Vulkan: Flush command buffers for queries less aggressively (#4387)

The AutoFlushCounter would flush command buffers on any attachment change (write mask or bindings change) if there was a pending query. This is to get query results as soon as possible for draw skips, but it's assuming that a full occlusion query _pass_ happened, that we want to flush it's data before getting onto draws, rather than the queries being randomly interspersed throughout a pass that also draws.

Xenoblade 2 repeatedly switches between performing a samples passed query and outputting to a render target on each draw, and flips the write mask to do so. Flushing the command buffer every 2 draws isn't ideal, so it's best that we only do this if the pattern matches the large block style of occlusion query.

This change makes this flush only happen after a few consecutive query reports. "Consecutive" is interrupted by attachment changes or command buffer flush.

This doesn't really solve the issue where it resets more queries than it uses, it just stops the game doing it as often. I'm not sure of the best way to do that. The cost of resetting could probably be reduced by using query pools with more than one element and resetting in bulk.

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -1,13 +1,8 @@
 name: Bug Report
 description: File a bug report
+title: "[Bug] <title>"
+labels: bug
 body:
-  - type: checkboxes
-    attributes:
-      label: Is there an existing issue for this?
-      description: Please search to see if an issue already exists for the bug you encountered.
-      options:
-      - label: I have searched the existing issues
-        required: true
   - type: textarea
     id: issue
     attributes:
@@ -53,17 +48,43 @@ body:
         - *(e.g. 1.1.1)*
     validations:
       required: false
-  - type: textarea
-    id: environment
+  - type: input
+    id: cpu
     attributes:
-      label: Environment?
-      value: |
-        - ##### CPU: *(e.g. i7-6700)*
-        - ##### GPU: *(e.g. NVIDIA RTX 2070)*
-        - ##### RAM: *(e.g. 16GB)*
-        - Applied Mods: [ Yes (Which ones) / No ]
+      label: CPU
+      placeholder: |
+        - *(e.g. i7-6700)*
     validations:
-      required: true
+      required: false
+  - type: input
+    id: gpu
+    attributes:
+      label: GPU
+      placeholder: |
+        - *(e.g. NVIDIA RTX 2070)*
+    validations:
+      required: false
+  - type: input
+    id: ram
+    attributes:
+      label: RAM
+      placeholder: |
+        - *(e.g. 16GB)*
+    validations:
+      required: false
+  - type: checkboxes
+    attributes:
+      label: Applied Mods?
+      options:
+      - label: "Yes"
+        required: false
+  - type: textarea
+    id: mods
+    attributes:
+      label: List of applied mods
+      placeholder: You can list applied mods here.
+    validations:
+      required: false
   - type: textarea
     id: additional-context
     attributes:
@@ -72,4 +93,4 @@ body:
         - Additional info about your environment:
         - Any other information relevant to your issue.
     validations:
-      required: false
+      required: false

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -1,13 +1,7 @@
 name: Feature Request
 description: Suggest a new feature for Ryujinx.
+title: "[Feature Request] <title>"
 body:
-  - type: checkboxes
-    attributes:
-      label: Is there an existing issue for this?
-      description: Please search to see if an issue already exists for the feature you are requesting.
-      options:
-      - label: I have searched the existing issues
-        required: true
   - type: textarea
     id: overview
     attributes:

.github/ISSUE_TEMPLATE/missing_cpu_instruction.yml

@@ -1,13 +1,8 @@
 name: Missing CPU Instruction
 description: CPU Instruction is missing in Ryujinx.
+title: "[CPU] <title>"
+labels: [cpu, not-implemented]
 body:
-  - type: checkboxes
-    attributes:
-      label: Is there an existing issue for this?
-      description: Please search for existing missing CPU instruction
-      options:
-      - label: I have searched the existing issues
-        required: true
   - type: textarea
     id: instruction
     attributes:

.github/ISSUE_TEMPLATE/missing_service_call.yml

@@ -1,13 +1,7 @@
 name: Missing Service Call
 description: Service call is missing in Ryujinx.
+labels: not-implemented
 body:
-  - type: checkboxes
-    attributes:
-      label: Is there an existing issue for this?
-      description: Please search for an [existing missing service call issue](https://github.com/Ryujinx/Ryujinx/issues) first.
-      options:
-      - label: I have searched the existing issues
-        required: true
   - type: textarea
     id: instruction
     attributes:

Directory.Packages.props

@@ -44,7 +44,7 @@
     <PackageVersion Include="SixLabors.ImageSharp.Drawing" Version="1.0.0-beta11" />
     <PackageVersion Include="SPB" Version="0.0.4-build28" />
     <PackageVersion Include="System.Drawing.Common" Version="7.0.0" />
-    <PackageVersion Include="System.IdentityModel.Tokens.Jwt" Version="6.26.0" />
+    <PackageVersion Include="System.IdentityModel.Tokens.Jwt" Version="6.26.1" />
     <PackageVersion Include="System.IO.FileSystem.Primitives" Version="4.3.0" />
     <PackageVersion Include="System.Management" Version="7.0.0" />
     <PackageVersion Include="System.Net.NameResolution" Version="4.3.0" />

Ryujinx.Graphics.Vulkan/AutoFlushCounter.cs

@@ -10,11 +10,13 @@ namespace Ryujinx.Graphics.Vulkan
         private readonly static long FramebufferFlushTimer = Stopwatch.Frequency / 1000;
 
         private const int MinDrawCountForFlush = 10;
+        private const int MinConsecutiveQueryForFlush = 10;
         private const int InitialQueryCountForFlush = 32;
 
         private long _lastFlush;
         private ulong _lastDrawCount;
         private bool _hasPendingQuery;
+        private int _consecutiveQueries;
         private int _queryCount;
 
         private int[] _queryCountHistory = new int[3];
@@ -27,11 +29,13 @@ namespace Ryujinx.Graphics.Vulkan
             _lastDrawCount = drawCount;
 
             _hasPendingQuery = false;
+            _consecutiveQueries = 0;
         }
 
         public bool RegisterPendingQuery()
         {
             _hasPendingQuery = true;
+            _consecutiveQueries++;
             _remainingQueries--;
 
             _queryCountHistory[_queryCountHistoryIndex]++;
@@ -65,15 +69,18 @@ namespace Ryujinx.Graphics.Vulkan
             return _hasPendingQuery;
         }
 
-        public bool ShouldFlush(ulong drawCount)
+        public bool ShouldFlushAttachmentChange(ulong drawCount)
         {
             _queryCount = 0;
 
-            if (_hasPendingQuery)
+            // Flush when there's an attachment change out of a large block of queries.
+            if (_consecutiveQueries > MinConsecutiveQueryForFlush)
             {
                 return true;
             }
 
+            _consecutiveQueries = 0;
+
             long draws = (long)(drawCount - _lastDrawCount);
 
             if (draws < MinDrawCountForFlush)

Ryujinx.Graphics.Vulkan/PipelineConverter.cs

@@ -27,6 +27,7 @@ namespace Ryujinx.Graphics.Vulkan
 
             int attachmentCount = 0;
             int colorCount = 0;
+            int maxColorAttachmentIndex = -1;
 
             for (int i = 0; i < state.AttachmentEnable.Length; i++)
             {
@@ -36,6 +37,7 @@ namespace Ryujinx.Graphics.Vulkan
 
                     attachmentIndices[attachmentCount++] = i;
                     colorCount++;
+                    maxColorAttachmentIndex = i;
                 }
             }
 
@@ -73,12 +75,11 @@ namespace Ryujinx.Graphics.Vulkan
 
                 if (colorAttachmentsCount != 0)
                 {
-                    int maxAttachmentIndex = Constants.MaxRenderTargets - 1;
-                    subpass.ColorAttachmentCount = (uint)maxAttachmentIndex + 1;
+                    subpass.ColorAttachmentCount = (uint)maxColorAttachmentIndex + 1;
                     subpass.PColorAttachments = &attachmentReferences[0];
 
                     // Fill with VK_ATTACHMENT_UNUSED to cover any gaps.
-                    for (int i = 0; i <= maxAttachmentIndex; i++)
+                    for (int i = 0; i <= maxColorAttachmentIndex; i++)
                     {
                         subpass.PColorAttachments[i] = new AttachmentReference(Vk.AttachmentUnused, ImageLayout.Undefined);
                     }

Ryujinx.Graphics.Vulkan/PipelineFull.cs

@@ -10,7 +10,7 @@ namespace Ryujinx.Graphics.Vulkan
     {
         private const ulong MinByteWeightForFlush = 256 * 1024 * 1024; // MiB
 
-        private readonly List<QueryPool> _activeQueries;
+        private readonly List<(QueryPool, bool)> _activeQueries;
         private CounterQueueEvent _activeConditionalRender;
 
         private readonly List<BufferedQuery> _pendingQueryCopies;
@@ -19,7 +19,7 @@ namespace Ryujinx.Graphics.Vulkan
 
         public PipelineFull(VulkanRenderer gd, Device device) : base(gd, device)
         {
-            _activeQueries = new List<QueryPool>();
+            _activeQueries = new List<(QueryPool, bool)>();
             _pendingQueryCopies = new();
 
             CommandBuffer = (Cbs = gd.CommandBufferPool.Rent()).CommandBuffer;
@@ -202,7 +202,7 @@ namespace Ryujinx.Graphics.Vulkan
             AutoFlush.RegisterFlush(DrawCount);
             EndRenderPass();
 
-            foreach (var queryPool in _activeQueries)
+            foreach ((var queryPool, _) in _activeQueries)
             {
                 Gd.Api.CmdEndQuery(CommandBuffer, queryPool, 0);
             }
@@ -220,10 +220,12 @@ namespace Ryujinx.Graphics.Vulkan
 
             // Restore per-command buffer state.
 
-            foreach (var queryPool in _activeQueries)
+            foreach ((var queryPool, var isOcclusion) in _activeQueries)
             {
+                bool isPrecise = Gd.Capabilities.SupportsPreciseOcclusionQueries && isOcclusion;
+
                 Gd.Api.CmdResetQueryPool(CommandBuffer, queryPool, 0, 1);
-                Gd.Api.CmdBeginQuery(CommandBuffer, queryPool, 0, Gd.Capabilities.SupportsPreciseOcclusionQueries ? QueryControlFlags.PreciseBit : 0);
+                Gd.Api.CmdBeginQuery(CommandBuffer, queryPool, 0, isPrecise ? QueryControlFlags.PreciseBit : 0);
             }
 
             Gd.ResetCounterPool();
@@ -231,7 +233,7 @@ namespace Ryujinx.Graphics.Vulkan
             Restore();
         }
 
-        public void BeginQuery(BufferedQuery query, QueryPool pool, bool needsReset, bool fromSamplePool)
+        public void BeginQuery(BufferedQuery query, QueryPool pool, bool needsReset, bool isOcclusion, bool fromSamplePool)
         {
             if (needsReset)
             {
@@ -247,16 +249,24 @@ namespace Ryujinx.Graphics.Vulkan
                 }
             }
 
-            Gd.Api.CmdBeginQuery(CommandBuffer, pool, 0, Gd.Capabilities.SupportsPreciseOcclusionQueries ? QueryControlFlags.PreciseBit : 0);
+            bool isPrecise = Gd.Capabilities.SupportsPreciseOcclusionQueries && isOcclusion;
+            Gd.Api.CmdBeginQuery(CommandBuffer, pool, 0, isPrecise ? QueryControlFlags.PreciseBit : 0);
 
-            _activeQueries.Add(pool);
+            _activeQueries.Add((pool, isOcclusion));
         }
 
         public void EndQuery(QueryPool pool)
         {
             Gd.Api.CmdEndQuery(CommandBuffer, pool, 0);
 
-            _activeQueries.Remove(pool);
+            for (int i = 0; i < _activeQueries.Count; i++)
+            {
+                if (_activeQueries[i].Item1.Handle == pool.Handle)
+                {
+                    _activeQueries.RemoveAt(i);
+                    break;
+                }
+            }
         }
 
         public void CopyQueryResults(BufferedQuery query)
@@ -271,7 +281,7 @@ namespace Ryujinx.Graphics.Vulkan
 
         protected override void SignalAttachmentChange()
         {
-            if (AutoFlush.ShouldFlush(DrawCount))
+            if (AutoFlush.ShouldFlushAttachmentChange(DrawCount))
             {
                 FlushCommandsImpl();
             }

Ryujinx.Graphics.Vulkan/Queries/BufferedQuery.cs

@@ -100,7 +100,8 @@ namespace Ryujinx.Graphics.Vulkan.Queries
             if (_isSupported)
             {
                 bool needsReset = resetSequence == null || _resetSequence == null || resetSequence.Value != _resetSequence.Value;
-                _pipeline.BeginQuery(this, _queryPool, needsReset, _type == CounterType.SamplesPassed && resetSequence != null);
+                bool isOcclusion = _type == CounterType.SamplesPassed;
+                _pipeline.BeginQuery(this, _queryPool, needsReset, isOcclusion, isOcclusion && resetSequence != null);
             }
             _resetSequence = null;
         }

Ryujinx.Graphics.Vulkan/TextureView.cs

@@ -90,7 +90,7 @@ namespace Ryujinx.Graphics.Vulkan
             var componentMapping = new ComponentMapping(swizzleR, swizzleG, swizzleB, swizzleA);
 
             var aspectFlags = info.Format.ConvertAspectFlags(info.DepthStencilMode);
-            var aspectFlagsDepth = info.Format.ConvertAspectFlags(DepthStencilMode.Depth);
+            var aspectFlagsDepth = info.Format.ConvertAspectFlags();
 
             var subresourceRange = new ImageSubresourceRange(aspectFlags, (uint)firstLevel, levels, (uint)firstLayer, layers);
             var subresourceRangeDepth = new ImageSubresourceRange(aspectFlagsDepth, (uint)firstLevel, levels, (uint)firstLayer, layers);

Ryujinx.HLE/FileSystem/ContentManager.cs

@@ -229,7 +229,7 @@ namespace Ryujinx.HLE.FileSystem
                     continue;
                 }
 
-                string ncaId = BitConverter.ToString(cnmt.ContentEntries[0].NcaId).Replace("-", "").ToLower();
+                string ncaId = Convert.ToHexString(cnmt.ContentEntries[0].NcaId).ToLower();
 
                 AddAocItem(cnmt.TitleId, containerPath, $"{ncaId}.nca", true);
             }

Ryujinx.HLE/HOS/ModLoader.cs

@@ -696,8 +696,8 @@ namespace Ryujinx.HLE.HOS
 
             var buildIds = programs.Select(p => p switch
             {
-                NsoExecutable nso => BitConverter.ToString(nso.BuildId.ItemsRo.ToArray()).Replace("-", "").TrimEnd('0'),
-                NroExecutable nro => BitConverter.ToString(nro.Header.BuildId).Replace("-", "").TrimEnd('0'),
+                NsoExecutable nso => Convert.ToHexString(nso.BuildId.ItemsRo.ToArray()).TrimEnd('0'),
+                NroExecutable nro => Convert.ToHexString(nro.Header.BuildId).TrimEnd('0'),
                 _ => string.Empty
             }).ToList();
 

Ryujinx.HLE/HOS/Services/Account/Acc/AccountService/ManagerServer.cs

@@ -51,11 +51,11 @@ namespace Ryujinx.HLE.HOS.Services.Account.Acc.AccountService
 
             var payload = new JwtPayload
             {
-                { "sub", BitConverter.ToString(rawUserId).Replace("-", "").ToLower() },
+                { "sub", Convert.ToHexString(rawUserId).ToLower() },
                 { "aud", "ed9e2f05d286f7b8" },
-                { "di", BitConverter.ToString(deviceId).Replace("-", "").ToLower() },
+                { "di", Convert.ToHexString(deviceId).ToLower() },
                 { "sn", "XAW10000000000" },
-                { "bs:did", BitConverter.ToString(deviceAccountId).Replace("-", "").ToLower() },
+                { "bs:did", Convert.ToHexString(deviceAccountId).ToLower() },
                 { "iss", "https://e0d67c509fb203858ebcb2fe3f88c2aa.baas.nintendo.com" },
                 { "typ", "id_token" },
                 { "iat", DateTimeOffset.UtcNow.ToUnixTimeSeconds() },

Ryujinx.HLE/HOS/Services/Caps/CaptureManager.cs

@@ -101,7 +101,7 @@ namespace Ryujinx.HLE.HOS.Services.Caps
                 };
 
                 // NOTE: The hex hash is a HMAC-SHA256 (first 32 bytes) using a hardcoded secret key over the titleId, we can simulate it by hashing the titleId instead.
-                string hash       = BitConverter.ToString(SHA256.HashData(BitConverter.GetBytes(titleId))).Replace("-", "").Remove(0x20);
+                string hash       = Convert.ToHexString(SHA256.HashData(BitConverter.GetBytes(titleId))).Remove(0x20);
                 string folderPath = Path.Combine(_sdCardPath, "Nintendo", "Album", currentDateTime.Year.ToString("00"), currentDateTime.Month.ToString("00"), currentDateTime.Day.ToString("00"));
                 string filePath   = GenerateFilePath(folderPath, applicationAlbumEntry, currentDateTime, hash);