JUST a quick replacement

src/video_core/buffer_cache/buffer_cache.h

@@ -1892,9 +1892,13 @@ bool BufferCache<P>::IntersectsInPlaceWrites(DAddr device_addr, u64 size) {
 template <class P>
 void BufferCache<P>::WaitForInPlaceWrites() {
     if constexpr (!IS_OPENGL) {
-        if (in_place_write_tick > runtime.KnownGpuTick()) {
-            runtime.Wait(in_place_write_tick);
+        if (in_place_write_tick <= in_place_host_available_tick) {
+            return;
         }
+        runtime.UnifiedMemoryWriteBarrier();
+        const u64 barrier_tick = runtime.CurrentTick();
+        runtime.Wait(barrier_tick);
+        in_place_host_available_tick = barrier_tick;
     }
 }
 

src/video_core/buffer_cache/buffer_cache_base.h

@@ -554,6 +554,7 @@ private:
     Common::RangeSet<DAddr> gpu_modified_ranges;
     Common::RangeSet<DAddr> in_place_gpu_written_ranges;
     u64 in_place_write_tick = 0;
+    u64 in_place_host_available_tick = 0;
     std::deque<Common::RangeSet<DAddr>> committed_gpu_modified_ranges;
 
     // Async Buffers