GPU: Cap batch sizes at 1024 vertices, flush if exceeded

src/pse/gpu_hw.cpp

@@ -377,8 +377,11 @@ void GPU_HW::DispatchRenderCommand(RenderCommand rc, u32 num_vertices)
   }
 
   // flush when the command changes
-  if (!m_batch_vertices.empty() && m_batch_command.bits != rc.bits)
+  if (!m_batch_vertices.empty())
+  {
+    if (m_batch_vertices.size() >= MAX_BATCH_VERTEX_COUNT || m_batch_command.bits != rc.bits)
       FlushRender();
+  }
 
   m_batch_command = rc;
   LoadVertices(rc, num_vertices);

src/pse/gpu_hw.h

@@ -11,6 +11,8 @@ public:
   virtual ~GPU_HW();
 
 protected:
+  static constexpr u32 MAX_BATCH_VERTEX_COUNT = 1024;
+
   struct HWVertex
   {
     s32 x;