fix crash when using -O3.

gcc will try to optimize the memfill with MMX opcodes, but those seem to crash if the memory isn't aligned to a 8-byte boundary.

CMakeLists.txt

@@ -21,7 +21,7 @@ else()
 endif()
 
 if(ENABLE_LTO)
-	add_compile_options(-O2 -flto)
+	add_compile_options(-O3 -flto)
 endif()
 
 add_compile_options(-fno-pic)

src/GPU2D.h

@@ -71,11 +71,11 @@ private:
 
     bool Accelerated;
 
-    u32 BGOBJLine[256*3];
+    u32 BGOBJLine[256*3] __attribute__((aligned (8)));
     u32* _3DLine;
 
-    u8 WindowMask[256];
-    u32 OBJLine[256];
+    u8 WindowMask[256] __attribute__((aligned (8)));
+    u32 OBJLine[256] __attribute__((aligned (8)));
 
     u16 DispFIFO[16];
     u32 DispFIFOReadPtr;