[pthread] Fix cond-var lock inversion and POSIX timedwait ABI (upstream #113) (#115)

Two follow-ups to upstream #102's condition-variable changes, as analyzed
in upstream issue #113:

- The pending-signal consume path reacquired the guest mutex while still
  holding the condition state lock, inverting lock order against
  cond-signal (mutex -> SyncRoot) and deadlocking both threads. Leave the
  condition lock before relocking, matching the normal wake path.
  This unfroze Dreaming Sarah (PPSA02929) at its title screen.

- pthread_cond_timedwait's third argument is a pointer to an absolute
  CLOCK_REALTIME timespec, not a relative microsecond count; the guest
  address was being truncated into a duration, yielding arbitrary
  timeouts. Read the timespec and convert to a relative wait.
  scePthreadCondTimedwait keeps its separate relative-time ABI.

src/SharpEmu.Libs/Agc/Gen5ShaderTranslator.cs

@@ -621,6 +621,8 @@ internal static class Gen5ShaderTranslator
             0x10 => "SSendmsg",
             0x16 => "STtraceData",
             0x20 => "SInstPrefetch",
+            0x21 => "SClause",
+            0x23 => "SWaitcntDepctr",
             _ => string.Empty,
         };
 

src/SharpEmu.Libs/Agc/Gen5SpirvTranslator.Alu.cs

@@ -1136,6 +1136,52 @@ internal static partial class Gen5SpirvTranslator
                                 left,
                                 right);
                             break;
+                        case "SMulHiU32":
+                        {
+                            var wideLeft = _module.AddInstruction(
+                                SpirvOp.UConvert,
+                                _ulongType,
+                                left);
+                            var wideRight = _module.AddInstruction(
+                                SpirvOp.UConvert,
+                                _ulongType,
+                                right);
+                            var product = _module.AddInstruction(
+                                SpirvOp.IMul,
+                                _ulongType,
+                                wideLeft,
+                                wideRight);
+                            result = _module.AddInstruction(
+                                SpirvOp.UConvert,
+                                _uintType,
+                                ShiftRightLogical64(
+                                    product,
+                                    _module.Constant64(_ulongType, 32)));
+                            break;
+                        }
+                        case "SMulHiI32":
+                        {
+                            var wideLeft = _module.AddInstruction(
+                                SpirvOp.SConvert,
+                                _longType,
+                                Bitcast(_intType, left));
+                            var wideRight = _module.AddInstruction(
+                                SpirvOp.SConvert,
+                                _longType,
+                                Bitcast(_intType, right));
+                            var product = _module.AddInstruction(
+                                SpirvOp.IMul,
+                                _longType,
+                                wideLeft,
+                                wideRight);
+                            result = _module.AddInstruction(
+                                SpirvOp.UConvert,
+                                _uintType,
+                                ShiftRightLogical64(
+                                    Bitcast(_ulongType, product),
+                                    _module.Constant64(_ulongType, 32)));
+                            break;
+                        }
                         case "SAndB32":
                             result = BitwiseAnd(left, right);
                             Store(_scc, IsNotZero(result));

src/SharpEmu.Libs/Kernel/KernelPthreadCompatExports.cs

@@ -349,7 +349,32 @@ public static class KernelPthreadCompatExports
         ExportName = "pthread_cond_timedwait",
         Target = Generation.Gen4 | Generation.Gen5,
         LibraryName = "libKernel")]
-    public static int PosixPthreadCondTimedwait(CpuContext ctx) => PthreadCondTimedwait(ctx);
+    public static int PosixPthreadCondTimedwait(CpuContext ctx)
+    {
+        // POSIX passes `const struct timespec *abstime` (absolute deadline), not relative microseconds (#113).
+        var abstimeAddress = ctx[CpuRegister.Rdx];
+        if (abstimeAddress == 0)
+        {
+            return (int)OrbisGen2Result.ORBIS_GEN2_ERROR_INVALID_ARGUMENT;
+        }
+
+        if (!ctx.TryReadUInt64(abstimeAddress, out var deadlineSeconds) ||
+            !ctx.TryReadUInt64(abstimeAddress + 8, out var deadlineNanoseconds))
+        {
+            return (int)OrbisGen2Result.ORBIS_GEN2_ERROR_MEMORY_FAULT;
+        }
+
+        var nowMicroseconds = DateTimeOffset.UtcNow.ToUnixTimeMilliseconds() * 1000L;
+        var deadlineMicroseconds =
+            deadlineSeconds >= long.MaxValue / 1_000_000UL
+                ? long.MaxValue
+                : (long)(deadlineSeconds * 1_000_000UL + Math.Min(deadlineNanoseconds, 999_999_999UL) / 1000UL);
+        var remainingMicroseconds = deadlineMicroseconds - nowMicroseconds;
+        var timeoutUsec = remainingMicroseconds <= 0
+            ? 1u
+            : (uint)Math.Min(remainingMicroseconds, uint.MaxValue);
+        return PthreadCondWaitCore(ctx, ctx[CpuRegister.Rdi], ctx[CpuRegister.Rsi], timed: true, timeoutUsec: timeoutUsec);
+    }
 
     [SysAbiExport(
         Nid = "kDh-NfxgMtE",
@@ -1185,10 +1210,19 @@ public static class KernelPthreadCompatExports
 
             if (consumedPendingSignal)
             {
+                // Leave SyncRoot before relocking the mutex to avoid lock-order inversion with cond-signal (#113).
                 state.Waiters = Math.Max(0, state.Waiters - 1);
                 TracePthreadCond("wait-wake-pending", condAddress, mutexAddress, state, timed, waitResult);
-                var pendingLockResult = PthreadMutexRelockForCondWait(ctx, mutexAddress, releasedRecursion);
-                return pendingLockResult != (int)OrbisGen2Result.ORBIS_GEN2_OK ? pendingLockResult : waitResult;
+                Monitor.Exit(state.SyncRoot);
+                try
+                {
+                    var pendingLockResult = PthreadMutexRelockForCondWait(ctx, mutexAddress, releasedRecursion);
+                    return pendingLockResult != (int)OrbisGen2Result.ORBIS_GEN2_OK ? pendingLockResult : waitResult;
+                }
+                finally
+                {
+                    Monitor.Enter(state.SyncRoot);
+                }
             }
 
             var scheduler = GuestThreadExecution.Scheduler;

src/SharpEmu.Libs/VideoOut/VideoOutExports.cs

@@ -45,6 +45,59 @@ public static class VideoOutExports
     private static readonly object _stateGate = new();
     private static readonly object _frameDumpGate = new();
     private static readonly Dictionary<int, VideoOutPortState> _ports = new();
+
+    // Hardware raises vblank autonomously; UE blocks its frame loop on it.
+    private static Timer? _vblankPumpTimer;
+    private static int _vblankPumpActive;
+    private const int VblankPumpIntervalMs = 16;
+
+    private static void EnsureVblankPumpStarted()
+    {
+        if (_vblankPumpTimer is not null)
+        {
+            return;
+        }
+
+        _vblankPumpTimer = new Timer(
+            static _ => PumpVblanks(),
+            null,
+            VblankPumpIntervalMs,
+            VblankPumpIntervalMs);
+    }
+
+    private static void PumpVblanks()
+    {
+        if (Interlocked.Exchange(ref _vblankPumpActive, 1) != 0)
+        {
+            return;
+        }
+
+        try
+        {
+            VideoOutPortState[] ports;
+            lock (_stateGate)
+            {
+                if (_ports.Count == 0)
+                {
+                    return;
+                }
+
+                // Signalling reaches WakeBlockedThreads -> Pump(), which serialises on one global
+                // flag. Waking an unwatched queue would hold it 60x/sec and starve guest threads.
+                ports = _ports.Values.Where(static port => port.VblankEvents.Count != 0).ToArray();
+            }
+
+            foreach (var port in ports)
+            {
+                SignalVblank(port);
+            }
+        }
+        finally
+        {
+            Volatile.Write(ref _vblankPumpActive, 0);
+        }
+    }
+
     private static readonly Dictionary<(int Handle, int BufferIndex, ulong Address), ulong> _lastFrameFingerprints = new();
     private static int _nextHandle = 1;
     private static int _frameDumpCount;
@@ -177,6 +230,7 @@ public static class VideoOutExports
             {
                 Handle = handle,
             };
+            EnsureVblankPumpStarted();
             return handle;
         }
     }