feat(hle): prefer loaded guest exports for selected registrations (#844)

Co-authored-by: Acelogic <miguelc4600@gmail.com>

src/SharpEmu.Core/Cpu/Native/DirectExecutionBackend.cs

@@ -8,6 +8,7 @@ using System.Diagnostics;
 using System.Linq;
 using System.Runtime.InteropServices;
 using System.Threading;
+using Iced.Intel;
 using SharpEmu.Core.Cpu;
 using SharpEmu.Core.Cpu.Debugging;
 using SharpEmu.Core.Loader;
@@ -1638,18 +1639,15 @@ public sealed unsafe partial class DirectExecutionBackend : INativeCpuBackend, I
 
 		if (_moduleManager.TryGetExport(nid, out ExportedFunction export))
 		{
-			if (IsKernelLibrary(export.LibraryName))
+			var preferLleForLibc = IsLibcLibrary(export.LibraryName) && PreferLleForLibcExport(export.Name);
+			if (!ShouldResolveRegisteredExportViaLle(export, preferLleForLibc))
 			{
-				if (_logAllImports)
+				if (_logAllImports && IsKernelLibrary(export.LibraryName))
 				{
 					Console.Error.WriteLine($"[LOADER][DEBUG] TryResolveDirectImportTarget: {nid} ({export.LibraryName}:{export.Name}) -> HLE (kernel library)");
 				}
 				return false;
 			}
-			if (!IsLibcLibrary(export.LibraryName) || !PreferLleForLibcExport(export.Name))
-			{
-				return false;
-			}
 			if (TryResolveRuntimeSymbolAddress(nid, out var value2) && IsDirectImportTargetUsable(value2))
 			{
 				targetAddress = value2;
@@ -1703,6 +1701,14 @@ public sealed unsafe partial class DirectExecutionBackend : INativeCpuBackend, I
 		return false;
 	}
 
+	internal static bool ShouldResolveRegisteredExportViaLle(
+		ExportedFunction export,
+		bool preferLleForLibc)
+	{
+		ArgumentNullException.ThrowIfNull(export);
+		return !IsKernelLibrary(export.LibraryName) && (export.PreferLle || preferLleForLibc);
+	}
+
 	private static bool IsHlePreferredNid(string nid)
 	{
 		return string.Equals(nid, "QrZZdJ8XsX0", StringComparison.Ordinal) ||
@@ -3200,7 +3206,7 @@ public sealed unsafe partial class DirectExecutionBackend : INativeCpuBackend, I
 				{
 					nint address = (nint)(ptr + i);
 					int remainingBytes = scanBytes - i;
-					if (TryPatchTlsLoadInstruction(address, ptr + i, remainingBytes))
+					if (TryPatchTlsLoadInstruction(address, ptr + i, remainingBytes, i))
 					{
 						num3++;
 					}
@@ -3343,13 +3349,19 @@ public sealed unsafe partial class DirectExecutionBackend : INativeCpuBackend, I
 		return true;
 	}
 
-	private unsafe bool TryPatchTlsLoadInstruction(nint address, byte* source, int availableLength)
+	private unsafe bool TryPatchTlsLoadInstruction(nint address, byte* source, int availableLength, int regionOffset)
 	{
 		if (availableLength < MinTlsPatchInstructionBytes)
 		{
 			return false;
 		}
 
+		var region = new ReadOnlySpan<byte>(source - regionOffset, regionOffset + availableLength);
+		if (IsTlsLoadCandidateInsideShortJump(region, regionOffset))
+		{
+			return false;
+		}
+
 		var offset = 0;
 		while (offset < availableLength && source[offset] == 0x66)
 		{
@@ -3402,6 +3414,76 @@ public sealed unsafe partial class DirectExecutionBackend : INativeCpuBackend, I
 		return PatchTlsLoadInstruction(address, instructionLength, destinationRegister);
 	}
 
+	internal static bool IsTlsLoadCandidateInsideShortJump(ReadOnlySpan<byte> region, int candidateOffset)
+	{
+		if ((uint)candidateOffset >= (uint)region.Length ||
+			candidateOffset < 1 ||
+			region[candidateOffset - 1] != 0xEB)
+		{
+			return false;
+		}
+
+		// Accept EB when it is an aligned rel8 operand.
+		if (IsRel8ControlFlowInstructionEndingAtCandidate(region, candidateOffset))
+		{
+			return false;
+		}
+
+		return true;
+	}
+
+	private static bool IsRel8ControlFlowInstructionEndingAtCandidate(
+		ReadOnlySpan<byte> region,
+		int candidateOffset)
+	{
+		if (candidateOffset < 2)
+		{
+			return false;
+		}
+
+		var branchOffset = candidateOffset - 2;
+		var opcode = region[branchOffset];
+		if (!((opcode >= 0x70 && opcode <= 0x7F) ||
+			opcode is >= 0xE0 and <= 0xE3 ||
+			opcode == 0xEB))
+		{
+			return false;
+		}
+
+		var branchTarget = candidateOffset + (sbyte)region[candidateOffset - 1];
+		if (branchTarget < 0 || branchTarget >= branchOffset)
+		{
+			return false;
+		}
+
+		// Require an aligned instruction stream.
+		var decoder = Decoder.Create(
+			64,
+			new ByteArrayCodeReader(region[branchTarget..candidateOffset].ToArray()));
+		decoder.IP = (ulong)branchTarget;
+		while (decoder.IP < (ulong)candidateOffset)
+		{
+			var instructionOffset = (int)decoder.IP;
+			decoder.Decode(out var instruction);
+			if (instruction.Code == Code.INVALID || instruction.Length <= 0)
+			{
+				return false;
+			}
+
+			if (instructionOffset == branchOffset)
+			{
+				return instruction.Length == 2 && decoder.IP == (ulong)candidateOffset;
+			}
+
+			if (decoder.IP > (ulong)branchOffset)
+			{
+				return false;
+			}
+		}
+
+		return false;
+	}
+
 	private unsafe bool PatchTlsLoadInstruction(nint address, int instructionLength, int destinationRegister)
 	{
 		uint flNewProtect = default(uint);

src/SharpEmu.Core/Memory/PhysicalVirtualMemory.cs

@@ -1,6 +1,7 @@
 // Copyright (C) 2026 SharpEmu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+using System.Collections.Concurrent;
 using System.Runtime.InteropServices;
 using SharpEmu.Core.Loader;
 using SharpEmu.HLE;
@@ -18,7 +19,7 @@ public sealed unsafe class PhysicalVirtualMemory : IVirtualMemory, IGuestMemoryA
     private readonly object _allocationSearchHintGate = new();
     private readonly List<MemoryRegion> _regions = new();
     private readonly Dictionary<(ulong DesiredAddress, ulong Alignment, bool Executable), ulong> _allocationSearchHints = new();
-    private readonly Dictionary<ulong, ProgramHeaderFlags> _pageProtections = new();
+    private readonly ConcurrentDictionary<ulong, ProgramHeaderFlags> _pageProtections = new();
     private bool _disposed;
 
     [ThreadStatic]
@@ -28,7 +29,13 @@ public sealed unsafe class PhysicalVirtualMemory : IVirtualMemory, IGuestMemoryA
     private const ulong PageSize = 0x1000;
     private const ulong HostAllocationGranularity = 0x10000;
     private const ulong GuestAllocationArenaAddress = 0x00006000_0000_0000;
-    private const ulong GuestAllocationArenaSize = 0x0100_0000;
+    // Full C++ runtimes can route large numbers of HLE-backed heap allocations
+    // through this arena. The original 16 MiB capacity could be exhausted
+    // during asset setup, turning a valid allocation request into a null
+    // pointer that failed later in unrelated guest code. Keep enough capacity
+    // for those workloads. Adapted from foufouadi's allocator-exhaustion
+    // investigation.
+    private const ulong GuestAllocationArenaSize = 0x2000_0000;
     private const ulong GuestAllocationArenaStartOffset = PageSize;
     private const ulong LargeDataReserveThreshold = 0x4000_0000UL; // 1 GiB
     private const ulong FullCommitRegionLimit = 4UL << 30;

src/SharpEmu.GUI/Languages/ee.json

@@ -0,0 +1,213 @@
+{
+  "_languageName": "Eesti",
+
+  "Page.Library": "Teek",
+  "Page.Options": "Seaded",
+  "Page.GameCount.One": "1 mäng",
+  "Page.GameCount.Other": "{0} mängu",
+
+  "Library.SearchWatermark": "Otsi teegist…",
+  "Library.AddFolder": "Lisa kaust",
+  "Library.OpenFile": "Ava fail…",
+  "Library.View.Grid": "Ruudustikvaade",
+  "Library.View.Carousel": "Karussellvaade",
+
+  "Library.Context.Launch": "Käivita",
+  "Library.Context.OpenFolder": "Ava mängukaust",
+  "Library.Context.CopyPath": "Kopeeri tee",
+  "Library.Context.CopyTitleId": "Kopeeri Title ID",
+  "Library.Context.GameSettings": "Mängu seaded…",
+  "Library.Context.Remove": "Eemalda teegist",
+
+  "Library.Empty.Title": "Sinu teek on tühi",
+  "Library.Empty.Hint": "Alustamiseks lisa kaust, kus sinu mängud asuvad.",
+  "Library.Empty.SearchTitle": "Otsingule ei vasta ükski mäng",
+  "Library.Empty.SearchHint": "Ükski teegis olev mäng ei vasta otsingule “{0}”.",
+  "Library.Empty.AddFolder": "＋ Lisa mängukaust",
+
+  "Library.Loading": "Teeki laaditakse…",
+  "Library.Stat.Version": "Versioon",
+  "Library.Stat.Installed": "Paigaldatud",
+  "Library.Stat.TitleId": "Title ID",
+  "Common.Back": "Tagasi",
+
+  "Options.General": "Üldine",
+  "Options.Logging": "Logimine",
+  "Options.Env.Tab": "Keskkond",
+  "Options.Section.Environment": "KESKKONNAMUUTUJAD",
+  "Options.Env.Desc": "Lülitid, mis edastatakse emulaatorile käivitamisel keskkonnamuutujatena.",
+  "Options.Env.Bthid.Desc": "Märgi Bluetooth HID kättesaamatuks mängude puhul, mille rooli/FFB-vahetarkvara küsib andmeid lõputult.\nTavaliselt jäta välja lülitatuks. Mõni mäng hangub, kui initsialiseerimine ebaõnnestub.",
+  "Options.Env.LoopGuard.Desc": "Ära sulge sundkorras mänge, mis kordavad sama kutset liiga kaua.\nProovi seda, kui mäng laadimise ajal iseenesest sulgub.",
+  "Options.Env.WritableApp0.Desc": "Luba mängudel luua ja kirjutada faile oma paigalduskausta.\nVajalik pakendamata koopiatele, mis salvestavad oma andmed või seadistused kausta /app0.",
+  "Options.Env.VkValidation.Desc": "Luba Vulkani valideerimiskihid GPU silumiseks.\nAeglane. Nõuab paigaldatud Vulkan SDK-d.",
+  "Options.Env.DumpSpirv.Desc": "Salvesta AGC-shaderid ja nende SPIR-V tõlked kausta shader-dumps.\nKasuta shaderi- või kuvamisvea teatamisel.",
+  "Options.Env.LogDirectMemory.Desc": "Logi konsooli otsese mälu eraldamised ja tõrked.\nKasuta siis, kui mäng katkeb või sulgub käivitamise ajal.",
+  "Options.Env.LogIo.Desc": "Logi konsooli failide avamise, lugemise ja tee lahendamise tegevused.\nKasuta siis, kui mäng ei leia käivitamisel oma andmefaile.",
+  "Options.Env.LogNp.Desc": "Logi konsooli NP (PlayStation Network) teegi kutsed.",
+  "Options.Env.Group.Debug": "Silumine",
+  "Options.Env.Group.General": "Üldine",
+  "Options.Env.RenderDoc.Desc": "Laadib RenderDoc'i rakendusesisese API, et kaadreid saaks hõivata otse emulaatorist.\nVajuta mängu käigus F10 ühe kaadri hõivamiseks; hõivatised salvestatakse kausta user/logs/capture_logs/<TITLE_ID>.\nNõuab paigaldatud RenderDoc'i. Aeglustab GPU-d ja põhjustab mõningate mängude hangumist, seega hoia välja lülitatuna, kui sa ei silu.",
+  "Options.Env.GuestImageCpuSync.Desc": "Laadi uuesti üles mängupinnad, mida selle enda CPU-kood ümber kirjutab.\nTavaliselt jäta välja lülitatuks. Lülita sisse mängude puhul, mille CPU-ga joonistatud pinnad ei jõua kunagi ekraanile.\nVähendab jõudlust ja halvendab mõne mängu töökindlust (nt GTA V).",
+  "Options.Env.ForceSubmitOrphanPreambles.Desc": "Edasta GPU käsupuhvrite preambulid isegi siis, kui nende sihtjärjekord neid kunagi ei võta.\nTavaliselt jäta välja lülitatuks. Lülita sisse mängude puhul, mis hanguvad oodates GPU-fence'i, mis kunagi ei rakendu.",
+  "Options.DefaultProfile.Label": "Vaikimisi profiili nimi",
+  "Options.DefaultProfile.Desc": "Nimi, mida kasutatakse, kui mäng küsib tekstisisestust. Vaikimisi Sharp.",
+  "Options.Section.Emulation": "EMULATSIOON",
+  "Options.Section.Logging": "LOGIMINE",
+  "Options.Section.Launcher": "KÄIVITI",
+  "Options.Section.Rendering": "RENDERDAMINE",
+  "Options.Section.Display": "KUVA",
+  "Options.Graphics": "Graafika",
+
+  "Options.RenderResolution.Label": "Sisemine eraldusvõime",
+  "Options.RenderResolution.Desc": "Renderdab ekraanivälised sihid natiivsest madalamal eraldusvõimel ja skaleerib kuvamisel üles. Madalamad väärtused annavad pildikvaliteedi arvelt GPU-le jõudlusvaru; rakendub järgmisel käivitamisel.",
+  "Options.RenderResolution.Native": "100% (natiivne)",
+  "Options.WindowMode.Label": "Aknarežiim",
+  "Options.WindowMode.Desc": "Tavaline aken, äärteta töölauavaade või eksklusiivne täisekraan.",
+  "Options.WindowMode.Windowed": "Aken",
+  "Options.WindowMode.Borderless": "Äärteta",
+  "Options.WindowMode.Exclusive": "Eksklusiivne",
+  "Options.Resolution.Label": "Eraldusvõime",
+  "Options.Resolution.Desc": "Algne akna suurus või eksklusiivse täisekraani eraldusvõime.",
+  "Options.Display.Label": "Kuvar",
+  "Options.Display.Desc": "Monitor, mida kasutatakse tsentreerimiseks ja täisekraanrežiimiks.",
+  "Options.RefreshRate.Label": "Värskendussagedus",
+  "Options.RefreshRate.Desc": "Eksklusiivse täisekraani värskendussagedus. Automaatne valib lähima režiimi.",
+  "Options.RefreshRate.Automatic": "Automaatne",
+  "Options.Scaling.Label": "Skaleerimine",
+  "Options.Scaling.Desc": "Skaleeri algset mängupilti muutmata selle sisemist eraldusvõimet.",
+  "Options.Scaling.Fit": "Mahuta",
+  "Options.Scaling.Cover": "Kata",
+  "Options.Scaling.Stretch": "Venita",
+  "Options.Scaling.Integer": "Täisarvuline",
+  "Options.VSync.Label": "VSync",
+  "Options.VSync.Desc": "Kasuta FIFO-esitust pildi rebimise vältimiseks.",
+  "Options.Hdr.Label": "HDR-väljund",
+  "Options.Hdr.Desc": "Kasuta HDR-i, kui valitud kuvar ja graafikaliides seda toetavad. Automaatrežiim lülitub tagasi SDR-ile.",
+  "Options.Hdr.Auto": "Automaatne",
+
+  "Options.CpuEngine.Label": "CPU-mootor",
+  "Options.CpuEngine.Desc": "Käitusmootor, mida kasutatakse mängukoodi käivitamiseks.",
+  "Options.CpuEngine.Native": "Natiivne",
+
+  "Options.Strict.Label": "Range dynlib-lahendamine",
+  "Options.Strict.Desc": "Katkesta käivitamine, kui imporditud sümbolit ei õnnestu lahendada.",
+
+  "Options.LogLevel.Label": "Logimise tase",
+  "Options.LogLevel.Desc": "Emulaatori konsooliväljundi üksikasjalikkus.",
+  "Options.LogLevel.Trace": "Trace (Jälitus)",
+  "Options.LogLevel.Debug": "Debug (Silumine)",
+  "Options.LogLevel.Info": "Info",
+  "Options.LogLevel.Warning": "Hoiatus",
+  "Options.LogLevel.Error": "Viga",
+  "Options.LogLevel.Critical": "Kriitiline",
+
+  "Options.TraceImports.Label": "Impordi jälituse piir",
+  "Options.TraceImports.Desc": "Jälita iga mooduli esimesi N importi (0 = väljas).",
+
+  "Options.LogToFile.Label": "Logi faili",
+  "Options.LogToFile.Desc": "Kopeeri emulaatori väljund logifaili.",
+
+  "Options.LogFilePath.Label": "Logifaili tee",
+  "Options.LogFilePath.Default": "Kohandatud tee puudub — logid salvestatakse emulaatori kõrvale kausta user/logs.",
+  "Options.LogFilePath.Select": "Vali…",
+
+  "Options.OverrideLogFile.Label": "Kirjuta logifail üle",
+  "Options.OverrideLogFile.Desc": "Kasuta täpset failiteed selle asemel, et lisada Title ID ja ajatempel.",
+
+  "Options.TitleMusic.Label": "Mängu muusika",
+  "Options.TitleMusic.Desc": "Esita valitud mängu eelvaatemuusikat teegis korduvalt.",
+
+  "Options.Discord.Label": "Discordi olek",
+  "Options.Discord.Desc": "Näita käimasolevat mängu oma Discordi profiilis.",
+
+  "Options.Language.Label": "Emulaatori keel",
+  "Options.Language.Desc": "Käivitis kasutatav keel. Rakendub kohe.",
+
+  "Common.On": "Sisse",
+  "Common.Off": "Välja",
+  "Common.Save": "Salvesta",
+  "Common.Cancel": "Tühista",
+
+
+  "Console.Title": "KONSOOL",
+  "Console.SearchWatermark": "Otsi…",
+  "Console.AutoScroll": "Automaatne kerimine",
+  "Console.Split": "Poolita",
+  "Console.Copy": "Kopeeri",
+  "Console.Clear": "Tühjenda",
+  "Console.WindowTitle": "SharpEmu konsool",
+
+  "Launch.NoGameSelected": "Mäng pole valitud",
+  "Launch.NoGameHint": "Vali mäng teegist või ava eboot.bin fail otse.",
+  "Launch.Idle": "Ootel",
+  "Launch.Console": "≡ Konsool",
+  "Launch.Launch": "▶  Käivita",
+  "Launch.Stop": "■  Peata",
+  "Launch.Running": "Käib — {0}",
+  "Launch.Stopping": "Peatamine…",
+  "Launch.Exited": "Väljus koodiga {0} ({1})",
+  "Launch.ExeNotFound": "SharpEmu käivitavat faili ei leitud. Ehita esmalt SharpEmu.CLI projekt (dotnet build).",
+  "Launch.LogFile": "Logifail: {0}",
+  "Launch.Command": "$ SharpEmu {0}",
+  "Launch.StartFailed": "Emulaatori käivitamine nurjus: {0}",
+  "Launch.ProcessExited": "Protsess väljus koodiga {0} ({1}).",
+
+  "Exit.Ok": "OK",
+  "Exit.InvalidArguments": "vigased argumendid",
+  "Exit.EbootNotFound": "eboot-faili ei leitud",
+  "Exit.RuntimeException": "käitusaegne erand",
+  "Exit.EmulationError": "emulatsioonitõrge",
+  "Exit.Unknown": "tundmatu",
+
+  "Status.EmulatorLocating": "Emulaator: otsimine…",
+  "Status.EmulatorPath": "Emulaator: {0}",
+  "Status.EmulatorNotFound": "Emulaator: SharpEmu käivitavat faili ei leitud — ehita esmalt SharpEmu.CLI.",
+  "Status.ScanningLibrary": "Teegi skaneerimine…",
+  "Status.AddFolderPrompt": "Lisa mängukaust teegi täitmiseks.",
+  "Status.LibraryScanned": "Teek skaneeritud: {0} mängu {1} kaustas.",
+  "Status.CouldNotOpenFolder": "Kausta ei saanud avada: {0}",
+  "Status.CopiedToClipboard": "{0} kopeeritud lõikelauale.",
+  "Status.RemovedFromLibrary": "Eemaldati “{0}” teegist. Taastamiseks lisa selle kaust uuesti.",
+  "Status.Running": "Käib: {0}",
+  "Status.Stopping": "Peatamine…",
+  "Status.Idle": "Ootel",
+
+  "Clipboard.Path": "Asukoht",
+  "Clipboard.TitleId": "Title ID",
+
+  "Discord.Playing": "Mängib: {0}",
+  "Discord.Browsing": "Sirvib teeki",
+
+  "Dialog.ChooseGameFolder": "Vali mänge sisaldav kaust",
+  "Dialog.OpenExecutable": "Ava käivitatav fail käivitamiseks",
+  "Dialog.PsExecutables": "PS käivitusfailid",
+  "Dialog.SaveLogFile": "Vali logifaili salvestamiskoht",
+  "Dialog.PlainTextFiles": "Tavatekstifailid",
+  "Dialog.LogFiles": "Logifailid",
+
+  "Options.About": "Teave",
+  "About.Github.Label": "GitHub",
+  "About.Github.Desc": "Lähtekood, vearaportid ja projekti arendus.",
+  "About.Github.LatestCommitLabel": "Viimane muudatus (commit)",
+  "About.Github.LatestCommitDescription": "Viimane muudatus peaharul",
+  "About.Discord.Label": "Discord",
+  "About.Discord.Desc": "Liitu kogukonnaga, saa tuge ja jälgi arendust.",
+  "About.GithubButton": "Löö GitHubis kaasa!",
+  "About.DiscordComingSoon": "Tulekul",
+
+  "Updater.Auto.Label": "Kontrolli värskendusi käivitamisel",
+  "Updater.Auto.Desc": "Kontrollib GitHubist värskendusi käivitamist viivitamata.",
+  "Updater.Label": "Värskendused",
+  "Updater.Check": "Kontrolli värskendusi",
+  "Updater.DownloadRestart": "Laadi alla ja taaskäivita",
+  "Updater.Status.Ready": "Praegune versioon: {0}",
+  "Updater.Status.Checking": "Värskenduste kontrollimine…",
+  "Updater.Status.Current": "Sinu versioon on ajakohane ({0}).",
+  "Updater.Status.Available": "Uus versioon on saadaval: {0}",
+  "Updater.Status.Downloading": "Värskenduse allalaadimine… {0}%",
+  "Updater.Status.Installing": "Värskenduse paigaldamine…",
+  "Updater.Status.Timeout": "Värskenduste kontroll aegus 10 sekundi järel.",
+  "Updater.Status.Failed": "Värskendusi ei saanud kontrollida.",
+  "Updater.Status.ChecksumFailed": "Allalaaditud värskendus ei läbinud SHA-256 kontrolli.",
+  "Updater.Status.Unsupported": "Automaatne värskendamine nõuab Windowsi, Linuxi või macOS-i x64 versiooni."
+}

src/SharpEmu.GUI/Languages/pl.json

@@ -0,0 +1,213 @@
+{
+  "_languageName": "Polski",
+
+  "Page.Library": "Biblioteka",
+  "Page.Options": "Opcje",
+  "Page.GameCount.One": "1 gra",
+  "Page.GameCount.Other": "{0} gier",
+
+  "Library.SearchWatermark": "Szukaj w bibliotece…",
+  "Library.AddFolder": "Dodaj folder",
+  "Library.OpenFile": "Otwórz plik…",
+  "Library.View.Grid": "Widok siatki",
+  "Library.View.Carousel": "Widok listy",
+
+  "Library.Context.Launch": "Uruchom",
+  "Library.Context.OpenFolder": "Otwórz folder gry",
+  "Library.Context.CopyPath": "Kopiuj ścieżkę",
+  "Library.Context.CopyTitleId": "Kopiuj numer seryjny",
+  "Library.Context.GameSettings": "Ustawienia gry…",
+  "Library.Context.Remove": "Usuń z biblioteki",
+
+  "Library.Empty.Title": "Twoja biblioteka jest pusta",
+  "Library.Empty.Hint": "Dodaj folder zawierający twoje gry, aby rozpocząć.",
+  "Library.Empty.SearchTitle": "Brak gier pasujących do twojego wyszukiwania",
+  "Library.Empty.SearchHint": "Nic w bibliotece nie pasuje do “{0}”.",
+  "Library.Empty.AddFolder": "＋ Dodaj folder z grami",
+
+  "Library.Loading": "Ładowanie biblioteki…",
+  "Library.Stat.Version": "Wersja",
+  "Library.Stat.Installed": "Zainstalowano",
+  "Library.Stat.TitleId": "Numer seryjny",
+  "Common.Back": "Wstecz",
+
+  "Options.General": "Ogólne",
+  "Options.Logging": "Dziennik zdarzeń",
+  "Options.Env.Tab": "Środowisko",
+  "Options.Section.Environment": "ZMIENNE ŚRODOWISKOWE",
+  "Options.Env.Desc": "Przełączniki przekazywane emulatorowi jako zmienne środowiskowe przy uruchomieniu.",
+  "Options.Env.Bthid.Desc": "Zgłaszaj Bluetooth HID jako niedostępny dla tytułów, których middleware kierownicy/FFB odpytuje w nieskończoność.\nZostaw wyłączone normalnie. Niektóre tytuły zawieszają się, gdy inicjalizacja zawiedzie.",
+  "Options.Env.LoopGuard.Desc": "Nie wymuszaj zamknięcia tytułów, które powtarzają to samo wywołanie zbyt długo.\nSpróbuj tego, gdy gra zamyka się sama podczas ładowania.",
+  "Options.Env.WritableApp0.Desc": "Pozwól tytułom tworzyć i zapisywać pliki wewnątrz folderu instalacji.\nPotrzebne w przypadku niespakowanych zrzutów, które zapisują dane zapisu lub konfiguracji w /app0.",
+  "Options.Env.VkValidation.Desc": "Włącz warstwy walidacji Vulkan do debugowania GPU.\nWolne. Wymaga zainstalowanego Vulkan SDK.",
+  "Options.Env.DumpSpirv.Desc": "Zrzuć shadery AGC i ich translacje SPIR-V do folderu shader-dumps.\nUżyj przy zgłaszaniu błędów shaderów lub renderowania.",
+  "Options.Env.LogDirectMemory.Desc": "Loguj alokacje pamięci bezpośredniej i błędy do konsoli.\nUżyj, gdy gra przerywa działanie lub zamyka się podczas uruchamiania.",
+  "Options.Env.LogIo.Desc": "Loguj otwieranie plików, odczyt i rozwiązywanie ścieżek do konsoli.\nUżyj, gdy gra nie może znaleźć swoich plików danych podczas uruchamiania.",
+  "Options.Env.LogNp.Desc": "Loguj wywołania biblioteki NP (PlayStation Network) do konsoli.",
+  "Options.Env.Group.Debug": "Debugowanie",
+  "Options.Env.Group.General": "Ogólne",
+  "Options.Env.RenderDoc.Desc": "Załaduj API RenderDoc w aplikacji, aby można było przechwytywać klatki z wnętrza emulatora.\nNaciśnij F10 podczas działania gry, aby przechwycić jedną klatkę; przechwycone klatki trafiają do user/logs/capture_logs/<TITLE_ID>.\nWymaga zainstalowanego RenderDoc. Spowalnia GPU i zawiesza niektóre tytuły, więc zostaw wyłączone, chyba że debugujesz.",
+  "Options.Env.GuestImageCpuSync.Desc": "Ponownie przesyłaj powierzchnie gościa, które procesor gry nadpisuje.\nZostaw wyłączone normalnie. Włącz dla tytułów, których powierzchnie rysowane przez CPU nigdy nie docierają na ekran.\nKosztuje wydajność i pogarsza działanie niektórych tytułów, takich jak GTA V.",
+  "Options.Env.ForceSubmitOrphanPreambles.Desc": "Dostarczaj preambuły buforów poleceń GPU nawet gdy ich docelowa kolejka ich nie odbiera.\nZostaw wyłączone normalnie. Włącz dla tytułów, które zawieszają się w oczekiwaniu na sygnał z GPU.",
+  "Options.DefaultProfile.Label": "Domyślna nazwa profilu",
+  "Options.DefaultProfile.Desc": "Nazwa używana, gdy gra prosi o wprowadzenie tekstu. Domyślnie Sharp.",
+  "Options.Section.Emulation": "EMULACJA",
+  "Options.Section.Logging": "DZIENNIK ZDARZEŃ",
+  "Options.Section.Launcher": "LAUNCHER",
+  "Options.Section.Rendering": "RENDEROWANIE",
+  "Options.Section.Display": "WYŚWIETLANIE",
+  "Options.Graphics": "Grafika",
+
+  "Options.RenderResolution.Label": "Rozdzielczość wewnętrzna",
+  "Options.RenderResolution.Desc": "Renderuj cele pozaekranowe poniżej natywnej rozdzielczości i skaluj w górę przy prezentacji. Niższe wartości zamieniają jakość obrazu na zapas GPU; działa od następnego uruchomienia.",
+  "Options.RenderResolution.Native": "100% (natywna)",
+  "Options.WindowMode.Label": "Tryb okna",
+  "Options.WindowMode.Desc": "Zwykłe okno, bezramkowe na pulpicie lub pełny ekran wyłączny.",
+  "Options.WindowMode.Windowed": "W oknie",
+  "Options.WindowMode.Borderless": "Bezramkowe",
+  "Options.WindowMode.Exclusive": "Wyłączne",
+  "Options.Resolution.Label": "Rozdzielczość",
+  "Options.Resolution.Desc": "Początkowy rozmiar okna lub rozdzielczość pełnego ekranu wyłącznego.",
+  "Options.Display.Label": "Wyświetlacz",
+  "Options.Display.Desc": "Monitor używany do centrowania i trybu pełnoekranowego.",
+  "Options.RefreshRate.Label": "Częstotliwość odświeżania",
+  "Options.RefreshRate.Desc": "Częstotliwość odświeżania w trybie pełnego ekranu wyłącznego. Automatyczna wybiera najbliższy tryb.",
+  "Options.RefreshRate.Automatic": "Automatyczna",
+  "Options.Scaling.Label": "Skalowanie",
+  "Options.Scaling.Desc": "Skaluj natywny obraz gościa bez zmiany jego wewnętrznej rozdzielczości.",
+  "Options.Scaling.Fit": "Dopasuj",
+  "Options.Scaling.Cover": "Wypełnij",
+  "Options.Scaling.Stretch": "Rozciągnij",
+  "Options.Scaling.Integer": "Całkowitoliczbowe",
+  "Options.VSync.Label": "VSync",
+  "Options.VSync.Desc": "Używaj prezentacji FIFO dla obrazu bez rozrywania.",
+  "Options.Hdr.Label": "Wyjście HDR",
+  "Options.Hdr.Desc": "Używaj HDR, gdy wybrany wyświetlacz i backend graficzny to obsługują. Automatyczny przełącza na SDR.",
+  "Options.Hdr.Auto": "Automatyczny",
+
+  "Options.CpuEngine.Label": "Silnik CPU",
+  "Options.CpuEngine.Desc": "Silnik wykonawczy używany do uruchamiania kodu gry.",
+  "Options.CpuEngine.Native": "Natywny",
+
+  "Options.Strict.Label": "Ścisłe rozwiązywanie dynlib",
+  "Options.Strict.Desc": "Przerwij uruchomienie, gdy importowany symbol nie może zostać rozwiązany.",
+
+  "Options.LogLevel.Label": "Poziom logowania",
+  "Options.LogLevel.Desc": "Szczegółowość wyjścia konsoli emulatora.",
+  "Options.LogLevel.Trace": "Śledzenie",
+  "Options.LogLevel.Debug": "Debugowanie",
+  "Options.LogLevel.Info": "Informacja",
+  "Options.LogLevel.Warning": "Ostrzeżenie",
+  "Options.LogLevel.Error": "Błąd",
+  "Options.LogLevel.Critical": "Krytyczny",
+
+  "Options.TraceImports.Label": "Limit śledzenia importów",
+  "Options.TraceImports.Desc": "Śledź pierwszych N importów na moduł (0 = wyłączone).",
+
+  "Options.LogToFile.Label": "Loguj do pliku",
+  "Options.LogToFile.Desc": "Kopiuj wyjście emulatora do pliku logu.",
+
+  "Options.LogFilePath.Label": "Ścieżka pliku logu",
+  "Options.LogFilePath.Default": "Brak niestandardowej ścieżki — logi trafiają do user/logs obok emulatora.",
+  "Options.LogFilePath.Select": "Wybierz…",
+
+  "Options.OverrideLogFile.Label": "Nadpisz plik logu",
+  "Options.OverrideLogFile.Desc": "Użyj dokładnej ścieżki pliku zamiast dopisywania ID tytułu i znacznika czasu.",
+
+  "Options.TitleMusic.Label": "Muzyka tytułowa",
+  "Options.TitleMusic.Desc": "Odtwarzaj w pętli muzykę podglądu wybranej gry w bibliotece.",
+
+  "Options.Discord.Label": "Discord Rich Presence",
+  "Options.Discord.Desc": "Pokazuj uruchomioną grę na swoim profilu Discord.",
+
+  "Options.Language.Label": "Język emulatora",
+  "Options.Language.Desc": "Język używany w całym launcherze. Działa natychmiast.",
+
+  "Common.On": "Włącz",
+  "Common.Off": "Wyłącz",
+  "Common.Save": "Zapisz",
+  "Common.Cancel": "Anuluj",
+
+
+  "Console.Title": "KONSOLA",
+  "Console.SearchWatermark": "Szukaj...",
+  "Console.AutoScroll": "Autoprzewijanie",
+  "Console.Split": "Podziel",
+  "Console.Copy": "Kopiuj",
+  "Console.Clear": "Wyczyść",
+  "Console.WindowTitle": "Konsola SharpEmu",
+
+  "Launch.NoGameSelected": "Nie wybrano gry",
+  "Launch.NoGameHint": "Wybierz grę z biblioteki lub otwórz bezpośrednio plik eboot.bin.",
+  "Launch.Idle": "Bezczynny",
+  "Launch.Console": "≡ Konsola",
+  "Launch.Launch": "▶  Uruchom",
+  "Launch.Stop": "■  Zatrzymaj",
+  "Launch.Running": "Działa — {0}",
+  "Launch.Stopping": "Zatrzymywanie…",
+  "Launch.Exited": "Zakończono z kodem {0} ({1})",
+  "Launch.ExeNotFound": "Nie znaleziono pliku wykonywalnego SharpEmu. Najpierw zbuduj projekt SharpEmu.CLI (dotnet build).",
+  "Launch.LogFile": "Plik logu: {0}",
+  "Launch.Command": "$ SharpEmu {0}",
+  "Launch.StartFailed": "Nie udało się uruchomić emulatora: {0}",
+  "Launch.ProcessExited": "Proces zakończony z kodem {0} ({1}).",
+
+  "Exit.Ok": "OK",
+  "Exit.InvalidArguments": "nieprawidłowe argumenty",
+  "Exit.EbootNotFound": "nie znaleziono eboot",
+  "Exit.RuntimeException": "wyjątek czasu wykonania",
+  "Exit.EmulationError": "błąd emulacji",
+  "Exit.Unknown": "nieznany",
+
+  "Status.EmulatorLocating": "Emulator: lokalizowanie…",
+  "Status.EmulatorPath": "Emulator: {0}",
+  "Status.EmulatorNotFound": "Emulator: nie znaleziono pliku wykonywalnego SharpEmu — najpierw zbuduj SharpEmu.CLI.",
+  "Status.ScanningLibrary": "Skanowanie biblioteki…",
+  "Status.AddFolderPrompt": "Dodaj folder z grami, aby wypełnić bibliotekę.",
+  "Status.LibraryScanned": "Biblioteka przeskanowana: {0} gier w {1} folderach.",
+  "Status.CouldNotOpenFolder": "Nie można otworzyć folderu: {0}",
+  "Status.CopiedToClipboard": "{0} skopiowano do schowka.",
+  "Status.RemovedFromLibrary": "Usunięto \u201c{0}\u201d z biblioteki. Dodaj ponownie folder, aby przywrócić.",
+  "Status.Running": "Uruchomiono {0}",
+  "Status.Stopping": "Zatrzymywanie…",
+  "Status.Idle": "Bezczynny",
+
+  "Clipboard.Path": "Ścieżka",
+  "Clipboard.TitleId": "ID tytułu",
+
+  "Discord.Playing": "Gra w {0}",
+  "Discord.Browsing": "Przegląda bibliotekę",
+
+  "Dialog.ChooseGameFolder": "Wybierz folder zawierający gry",
+  "Dialog.OpenExecutable": "Otwórz plik wykonywalny do uruchomienia",
+  "Dialog.PsExecutables": "Pliki wykonywalne PS",
+  "Dialog.SaveLogFile": "Wybierz, gdzie zapisać plik logu",
+  "Dialog.PlainTextFiles": "Pliki tekstowe",
+  "Dialog.LogFiles": "Pliki logów",
+
+  "Options.About" : "O programie",
+  "About.Github.Label": "GitHub",
+  "About.Github.Desc": "Kod źródłowy, zgłoszenia i rozwój projektu.",
+  "About.Github.LatestCommitLabel": "Najnowszy commit",
+  "About.Github.LatestCommitDescription": "Najnowszy commit na gałęzi głównej",
+  "About.Discord.Label": "Discord",
+  "About.Discord.Desc": "Dołącz do społeczności, uzyskaj wsparcie i śledź rozwój.",
+  "About.GithubButton": "Wspieraj na GitHubie!",
+  "About.DiscordComingSoon": "Wkrótce",
+
+  "Updater.Auto.Label": "Sprawdzaj aktualizacje przy uruchomieniu",
+  "Updater.Auto.Desc": "Sprawdza GitHub bez opóźniania uruchomienia.",
+  "Updater.Label": "Aktualizacje",
+  "Updater.Check": "Sprawdź aktualizacje",
+  "Updater.DownloadRestart": "Pobierz i uruchom ponownie",
+  "Updater.Status.Ready": "Obecna wersja: {0}",
+  "Updater.Status.Checking": "Sprawdzanie aktualizacji…",
+  "Updater.Status.Current": "Masz najnowszą wersję ({0}).",
+  "Updater.Status.Available": "Dostępna nowa wersja: {0}",
+  "Updater.Status.Downloading": "Pobieranie aktualizacji… {0}%",
+  "Updater.Status.Installing": "Instalowanie aktualizacji…",
+  "Updater.Status.Timeout": "Sprawdzanie aktualizacji przekroczyło limit 10 sekund.",
+  "Updater.Status.Failed": "Nie udało się sprawdzić aktualizacji.",
+  "Updater.Status.ChecksumFailed": "Pobrana aktualizacja nie przeszła weryfikacji SHA-256.",
+  "Updater.Status.Unsupported": "Automatyczna aktualizacja wymaga wersji dla Windows, Linux lub macOS x64."
+}

src/SharpEmu.HLE/ExportedFunction.cs

@@ -5,7 +5,13 @@ namespace SharpEmu.HLE;
 
 public sealed class ExportedFunction
 {
-    public ExportedFunction(string libraryName, string nid, string name, Generation target, SysAbiFunction function)
+    public ExportedFunction(
+        string libraryName,
+        string nid,
+        string name,
+        Generation target,
+        SysAbiFunction function,
+        bool preferLle = false)
     {
         ArgumentException.ThrowIfNullOrWhiteSpace(libraryName);
         ArgumentException.ThrowIfNullOrWhiteSpace(nid);
@@ -17,6 +23,7 @@ public sealed class ExportedFunction
         Name = name;
         Target = target;
         Function = function;
+        PreferLle = preferLle;
     }
 
     public string LibraryName { get; }
@@ -28,4 +35,10 @@ public sealed class ExportedFunction
     public Generation Target { get; }
 
     public SysAbiFunction Function { get; }
+
+    /// <summary>
+    /// A loaded guest export is authoritative for this registration. The HLE function
+    /// remains available as an explicit fallback when no usable guest target exists.
+    /// </summary>
+    public bool PreferLle { get; }
 }

src/SharpEmu.HLE/SysAbiExportAttribute.cs

@@ -3,7 +3,7 @@
 
 namespace SharpEmu.HLE;
 
-[AttributeUsage(AttributeTargets.Method, Inherited = false, AllowMultiple = false)]
+[AttributeUsage(AttributeTargets.Method, Inherited = false, AllowMultiple = true)]
 public sealed class SysAbiExportAttribute : Attribute
 {
     public string LibraryName { get; set; } = "libKernel";
@@ -13,4 +13,11 @@ public sealed class SysAbiExportAttribute : Attribute
     public string ExportName { get; set; } = string.Empty;
 
     public Generation Target { get; set; } = Generation.None;
+
+    /// <summary>
+    /// Prefer a matching export from a loaded guest module and use this handler only
+    /// as the explicit fallback when that LLE provider is unavailable. Individual
+    /// handlers define whether that fallback is fail-closed or compatibility behavior.
+    /// </summary>
+    public bool PreferLle { get; set; }
 }

src/SharpEmu.Libs/Agc/AgcExports.cs

@@ -118,6 +118,29 @@ public static partial class AgcExports
     // Multiple producers can share one target label; last-writer-wins would
     // starve waits on the others.
     private static readonly Dictionary<ulong, List<ulong>> _cbReleaseMemTargets = new();
+    // CMASK meta-state tracking: maps colour-buffer addresses to their
+    // compression metadata.  Keyed by colour-buffer base address so the
+    // consumption path (which only knows the surface address) can query
+    // directly without a reverse lookup.
+    private record struct MetaSurfaceInfo(
+        ulong CmaskAddress,
+        uint ClearWord0,
+        uint ClearWord1,
+        bool IsCleared);
+    private static readonly Dictionary<ulong, MetaSurfaceInfo> _metaSurfaces = new();
+    // Reverse map: CMASK address → colour-buffer address.  Needed so
+    // CheckCmaskWrite (which only sees the write target address) can
+    // find the owning surface.
+    private static readonly Dictionary<ulong, ulong> _cmaskToColorBuffer = new();
+    // Guards _metaSurfaces and _cmaskToColorBuffer.  Two threads touch them:
+    // the parse thread (registration in TrackCmaskAddresses, CheckCmaskWrite
+    // from DMA/compute writes, EFC consumption) and the render thread
+    // (MarkAllSurfacesCleared at guest flip, IsMetaClearedForSurface /
+    // ConsumeMetaClear / GetMetaClearValue at pass-record time).  Plain
+    // Dictionaries corrupt under concurrent write, so every access below
+    // holds this gate.  Keep the critical sections tiny and never block on
+    // anything external while holding it.
+    private static readonly object _metaSurfaceGate = new();
     // header -> {ring base, write cursor} of the last submitted slice.
     // Submissions stay cursor-bounded since rings aren't zeroed. Lap
     // distinguishes a stale cursor from a previous pass over the same base.
@@ -1095,15 +1118,20 @@ public static partial class AgcExports
     private const uint CbColor0Base = 0x318;
     private const uint CbColorRegisterStride = 15;
     private const uint CbColor0Info = 0x31C;
+    private const uint CbColor0Cmask = 0x31F;
     private const uint CbColor0ClearWord0 = 0x323;
     private const uint CbColor0ClearWord1 = 0x324;
+    private const uint CbColor0DccBase = 0x325;
     private const uint CbColor0BaseExt = 0x390;
+    private const uint CbColor0CmaskBaseExt = 0x398;
+    private const uint CbColor0DccBaseExt = 0x3A8;
     private const uint CbColor0Attrib2 = 0x3B0;
     private const uint CbColor0Attrib3 = 0x3B8;
     // CB_COLORn_INFO.DCC_ENABLE (gc_10_1_0_sh_mask.h). On GFX10 the legacy
     // FAST_CLEAR and COMPRESSION bits stay clear because DCC, not CMASK,
     // carries the compression.
     private const uint CbColorInfoDccEnableMask = 1u << 28;
+    private const uint CbColorInfoFastClearEnableMask = 1u << 12;
     private const uint CbBlend0Control = 0x1E0;
     private const uint PaScModeCntl0 = 0x292;
     // GFX10 DB context registers (register byte address minus 0x28000, / 4).
@@ -1123,8 +1151,8 @@ public static partial class AgcExports
     private const uint EsUserDataRegister = 0xCC;
     private const uint ComputeUserDataRegister = 0x240;
     private const uint NggUserDataScalarRegisterBase = 8;
-    private const uint Gen5TextureFormatR8G8B8A8Unorm = 10;
-    private const uint Gen5TextureFormatR16G16B16A16Float = 12;
+    internal const uint Gen5TextureFormatR8G8B8A8Unorm = 10;
+    internal const uint Gen5TextureFormatR16G16B16A16Float = 12;
     private const uint Gen5TextureType1D = 8;
     private const uint Gen5TextureType2D = 9;
     private const uint Gen5TextureType3D = 10;
@@ -5083,6 +5111,10 @@ public static partial class AgcExports
 
             if (op == ItNop && register == RDmaData && length >= 7)
             {
+                // Ensure CMASK addresses are tracked before DMA fills
+                var tempTargets = GetRenderTargets(state.CxRegisters);
+                TrackCmaskAddresses(state.CxRegisters, tempTargets);
+
                 ApplySubmittedDmaData(
                     ctx,
                     gpuState,
@@ -5243,6 +5275,7 @@ public static partial class AgcExports
             {
                 TraceFramePacketSummary(state);
                 SyncCpuWrittenGuestImages(ctx);
+                GpuWaitRegistry.AdvanceFrame();
                 if (!TryReadUInt32(ctx, currentAddress + 4, out var videoOutHandle) ||
                     !TryReadUInt32(ctx, currentAddress + 8, out var displayBufferIndexRaw) ||
                     !TryReadUInt32(ctx, currentAddress + 12, out var flipMode) ||
@@ -6264,6 +6297,13 @@ public static partial class AgcExports
         ulong byteCount,
         uint? fillValue)
     {
+        // Check if this DMA write targets a CMASK address (shadPS4's FillBuffer
+        // logic: when a buffer fill targets CMASK metadata, mark it as "all clear")
+        if (fillValue is { } fillVal && fillVal == 0)
+        {
+            CheckCmaskWrite(destinationAddress, null);
+        }
+
         var hasImage = GuestGpu.Current.TryGetGuestImageExtent(
             destinationAddress,
             out var width,
@@ -6478,6 +6518,25 @@ public static partial class AgcExports
                     var targetAddress = destinationAddress +
                         (incrementAddress ? (ulong)index * sizeof(uint) : 0);
                     wroteData = TryWriteUInt32(ctx, targetAddress, values[index]);
+                    if (wroteData)
+                    {
+                        GpuWaitRegistry.RecordProduced(
+                            ctx.Memory, targetAddress, values[index]);
+                    }
+                }
+
+                // Like ReleaseMem dataSel=2: a 64-bit WAIT_REG_MEM watches an
+                // 8-byte label written as two 32-bit dwords. Record the combined
+                // 64-bit value so a 64-bit EQ can latch even though the writes
+                // landed as two 32-bit stores.
+                if (wroteData && dwordCount >= 2 && incrementAddress)
+                {
+                    var combined = ((ulong)values[1] << 32) | values[0];
+                    GpuWaitRegistry.RecordProduced(
+                        ctx.Memory, destinationAddress, combined);
+                    // Also latch the high half's address for symmetry: a stray
+                    // 32-bit wait on the high dword should not be confused, but
+                    // recording it does not hurt and mirrors the per-dword stores.
                 }
 
                 if (tracePacket)
@@ -7053,7 +7112,13 @@ public static partial class AgcExports
 
         if (hasCurrent && GpuWaitRegistry.Compare(waiter, currentValue))
         {
-            return false; // already satisfied — keep parsing
+            // Value satisfies the condition, but only bypass if the label was
+            // written in the current frame. A stale label from a previous frame
+            // means the producer hasn't written yet this frame — must wait.
+            if (GpuWaitRegistry.IsLabelFresh(ctx.Memory, waitAddress))
+            {
+                return false; // satisfied by current-frame write — keep parsing
+            }
         }
 
         if (!_gpuWaitSuspendEnabled)
@@ -8044,7 +8109,8 @@ public static partial class AgcExports
         var hasPsInputEna = state.CxRegisters.TryGetValue(SpiPsInputEna, out var psInputEna);
         var hasPsInputAddr = state.CxRegisters.TryGetValue(SpiPsInputAddr, out var psInputAddr);
         state.UcRegisters.TryGetValue(VgtPrimitiveType, out var primitiveType);
-        var renderTargets = GetRenderTargets(state.CxRegisters);
+var renderTargets = GetRenderTargets(state.CxRegisters);
+        TrackCmaskAddresses(state.CxRegisters, renderTargets);
         var drawSequence = ++gpuState.WorkSequence;
         if (state.PendingTargetlessDraw is { } stalePendingDraw)
         {
@@ -8064,6 +8130,31 @@ public static partial class AgcExports
         if (TryGetCbColorControlMode(state.CxRegisters, out var cbMode) &&
             IsCbMetadataColorMode(cbMode))
         {
+            // EliminateFastClear: the game explicitly asks the CB to clear
+            // the fast-clear metadata and the colour buffer.
+            if (cbMode == (uint)CbColorMode.EliminateFastClear &&
+                renderTargets.Count > 0 &&
+                renderTargets[0].Address != 0)
+            {
+                var targetAddr = renderTargets[0].Address;
+                bool requestClear;
+                lock (_metaSurfaceGate)
+                {
+                    requestClear =
+                        _metaSurfaces.TryGetValue(targetAddr, out var meta) &&
+                        meta.IsCleared;
+                    if (requestClear)
+                    {
+                        _metaSurfaces[targetAddr] = meta with { IsCleared = false };
+                    }
+                }
+
+                if (requestClear)
+                {
+                    VulkanVideoPresenter.RequestGuestColorClear(targetAddr);
+                }
+            }
+
             if (_traceAgcShader || ShouldTraceHotPath(ref _cbMetadataSkipTraceCount))
             {
                 TraceAgcShader(
@@ -8278,6 +8369,34 @@ public static partial class AgcExports
                 return;
             }
 
+            // DbRenderControl CLEARON (bit0): when set, the CB clears color
+            // targets on first draw. Handle color targets (depth is already
+            // handled by DecodeDepthState).
+            if (state.CxRegisters.TryGetValue(DbRenderControl, out var rc) && (rc & 0x1u) != 0)
+            {
+                foreach (var rt in translatedDraw.RenderTargets)
+                {
+                    if (rt.Address != 0)
+                    {
+                        VulkanVideoPresenter.RequestGuestColorClear(rt.Address);
+                    }
+                }
+            }
+
+            // CMASK fast clear: CB_COLORn_INFO.FAST_CLEAR (bit12) set on
+            // one or more targets. The CB clears via CMASK before the draw
+            // writes; mark targets for clear-on-first-use.
+            if (IsCmaskFastClearDraw(state.CxRegisters, translatedDraw.RenderTargets))
+            {
+                foreach (var rt in translatedDraw.RenderTargets)
+                {
+                    if (rt.Address != 0)
+                    {
+                        VulkanVideoPresenter.RequestGuestColorClear(rt.Address);
+                    }
+                }
+            }
+
             var firstTarget = translatedDraw.RenderTargets.FirstOrDefault();
             if (firstTarget.Address != 0)
             {
@@ -9570,6 +9689,193 @@ public static partial class AgcExports
             CoversClipSpace(vertexInputs, vertexCount);
     }
 
+    /// <summary>
+    /// GFX10 CMASK fast clear: CB_COLORn_INFO.FAST_CLEAR (bit 12) set on
+    /// one or more targets. The CB clears via CMASK before the draw writes;
+    /// mark targets for clear-on-first-use. Unlike DCC, the draw content
+    /// IS written (not dropped). Dead Cells uses DbRenderControl CLEARON
+    /// instead (bit0), not this mechanism.
+    /// </summary>
+    private static bool IsCmaskFastClearDraw(
+        IReadOnlyDictionary<uint, uint> registers,
+        IReadOnlyList<RenderTargetDescriptor> renderTargets)
+    {
+        foreach (var rt in renderTargets)
+        {
+            var stride = rt.Slot * CbColorRegisterStride;
+            if (registers.TryGetValue(CbColor0Info + stride, out var info) &&
+                (info & CbColorInfoFastClearEnableMask) != 0)
+            {
+                return true;
+            }
+        }
+
+        return false;
+    }
+
+    /// <summary>
+    /// Registers the CMASK metadata mapping for each colour buffer.
+    /// Does NOT mark as cleared — clearing only happens on actual clear
+    /// events (DMA fill, compute write, EFC draw).
+    /// </summary>
+    private static void TrackCmaskAddresses(
+        IReadOnlyDictionary<uint, uint> registers,
+        IReadOnlyList<RenderTargetDescriptor> renderTargets)
+    {
+        foreach (var rt in renderTargets)
+        {
+            var stride = rt.Slot * CbColorRegisterStride;
+
+            // CMASK metadata address (legacy GCN path).
+            var cmaskRegAddr = CbColor0Cmask + stride;
+            registers.TryGetValue(cmaskRegAddr, out var cmaskLow);
+            var cmaskExtAddr = CbColor0CmaskBaseExt + rt.Slot;
+            registers.TryGetValue(cmaskExtAddr, out var cmaskExt);
+            var cmaskAddress = ((ulong)(cmaskExt & 0xFFu) << 40) |
+                               ((ulong)(cmaskLow & 0x1FFFFFFFu) << 8);
+
+            // DCC metadata address (GFX10+ primary path).
+            var dccRegAddr = CbColor0DccBase + stride;
+            registers.TryGetValue(dccRegAddr, out var dccLow);
+            var dccExtAddr = CbColor0DccBaseExt + rt.Slot;
+            registers.TryGetValue(dccExtAddr, out var dccExt);
+            var dccAddress = ((ulong)(dccExt & 0xFFu) << 40) |
+                             ((ulong)(dccLow & 0x1FFFFFFFu) << 8);
+
+            // Prefer CMASK if present; fall back to DCC.
+            var metaAddress = cmaskAddress != 0 ? cmaskAddress : dccAddress;
+
+            var cw0Addr = CbColor0ClearWord0 + stride;
+            var cw1Addr = CbColor0ClearWord1 + stride;
+            registers.TryGetValue(cw0Addr, out var cw0);
+            registers.TryGetValue(cw1Addr, out var cw1);
+
+            lock (_metaSurfaceGate)
+            {
+                _metaSurfaces[rt.Address] = new MetaSurfaceInfo(
+                    metaAddress, cw0, cw1,
+                    // Re-registration runs on every draw; keep the cleared state
+                    // so a mark-clear event survives until the pass consumes it.
+                    // If the metadata binding changed, the old state refers to
+                    // the old metadata and must be reset.
+                    IsCleared: _metaSurfaces.TryGetValue(rt.Address, out var prev) &&
+                               prev.IsCleared &&
+                               prev.CmaskAddress == metaAddress);
+                if (metaAddress != 0)
+                {
+                    _cmaskToColorBuffer[metaAddress] = rt.Address;
+                }
+            }
+        }
+    }
+
+    /// <summary>
+    /// Checks if a write targets a registered CMASK address. If so,
+    /// marks the owning colour buffer's metadata as "all clear".
+    /// </summary>
+    private static void CheckCmaskWrite(
+        ulong writeAddress,
+        SubmittedGpuState? gpuState)
+    {
+        if (writeAddress == 0)
+        {
+            return;
+        }
+
+        lock (_metaSurfaceGate)
+        {
+            // Exact match: write directly to a registered CMASK address.
+            if (_cmaskToColorBuffer.TryGetValue(writeAddress, out var cbAddr))
+            {
+                if (_metaSurfaces.TryGetValue(cbAddr, out var meta))
+                {
+                    _metaSurfaces[cbAddr] = meta with { IsCleared = true };
+                }
+
+                return;
+            }
+
+            // CMASK surfaces are small (typically ≤ 4 KiB).  Check the ±1024
+            // window around each registered address to catch partial writes.
+            foreach (var (cmaskAddr, colorBufAddr) in _cmaskToColorBuffer)
+            {
+                if (writeAddress >= cmaskAddr && writeAddress < cmaskAddr + 1024)
+                {
+                    if (_metaSurfaces.TryGetValue(colorBufAddr, out var meta))
+                    {
+                        _metaSurfaces[colorBufAddr] = meta with { IsCleared = true };
+                    }
+
+                    return;
+                }
+            }
+        }
+    }
+
+    /// <summary>
+    /// Returns true if the colour buffer at <paramref name="colorBufferAddress"/>
+    /// has pending CMASK "all clear" metadata — i.e. the surface was fast-cleared
+    /// but not yet rendered into.
+    /// </summary>
+    internal static bool IsMetaClearedForSurface(ulong colorBufferAddress)
+    {
+        lock (_metaSurfaceGate)
+        {
+            return _metaSurfaces.TryGetValue(colorBufferAddress, out var meta) &&
+                   meta.IsCleared;
+        }
+    }
+
+    /// <summary>
+    /// Consumes the "all clear" state for the given surface, marking it dirty.
+    /// Called after the first render pass uses LoadOp.Clear.
+    /// </summary>
+    internal static void ConsumeMetaClear(ulong colorBufferAddress)
+    {
+        lock (_metaSurfaceGate)
+        {
+            if (_metaSurfaces.TryGetValue(colorBufferAddress, out var meta))
+            {
+                _metaSurfaces[colorBufferAddress] = meta with { IsCleared = false };
+            }
+        }
+    }
+
+    /// <summary>
+    /// Returns the CB clear word values for the given colour buffer.
+    /// </summary>
+    internal static (uint Cw0, uint Cw1) GetMetaClearValue(ulong colorBufferAddress)
+    {
+        lock (_metaSurfaceGate)
+        {
+            if (_metaSurfaces.TryGetValue(colorBufferAddress, out var meta))
+            {
+                return (meta.ClearWord0, meta.ClearWord1);
+            }
+        }
+
+        return (0, 0);
+    }
+
+    /// <summary>
+    /// Marks all registered surfaces as "all clear".  Called at guest flip
+    /// (frame boundary).  Real hardware applies a fast clear / load-clear to
+    /// its per-frame surfaces every frame; the emulator restores that
+    /// per-frame clear here, per surface, at flip time.  This is the
+    /// per-surface successor of the removed flip-arm heuristic (which reset
+    /// only the first multi-attachment group).
+    /// </summary>
+    internal static void MarkAllSurfacesCleared()
+    {
+        lock (_metaSurfaceGate)
+        {
+            foreach (var (addr, meta) in _metaSurfaces)
+            {
+                _metaSurfaces[addr] = meta with { IsCleared = true };
+            }
+        }
+    }
+
     /// <summary>
     /// True when the draw's float32x3 position stream spans the full clip
     /// rectangle, i.e. x and y both reach -1 and +1.
@@ -9982,7 +10288,7 @@ public static partial class AgcExports
     private static readonly HashSet<ulong> _sampledRenderTargets = new();
     private static readonly object _renderTargetProbeGate = new();
     private static long _renderTargetSampleTraceCount;
-    private static long _indirectDrawProbeCount;
+private static long _indirectDrawProbeCount;
     private static long _indirectDrawEmitCount;
     private static long _indirectDrawEmitRejectCount;
     private static long _indirectMultiProbeCount;
@@ -12174,15 +12480,18 @@ public static partial class AgcExports
 
         if (dispatchEndX == 0 || dispatchEndY == 0 || dispatchEndZ == 0)
         {
-            // Indirect dispatches read their dimensions from a guest buffer a
-            // prior GPU dispatch fills. Zero here means that producer has not run
-            // yet — signal the caller to suspend on the dims buffer and retry,
-            // rather than dropping the work (which black-screens GPU-driven games
-            // like Astro Bot). Direct dispatches carry dims inline, so a zero is
-            // genuinely malformed and still rejected.
-            if (opcode == ItDispatchIndirect)
+            // For indirect dispatches (both absolute and base), zero dimensions are a valid outcome
+            // of GPU culling passes (0 workgroups). VulkanVideoPresenter handles groupCount = 0 as a clean no-op.
+            if (opcode == ItDispatchIndirect || dispatchSource is "absolute-indirect" or "base-indirect")
             {
-                indirectDimsRetryAddress = dimensionsAddress;
+                var waveCount = (initiator & (1u << 15)) != 0 ? 32u : 64u;
+                dispatch = new ComputeDispatch(
+                    0, 0, 0,
+                    0, 0, 0,
+                    waveCount,
+                    IsIndirect: true,
+                    0, 0, 0);
+                return true;
             }
 
             return RejectComputeDispatch(
@@ -12465,6 +12774,10 @@ public static partial class AgcExports
                     shaderAddress,
                     binding.Opcode);
 
+                // Check if this compute shader writes to a CMASK address
+                // (shadPS4's IsComputeMetaClear logic)
+                CheckCmaskWrite(texture.Address, gpuState);
+
                 TraceAgcShader(
                     $"agc.compute_writer addr=0x{texture.Address:X16} " +
                     $"fmt={texture.Format} num={texture.NumberType} tile={texture.TileMode} " +
@@ -13060,11 +13373,14 @@ public static partial class AgcExports
                     return;
                 }
 
-                GuestImageWriteTracker.Track(
+GuestImageWriteTracker.Track(
                     destinationAddress,
                     (ulong)output.Length,
                     VulkanVideoPresenter.CurrentGuestWorkSequenceForDiagnostics,
                     "agc.constant-fill");
+
+                VulkanVideoPresenter.RequestGuestColorClear(destinationAddress);
+
             },
             $"constant_fill dst=0x{destinationAddress:X16} bytes={output.Length}");
         description =

src/SharpEmu.Libs/Agc/AgcVertexMetadata.cs

@@ -238,10 +238,12 @@ internal static class AgcVertexMetadata
     }
 
     /// <summary>
-    /// Patch IR-discovered fetches from the attrib table onto the V# format/offset.
+    /// Patch IR-discovered fetches from the attrib table onto the V# layout.
     /// Prefer 1:1 Location pairing when counts match on one interleaved stream
-    /// (GTA UI glyphs). Otherwise match by stride + byte offset. Never rebases
-    /// BaseAddress/Data/Location/Pc/PerInstance.
+    /// (GTA UI glyphs). Otherwise match by the effective captured byte offset.
+    /// Never rebases BaseAddress/Data/Location/Pc or overwrites a discovered
+    /// offset: metadata may refine the format, stride and instance rate only
+    /// after both address keys independently resolve to the same attribute.
     /// </summary>
     internal static IReadOnlyList<Gen5VertexInputBinding> MergeVertexInputsFromMetadata(
         CpuContext ctx,
@@ -269,13 +271,13 @@ internal static class AgcVertexMetadata
         var changed = false;
         foreach (var input in discovered)
         {
-            if (!TryMatchMetadataResource(input, resources, usedResources, out var resource, out var fillOffset))
+            if (!TryMatchMetadataResource(input, resources, usedResources, out var resource))
             {
                 merged.Add(input);
                 continue;
             }
 
-            var refined = ApplyMetadataFormat(input, resource, fillOffset);
+            var refined = ApplyMetadataFormat(input, resource);
             changed |= refined != input;
             merged.Add(refined);
         }
@@ -286,7 +288,7 @@ internal static class AgcVertexMetadata
     /// <summary>
     /// When discovery and metadata describe the same interleaved stream with
     /// equal attribute counts, pair by sorted Location (semantic order).
-    /// Keeps each binding's Pc/Location for SPIR-V; overlays format + offset.
+    /// Keeps each binding's Pc/Location/address for SPIR-V and overlays layout.
     /// </summary>
     private static bool TryMergeByLocationPairing(
         IReadOnlyList<Gen5VertexInputBinding> discovered,
@@ -299,34 +301,36 @@ internal static class AgcVertexMetadata
             return false;
         }
 
-        var orderedInputs = discovered.OrderBy(static input => input.Location).ToArray();
+        var orderedInputs = discovered
+            .Select(static (input, originalIndex) => (Input: input, OriginalIndex: originalIndex))
+            .OrderBy(static entry => entry.Input.Location)
+            .ThenBy(static entry => entry.OriginalIndex)
+            .ToArray();
         var orderedResources = resources.OrderBy(static resource => resource.Location).ToArray();
         var streamBase = orderedResources[0].SharpBase;
         var streamStride = orderedResources[0].Stride;
         for (var index = 0; index < orderedResources.Length; index++)
         {
             var resource = orderedResources[index];
-            var input = orderedInputs[index];
+            var input = orderedInputs[index].Input;
             if (resource.SharpBase != streamBase ||
                 resource.Stride != streamStride ||
-                (input.Stride != 0 && input.Stride != streamStride) ||
-                !IsSameVertexStream(input, resource))
+                !TryGetMetadataOffset(input, resource, out var resolvedOffset) ||
+                resolvedOffset != input.OffsetBytes)
             {
                 return false;
             }
         }
 
-        var byPc = new Dictionary<uint, Gen5VertexInputBinding>(discovered.Count);
+        var result = discovered.ToArray();
         var changed = false;
         for (var index = 0; index < orderedInputs.Length; index++)
         {
-            var input = orderedInputs[index];
+            var input = orderedInputs[index].Input;
             var resource = orderedResources[index];
-            var fillOffset = input.BaseAddress == resource.SharpBase ||
-                             IsAddressInsideCapturedSpan(input, resource.SharpBase);
-            var refined = ApplyMetadataFormat(input, resource, fillOffset);
+            var refined = ApplyMetadataFormat(input, resource);
             changed |= refined != input;
-            byPc[input.Pc] = refined;
+            result[orderedInputs[index].OriginalIndex] = refined;
         }
 
         if (!changed)
@@ -334,20 +338,13 @@ internal static class AgcVertexMetadata
             return false;
         }
 
-        var result = new Gen5VertexInputBinding[discovered.Count];
-        for (var index = 0; index < discovered.Count; index++)
-        {
-            result[index] = byPc[discovered[index].Pc];
-        }
-
         merged = result;
         return true;
     }
 
     private static Gen5VertexInputBinding ApplyMetadataFormat(
         Gen5VertexInputBinding input,
-        MetadataVertexResource resource,
-        bool fillOffsetBytes)
+        MetadataVertexResource resource)
     {
         var components = input.ComponentCount != 0 &&
                          input.ComponentCount < resource.ComponentCount
@@ -359,7 +356,8 @@ internal static class AgcVertexMetadata
             DataFormat = resource.DataFormat,
             NumberFormat = resource.NumberFormat,
             ComponentCount = components,
-            OffsetBytes = fillOffsetBytes ? resource.OffsetBytes : input.OffsetBytes,
+            Stride = resource.Stride,
+            PerInstance = resource.PerInstance,
         };
     }
 
@@ -434,14 +432,11 @@ internal static class AgcVertexMetadata
         Gen5VertexInputBinding input,
         IReadOnlyList<MetadataVertexResource> resources,
         bool[] usedResources,
-        out MetadataVertexResource resource,
-        out bool fillOffsetBytes)
+        out MetadataVertexResource resource)
     {
         resource = default;
-        fillOffsetBytes = false;
         var bestScore = int.MinValue;
         var bestIndex = -1;
-        var bestFillOffset = false;
         for (var index = 0; index < resources.Count; index++)
         {
             if (usedResources[index])
@@ -450,100 +445,64 @@ internal static class AgcVertexMetadata
             }
 
             var candidate = resources[index];
-            if (candidate.Stride != 0 &&
-                input.Stride != 0 &&
-                candidate.Stride != input.Stride)
+            if (!TryGetMetadataOffset(input, candidate, out var resolvedOffset) ||
+                resolvedOffset != input.OffsetBytes)
             {
                 continue;
             }
 
-            if (!IsSameVertexStream(input, candidate))
-            {
-                continue;
-            }
+            // The effective captured offset (including any base rebasing done
+            // while coalescing adjacent vertex streams) is the primary key.
+            var score = 400;
 
-            var attrAddress = candidate.SharpBase + candidate.OffsetBytes;
-            var score = int.MinValue;
-            var fillOffset = false;
-
-            // Post-capture interleaved: shared BaseAddress, distinct OffsetBytes.
-            if (input.OffsetBytes == candidate.OffsetBytes &&
-                (input.BaseAddress == candidate.SharpBase ||
-                 IsAddressInsideCapturedSpan(input, candidate.SharpBase)))
+            // Discovery can carry a stale inferred stride (notably 32 for a
+            // real stride-40 interleaved layout). Prefer a matching stride
+            // when candidates are otherwise equivalent, but do not reject an
+            // unambiguous metadata match: the V# descriptor is authoritative.
+            if (input.Stride == candidate.Stride)
             {
-                score = 400;
-            }
-            // IR prolog baked attrib offset into the V# base.
-            else if (input.BaseAddress == attrAddress)
-            {
-                score = 350;
-            }
-            // Discovery never saw the attrib offset — only safe when this
-            // resource's offset uniquely identifies it among unused entries.
-            else if (input.BaseAddress == candidate.SharpBase &&
-                     input.OffsetBytes == 0 &&
-                     candidate.OffsetBytes != 0 &&
-                     IsUniqueUnusedOffset(resources, usedResources, candidate.OffsetBytes, index))
-            {
-                score = 300;
-                fillOffset = true;
-            }
-            else if (input.BaseAddress == candidate.SharpBase &&
-                     input.OffsetBytes == 0 &&
-                     candidate.OffsetBytes == 0)
-            {
-                score = 250;
+                score += 25;
             }
 
             if (score > bestScore)
             {
                 bestScore = score;
                 bestIndex = index;
-                bestFillOffset = fillOffset;
             }
         }
 
-        // Require an offset-aware match. Bare SharpBase ties (score 250) are
-        // only accepted when a single unused resource remains for that stream.
-        if (bestIndex < 0 || bestScore < 300)
+        if (bestIndex < 0)
         {
-            if (bestIndex < 0 || bestScore < 250)
-            {
-                return false;
-            }
-
-            var unusedSameStream = 0;
-            for (var index = 0; index < resources.Count; index++)
-            {
-                if (!usedResources[index] && IsSameVertexStream(input, resources[index]))
-                {
-                    unusedSameStream++;
-                }
-            }
-
-            if (unusedSameStream != 1)
-            {
-                return false;
-            }
+            return false;
         }
 
         usedResources[bestIndex] = true;
         resource = resources[bestIndex];
-        fillOffsetBytes = bestFillOffset;
         return true;
     }
 
-    private static bool IsSameVertexStream(
+    private static bool TryGetMetadataOffset(
         Gen5VertexInputBinding input,
-        MetadataVertexResource resource)
+        MetadataVertexResource resource,
+        out uint offsetBytes)
     {
-        if (input.BaseAddress == resource.SharpBase ||
-            input.BaseAddress == resource.SharpBase + resource.OffsetBytes)
+        offsetBytes = input.OffsetBytes;
+        if (resource.SharpBase < input.BaseAddress ||
+            (!IsAddressInsideCapturedSpan(input, resource.SharpBase) &&
+             resource.SharpBase != input.BaseAddress))
         {
-            return true;
+            return false;
         }
 
-        return IsAddressInsideCapturedSpan(input, resource.SharpBase);
+        var relativeBase = resource.SharpBase - input.BaseAddress;
+        var resolvedOffset = relativeBase + resource.OffsetBytes;
+        if (resolvedOffset > uint.MaxValue)
+        {
+            return false;
+        }
+
+        offsetBytes = (uint)resolvedOffset;
+        return true;
     }
 
     private static bool IsAddressInsideCapturedSpan(
@@ -553,28 +512,6 @@ internal static class AgcVertexMetadata
         address >= input.BaseAddress &&
         address < input.BaseAddress + (ulong)input.DataLength;
 
-    private static bool IsUniqueUnusedOffset(
-        IReadOnlyList<MetadataVertexResource> resources,
-        bool[] usedResources,
-        uint offsetBytes,
-        int candidateIndex)
-    {
-        for (var index = 0; index < resources.Count; index++)
-        {
-            if (index == candidateIndex || usedResources[index])
-            {
-                continue;
-            }
-
-            if (resources[index].OffsetBytes == offsetBytes)
-            {
-                return false;
-            }
-        }
-
-        return true;
-    }
-
     /// <summary>
     /// Attrib-table format
     /// fields are VertexAttribFormat; V# / Vulkan paths need BufferFormat.

src/SharpEmu.Libs/Agc/GpuWaitRegistry.cs

@@ -59,6 +59,10 @@ internal static class GpuWaitRegistry
     // cycle forever even though a real producer did signal it. Keyed by (memory,
     // address) so distinct guest processes never alias.
     private static readonly Dictionary<(object, ulong), ulong> _lastProduced = new();
+    // Frame-staleness guard: tracks the frame ID of each label write so that
+    // WAIT_REG_MEM in frame N+1 is not satisfied by a stale write from frame N.
+    private static readonly Dictionary<(object, ulong), long> _labelFrameIds = new();
+    private static long _currentFrameId;
 
   
     private static object? Canonicalize(object? memory)
@@ -71,6 +75,34 @@ internal static class GpuWaitRegistry
         return memory;
     }
 
+    /// <summary>
+    /// Advances the frame counter. Called at each frame boundary (flip) so that
+    /// stale label writes from previous frames cannot satisfy WAIT_REG_MEM.
+    /// </summary>
+    public static void AdvanceFrame()
+    {
+        System.Threading.Interlocked.Increment(ref _currentFrameId);
+    }
+
+    /// <summary>
+    /// Returns true if the label at (memory, address) was written in the
+    /// current frame, or has never been written (uninitialized).
+    /// Only labels written in a PREVIOUS frame are considered stale.
+    /// </summary>
+    public static bool IsLabelFresh(object memory, ulong address)
+    {
+        memory = Canonicalize(memory)!;
+        lock (_gate)
+        {
+            if (!_labelFrameIds.TryGetValue((memory, address), out var frameId))
+            {
+                return true; // never written — treat as fresh (not stale)
+            }
+
+            return frameId >= System.Threading.Volatile.Read(ref _currentFrameId);
+        }
+    }
+
     public static int Count
     {
         get
@@ -576,6 +608,7 @@ internal static class GpuWaitRegistry
             }
 
             _lastProduced[(memory, address)] = value;
+            _labelFrameIds[(memory, address)] = System.Threading.Volatile.Read(ref _currentFrameId);
         }
 
         return LatchSatisfiedByValue(memory, address, value);

src/SharpEmu.Libs/VideoOut/VulkanVideoPresenter.cs

@@ -5884,6 +5884,7 @@ internal static unsafe class VulkanVideoPresenter
 
         private void ExecuteOrderedGuestFlip(VulkanOrderedGuestFlip work)
         {
+            Agc.AgcExports.MarkAllSurfacesCleared();
             FlushBatchedGuestCommands();
             _guestImages.TryGetValue(work.Address, out var source);
             if (_deviceLost ||
@@ -12770,6 +12771,18 @@ internal static unsafe class VulkanVideoPresenter
                     targets[index].Initialized = false;
                 }
 
+                // CMASK meta-state: if the surface's metadata says "all clear",
+                // start this pass from LoadOp.Clear and consume the state.
+                // CPU-backed targets are skipped (their guest memory contents
+                // are uploaded, not cleared) — same rule the flip-arm used.
+                if (work.Targets[index].Address != 0 &&
+                    !targets[index].IsCpuBacked &&
+                    Agc.AgcExports.IsMetaClearedForSurface(work.Targets[index].Address))
+                {
+                    targets[index].Initialized = false;
+                    Agc.AgcExports.ConsumeMetaClear(work.Targets[index].Address);
+                }
+
                 if (work.Targets[index].Address != 0 &&
                     TakeGuestImageInitialData(work.Targets[index].Address) is { } initialData &&
                     !targets[index].Initialized &&
@@ -13031,13 +13044,31 @@ internal static unsafe class VulkanVideoPresenter
                         &toDepthAttachment);
                 }
 
+                ClearColorValue[]? metaClearValues = null;
+                for (var ci = 0; ci < targets.Length; ci++)
+                {
+                    if (!targets[ci].Initialized &&
+                        work.Targets[ci].Address != 0)
+                    {
+                        var (cw0, cw1) = Agc.AgcExports.GetMetaClearValue(
+                            work.Targets[ci].Address);
+                        if (cw0 != 0 || cw1 != 0)
+                        {
+                            metaClearValues ??= new ClearColorValue[targets.Length];
+                            metaClearValues[ci] = UnpackMetaClearValue(
+                                work.Targets[ci].Format, cw0, cw1);
+                        }
+                    }
+                }
+
                 BeginTranslatedRenderPass(
                     renderPass,
                     framebuffer,
                     extent,
                     colorAttachmentCount: targets.Length,
                     hasDepthAttachment: depth is not null && !clearDepthSeparately,
-                    clearDepth: depth?.ClearDepth ?? 1f);
+                    clearDepth: depth?.ClearDepth ?? 1f,
+                    colorClearValues: metaClearValues);
                 RecordTranslatedDrawInPass(resources, extent);
                 _vk.CmdEndRenderPass(_commandBuffer);
 
@@ -13828,16 +13859,10 @@ internal static unsafe class VulkanVideoPresenter
                     existing.LogicalDepth == depth &&
                     existing.Type == type &&
                     existing.MipLevels == mipLevels &&
+                    (!requiresStorage || existing.SupportsStorageUsage) &&
                     (exactFormatMatch ||
-                     (IsAliasableGuestImageFormat(existing.Format, format) &&
-                      (!requiresStorage || existing.SupportsStorageUsage))))
+                    IsAliasableGuestImageFormat(existing.Format, format)))
                 {
-                    if (requiresStorage && !existing.SupportsStorageUsage)
-                    {
-                        throw new InvalidOperationException(
-                            $"Guest image 0x{target.Address:X16} was created without storage usage.");
-                    }
-
                     existing.IsCpuBacked = false;
                     existing.CpuContentFingerprint = 0;
                     if (existing.RenderPass.Handle == 0 &&
@@ -13870,14 +13895,9 @@ internal static unsafe class VulkanVideoPresenter
                 if (existing.Width == target.Width &&
                     existing.Height == target.Height &&
                     existing.MipLevels == mipLevels &&
+                    (!requiresStorage || existing.SupportsStorageUsage) &&
                     IsCompatibleViewFormat(existing.Format, format))
                 {
-                    if (requiresStorage && !existing.SupportsStorageUsage)
-                    {
-                        throw new InvalidOperationException(
-                            $"Guest image 0x{target.Address:X16} was created without storage usage.");
-                    }
-
                     if (_traceGuestImageEvents)
                     {
                         Console.Error.WriteLine(
@@ -13952,50 +13972,52 @@ internal static unsafe class VulkanVideoPresenter
             {
                 if (requiresStorage && !retained.SupportsStorageUsage)
                 {
-                    throw new InvalidOperationException(
-                        $"Retained guest image 0x{target.Address:X16} was created without storage usage.");
+                    // Do not reuse retained image if it lacks required storage usage
+                    DestroyGuestImage(retained);
                 }
-
-                retained.IsCpuBacked = false;
-                retained.CpuContentFingerprint = 0;
-                _guestImages.Add(target.Address, retained);
-                var retainedByteCount = GetTextureByteCount(
-                    target.Format,
-                    target.Width,
-                    target.Height,
-                    depth);
-                lock (_gate)
+                else
                 {
-                    _cpuBackedUploadGenerations.Remove(target.Address);
-                    _guestImageExtents[target.Address] = (
+                    retained.IsCpuBacked = false;
+                    retained.CpuContentFingerprint = 0;
+                    _guestImages.Add(target.Address, retained);
+                    var retainedByteCount = GetTextureByteCount(
+                        target.Format,
                         target.Width,
                         target.Height,
-                        retainedByteCount);
-                }
+                        depth);
+                    lock (_gate)
+                    {
+                        _cpuBackedUploadGenerations.Remove(target.Address);
+                        _guestImageExtents[target.Address] = (
+                            target.Width,
+                            target.Height,
+                            retainedByteCount);
+                    }
 
-                // Arm the exact extent the flip/acquire sync path would read
-                // back, budgeted by bytes rather than by resolution: the old
-                // 1920x1080 cap left every 4K surface permanently
-                // un-invalidated, so a guest CPU rewrite of one was never
-                // reflected and the sample served stale bytes.
-                if (ShouldTrackGuestImageWrites(retainedByteCount))
-                {
-                    SharpEmu.HLE.GuestImageWriteTracker.Track(
-                        target.Address,
-                        retainedByteCount,
-                        CurrentGuestWorkSequenceForDiagnostics,
-                        "vulkan.render-target");
-                }
+                    // Arm the exact extent the flip/acquire sync path would read
+                    // back, budgeted by bytes rather than by resolution: the old
+                    // 1920x1080 cap left every 4K surface permanently
+                    // un-invalidated, so a guest CPU rewrite of one was never
+                    // reflected and the sample served stale bytes.
+                    if (ShouldTrackGuestImageWrites(retainedByteCount))
+                    {
+                        SharpEmu.HLE.GuestImageWriteTracker.Track(
+                            target.Address,
+                            retainedByteCount,
+                            CurrentGuestWorkSequenceForDiagnostics,
+                            "vulkan.render-target");
+                    }
 
-                if (_traceGuestImageEvents)
-                {
-                    Console.Error.WriteLine(
-                        $"[GIMG] retained addr=0x{target.Address:X} " +
-                        $"{target.Width}x{target.Height} fmt={format} " +
-                        $"initialized={retained.Initialized}");
-                }
+                    if (_traceGuestImageEvents)
+                    {
+                        Console.Error.WriteLine(
+                            $"[GIMG] retained addr=0x{target.Address:X} " +
+                            $"{target.Width}x{target.Height} fmt={format} " +
+                            $"initialized={retained.Initialized}");
+                    }
 
-                return retained;
+                    return retained;
+                }
             }
 
             var imageInfo = new ImageCreateInfo
@@ -17697,20 +17719,67 @@ internal static unsafe class VulkanVideoPresenter
             _vk.CmdEndRenderPass(_commandBuffer);
         }
 
+        /// <summary>
+        /// Decodes the CB CLEAR_WORD0/1 pair into a float RGBA clear value
+        /// according to the surface pixel format.  CLEAR_WORD holds the clear
+        /// colour packed in the surface's native layout, so the two 32-bit
+        /// words must be unpacked channel-by-channel; passing the raw word as
+        /// a single float channel clears to a garbage colour.
+        /// </summary>
+        private static ClearColorValue UnpackMetaClearValue(
+            uint format, uint cw0, uint cw1)
+        {
+            switch (format)
+            {
+                // Gen5 8_8_8_8 (R8G8B8A8): four UNORM bytes packed in WORD0,
+                // little-endian channel order R,G,B,A.
+                case Agc.AgcExports.Gen5TextureFormatR8G8B8A8Unorm:
+                    return new ClearColorValue(
+                        float32_0: ((cw0 >> 0) & 0xFF) / 255f,
+                        float32_1: ((cw0 >> 8) & 0xFF) / 255f,
+                        float32_2: ((cw0 >> 16) & 0xFF) / 255f,
+                        float32_3: ((cw0 >> 24) & 0xFF) / 255f);
+
+                // Gen5 16_16_16_16 float (R16G16B16A16F): R,G as halfs in
+                // WORD0 and B,A as halfs in WORD1.
+                case Agc.AgcExports.Gen5TextureFormatR16G16B16A16Float:
+                    return new ClearColorValue(
+                        float32_0: HalfToFloat((ushort)(cw0 >> 0)),
+                        float32_1: HalfToFloat((ushort)(cw0 >> 16)),
+                        float32_2: HalfToFloat((ushort)(cw1 >> 0)),
+                        float32_3: HalfToFloat((ushort)(cw1 >> 16)));
+
+                default:
+                    // Unknown format: fall back to the common 8_8_8_8 layout.
+                    return new ClearColorValue(
+                        float32_0: ((cw0 >> 0) & 0xFF) / 255f,
+                        float32_1: ((cw0 >> 8) & 0xFF) / 255f,
+                        float32_2: ((cw0 >> 16) & 0xFF) / 255f,
+                        float32_3: ((cw0 >> 24) & 0xFF) / 255f);
+            }
+        }
+
+        private static float HalfToFloat(ushort halfBits) =>
+            (float)BitConverter.UInt16BitsToHalf(halfBits);
+
         private void BeginTranslatedRenderPass(
             RenderPass renderPass,
             Framebuffer framebuffer,
             Extent2D extent,
             int colorAttachmentCount = 1,
             bool hasDepthAttachment = false,
-            float clearDepth = 1f)
+            float clearDepth = 1f,
+            ClearColorValue[]? colorClearValues = null)
         {
             colorAttachmentCount = Math.Max(colorAttachmentCount, 1);
             var clearValueCount = colorAttachmentCount + (hasDepthAttachment ? 1 : 0);
             var clearValues = stackalloc ClearValue[clearValueCount];
             for (var index = 0; index < colorAttachmentCount; index++)
             {
-                clearValues[index] = default;
+                clearValues[index] = colorClearValues is not null &&
+                    index < colorClearValues.Length
+                        ? new ClearValue { Color = colorClearValues[index] }
+                        : default;
             }
             // Reverse-Z is not assumed; clear depth to 1.0 (far) so a standard
             // LessOrEqual/Less test keeps the nearest fragment.

src/SharpEmu.ShaderCompiler.Metal/Gen5MslTranslator.Alu.cs

@@ -144,11 +144,35 @@ public static partial class Gen5MslTranslator
 
                 // ---- float arithmetic ----
                 "VAddF32" => FloatResult(instruction, $"{F(instruction, 0)} + {F(instruction, 1)}"),
+                "VAddF16" => Float16Result(
+                    instruction,
+                    destination,
+                    $"{F16(instruction, 0)} + {F16(instruction, 1)}"),
                 "VSubF32" => FloatResult(instruction, $"{F(instruction, 0)} - {F(instruction, 1)}"),
                 "VSubrevF32" => FloatResult(instruction, $"{F(instruction, 1)} - {F(instruction, 0)}"),
+                "VSubF16" => Float16Result(
+                    instruction,
+                    destination,
+                    $"{F16(instruction, 0)} - {F16(instruction, 1)}"),
+                "VSubrevF16" => Float16Result(
+                    instruction,
+                    destination,
+                    $"{F16(instruction, 1)} - {F16(instruction, 0)}"),
                 "VMulF32" => FloatResult(instruction, $"{F(instruction, 0)} * {F(instruction, 1)}"),
+                "VMulF16" => Float16Result(
+                    instruction,
+                    destination,
+                    $"{F16(instruction, 0)} * {F16(instruction, 1)}"),
                 "VMinF32" => FloatResult(instruction, $"fmin({F(instruction, 0)}, {F(instruction, 1)})"),
                 "VMaxF32" => FloatResult(instruction, $"fmax({F(instruction, 0)}, {F(instruction, 1)})"),
+                "VMinF16" => Float16Result(
+                    instruction,
+                    destination,
+                    $"fmin({F16(instruction, 0)}, {F16(instruction, 1)})"),
+                "VMaxF16" => Float16Result(
+                    instruction,
+                    destination,
+                    $"fmax({F16(instruction, 0)}, {F16(instruction, 1)})"),
                 // The decoder normalizes mk/ak literal placement, so every MAD/FMA
                 // form is fma(src0, src1, src2) exactly like the SPIR-V translator.
                 "VFmaF32" or "VMadF32" or "VMadAkF32" or "VMadMkF32" or "VFmaAkF32" or "VFmaMkF32" =>
@@ -578,23 +602,46 @@ public static partial class Gen5MslTranslator
             {
                 condition = EmitCompareClass(instruction);
             }
-            else if (opcode is "VCmpTruF32" or "VCmpxTruF32" or "VCmpTI32" or "VCmpTU32")
+            else if (opcode is
+                     "VCmpTruF32" or "VCmpxTruF32" or
+                     "VCmpTruF16" or "VCmpxTruF16" or
+                     "VCmpTI32" or "VCmpTU32")
             {
                 condition = "true";
             }
-            else if (opcode is "VCmpFF32" or "VCmpxFF32" or "VCmpFI32" or "VCmpFU32")
+            else if (opcode is
+                     "VCmpFF32" or "VCmpxFF32" or
+                     "VCmpFF16" or "VCmpxFF16" or
+                     "VCmpFI32" or "VCmpFU32")
             {
                 condition = "false";
             }
-            else if (opcode is "VCmpOF32" or "VCmpxOF32")
+            else if (opcode is
+                     "VCmpOF32" or "VCmpxOF32" or
+                     "VCmpOF16" or "VCmpxOF16")
             {
-                condition = $"(!isnan({F(instruction, 0)}) && !isnan({F(instruction, 1)}))";
+                var left = opcode.EndsWith("F16", StringComparison.Ordinal)
+                    ? F16(instruction, 0)
+                    : F(instruction, 0);
+                var right = opcode.EndsWith("F16", StringComparison.Ordinal)
+                    ? F16(instruction, 1)
+                    : F(instruction, 1);
+                condition = $"(!isnan({left}) && !isnan({right}))";
             }
-            else if (opcode is "VCmpUF32" or "VCmpxUF32")
+            else if (opcode is
+                     "VCmpUF32" or "VCmpxUF32" or
+                     "VCmpUF16" or "VCmpxUF16")
             {
-                condition = $"(isnan({F(instruction, 0)}) || isnan({F(instruction, 1)}))";
+                var left = opcode.EndsWith("F16", StringComparison.Ordinal)
+                    ? F16(instruction, 0)
+                    : F(instruction, 0);
+                var right = opcode.EndsWith("F16", StringComparison.Ordinal)
+                    ? F16(instruction, 1)
+                    : F(instruction, 1);
+                condition = $"(isnan({left}) || isnan({right}))";
             }
-            else if (opcode.EndsWith("F32", StringComparison.Ordinal))
+            else if (opcode.EndsWith("F32", StringComparison.Ordinal) ||
+                     opcode.EndsWith("F16", StringComparison.Ordinal))
             {
                 // Ordered compares are the plain C operators (false on NaN);
                 // the Nxx forms are their unordered negations (true on NaN).
@@ -620,7 +667,13 @@ public static partial class Gen5MslTranslator
                     return false;
                 }
 
-                var comparison = $"({F(instruction, 0)} {op} {F(instruction, 1)})";
+                var left = opcode.EndsWith("F16", StringComparison.Ordinal)
+                    ? F16(instruction, 0)
+                    : F(instruction, 0);
+                var right = opcode.EndsWith("F16", StringComparison.Ordinal)
+                    ? F16(instruction, 1)
+                    : F(instruction, 1);
+                var comparison = $"({left} {op} {right})";
                 condition = unordered ? $"(!{comparison})" : comparison;
             }
             else
@@ -1570,6 +1623,80 @@ public static partial class Gen5MslTranslator
             return expression;
         }
 
+        /// <summary>Reads the selected 16-bit half as a widened float.</summary>
+        private string F16(Gen5ShaderInstruction instruction, int sourceIndex)
+        {
+            var operand = instruction.Sources[sourceIndex];
+            string expression;
+            if (operand.Kind == Gen5OperandKind.EncodedConstant &&
+                Gen5InlineConstants.TryDecode(operand.Value, out var inline))
+            {
+                expression = operand.Value switch
+                {
+                    >= 128 and <= 192 => $"{operand.Value - 128}.0f",
+                    >= 193 and <= 208 => $"(-{operand.Value - 192}.0f)",
+                    _ => AsFloat(FormatUInt(inline)),
+                };
+            }
+            else
+            {
+                var raw = RawSource(
+                    instruction,
+                    sourceIndex,
+                    applySdwaIntegerModifiers: false);
+                var shift = instruction.Control is Gen5Vop3Control control &&
+                    (control.OperandSelect & (1u << sourceIndex)) != 0
+                        ? 16
+                        : 0;
+                expression =
+                    $"(float)as_type<half>((ushort)((({raw}) >> {shift}) & 0xFFFFu))";
+            }
+
+            var (absoluteMask, negateMask) = instruction.Control switch
+            {
+                Gen5Vop3Control control => (control.AbsoluteMask, control.NegateMask),
+                Gen5SdwaControl control => (control.AbsoluteMask, control.NegateMask),
+                Gen5DppControl control => (control.AbsoluteMask, control.NegateMask),
+                _ => (0u, 0u),
+            };
+            if ((absoluteMask & (1u << sourceIndex)) != 0)
+            {
+                expression = $"fabs({expression})";
+            }
+
+            if ((negateMask & (1u << sourceIndex)) != 0)
+            {
+                expression = $"(-{expression})";
+            }
+
+            return expression;
+        }
+
+        /// <summary>Rounds to f16 and preserves the unselected VGPR half.</summary>
+        private string Float16Result(
+            Gen5ShaderInstruction instruction,
+            uint destination,
+            string expression)
+        {
+            var control = instruction.Control as Gen5Vop3Control;
+            expression = (control?.OutputModifier ?? 0) switch
+            {
+                1 => $"(({expression}) * 2.0f)",
+                2 => $"(({expression}) * 4.0f)",
+                3 => $"(({expression}) * 0.5f)",
+                _ => expression,
+            };
+            if (control?.Clamp == true)
+            {
+                expression = $"clamp({expression}, 0.0f, 1.0f)";
+            }
+
+            var packed = $"(uint)as_type<ushort>(half({expression}))";
+            return ((control?.OperandSelect ?? 0) & 8) != 0
+                ? $"((v[{destination}] & 0x0000FFFFu) | (({packed}) << 16))"
+                : $"((v[{destination}] & 0xFFFF0000u) | ({packed}))";
+        }
+
         /// <summary>
         /// Wraps a float expression with VOP3/SDWA output modifiers and clamp,
         /// then bitcasts back to the register file's uint domain.

src/SharpEmu.ShaderCompiler.Vulkan/Gen5SpirvTranslator.Alu.cs

@@ -340,21 +340,43 @@ public static partial class Gen5SpirvTranslator
                 case "VAddF32":
                     result = EmitFloatBinary(instruction, SpirvOp.FAdd);
                     break;
+                case "VAddF16":
+                    result = EmitFloat16Binary(instruction, destination, SpirvOp.FAdd);
+                    break;
                 case "VSubF32":
                     result = EmitFloatBinary(instruction, SpirvOp.FSub);
                     break;
                 case "VSubrevF32":
                     result = EmitFloatBinary(instruction, SpirvOp.FSub, reverse: true);
                     break;
+                case "VSubF16":
+                    result = EmitFloat16Binary(instruction, destination, SpirvOp.FSub);
+                    break;
+                case "VSubrevF16":
+                    result = EmitFloat16Binary(
+                        instruction,
+                        destination,
+                        SpirvOp.FSub,
+                        reverse: true);
+                    break;
                 case "VMulF32":
                     result = EmitFloatBinary(instruction, SpirvOp.FMul);
                     break;
+                case "VMulF16":
+                    result = EmitFloat16Binary(instruction, destination, SpirvOp.FMul);
+                    break;
                 case "VMinF32":
                     result = EmitFloatExtBinary(instruction, 37);
                     break;
                 case "VMaxF32":
                     result = EmitFloatExtBinary(instruction, 40);
                     break;
+                case "VMinF16":
+                    result = EmitFloat16ExtBinary(instruction, destination, 37);
+                    break;
+                case "VMaxF16":
+                    result = EmitFloat16ExtBinary(instruction, destination, 40);
+                    break;
                 case "VMadF32":
                 case "VFmaF32":
                 case "VMadMkF32":
@@ -1609,29 +1631,72 @@ public static partial class Gen5SpirvTranslator
                     condition,
                     SignedClass(0x020, 0x040, zero));
             }
-            else if (opcode is "VCmpFF32" or "VCmpxFF32" or "VCmpFI32" or "VCmpFU32")
+            else if (opcode is
+                     "VCmpFF32" or "VCmpxFF32" or
+                     "VCmpFF16" or "VCmpxFF16" or
+                     "VCmpFI32" or "VCmpFU32")
             {
                 condition = _module.ConstantBool(false);
             }
-            else if (opcode is "VCmpTruF32" or "VCmpxTruF32" or "VCmpTI32" or "VCmpTU32")
+            else if (opcode is
+                     "VCmpTruF32" or "VCmpxTruF32" or
+                     "VCmpTruF16" or "VCmpxTruF16" or
+                     "VCmpTI32" or "VCmpTU32")
             {
                 condition = _module.ConstantBool(true);
             }
             else if (opcode is
                      "VCmpOF32" or "VCmpxOF32" or
-                     "VCmpUF32" or "VCmpxUF32")
+                     "VCmpUF32" or "VCmpxUF32" or
+                     "VCmpOF16" or "VCmpxOF16" or
+                     "VCmpUF16" or "VCmpxUF16")
             {
-                var left = GetFloatSource(instruction, 0);
-                var right = GetFloatSource(instruction, 1);
+                var isHalf = opcode.EndsWith("F16", StringComparison.Ordinal);
+                var left = isHalf
+                    ? GetFloat16Source(instruction, 0)
+                    : GetFloatSource(instruction, 0);
+                var right = isHalf
+                    ? GetFloat16Source(instruction, 1)
+                    : GetFloatSource(instruction, 1);
                 var unordered = _module.AddInstruction(
                     SpirvOp.LogicalOr,
                     _boolType,
                     _module.AddInstruction(SpirvOp.IsNan, _boolType, left),
                     _module.AddInstruction(SpirvOp.IsNan, _boolType, right));
-                condition = opcode is "VCmpUF32" or "VCmpxUF32"
+                condition = opcode is
+                    "VCmpUF32" or "VCmpxUF32" or
+                    "VCmpUF16" or "VCmpxUF16"
                     ? unordered
                     : _module.AddInstruction(SpirvOp.LogicalNot, _boolType, unordered);
             }
+            else if (opcode.EndsWith("F16", StringComparison.Ordinal))
+            {
+                var left = GetFloat16Source(instruction, 0);
+                var right = GetFloat16Source(instruction, 1);
+                var operation = opcode switch
+                {
+                    "VCmpLtF16" or "VCmpxLtF16" => SpirvOp.FOrdLessThan,
+                    "VCmpEqF16" or "VCmpxEqF16" => SpirvOp.FOrdEqual,
+                    "VCmpLeF16" or "VCmpxLeF16" => SpirvOp.FOrdLessThanEqual,
+                    "VCmpGtF16" or "VCmpxGtF16" => SpirvOp.FOrdGreaterThan,
+                    "VCmpLgF16" or "VCmpxLgF16" => SpirvOp.FOrdNotEqual,
+                    "VCmpGeF16" or "VCmpxGeF16" => SpirvOp.FOrdGreaterThanEqual,
+                    "VCmpNeqF16" or "VCmpxNeqF16" => SpirvOp.FUnordNotEqual,
+                    "VCmpNltF16" or "VCmpxNltF16" => SpirvOp.FUnordGreaterThanEqual,
+                    "VCmpNleF16" or "VCmpxNleF16" => SpirvOp.FUnordGreaterThan,
+                    "VCmpNgtF16" or "VCmpxNgtF16" => SpirvOp.FUnordLessThanEqual,
+                    "VCmpNgeF16" or "VCmpxNgeF16" => SpirvOp.FUnordLessThan,
+                    "VCmpNlgF16" or "VCmpxNlgF16" => SpirvOp.FUnordEqual,
+                    _ => SpirvOp.Nop,
+                };
+                if (operation == SpirvOp.Nop)
+                {
+                    error = $"unsupported half compare {opcode}";
+                    return false;
+                }
+
+                condition = _module.AddInstruction(operation, _boolType, left, right);
+            }
             else if (opcode is not ("VCmpClassF32" or "VCmpxClassF32") &&
                      opcode.EndsWith("F32", StringComparison.Ordinal))
             {
@@ -3108,6 +3173,70 @@ public static partial class Gen5SpirvTranslator
                     sourceAllowsWrite));
         }
 
+        private uint GetFloat16Source(
+            Gen5ShaderInstruction instruction,
+            int sourceIndex)
+        {
+            var operand = instruction.Sources[sourceIndex];
+            uint value;
+            if (operand.Kind == Gen5OperandKind.EncodedConstant &&
+                operand.Value is >= 128 and <= 192)
+            {
+                value = Float(operand.Value - 128);
+            }
+            else if (operand.Kind == Gen5OperandKind.EncodedConstant &&
+                     operand.Value is >= 193 and <= 208)
+            {
+                value = Float(-(operand.Value - 192));
+            }
+            else if (operand.Kind == Gen5OperandKind.EncodedConstant &&
+                     Gen5InlineConstants.TryDecode(operand.Value, out var inline))
+            {
+                value = Bitcast(_floatType, UInt(inline));
+            }
+            else
+            {
+                var raw = GetRawSource(
+                    instruction,
+                    sourceIndex,
+                    applySdwaIntegerModifiers: false);
+                if (instruction.Control is Gen5Vop3Control control &&
+                    (control.OperandSelect & (1u << sourceIndex)) != 0)
+                {
+                    raw = ShiftRightLogical(raw, UInt(16));
+                }
+
+                value = Bitcast(_floatType, EmitHalfToFloat(raw));
+            }
+
+            uint absoluteMask = 0;
+            uint negateMask = 0;
+            switch (instruction.Control)
+            {
+                case Gen5Vop3Control control:
+                    absoluteMask = control.AbsoluteMask;
+                    negateMask = control.NegateMask;
+                    break;
+                case Gen5SdwaControl control:
+                    absoluteMask = control.AbsoluteMask;
+                    negateMask = control.NegateMask;
+                    break;
+                case Gen5DppControl control:
+                    absoluteMask = control.AbsoluteMask;
+                    negateMask = control.NegateMask;
+                    break;
+            }
+
+            if ((absoluteMask & (1u << sourceIndex)) != 0)
+            {
+                value = Ext(4, _floatType, value);
+            }
+
+            return (negateMask & (1u << sourceIndex)) != 0
+                ? _module.AddInstruction(SpirvOp.FNegate, _floatType, value)
+                : value;
+        }
+
         private uint GetFloatSource(
             Gen5ShaderInstruction instruction,
             int sourceIndex)
@@ -3232,6 +3361,33 @@ public static partial class Gen5SpirvTranslator
                 _module.AddInstruction(SpirvOp.UConvert, _uintType, high));
         }
 
+        private uint EmitFloat16Binary(
+            Gen5ShaderInstruction instruction,
+            uint destination,
+            SpirvOp operation,
+            bool reverse = false)
+        {
+            var left = GetFloat16Source(instruction, reverse ? 1 : 0);
+            var right = GetFloat16Source(instruction, reverse ? 0 : 1);
+            return EmitFloat16Result(
+                instruction,
+                destination,
+                _module.AddInstruction(operation, _floatType, left, right));
+        }
+
+        private uint EmitFloat16ExtBinary(
+            Gen5ShaderInstruction instruction,
+            uint destination,
+            uint operation) =>
+            EmitFloat16Result(
+                instruction,
+                destination,
+                Ext(
+                    operation,
+                    _floatType,
+                    GetFloat16Source(instruction, 0),
+                    GetFloat16Source(instruction, 1)));
+
         private uint EmitFloatBinary(
             Gen5ShaderInstruction instruction,
             SpirvOp operation,
@@ -3753,6 +3909,35 @@ public static partial class Gen5SpirvTranslator
                 UInt(0));
         }
 
+        private uint EmitFloat16Result(
+            Gen5ShaderInstruction instruction,
+            uint destination,
+            uint value)
+        {
+            var control = instruction.Control as Gen5Vop3Control;
+            value = (control?.OutputModifier ?? 0) switch
+            {
+                1 => _module.AddInstruction(SpirvOp.FMul, _floatType, value, Float(2)),
+                2 => _module.AddInstruction(SpirvOp.FMul, _floatType, value, Float(4)),
+                3 => _module.AddInstruction(SpirvOp.FMul, _floatType, value, Float(0.5f)),
+                _ => value,
+            };
+            if (control?.Clamp == true)
+            {
+                value = Ext(43, _floatType, value, Float(0), Float(1));
+            }
+
+            var half = EmitFloatToHalf(Bitcast(_uintType, value));
+            var current = LoadV(destination);
+            return ((control?.OperandSelect ?? 0) & 8) != 0
+                ? BitwiseOr(
+                    BitwiseAnd(current, UInt(0x0000_FFFF)),
+                    ShiftLeftLogical(half, UInt(16)))
+                : BitwiseOr(
+                    BitwiseAnd(current, UInt(0xFFFF_0000)),
+                    half);
+        }
+
         private uint EmitFloatResult(
             Gen5ShaderInstruction instruction,
             uint value)

src/SharpEmu.ShaderCompiler.Vulkan/Gen5SpirvTranslator.cs

@@ -2366,17 +2366,16 @@ public static partial class Gen5SpirvTranslator
                         return;
                     }
 
+                    // GLOBAL_STORE/LOAD_DWORD(x2/x3/x4) are dword-aligned by the GCN ISA, so read/write dwords directly instead of the per-byte loop.
                     for (uint index = 0; index < control.DwordCount; index++)
                     {
-                        var address = index == 0
-                            ? byteAddress
-                            : IAdd(byteAddress, UInt(index * sizeof(uint)));
-                        StoreBufferBytes(
+                        var indexedDwordAddress = index == 0
+                            ? dwordAddress
+                            : IAdd(dwordAddress, UInt(index));
+                        StoreBufferWord(
                             bindingIndex,
-                            address,
-                            LoadV(control.VectorData + index),
-                            sizeof(uint),
-                            0);
+                            indexedDwordAddress,
+                            LoadV(control.VectorData + index));
                     }
                 });
                 return true;
@@ -2404,12 +2403,12 @@ public static partial class Gen5SpirvTranslator
 
             for (uint index = 0; index < control.DwordCount; index++)
             {
-                var address = index == 0
-                    ? byteAddress
-                    : IAdd(byteAddress, UInt(index * sizeof(uint)));
+                var indexedDwordAddress = index == 0
+                    ? dwordAddress
+                    : IAdd(dwordAddress, UInt(index));
                 StoreV(
                     control.VectorData + index,
-                    LoadUnalignedBufferWord(bindingIndex, address));
+                    LoadBufferWord(bindingIndex, indexedDwordAddress));
             }
 
             return true;
@@ -2510,17 +2509,16 @@ public static partial class Gen5SpirvTranslator
                         return;
                     }
 
+                    // BUFFER_STORE/LOAD_DWORD(x2/x3/x4) are dword-aligned by the GCN ISA, same as the GLOBAL case above — no per-byte reassembly needed.
                     for (uint index = 0; index < control.DwordCount; index++)
                     {
-                        var address = index == 0
-                            ? byteAddress
-                            : IAdd(byteAddress, UInt(index * sizeof(uint)));
-                        StoreBufferBytes(
+                        var indexedDwordAddress = index == 0
+                            ? dwordAddress
+                            : IAdd(dwordAddress, UInt(index));
+                        StoreBufferWord(
                             bindingIndex,
-                            address,
-                            LoadV(control.VectorData + index),
-                            sizeof(uint),
-                            0);
+                            indexedDwordAddress,
+                            LoadV(control.VectorData + index));
                     }
                 });
 
@@ -2576,12 +2574,12 @@ public static partial class Gen5SpirvTranslator
 
             for (uint index = 0; index < control.DwordCount; index++)
             {
-                var address = index == 0
-                    ? byteAddress
-                    : IAdd(byteAddress, UInt(index * sizeof(uint)));
+                var indexedDwordAddress = index == 0
+                    ? dwordAddress
+                    : IAdd(dwordAddress, UInt(index));
                 StoreV(
                     control.VectorData + index,
-                    LoadUnalignedBufferWord(bindingIndex, address));
+                    LoadBufferWord(bindingIndex, indexedDwordAddress));
             }
 
             return true;

src/SharpEmu.ShaderCompiler/Gen5ShaderTranslator.cs

@@ -1015,6 +1015,12 @@ public static class Gen5ShaderTranslator
             0x2F => "VCvtPkrtzF16F32",
             0x30 => "VCvtPkU16U32",
             0x31 => "VCvtPkI16I32",
+            0x32 => "VAddF16",
+            0x33 => "VSubF16",
+            0x34 => "VSubrevF16",
+            0x35 => "VMulF16",
+            0x39 => "VMaxF16",
+            0x3A => "VMinF16",
             _ => string.Empty,
         };
 
@@ -1086,6 +1092,14 @@ public static class Gen5ShaderTranslator
             0xC5 => "VCmpNeU32",
             0xC6 => "VCmpGeU32",
             0xC7 => "VCmpTU32",
+            0xC8 => "VCmpFF16",
+            0xC9 => "VCmpLtF16",
+            0xCA => "VCmpEqF16",
+            0xCB => "VCmpLeF16",
+            0xCC => "VCmpGtF16",
+            0xCD => "VCmpLgF16",
+            0xCE => "VCmpGeF16",
+            0xCF => "VCmpOF16",
             0xD0 => "VCmpxFU32",
             0xD1 => "VCmpxLtU32",
             0xD2 => "VCmpxEqU32",
@@ -1094,6 +1108,30 @@ public static class Gen5ShaderTranslator
             0xD5 => "VCmpxNeU32",
             0xD6 => "VCmpxGeU32",
             0xD7 => "VCmpxTU32",
+            0xD8 => "VCmpxFF16",
+            0xD9 => "VCmpxLtF16",
+            0xDA => "VCmpxEqF16",
+            0xDB => "VCmpxLeF16",
+            0xDC => "VCmpxGtF16",
+            0xDD => "VCmpxLgF16",
+            0xDE => "VCmpxGeF16",
+            0xDF => "VCmpxOF16",
+            0xE8 => "VCmpUF16",
+            0xE9 => "VCmpNgeF16",
+            0xEA => "VCmpNlgF16",
+            0xEB => "VCmpNgtF16",
+            0xEC => "VCmpNleF16",
+            0xED => "VCmpNeqF16",
+            0xEE => "VCmpNltF16",
+            0xEF => "VCmpTruF16",
+            0xF8 => "VCmpxUF16",
+            0xF9 => "VCmpxNgeF16",
+            0xFA => "VCmpxNlgF16",
+            0xFB => "VCmpxNgtF16",
+            0xFC => "VCmpxNleF16",
+            0xFD => "VCmpxNeqF16",
+            0xFE => "VCmpxNltF16",
+            0xFF => "VCmpxTruF16",
             _ => string.Empty,
         };
 

src/SharpEmu.SourceGenerators/SysAbiExportAnalyzer.cs

@@ -102,17 +102,16 @@ public sealed class SysAbiExportAnalyzer : DiagnosticAnalyzer
         ConcurrentDictionary<string, IMethodSymbol> exportsByNid)
     {
         var method = (IMethodSymbol)context.Symbol;
-        AttributeData? exportAttribute = null;
+        var exportAttributes = ImmutableArray.CreateBuilder<AttributeData>();
         foreach (var attribute in method.GetAttributes())
         {
             if (SysAbiExportShape.IsSysAbiExportAttribute(attribute.AttributeClass))
             {
-                exportAttribute = attribute;
-                break;
+                exportAttributes.Add(attribute);
             }
         }
 
-        if (exportAttribute is null)
+        if (exportAttributes.Count == 0)
         {
             return;
         }
@@ -135,6 +134,28 @@ public sealed class SysAbiExportAnalyzer : DiagnosticAnalyzer
                 SysAbiDiagnostics.HandlerNotAccessible, location, methodDisplay));
         }
 
+        foreach (var exportAttribute in exportAttributes)
+        {
+            AnalyzeExportAttribute(
+                context,
+                catalogNames,
+                exportsByNid,
+                method,
+                exportAttribute,
+                location,
+                methodDisplay);
+        }
+    }
+
+    private static void AnalyzeExportAttribute(
+        SymbolAnalysisContext context,
+        HashSet<string>? catalogNames,
+        ConcurrentDictionary<string, IMethodSymbol> exportsByNid,
+        IMethodSymbol method,
+        AttributeData exportAttribute,
+        Location location,
+        string methodDisplay)
+    {
         var arguments = SysAbiExportShape.ReadArguments(exportAttribute);
         var hasNid = !string.IsNullOrWhiteSpace(arguments.Nid);
         var hasName = !string.IsNullOrWhiteSpace(arguments.ExportName);
@@ -188,9 +209,9 @@ public sealed class SysAbiExportAnalyzer : DiagnosticAnalyzer
             }
         }
 
-        var existing = exportsByNid.GetOrAdd(effectiveNid, method);
-        if (!SymbolEqualityComparer.Default.Equals(existing, method))
+        if (!exportsByNid.TryAdd(effectiveNid, method))
         {
+            var existing = exportsByNid[effectiveNid];
             context.ReportDiagnostic(Diagnostic.Create(
                 SysAbiDiagnostics.DuplicateNid,
                 location,

src/SharpEmu.SourceGenerators/SysAbiExportGenerator.cs

@@ -27,7 +27,16 @@ public sealed class SysAbiExportGenerator : IIncrementalGenerator
 
     private sealed class ExportModel : IEquatable<ExportModel>
     {
-        public ExportModel(string containingType, string methodName, SysAbiExportShape.HandlerShape shape, string typedParameterKinds, string libraryName, string nid, string exportName, int target)
+        public ExportModel(
+            string containingType,
+            string methodName,
+            SysAbiExportShape.HandlerShape shape,
+            string typedParameterKinds,
+            string libraryName,
+            string nid,
+            string exportName,
+            int target,
+            bool preferLle)
         {
             ContainingType = containingType;
             MethodName = methodName;
@@ -37,6 +46,7 @@ public sealed class SysAbiExportGenerator : IIncrementalGenerator
             Nid = nid;
             ExportName = exportName;
             Target = target;
+            PreferLle = preferLle;
         }
 
         public string ContainingType { get; }
@@ -51,6 +61,7 @@ public sealed class SysAbiExportGenerator : IIncrementalGenerator
         public string Nid { get; }
         public string ExportName { get; }
         public int Target { get; }
+        public bool PreferLle { get; }
 
         public bool Equals(ExportModel? other) =>
             other is not null &&
@@ -61,7 +72,8 @@ public sealed class SysAbiExportGenerator : IIncrementalGenerator
             LibraryName == other.LibraryName &&
             Nid == other.Nid &&
             ExportName == other.ExportName &&
-            Target == other.Target;
+            Target == other.Target &&
+            PreferLle == other.PreferLle;
 
         public override bool Equals(object? obj) => Equals(obj as ExportModel);
 
@@ -73,6 +85,7 @@ public sealed class SysAbiExportGenerator : IIncrementalGenerator
                 hash = (hash * 31) + ContainingType.GetHashCode();
                 hash = (hash * 31) + MethodName.GetHashCode();
                 hash = (hash * 31) + Nid.GetHashCode();
+                hash = (hash * 31) + PreferLle.GetHashCode();
                 return hash;
             }
         }
@@ -80,80 +93,90 @@ public sealed class SysAbiExportGenerator : IIncrementalGenerator
 
     public void Initialize(IncrementalGeneratorInitializationContext context)
     {
-        var exports = context.SyntaxProvider
+        var exportGroups = context.SyntaxProvider
             .ForAttributeWithMetadataName(
                 AttributeMetadataName,
                 static (node, _) => node is MethodDeclarationSyntax,
-                static (attributeContext, _) => CreateModel(attributeContext))
-            .Where(static model => model is not null)
+                static (attributeContext, _) => CreateModels(attributeContext))
+            .Where(static models => !models.IsDefaultOrEmpty)
             .Collect();
 
         var assemblyName = context.CompilationProvider
             .Select(static (compilation, _) => compilation.AssemblyName ?? "Assembly");
 
         context.RegisterSourceOutput(
-            exports.Combine(assemblyName),
+            exportGroups.Combine(assemblyName),
             static (productionContext, source) => Emit(productionContext, source.Left!, source.Right));
     }
 
-    private static ExportModel? CreateModel(GeneratorAttributeSyntaxContext context)
+    private static ImmutableArray<ExportModel> CreateModels(GeneratorAttributeSyntaxContext context)
     {
         if (context.TargetSymbol is not IMethodSymbol method ||
             !SysAbiExportShape.IsAccessibleFromGeneratedCode(method))
         {
-            return null;
+            return ImmutableArray<ExportModel>.Empty;
         }
 
         var shape = SysAbiExportShape.Classify(method, out var typedParameterKinds);
         if (shape == SysAbiExportShape.HandlerShape.Invalid)
         {
-            return null;
+            return ImmutableArray<ExportModel>.Empty;
         }
 
-        var attribute = context.Attributes[0];
-        var arguments = SysAbiExportShape.ReadArguments(attribute);
-        var nid = arguments.Nid;
-        var exportName = arguments.ExportName;
-
-        // Mirror ModuleManager.ResolveExportInfo: a missing NID resolves from the export
-        // name (algorithmically — equivalent to the runtime catalog lookup, which was
-        // built with the same computation); a missing name falls back to the method name.
-        if (string.IsNullOrWhiteSpace(nid) && !string.IsNullOrWhiteSpace(exportName))
+        var models = ImmutableArray.CreateBuilder<ExportModel>(context.Attributes.Length);
+        foreach (var attribute in context.Attributes)
         {
-            nid = Ps5Nid.Compute(exportName);
+            var arguments = SysAbiExportShape.ReadArguments(attribute);
+            var nid = arguments.Nid;
+            var exportName = arguments.ExportName;
+
+            // Mirror ModuleManager.ResolveExportInfo: a missing NID resolves from the
+            // export name. A missing name falls back to the method name.
+            if (string.IsNullOrWhiteSpace(nid) && !string.IsNullOrWhiteSpace(exportName))
+            {
+                nid = Ps5Nid.Compute(exportName);
+            }
+
+            if (string.IsNullOrWhiteSpace(nid))
+            {
+                continue;
+            }
+
+            if (string.IsNullOrWhiteSpace(exportName))
+            {
+                exportName = method.Name;
+            }
+
+            var libraryName = string.IsNullOrWhiteSpace(arguments.LibraryName) ? "libKernel" : arguments.LibraryName;
+            models.Add(new ExportModel(
+                method.ContainingType.ToDisplayString(SymbolDisplayFormat.FullyQualifiedFormat),
+                method.Name,
+                shape,
+                typedParameterKinds,
+                libraryName,
+                nid!,
+                exportName!,
+                arguments.Target,
+                arguments.PreferLle));
         }
 
-        if (string.IsNullOrWhiteSpace(nid))
-        {
-            return null;
-        }
-
-        if (string.IsNullOrWhiteSpace(exportName))
-        {
-            exportName = method.Name;
-        }
-
-        var libraryName = string.IsNullOrWhiteSpace(arguments.LibraryName) ? "libKernel" : arguments.LibraryName;
-        return new ExportModel(
-            method.ContainingType.ToDisplayString(SymbolDisplayFormat.FullyQualifiedFormat),
-            method.Name,
-            shape,
-            typedParameterKinds,
-            libraryName,
-            nid!,
-            exportName!,
-            arguments.Target);
+        return models.ToImmutable();
     }
 
     private static void Emit(
         SourceProductionContext context,
-        ImmutableArray<ExportModel?> exports,
+        ImmutableArray<ImmutableArray<ExportModel>> exportGroups,
         string assemblyName)
     {
         // No exports, no registry: an assembly that merely references the analyzer
         // (e.g. SharpEmu.HLE itself) must not mint a colliding
         // SharpEmu.Generated.SysAbiExportRegistry type.
-        if (exports.IsDefaultOrEmpty)
+        var exportCount = 0;
+        foreach (var group in exportGroups)
+        {
+            exportCount += group.Length;
+        }
+        if (exportCount == 0)
         {
             return;
         }
@@ -175,24 +198,23 @@ public sealed class SysAbiExportGenerator : IIncrementalGenerator
         builder.AppendLine("    public static global::System.Collections.Generic.IReadOnlyList<global::SharpEmu.HLE.ExportedFunction> CreateExports(");
         builder.AppendLine("        global::SharpEmu.HLE.Generation registrationGeneration)");
         builder.AppendLine("    {");
-        builder.AppendLine($"        var exports = new global::System.Collections.Generic.List<global::SharpEmu.HLE.ExportedFunction>({exports.Length});");
+        builder.AppendLine($"        var exports = new global::System.Collections.Generic.List<global::SharpEmu.HLE.ExportedFunction>({exportCount});");
 
-        foreach (var export in exports)
+        foreach (var group in exportGroups)
         {
-            if (export is null)
+            foreach (var export in group)
             {
-                continue;
+                var function = export.Shape switch
+                {
+                    SysAbiExportShape.HandlerShape.ContextOnly => $"{export.ContainingType}.{export.MethodName}",
+                    SysAbiExportShape.HandlerShape.Parameterless => $"static _ => {export.ContainingType}.{export.MethodName}()",
+                    _ => TypedThunk(export),
+                };
+                builder.AppendLine(
+                    $"        Add(exports, registrationGeneration, {Literal(export.LibraryName)}, {Literal(export.Nid)}, " +
+                    $"{Literal(export.ExportName)}, (global::SharpEmu.HLE.Generation){export.Target}, " +
+                    $"{(export.PreferLle ? "true" : "false")}, {function});");
             }
-
-            var function = export.Shape switch
-            {
-                SysAbiExportShape.HandlerShape.ContextOnly => $"{export.ContainingType}.{export.MethodName}",
-                SysAbiExportShape.HandlerShape.Parameterless => $"static _ => {export.ContainingType}.{export.MethodName}()",
-                _ => TypedThunk(export),
-            };
-            builder.AppendLine(
-                $"        Add(exports, registrationGeneration, {Literal(export.LibraryName)}, {Literal(export.Nid)}, " +
-                $"{Literal(export.ExportName)}, (global::SharpEmu.HLE.Generation){export.Target}, {function});");
         }
 
         builder.AppendLine("        return exports;");
@@ -205,6 +227,7 @@ public sealed class SysAbiExportGenerator : IIncrementalGenerator
         builder.AppendLine("        string nid,");
         builder.AppendLine("        string exportName,");
         builder.AppendLine("        global::SharpEmu.HLE.Generation attributeTarget,");
+        builder.AppendLine("        bool preferLle,");
         builder.AppendLine("        global::SharpEmu.HLE.SysAbiFunction function)");
         builder.AppendLine("    {");
         builder.AppendLine("        var target = attributeTarget == global::SharpEmu.HLE.Generation.None ? registrationGeneration : attributeTarget;");
@@ -213,7 +236,7 @@ public sealed class SysAbiExportGenerator : IIncrementalGenerator
         builder.AppendLine("            return;");
         builder.AppendLine("        }");
         builder.AppendLine();
-        builder.AppendLine("        exports.Add(new global::SharpEmu.HLE.ExportedFunction(libraryName, nid, exportName, target, function));");
+        builder.AppendLine("        exports.Add(new global::SharpEmu.HLE.ExportedFunction(libraryName, nid, exportName, target, function, preferLle));");
         builder.AppendLine("    }");
         builder.AppendLine("}");
 

src/SharpEmu.SourceGenerators/SysAbiExportShape.cs

@@ -18,18 +18,20 @@ public static class SysAbiExportShape
 
     public readonly struct Arguments
     {
-        public Arguments(string libraryName, string nid, string exportName, int target)
+        public Arguments(string libraryName, string nid, string exportName, int target, bool preferLle)
         {
             LibraryName = libraryName;
             Nid = nid;
             ExportName = exportName;
             Target = target;
+            PreferLle = preferLle;
         }
 
         public string LibraryName { get; }
         public string Nid { get; }
         public string ExportName { get; }
         public int Target { get; }
+        public bool PreferLle { get; }
     }
 
     /// <summary>
@@ -205,6 +207,7 @@ public static class SysAbiExportShape
         var nid = string.Empty;
         var exportName = string.Empty;
         var target = 0;
+        var preferLle = false;
         foreach (var argument in attribute.NamedArguments)
         {
             switch (argument.Key)
@@ -221,9 +224,12 @@ public static class SysAbiExportShape
                 case "Target":
                     target = argument.Value.Value is int value ? value : 0;
                     break;
+                case "PreferLle":
+                    preferLle = argument.Value.Value is bool boolValue && boolValue;
+                    break;
             }
         }
 
-        return new Arguments(libraryName, nid, exportName, target);
+        return new Arguments(libraryName, nid, exportName, target, preferLle);
     }
 }

tests/SharpEmu.Libs.Tests/Agc/AgcVertexMetadataTests.cs

@@ -74,7 +74,7 @@ public sealed class AgcVertexMetadataTests
     }
 
     [Fact]
-    public void MergeVertexInputs_OverlaysFormatWithoutRebasingCapture()
+    public void MergeVertexInputs_OverlaysLayoutWithoutRebasingCapture()
     {
         const ulong memoryBase = 0x1_0000_0000;
         var memory = new FakeCpuMemory(memoryBase, 0x2000);
@@ -114,7 +114,7 @@ public sealed class AgcVertexMetadataTests
                 NumberFormat: 7,
                 BaseAddress: sharpBase,
                 Stride: 16,
-                OffsetBytes: 0,
+                OffsetBytes: 12,
                 Data: data,
                 DataLength: data.Length,
                 DataPooled: false),
@@ -135,6 +135,137 @@ public sealed class AgcVertexMetadataTests
         Assert.Equal(0x40u, merged[0].Pc);
     }
 
+    [Fact]
+    public void MergeVertexInputs_MetadataCorrectsStaleStride40()
+    {
+        const ulong memoryBase = 0x1_0000_0000;
+        var memory = new FakeCpuMemory(memoryBase, 0x2000);
+        var ctx = new CpuContext(memory, Generation.Gen5);
+
+        const ulong semanticsAddress = memoryBase + 0x100;
+        const ulong attribTable = memoryBase + 0x200;
+        const ulong bufferTable = memoryBase + 0x300;
+        const ulong sharpBase = memoryBase + 0x800;
+
+        WriteUInt32(memory, semanticsAddress, 0u | (0u << 8) | (4u << 16));
+        WriteUInt32(memory, attribTable, 0u | (56u << 5) | (12u << 14));
+        WriteUInt32(memory, bufferTable, (uint)(sharpBase & 0xFFFF_FFFFUL));
+        WriteUInt32(memory, bufferTable + 4, (uint)(sharpBase >> 32) | (40u << 16));
+
+        var scalars = new uint[32];
+        scalars[4] = (uint)(attribTable & 0xFFFF_FFFFUL);
+        scalars[5] = (uint)(attribTable >> 32);
+        scalars[6] = (uint)(bufferTable & 0xFFFF_FFFFUL);
+        scalars[7] = (uint)(bufferTable >> 32);
+        var tables = new AgcVertexMetadata.VertexTableRegisters(
+            VertexBufferReg: 6,
+            VertexAttribReg: 4,
+            InputSemanticsCount: 1,
+            InputSemanticsAddress: semanticsAddress);
+
+        var data = new byte[160];
+        var discovered = new[]
+        {
+            new Gen5VertexInputBinding(
+                0x40, 0, 4, 14, 7, sharpBase, 32, 12, data, data.Length, false),
+        };
+
+        var merged = AgcVertexMetadata.MergeVertexInputsFromMetadata(
+            ctx,
+            scalars,
+            tables,
+            discovered);
+
+        Assert.Single(merged);
+        Assert.Equal(40u, merged[0].Stride);
+        Assert.Equal(12u, merged[0].OffsetBytes);
+        Assert.Equal(sharpBase, merged[0].BaseAddress);
+        Assert.Same(data, merged[0].Data);
+        Assert.Equal(0x40u, merged[0].Pc);
+    }
+
+    [Fact]
+    public void MergeVertexInputs_ConflictingMetadataOffsetDoesNotMoveBinding()
+    {
+        const ulong memoryBase = 0x1_0000_0000;
+        var memory = new FakeCpuMemory(memoryBase, 0x2000);
+        var ctx = new CpuContext(memory, Generation.Gen5);
+
+        const ulong semanticsAddress = memoryBase + 0x100;
+        const ulong attribTable = memoryBase + 0x200;
+        const ulong bufferTable = memoryBase + 0x300;
+        const ulong sharpBase = memoryBase + 0x800;
+
+        WriteUInt32(memory, semanticsAddress, 0u | (0u << 8) | (4u << 16));
+        WriteUInt32(memory, attribTable, 0u | (56u << 5) | (12u << 14));
+        WriteUInt32(memory, bufferTable, (uint)(sharpBase & 0xFFFF_FFFFUL));
+        WriteUInt32(memory, bufferTable + 4, (uint)(sharpBase >> 32) | (40u << 16));
+
+        var scalars = new uint[32];
+        scalars[4] = (uint)(attribTable & 0xFFFF_FFFFUL);
+        scalars[5] = (uint)(attribTable >> 32);
+        scalars[6] = (uint)(bufferTable & 0xFFFF_FFFFUL);
+        scalars[7] = (uint)(bufferTable >> 32);
+        var tables = new AgcVertexMetadata.VertexTableRegisters(
+            VertexBufferReg: 6,
+            VertexAttribReg: 4,
+            InputSemanticsCount: 1,
+            InputSemanticsAddress: semanticsAddress);
+
+        var original = new Gen5VertexInputBinding(
+            0x40, 0, 4, 14, 7, sharpBase, 32, 0, new byte[160], 160, false);
+        var merged = AgcVertexMetadata.MergeVertexInputsFromMetadata(
+            ctx,
+            scalars,
+            tables,
+            [original]);
+
+        Assert.Same(original, Assert.Single(merged));
+    }
+
+    [Fact]
+    public void MergeVertexInputs_UsesOffsetRelativeToCapturedBase()
+    {
+        const ulong memoryBase = 0x1_0000_0000;
+        var memory = new FakeCpuMemory(memoryBase, 0x2000);
+        var ctx = new CpuContext(memory, Generation.Gen5);
+
+        const ulong semanticsAddress = memoryBase + 0x100;
+        const ulong attribTable = memoryBase + 0x200;
+        const ulong bufferTable = memoryBase + 0x300;
+        const ulong capturedBase = memoryBase + 0x7F8;
+        const ulong sharpBase = memoryBase + 0x800;
+
+        WriteUInt32(memory, semanticsAddress, 0u | (0u << 8) | (4u << 16));
+        WriteUInt32(memory, attribTable, 0u | (56u << 5) | (12u << 14));
+        WriteUInt32(memory, bufferTable, (uint)(sharpBase & 0xFFFF_FFFFUL));
+        WriteUInt32(memory, bufferTable + 4, (uint)(sharpBase >> 32) | (40u << 16));
+
+        var scalars = new uint[32];
+        scalars[4] = (uint)(attribTable & 0xFFFF_FFFFUL);
+        scalars[5] = (uint)(attribTable >> 32);
+        scalars[6] = (uint)(bufferTable & 0xFFFF_FFFFUL);
+        scalars[7] = (uint)(bufferTable >> 32);
+        var tables = new AgcVertexMetadata.VertexTableRegisters(
+            VertexBufferReg: 6,
+            VertexAttribReg: 4,
+            InputSemanticsCount: 1,
+            InputSemanticsAddress: semanticsAddress);
+
+        var data = new byte[160];
+        var merged = AgcVertexMetadata.MergeVertexInputsFromMetadata(
+            ctx,
+            scalars,
+            tables,
+            [new Gen5VertexInputBinding(
+                0x40, 0, 4, 14, 7, capturedBase, 32, 20, data, data.Length, false)]);
+
+        Assert.Equal(40u, Assert.Single(merged).Stride);
+        Assert.Equal(20u, merged[0].OffsetBytes);
+        Assert.Equal(capturedBase, merged[0].BaseAddress);
+        Assert.Same(data, merged[0].Data);
+    }
+
     [Fact]
     public void MergeVertexInputs_AcceptsVertexAttribFormatEnums()
     {

tests/SharpEmu.Libs.Tests/Cpu/DirectExecutionBackendLlePreferenceTests.cs

@@ -0,0 +1,60 @@
+// Copyright (C) 2026 SharpEmu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+using SharpEmu.Core.Cpu.Native;
+using SharpEmu.HLE;
+using Xunit;
+
+namespace SharpEmu.Libs.Tests.Cpu;
+
+public sealed class DirectExecutionBackendLlePreferenceTests
+{
+    [Fact]
+    public void ExplicitLlePreference_AllowsNonKernelRegisteredExport()
+    {
+        var export = Export("libSceNpCppWebApi", preferLle: true);
+
+        Assert.True(DirectExecutionBackend.ShouldResolveRegisteredExportViaLle(
+            export,
+            preferLleForLibc: false));
+    }
+
+    [Fact]
+    public void ExplicitLlePreference_CannotOverrideKernelHleBoundary()
+    {
+        var export = Export("libKernel", preferLle: true);
+
+        Assert.False(DirectExecutionBackend.ShouldResolveRegisteredExportViaLle(
+            export,
+            preferLleForLibc: true));
+    }
+
+    [Fact]
+    public void RegisteredExportWithoutLlePreference_RemainsHle()
+    {
+        var export = Export("libSceNpCppWebApi", preferLle: false);
+
+        Assert.False(DirectExecutionBackend.ShouldResolveRegisteredExportViaLle(
+            export,
+            preferLleForLibc: false));
+    }
+
+    [Fact]
+    public void ExistingLibcPolicy_CanStillSelectRegisteredFirmwareExport()
+    {
+        var export = Export("libSceLibcInternal", preferLle: false);
+
+        Assert.True(DirectExecutionBackend.ShouldResolveRegisteredExportViaLle(
+            export,
+            preferLleForLibc: true));
+    }
+
+    private static ExportedFunction Export(string libraryName, bool preferLle) =>
+        new(
+            libraryName,
+            "Zxa0VhQVTsk",
+            "sceKernelWaitSema",
+            Generation.Gen5,
+            static _ => 0,
+            preferLle);
+}

tests/SharpEmu.Libs.Tests/Cpu/TlsLoadPatchBoundaryTests.cs

@@ -0,0 +1,75 @@
+// Copyright (C) 2026 SharpEmu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+using SharpEmu.Core.Cpu.Native;
+using Xunit;
+
+namespace SharpEmu.Libs.Tests.Cpu;
+
+public sealed class TlsLoadPatchBoundaryTests
+{
+    [Fact]
+    public void RejectsGtaShortJumpDisplacementAsTlsPrefix()
+    {
+        byte[] code =
+        [
+            0x90,
+            0xEB, 0x66,
+            0x66, 0x64, 0x48, 0x8B, 0x04, 0x25, 0x00, 0x00, 0x00, 0x00,
+        ];
+
+        Assert.True(DirectExecutionBackend.IsTlsLoadCandidateInsideShortJump(code, candidateOffset: 2));
+        Assert.False(DirectExecutionBackend.IsTlsLoadCandidateInsideShortJump(code, candidateOffset: 3));
+    }
+
+    [Fact]
+    public void KeepsDreamingSarahTlsInstructionAfterBackwardJnz()
+    {
+        byte[] code =
+        [
+            0x48, 0x8B, 0x1C, 0xD0,
+            0x4C, 0x39, 0x2B,
+            0x0F, 0x84, 0x10, 0x01, 0x00, 0x00,
+            0x48, 0xFF, 0xC2,
+            0x48, 0x39, 0xD1,
+            0x75, 0xEB,
+            0x66, 0x66, 0x66, 0x64, 0x48, 0x8B, 0x04, 0x25, 0x00, 0x00, 0x00, 0x00,
+        ];
+
+        Assert.False(DirectExecutionBackend.IsTlsLoadCandidateInsideShortJump(code, candidateOffset: 21));
+    }
+
+    [Theory]
+    [InlineData(0x70)]
+    [InlineData(0x7F)]
+    [InlineData(0xE0)]
+    [InlineData(0xE3)]
+    [InlineData(0xEB)]
+    public void KeepsTlsInstructionAfterRel8ControlFlow(byte opcode)
+    {
+        byte[] code =
+        [
+            0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90,
+            0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90,
+            opcode, 0xEB,
+            0x66, 0x64, 0x48, 0x8B, 0x04, 0x25, 0x00, 0x00, 0x00, 0x00,
+        ];
+
+        Assert.False(DirectExecutionBackend.IsTlsLoadCandidateInsideShortJump(code, candidateOffset: 21));
+    }
+
+    [Fact]
+    public void Rel8OpcodeByteInsidePreviousInstructionDoesNotBypassGuard()
+    {
+        byte[] code =
+        [
+            0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90,
+            0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90, 0x90,
+            0x6A, 0x75,
+            0xEB, 0x66,
+            0x66, 0x64, 0x48, 0x8B, 0x04, 0x25, 0x00, 0x00, 0x00, 0x00,
+        ];
+
+        Assert.True(DirectExecutionBackend.IsTlsLoadCandidateInsideShortJump(code, candidateOffset: 21));
+    }
+}

tests/SharpEmu.Libs.Tests/Memory/GuestMemoryAllocatorTests.cs

@@ -14,7 +14,7 @@ public sealed class GuestMemoryAllocatorTests
     public void FreedRangesAreReusedAndCoalesced()
     {
         using var memory = new PhysicalVirtualMemory(new FakeHostMemory());
-        const ulong usableArenaSize = 0x0100_0000 - 0x1000;
+        const ulong usableArenaSize = 0x2000_0000 - 0x1000;
 
         Assert.True(memory.TryAllocateGuestMemory(0x4000, 0x1000, out var first));
         Assert.True(memory.TryAllocateGuestMemory(0x8000, 0x1000, out var second));
@@ -34,6 +34,16 @@ public sealed class GuestMemoryAllocatorTests
         Assert.Equal(first, coalesced);
     }
 
+    [Fact]
+    public void ArenaSupportsAllocationsBeyondLegacySixteenMiBLimit()
+    {
+        using var memory = new PhysicalVirtualMemory(new FakeHostMemory());
+
+        Assert.True(memory.TryAllocateGuestMemory(0x0100_0000, 0x1000, out var first));
+        Assert.True(memory.TryAllocateGuestMemory(0x0020_0000, 0x1000, out var beyondLegacyLimit));
+        Assert.Equal(first + 0x0100_0000, beyondLegacyLimit);
+    }
+
     [Fact]
     public void SegmentProtectionIsAppliedInContiguousRuns()
     {

tests/SharpEmu.ShaderCompiler.Metal.Tests/Gen5MslF16CompareTests.cs

@@ -0,0 +1,85 @@
+// Copyright (C) 2026 SharpEmu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+using SharpEmu.ShaderCompiler;
+using SharpEmu.ShaderCompiler.Metal;
+using Xunit;
+
+namespace SharpEmu.ShaderCompiler.Metal.Tests;
+
+public sealed class Gen5MslF16CompareTests
+{
+    public static TheoryData<string> Opcodes = new()
+    {
+        "VCmpFF16",
+        "VCmpLtF16",
+        "VCmpEqF16",
+        "VCmpLeF16",
+        "VCmpGtF16",
+        "VCmpLgF16",
+        "VCmpGeF16",
+        "VCmpOF16",
+        "VCmpxFF16",
+        "VCmpxLtF16",
+        "VCmpxEqF16",
+        "VCmpxLeF16",
+        "VCmpxGtF16",
+        "VCmpxLgF16",
+        "VCmpxGeF16",
+        "VCmpxOF16",
+        "VCmpUF16",
+        "VCmpNgeF16",
+        "VCmpNlgF16",
+        "VCmpNgtF16",
+        "VCmpNleF16",
+        "VCmpNeqF16",
+        "VCmpNltF16",
+        "VCmpTruF16",
+        "VCmpxUF16",
+        "VCmpxNgeF16",
+        "VCmpxNlgF16",
+        "VCmpxNgtF16",
+        "VCmpxNleF16",
+        "VCmpxNeqF16",
+        "VCmpxNltF16",
+        "VCmpxTruF16",
+    };
+
+    [Theory]
+    [MemberData(nameof(Opcodes))]
+    public void F16CompareOpcodeLowersToMsl(string opcode)
+    {
+        var compare = new Gen5ShaderInstruction(
+            0,
+            Gen5ShaderEncoding.Vopc,
+            opcode,
+            [0u],
+            [Gen5Operand.Vector(0), Gen5Operand.Vector(1)],
+            [],
+            null);
+        var state = new Gen5ShaderState(
+            new Gen5ShaderProgram(0x1000, [compare]),
+            [],
+            null);
+        var scalars = new uint[256];
+        var evaluation = new Gen5ShaderEvaluation(scalars, scalars, [], []);
+
+        Assert.True(
+            Gen5MslTranslator.TryCompileComputeShader(
+                state,
+                evaluation,
+                1,
+                1,
+                1,
+                out var shader,
+                out var error),
+            error);
+        Assert.NotEmpty(shader.Source);
+        if (opcode is not (
+            "VCmpFF16" or "VCmpxFF16" or
+            "VCmpTruF16" or "VCmpxTruF16"))
+        {
+            Assert.Contains("half", shader.Source, StringComparison.Ordinal);
+        }
+    }
+}

tests/SharpEmu.ShaderCompiler.Metal.Tests/MslFloat16ArithmeticTests.cs

@@ -0,0 +1,34 @@
+// Copyright (C) 2026 SharpEmu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+using Xunit;
+
+namespace SharpEmu.ShaderCompiler.Metal.Tests;
+
+public sealed class MslFloat16ArithmeticTests
+{
+    [Fact]
+    public void CompactFloat16ArithmeticUsesHalfOperandsAndPreservesRegisterShape()
+    {
+        var fixture = new Gen5ComputeFixture(
+            "compact-f16-arithmetic",
+            [
+                0x64000501,
+                0x66060B04,
+                0x680C1107,
+                0x6A12170A,
+                0x72181D0D,
+                0x741E2310,
+                0xBF810000,
+            ],
+            StoreScalarResourceBase: 0,
+            StoreBackingBytes: 0);
+
+        var shader = Gen5ComputeFixtures.CompileOrThrow(fixture);
+
+        Assert.Contains("as_type<half>", shader.Source, StringComparison.Ordinal);
+        Assert.Contains("fmin(", shader.Source, StringComparison.Ordinal);
+        Assert.Contains("fmax(", shader.Source, StringComparison.Ordinal);
+        Assert.Contains("& 0xFFFF0000u", shader.Source, StringComparison.Ordinal);
+    }
+}

tests/SharpEmu.ShaderCompiler.Tests/Gen5Float16ArithmeticTests.cs

@@ -0,0 +1,153 @@
+// Copyright (C) 2026 SharpEmu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+using System.Buffers.Binary;
+using SharpEmu.HLE;
+using SharpEmu.ShaderCompiler.Vulkan;
+using Xunit;
+
+namespace SharpEmu.ShaderCompiler.Tests;
+
+public sealed class Gen5Float16ArithmeticTests
+{
+    private const ulong ShaderAddress = 0x1_0000_0000;
+    private const uint SEndpgm = 0xBF810000;
+
+    [Fact]
+    public void CompactFloat16ArithmeticDecodesAndCompilesWithoutNativeFloat16()
+    {
+        var program = Decode(
+        [
+            0x64000501, // v_add_f16 v0, v1, v2
+            0x66060B04, // v_sub_f16 v3, v4, v5
+            0x680C1107, // v_subrev_f16 v6, v7, v8
+            0x6A12170A, // v_mul_f16 v9, v10, v11
+            0x72181D0D, // v_max_f16 v12, v13, v14
+            0x741E2310, // v_min_f16 v15, v16, v17
+            SEndpgm,
+        ]);
+
+        Assert.Equal(
+            ["VAddF16", "VSubF16", "VSubrevF16", "VMulF16", "VMaxF16", "VMinF16", "SEndpgm"],
+            program.Instructions.Select(instruction => instruction.Opcode));
+
+        var state = new Gen5ShaderState(program, [], null);
+        var scalarRegisters = new uint[256];
+        var evaluation = new Gen5ShaderEvaluation(
+            scalarRegisters,
+            scalarRegisters,
+            [],
+            []);
+
+        Assert.True(
+            Gen5SpirvTranslator.TryCompileComputeShader(
+                state,
+                evaluation,
+                1,
+                1,
+                1,
+                out var shader,
+                out var error),
+            error);
+
+        var opcodes = ReadOpcodes(shader.Spirv);
+        Assert.Contains((ushort)SpirvOp.FAdd, opcodes);
+        Assert.Contains((ushort)SpirvOp.FSub, opcodes);
+        Assert.Contains((ushort)SpirvOp.FMul, opcodes);
+        Assert.True(opcodes.Count(opcode => opcode == (ushort)SpirvOp.ExtInst) >= 2);
+        Assert.DoesNotContain((ushort)SpirvCapability.Float16, ReadCapabilities(shader.Spirv));
+    }
+
+    private static Gen5ShaderProgram Decode(IReadOnlyList<uint> words)
+    {
+        var memory = new TestCpuMemory(ShaderAddress, words.Count * sizeof(uint));
+        var bytes = new byte[words.Count * sizeof(uint)];
+        for (var index = 0; index < words.Count; index++)
+        {
+            BinaryPrimitives.WriteUInt32LittleEndian(
+                bytes.AsSpan(index * sizeof(uint)),
+                words[index]);
+        }
+
+        Assert.True(memory.TryWrite(ShaderAddress, bytes));
+        var context = new CpuContext(memory, Generation.Gen5);
+        Assert.True(
+            Gen5ShaderTranslator.TryDecodeProgram(
+                context,
+                ShaderAddress,
+                out var program,
+                out var error),
+            error);
+        return program;
+    }
+
+    private static IReadOnlyList<ushort> ReadOpcodes(byte[] spirv) =>
+        ReadInstructions(spirv)
+            .Select(instruction => instruction.Opcode)
+            .ToArray();
+
+    private static IReadOnlyList<ushort> ReadCapabilities(byte[] spirv) =>
+        ReadInstructions(spirv)
+            .Where(instruction => instruction.Opcode == (ushort)SpirvOp.Capability)
+            .Select(instruction => (ushort)instruction.FirstOperand)
+            .ToArray();
+
+    private static IReadOnlyList<(ushort Opcode, uint FirstOperand)> ReadInstructions(
+        byte[] spirv)
+    {
+        Assert.Equal(0x07230203u, BinaryPrimitives.ReadUInt32LittleEndian(spirv));
+        var instructions = new List<(ushort Opcode, uint FirstOperand)>();
+        for (var offset = 5 * sizeof(uint); offset < spirv.Length;)
+        {
+            var header = BinaryPrimitives.ReadUInt32LittleEndian(spirv.AsSpan(offset));
+            var wordCount = checked((int)(header >> 16));
+            Assert.InRange(wordCount, 1, (spirv.Length - offset) / sizeof(uint));
+            var firstOperand = wordCount > 1
+                ? BinaryPrimitives.ReadUInt32LittleEndian(spirv.AsSpan(offset + sizeof(uint)))
+                : 0;
+            instructions.Add(((ushort)header, firstOperand));
+            offset += wordCount * sizeof(uint);
+        }
+
+        return instructions;
+    }
+
+    private sealed class TestCpuMemory(ulong baseAddress, int size) : ICpuMemory
+    {
+        private readonly byte[] _storage = new byte[size];
+
+        public bool TryRead(ulong virtualAddress, Span<byte> destination)
+        {
+            if (!TryResolve(virtualAddress, destination.Length, out var offset))
+            {
+                return false;
+            }
+
+            _storage.AsSpan(offset, destination.Length).CopyTo(destination);
+            return true;
+        }
+
+        public bool TryWrite(ulong virtualAddress, ReadOnlySpan<byte> source)
+        {
+            if (!TryResolve(virtualAddress, source.Length, out var offset))
+            {
+                return false;
+            }
+
+            source.CopyTo(_storage.AsSpan(offset, source.Length));
+            return true;
+        }
+
+        private bool TryResolve(ulong address, int length, out int offset)
+        {
+            offset = 0;
+            if (address < baseAddress || address - baseAddress > int.MaxValue)
+            {
+                return false;
+            }
+
+            offset = (int)(address - baseAddress);
+            return offset <= _storage.Length - length;
+        }
+    }
+}

tests/SharpEmu.ShaderCompiler.Tests/Gen5VopcF16Tests.cs

@@ -0,0 +1,154 @@
+// Copyright (C) 2026 SharpEmu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+using System.Buffers.Binary;
+using SharpEmu.HLE;
+using SharpEmu.ShaderCompiler;
+using SharpEmu.ShaderCompiler.Vulkan;
+using Xunit;
+
+namespace SharpEmu.ShaderCompiler.Tests;
+
+public sealed class Gen5VopcF16Tests
+{
+    private const ulong ShaderAddress = 0x1_0000_0000;
+    private const uint SEndpgm = 0xBF810000;
+
+    public static TheoryData<uint, string> Opcodes = new()
+    {
+        { 0xC8, "VCmpFF16" },
+        { 0xC9, "VCmpLtF16" },
+        { 0xCA, "VCmpEqF16" },
+        { 0xCB, "VCmpLeF16" },
+        { 0xCC, "VCmpGtF16" },
+        { 0xCD, "VCmpLgF16" },
+        { 0xCE, "VCmpGeF16" },
+        { 0xCF, "VCmpOF16" },
+        { 0xD8, "VCmpxFF16" },
+        { 0xD9, "VCmpxLtF16" },
+        { 0xDA, "VCmpxEqF16" },
+        { 0xDB, "VCmpxLeF16" },
+        { 0xDC, "VCmpxGtF16" },
+        { 0xDD, "VCmpxLgF16" },
+        { 0xDE, "VCmpxGeF16" },
+        { 0xDF, "VCmpxOF16" },
+        { 0xE8, "VCmpUF16" },
+        { 0xE9, "VCmpNgeF16" },
+        { 0xEA, "VCmpNlgF16" },
+        { 0xEB, "VCmpNgtF16" },
+        { 0xEC, "VCmpNleF16" },
+        { 0xED, "VCmpNeqF16" },
+        { 0xEE, "VCmpNltF16" },
+        { 0xEF, "VCmpTruF16" },
+        { 0xF8, "VCmpxUF16" },
+        { 0xF9, "VCmpxNgeF16" },
+        { 0xFA, "VCmpxNlgF16" },
+        { 0xFB, "VCmpxNgtF16" },
+        { 0xFC, "VCmpxNleF16" },
+        { 0xFD, "VCmpxNeqF16" },
+        { 0xFE, "VCmpxNltF16" },
+        { 0xFF, "VCmpxTruF16" },
+    };
+
+    [Theory]
+    [MemberData(nameof(Opcodes))]
+    public void F16CompareOpcodeDecodes(uint opcode, string expectedName)
+    {
+        var memory = new TestCpuMemory(ShaderAddress, 0x100);
+        Span<byte> shader = stackalloc byte[2 * sizeof(uint)];
+        var word = (0x3Eu << 25) | (opcode << 17) | (1u << 9);
+        BinaryPrimitives.WriteUInt32LittleEndian(shader, word);
+        BinaryPrimitives.WriteUInt32LittleEndian(shader[sizeof(uint)..], SEndpgm);
+        Assert.True(memory.TryWrite(ShaderAddress, shader));
+
+        var ctx = new CpuContext(memory, Generation.Gen5);
+        Assert.True(
+            Gen5ShaderTranslator.TryDecodeProgram(
+                ctx,
+                ShaderAddress,
+                out var program,
+                out var error),
+            error);
+        var instruction = Assert.Single(
+            program.Instructions,
+            candidate => candidate.Encoding == Gen5ShaderEncoding.Vopc);
+        Assert.Equal(expectedName, instruction.Opcode);
+    }
+
+    [Theory]
+    [MemberData(nameof(Opcodes))]
+    public void F16CompareOpcodeLowersToSpirv(uint _, string opcode)
+    {
+        var compare = new Gen5ShaderInstruction(
+            0,
+            Gen5ShaderEncoding.Vopc,
+            opcode,
+            [0u],
+            [Gen5Operand.Vector(0), Gen5Operand.Vector(1)],
+            [],
+            null);
+        var state = new Gen5ShaderState(
+            new Gen5ShaderProgram(ShaderAddress, [compare]),
+            [],
+            null);
+        var scalars = new uint[256];
+        var evaluation = new Gen5ShaderEvaluation(scalars, scalars, [], []);
+
+        Assert.True(
+            Gen5SpirvTranslator.TryCompileComputeShader(
+                state,
+                evaluation,
+                1,
+                1,
+                1,
+                out var shader,
+                out var error),
+            error);
+        Assert.NotEmpty(shader.Spirv);
+    }
+
+    private sealed class TestCpuMemory(ulong baseAddress, int size) : ICpuMemory
+    {
+        private readonly byte[] _storage = new byte[size];
+
+        public bool TryRead(ulong virtualAddress, Span<byte> destination)
+        {
+            if (!TryResolve(virtualAddress, destination.Length, out var offset))
+            {
+                return false;
+            }
+
+            _storage.AsSpan(offset, destination.Length).CopyTo(destination);
+            return true;
+        }
+
+        public bool TryWrite(ulong virtualAddress, ReadOnlySpan<byte> source)
+        {
+            if (!TryResolve(virtualAddress, source.Length, out var offset))
+            {
+                return false;
+            }
+
+            source.CopyTo(_storage.AsSpan(offset, source.Length));
+            return true;
+        }
+
+        private bool TryResolve(ulong virtualAddress, int length, out int offset)
+        {
+            offset = 0;
+            if (virtualAddress < baseAddress)
+            {
+                return false;
+            }
+
+            var relative = virtualAddress - baseAddress;
+            if (relative + (ulong)length > (ulong)_storage.Length)
+            {
+                return false;
+            }
+
+            offset = (int)relative;
+            return true;
+        }
+    }
+}

tests/SharpEmu.SourceGenerators.Tests/SysAbiExportAnalyzerTests.cs

@@ -53,6 +53,40 @@ public sealed class SysAbiExportAnalyzerTests
         AssertSingle(diagnostics, "SHEM001");
     }
 
+    [Fact]
+    public void DuplicateNidOnTheSameMultiAttributeHandlerIsReported()
+    {
+        var diagnostics = Analyze("""
+            using SharpEmu.HLE;
+
+            public static class Exports
+            {
+                [SysAbiExport(Nid = "Zxa0VhQVTsk", ExportName = "sceKernelWaitSema")]
+                [SysAbiExport(Nid = "Zxa0VhQVTsk", ExportName = "sceKernelWaitSema")]
+                public static int Shared(CpuContext ctx) => 0;
+            }
+            """);
+
+        AssertSingle(diagnostics, "SHEM001");
+    }
+
+    [Fact]
+    public void EveryAttributeOnAMultiAttributeHandlerIsAnalyzed()
+    {
+        var diagnostics = Analyze("""
+            using SharpEmu.HLE;
+
+            public static class Exports
+            {
+                [SysAbiExport(Nid = "Zxa0VhQVTsk", ExportName = "sceKernelWaitSema")]
+                [SysAbiExport(Nid = "not_a_nid")]
+                public static int Shared(CpuContext ctx) => 0;
+            }
+            """);
+
+        AssertSingle(diagnostics, "SHEM002");
+    }
+
     [Fact]
     public void MalformedNidIsReported()
     {

tests/SharpEmu.SourceGenerators.Tests/SysAbiExportGeneratorTests.cs

@@ -36,6 +36,11 @@ public sealed class SysAbiExportGeneratorTests
             // Guest string marshalling: the thunk reads the pointer before the handler.
             [SysAbiExport(Nid = "1G3lF1Gg1k8", ExportName = "sceKernelOpen")]
             public static int KernelOpen(CpuContext ctx, [GuestCString(4096)] string path, int flags) => 0;
+
+            // A single fail-closed handler may back a catalog of LLE-preferred exports.
+            [SysAbiExport(Nid = "5fbPUzoA2fM", ExportName = "sceLleFirst", Target = Generation.Gen5, LibraryName = "libSceLle", PreferLle = true)]
+            [SysAbiExport(Nid = "L9NfM+f4f1Y", ExportName = "sceLleSecond", Target = Generation.Gen5, LibraryName = "libSceLle", PreferLle = true)]
+            public static int LleFallback(CpuContext ctx) => -1;
         }
         """;
 
@@ -123,6 +128,22 @@ public sealed class SysAbiExportGeneratorTests
         Assert.Contains("(target & registrationGeneration) == 0", generated, StringComparison.Ordinal);
     }
 
+    [Fact]
+    public void MultipleLlePreferredAttributesShareOneFailClosedHandler()
+    {
+        var (_, generated) = RoslynTestHost.RunGenerator(RoslynTestHost.Compile(HandlerSource));
+
+        Assert.Contains("\"5fbPUzoA2fM\"", generated, StringComparison.Ordinal);
+        Assert.Contains("\"L9NfM+f4f1Y\"", generated, StringComparison.Ordinal);
+        Assert.Equal(
+            2,
+            generated.Split("global::TestExports.SampleExports.LleFallback", StringSplitOptions.None).Length - 1);
+        Assert.Contains(
+            ", true, global::TestExports.SampleExports.LleFallback",
+            generated,
+            StringComparison.Ordinal);
+    }
+
     [Fact]
     public void AssemblyWithoutExportsEmitsNoRegistry()
     {

tools/SharpEmu.Tools.ShaderDump/Program.cs

@@ -17,7 +17,9 @@
 // failure that must stay loud. Any unexpected outcome makes the tool exit
 // non-zero, so it can gate scripts/CI.
 //
-// Usage: SharpEmu.Tools.ShaderDump [output-directory]
+// Usage:
+//   SharpEmu.Tools.ShaderDump [output-directory]
+//   SharpEmu.Tools.ShaderDump --inspect <shader.bin> [byte-count]
 
 using System.Buffers.Binary;
 using SharpEmu.HLE;
@@ -26,6 +28,76 @@ using SharpEmu.ShaderCompiler.Vulkan;
 
 const ulong ProgramAddress = 0x100000;
 
+if (args.Length >= 1 && string.Equals(args[0], "--inspect", StringComparison.Ordinal))
+{
+    if (args.Length is < 2 or > 3)
+    {
+        Console.Error.WriteLine(
+            "Usage: SharpEmu.Tools.ShaderDump --inspect <shader.bin> [byte-count]");
+        Environment.ExitCode = 2;
+        return;
+    }
+
+    var inputPath = Path.GetFullPath(args[1]);
+    var input = File.ReadAllBytes(inputPath);
+    var requestedByteCount = input.Length;
+    if (args.Length >= 3)
+    {
+        var value = args[2];
+        requestedByteCount = value.StartsWith("0x", StringComparison.OrdinalIgnoreCase)
+            ? Convert.ToInt32(value[2..], 16)
+            : Convert.ToInt32(value, 10);
+    }
+
+    if (requestedByteCount <= 0 ||
+        requestedByteCount > input.Length ||
+        requestedByteCount % sizeof(uint) != 0)
+    {
+        Console.Error.WriteLine(
+            $"Invalid byte count {requestedByteCount}; expected a positive, " +
+            $"4-byte-aligned value no larger than {input.Length}.");
+        Environment.ExitCode = 2;
+        return;
+    }
+
+    var words = new uint[requestedByteCount / sizeof(uint)];
+    for (var index = 0; index < words.Length; index++)
+    {
+        words[index] = BinaryPrimitives.ReadUInt32LittleEndian(
+            input.AsSpan(index * sizeof(uint), sizeof(uint)));
+    }
+
+    var memory = new FakeMemory();
+    memory.AddRegion(ProgramAddress, words);
+    var ctx = new CpuContext(memory, Generation.Gen5);
+    if (!Gen5ShaderTranslator.TryDecodeProgram(
+            ctx,
+            ProgramAddress,
+            out var program,
+            out var decodeError))
+    {
+        Console.Error.WriteLine($"Decode failed: {decodeError}");
+        Environment.ExitCode = 1;
+        return;
+    }
+
+    Console.WriteLine(
+        $"path={inputPath} bytes={requestedByteCount} " +
+        $"instructions={program!.Instructions.Count}");
+    foreach (var instruction in program.Instructions)
+    {
+        Console.WriteLine(
+            $"pc=0x{instruction.Pc:X} enc={instruction.Encoding} " +
+            $"op={instruction.Opcode} " +
+            $"words={string.Join(',', instruction.Words.Select(word => $"{word:X8}"))} " +
+            $"src={string.Join('/', instruction.Sources)} " +
+            $"dst={string.Join('/', instruction.Destinations)} " +
+            $"control={instruction.Control?.ToString() ?? "-"}");
+    }
+
+    return;
+}
+
 (string Name, bool ExpectTranslate, uint[] Words)[] testPrograms =
 [
     ("fmac", true, [