diff --git a/CHANGES b/CHANGES
index 1015a1321..deb3e8b9f 100644
--- a/CHANGES
+++ b/CHANGES
@@ -27,6 +27,7 @@ Misc:
  - 3DS: Port to using citro3D
  - 3DS: Use system font for menus
  - PSP2: Use system font for menus
+ - All: Faster memory read/write
 
 0.4.1: (2016-07-11)
 Bugfixes:
diff --git a/src/gba/memory.c b/src/gba/memory.c
index 153d1249d..f3ac0128a 100644
--- a/src/gba/memory.c
+++ b/src/gba/memory.c
@@ -384,7 +384,7 @@ static void GBASetActiveRegion(struct ARMCore* cpu, uint32_t address) {
 	if ((address & (SIZE_CART0 - 1)) < memory->romSize) { \
 		LOAD_32(value, address & (SIZE_CART0 - 4), memory->rom); \
 	} else if (memory->mirroring && (address & memory->romMask) < memory->romSize) { \
-		LOAD_32(value, address & memory->romMask, memory->rom); \
+		LOAD_32(value, address & memory->romMask & -4, memory->rom); \
 	} else if (memory->vfame.cartType) { \
 		value = GBAVFameGetPatternValue(address, 32); \
 	} else { \
diff --git a/src/util/common.h b/src/util/common.h
index 7dd617e5c..25993cd26 100644
--- a/src/util/common.h
+++ b/src/util/common.h
@@ -93,12 +93,12 @@ typedef intptr_t ssize_t;
 #error Big endian build not supported on this platform.
 #endif
 #else
-#define LOAD_64LE(DEST, ADDR, ARR) DEST = ((uint64_t*) ARR)[(ADDR) >> 3]
-#define LOAD_32LE(DEST, ADDR, ARR) DEST = ((uint32_t*) ARR)[(ADDR) >> 2]
-#define LOAD_16LE(DEST, ADDR, ARR) DEST = ((uint16_t*) ARR)[(ADDR) >> 1]
-#define STORE_64LE(SRC, ADDR, ARR) ((uint64_t*) ARR)[(ADDR) >> 3] = SRC
-#define STORE_32LE(SRC, ADDR, ARR) ((uint32_t*) ARR)[(ADDR) >> 2] = SRC
-#define STORE_16LE(SRC, ADDR, ARR) ((uint16_t*) ARR)[(ADDR) >> 1] = SRC
+#define LOAD_64LE(DEST, ADDR, ARR) DEST = *(uint64_t*) ((uintptr_t) (ARR) + (size_t) (ADDR))
+#define LOAD_32LE(DEST, ADDR, ARR) DEST = *(uint32_t*) ((uintptr_t) (ARR) + (size_t) (ADDR))
+#define LOAD_16LE(DEST, ADDR, ARR) DEST = *(uint16_t*) ((uintptr_t) (ARR) + (size_t) (ADDR))
+#define STORE_64LE(SRC, ADDR, ARR) *(uint64_t*) ((uintptr_t) (ARR) + (size_t) (ADDR)) = SRC
+#define STORE_32LE(SRC, ADDR, ARR) *(uint32_t*) ((uintptr_t) (ARR) + (size_t) (ADDR)) = SRC
+#define STORE_16LE(SRC, ADDR, ARR) *(uint16_t*) ((uintptr_t) (ARR) + (size_t) (ADDR)) = SRC
 #endif
 
 #define MAKE_MASK(START, END) (((1 << ((END) - (START))) - 1) << (START))