader/ShaderCache.cs
index c998fe09..1803dae6 100644
--- a/Ryujinx.Graphics.Gpu/Shader/ShaderCache.cs
+++ b/Ryujinx.Graphics.Gpu/Shader/ShaderCache.cs
@@ -520,7 +520,9 @@ namespace Ryujinx.Graphics.Gpu.Shader
                 }
             }
 
-            return gpShaders.SpecializationState.MatchesGraphics(channel, poolState, graphicsState, true);
+            bool usesDrawParameters = gpShaders.Shaders[1]?.Info.UsesDrawParameters ?? false;
+
+            return gpShaders.SpecializationState.MatchesGraphics(channel, poolState, graphicsState, usesDrawParameters, true);
         }
 
         /// <summary>
diff --git a/Ryujinx.Graphics.Gpu/Shader/ShaderSpecializationList.cs b/Ryujinx.Graphics.Gpu/Shader/ShaderSpecializationList.cs
index 43ccd892..abc9d913 100644
--- a/Ryujinx.Graphics.Gpu/Shader/ShaderSpecializationList.cs
+++ b/Ryujinx.Graphics.Gpu/Shader/ShaderSpecializationList.cs
@@ -35,7 +35,9 @@ namespace Ryujinx.Graphics.Gpu.Shader
         {
             foreach (var entry in _entries)
             {
-                if (entry.SpecializationState.MatchesGraphics(channel, poolState, graphicsState, true))
+                bool usesDrawParameters = entry.Shaders[1]?.Info.UsesDrawParameters ?? false;
+
+                if (entry.SpecializationState.MatchesGraphics(channel, poolState, graphicsState, usesDrawParameters, true))
                 {
                     program = entry;
                     return true;
diff --git a/Ryujinx.Graphics.Gpu/Shader/ShaderSpecializationState.cs b/Ryujinx.Graphics.Gpu/Shader/ShaderSpecializationState.cs
index 28818304..0aecc5b7 100644
--- a/Ryujinx.Graphics.Gpu/Shader/ShaderSpecializationState.cs
+++ b/Ryujinx.Graphics.Gpu/Shader/ShaderSpecializationState.cs
@@ -481,9 +481,15 @@ namespace Ryujinx.Graphics.Gpu.Shader
         /// <param name="channel">GPU channel</param>
         /// <param name="poolState">Texture pool state</param>
         /// <param name="graphicsState">Graphics state</param>
+        /// <param name="usesDrawParameters">Indicates whether the vertex shader accesses draw parameters</param>
         /// <param name="checkTextures">Indicates whether texture descriptors should be checked</param>
         /// <returns>True if the state matches, false otherwise</returns>
-        public bool MatchesGraphics(GpuChannel channel, GpuChannelPoolState poolState, GpuChannelGraphicsState graphicsState, bool checkTextures)
+        public bool MatchesGraphics(
+            GpuChannel channel,
+            GpuChannelPoolState poolState,
+            GpuChannelGraphicsState graphicsState,
+            bool usesDrawParameters,
+            bool checkTextures)
         {
             if (graphicsState.ViewportTransformDisable != GraphicsState.ViewportTransformDisable)
             {
@@ -520,6 +526,11 @@ namespace Ryujinx.Graphics.Gpu.Shader
                 return false;
             }
 
+            if (usesDrawParameters && graphicsState.HasConstantBufferDrawParameters != GraphicsState.HasConstantBufferDrawParameters)
+            {
+                return false;
+            }
+
             return Matches(channel, poolState, checkTextures, isCompute: false);
         }