From db32a0cb82dd03bee31b934380cae9ba922f7bda Mon Sep 17 00:00:00 2001
From: pineappleEA <pineaea@gmail.com>
Date: Fri, 22 Jan 2021 02:50:09 +0100
Subject: [PATCH] early-access version 1356

---
 README.md                                     |  2 +-
 src/video_core/memory_manager.cpp             | 26 ++++++++++++++-----
 src/video_core/memory_manager.h               |  3 ++-
 .../renderer_opengl/gl_shader_decompiler.cpp  |  3 ++-
 4 files changed, 24 insertions(+), 10 deletions(-)

diff --git a/README.md b/README.md
index b9aa444a9..ad3077f25 100755
--- a/README.md
+++ b/README.md
@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 1355.
+This is the source code for early-access 1356.
 
 ## Legal Notice
 
diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp
index 20e90be3a..fed2a5ea0 100755
--- a/src/video_core/memory_manager.cpp
+++ b/src/video_core/memory_manager.cpp
@@ -333,17 +333,29 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buf
     }
 }
 
+void MemoryManager::FlushRegion(GPUVAddr gpu_addr, size_t size) const {
+    size_t remaining_size{size};
+    size_t page_index{gpu_addr >> page_bits};
+    size_t page_offset{gpu_addr & page_mask};
+    while (remaining_size > 0) {
+        const size_t num_bytes{std::min(page_size - page_offset, remaining_size)};
+        if (const auto page_addr{GpuToCpuAddress(page_index << page_bits)}; page_addr) {
+            rasterizer->FlushRegion(*page_addr + page_offset, num_bytes);
+        }
+        ++page_index;
+        page_offset = 0;
+        remaining_size -= num_bytes;
+    }
+}
+
 void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size) {
     std::vector<u8> tmp_buffer(size);
     ReadBlock(gpu_src_addr, tmp_buffer.data(), size);
-    WriteBlock(gpu_dest_addr, tmp_buffer.data(), size);
-}
 
-void MemoryManager::CopyBlockUnsafe(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr,
-                                    std::size_t size) {
-    std::vector<u8> tmp_buffer(size);
-    ReadBlockUnsafe(gpu_src_addr, tmp_buffer.data(), size);
-    WriteBlockUnsafe(gpu_dest_addr, tmp_buffer.data(), size);
+    // The output block must be flushed in case it has data modified from the GPU.
+    // Fixes NPC geometry in Zombie Panic in Wonderland DX
+    FlushRegion(gpu_dest_addr, size);
+    WriteBlock(gpu_dest_addr, tmp_buffer.data(), size);
 }
 
 bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) const {
diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h
index a57d66022..1e9f49393 100755
--- a/src/video_core/memory_manager.h
+++ b/src/video_core/memory_manager.h
@@ -107,7 +107,6 @@ public:
      */
     void ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size) const;
     void WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size);
-    void CopyBlockUnsafe(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size);
 
     /**
      * IsGranularRange checks if a gpu region can be simply read with a pointer.
@@ -139,6 +138,8 @@ private:
     void TryLockPage(PageEntry page_entry, std::size_t size);
     void TryUnlockPage(PageEntry page_entry, std::size_t size);
 
+    void FlushRegion(GPUVAddr gpu_addr, size_t size) const;
+
     [[nodiscard]] static constexpr std::size_t PageEntryIndex(GPUVAddr gpu_addr) {
         return (gpu_addr >> page_bits) & page_table_mask;
     }
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index db30dcd65..ac78d344c 100755
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -14,6 +14,7 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/common_types.h"
+#include "common/div_ceil.h"
 #include "common/logging/log.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/engines/shader_type.h"
@@ -855,7 +856,7 @@ private:
     void DeclareConstantBuffers() {
         u32 binding = device.GetBaseBindings(stage).uniform_buffer;
         for (const auto& [index, info] : ir.GetConstantBuffers()) {
-            const u32 num_elements = Common::AlignUp(info.GetSize(), 4) / 4;
+            const u32 num_elements = Common::DivCeil(info.GetSize(), 4 * sizeof(u32));
             const u32 size = info.IsIndirect() ? MAX_CONSTBUFFER_ELEMENTS : num_elements;
             code.AddLine("layout (std140, binding = {}) uniform {} {{", binding++,
                          GetConstBufferBlock(index));