From 3a744c991efbbcb942aca7b1d49a52ee84afd0d8 Mon Sep 17 00:00:00 2001
From: angleyanalbedo <100198247+angleyanalbedo@users.noreply.github.com>
Date: Sat, 22 Aug 2026 01:18:30 +0800
Subject: [PATCH] fix(agc): record write_data produced labels so WAIT_REG_MEM
 can resume (#834)

ApplySubmittedWriteData wrote label values to guest memory but never called
GpuWaitRegistry.RecordProduced, unlike ApplySubmittedReleaseMem. A suspended
WAIT_REG_MEM on a write_data label (e.g. Dead Cells frame fence 0x10243CFD8)
then could never be latched or deadlock-broken: guest memory is reset to 0 for
frame reuse before the next re-check, and _lastProduced had no value to replay.
Record each written dword and, for increment count2, the combined 64-bit value
so 32/64-bit waits latch like ReleaseMem dataSel=2.
---
 src/SharpEmu.Libs/Agc/AgcExports.cs | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/src/SharpEmu.Libs/Agc/AgcExports.cs b/src/SharpEmu.Libs/Agc/AgcExports.cs
index 53ed15ba..d5856914 100644
--- a/src/SharpEmu.Libs/Agc/AgcExports.cs
+++ b/src/SharpEmu.Libs/Agc/AgcExports.cs
@@ -6478,6 +6478,25 @@ 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)