diff --git a/desmume/src/FIFO.c b/desmume/src/FIFO.c
index b49de3c87..b1cbdc2b0 100644
--- a/desmume/src/FIFO.c
+++ b/desmume/src/FIFO.c
@@ -27,7 +27,7 @@ void FIFOInit(FIFO * fifo)
 
 	fifo->begin = 0;
         fifo->end = 0;
-        for(i = 0; i<0x2000; ++i)
+        for(i = 0; i<0x8000; ++i)
         	fifo->data[i] = 0;
         fifo->full = FALSE;
         fifo->empty = TRUE;
@@ -42,7 +42,7 @@ void FIFOAdd(FIFO * fifo, u32 v)
                 return;
         }
         fifo->data[fifo->end] = v;
-        fifo->end = (fifo->end + 1)& 0x1FFF;
+        fifo->end = (fifo->end + 1)& 0x7FFF;
         fifo->full = (fifo->end == fifo->begin);
         fifo->empty = FALSE;
 }
@@ -57,7 +57,7 @@ u32 FIFOValue(FIFO * fifo)
                 return 0;
         }
         v = fifo->data[fifo->begin];
-        fifo->begin = (fifo->begin + 1)& 0x1FFF;
+        fifo->begin = (fifo->begin + 1)& 0x7FFF;
         fifo->empty = (fifo->begin == fifo->end);
         return v;
 }
diff --git a/desmume/src/FIFO.h b/desmume/src/FIFO.h
index 0bc6eb095..5010819ee 100644
--- a/desmume/src/FIFO.h
+++ b/desmume/src/FIFO.h
@@ -30,7 +30,7 @@ extern "C" {
 
 typedef struct
 {
-       u32 data[0x2000];
+       u32 data[0x8000];
        u32 begin;
        u32 end;
        BOOL full;
diff --git a/desmume/src/GPU.c b/desmume/src/GPU.c
index 1fca45e5f..ed21e1117 100644
--- a/desmume/src/GPU.c
+++ b/desmume/src/GPU.c
@@ -24,6 +24,7 @@
 #include <string.h>
 #include <stdlib.h>
 
+#include "MMU.h"
 #include "GPU.h"
 #include "debug.h"
 
diff --git a/desmume/src/GPU.h b/desmume/src/GPU.h
index 97a497e0a..e75c22c3a 100644
--- a/desmume/src/GPU.h
+++ b/desmume/src/GPU.h
@@ -270,6 +270,8 @@ void Screen_Init(void);
 void Screen_Reset(void);
 void Screen_DeInit(void);
 
+extern MMU_struct MMU;
+
 static INLINE void GPU_ligne(Screen * screen, u16 l)
 {
      GPU * gpu = screen->gpu;
@@ -307,11 +309,13 @@ static INLINE void GPU_ligne(Screen * screen, u16 l)
            }
            return;
         }
-        case 3: 
-		// it is just a wild wild guess (WWG)
-		// I dont need a game using this mode ...
+        case 3:
+		//	Read from FIFO MAIN_MEMORY_DISP_FIFO, two pixels
+		//  at once format is 5bit per component, bit15 unused
+		//    Reference:  http://nocash.emubase.de/gbatek.htm#dsvideocaptureandmainmemorydisplaymode
+	    //       (under DISP_MMEM_FIFO)
 		for (i=0; i<256;) {
-			c = FIFOValue(REG_DISPA_DISPMMEMFIFO);
+			c = FIFOValue(MMU.fifos + MAIN_MEMORY_DISP_FIFO);
 			T2WriteWord(dst, i << 1, c&0xFFFF); i++;
 			T2WriteWord(dst, i << 1, c>>16); i++;
 		}
diff --git a/desmume/src/MMU.c b/desmume/src/MMU.c
index 191a8f4b3..608961a81 100644
--- a/desmume/src/MMU.c
+++ b/desmume/src/MMU.c
@@ -1944,7 +1944,7 @@ void FASTCALL MMU_write32(u32 proc, u32 adr, u32 val)
 				MMU.DMAStartTime[proc][0] = (proc ? (val>>28) & 0x3 : (val>>27) & 0x7);
 				MMU.DMACrt[proc][0] = val;
 				T1WriteLong(MMU.MMU_MEM[proc][0x40], 0xB8, val);
-				if(MMU.DMAStartTime[proc][0] == 0)
+				if( MMU.DMAStartTime[proc][0] == 0)		// Start Immediately
 					MMU_doDMA(proc, 0);
 				#ifdef LOG_DMA2
 				else
@@ -1961,7 +1961,7 @@ void FASTCALL MMU_write32(u32 proc, u32 adr, u32 val)
 				MMU.DMAStartTime[proc][1] = (proc ? (val>>28) & 0x3 : (val>>27) & 0x7);
 				MMU.DMACrt[proc][1] = val;
 				T1WriteLong(MMU.MMU_MEM[proc][0x40], 0xC4, val);
-				if(MMU.DMAStartTime[proc][1] == 0)
+				if(MMU.DMAStartTime[proc][1] == 0)		// Start Immediately
 					MMU_doDMA(proc, 1);
 				#ifdef LOG_DMA2
 				else
@@ -1977,7 +1977,7 @@ void FASTCALL MMU_write32(u32 proc, u32 adr, u32 val)
 				MMU.DMAStartTime[proc][2] = (proc ? (val>>28) & 0x3 : (val>>27) & 0x7);
 				MMU.DMACrt[proc][2] = val;
 				T1WriteLong(MMU.MMU_MEM[proc][0x40], 0xD0, val);
-				if(MMU.DMAStartTime[proc][2] == 0)
+				if(MMU.DMAStartTime[proc][2] == 0)		// Start Immediately
 					MMU_doDMA(proc, 2);
 				#ifdef LOG_DMA2
 				else
@@ -1993,7 +1993,7 @@ void FASTCALL MMU_write32(u32 proc, u32 adr, u32 val)
 				MMU.DMAStartTime[proc][3] = (proc ? (val>>28) & 0x3 : (val>>27) & 0x7);
 				MMU.DMACrt[proc][3] = val;
 				T1WriteLong(MMU.MMU_MEM[proc][0x40], 0xDC, val);
-				if(MMU.DMAStartTime[proc][3] == 0)
+				if(	MMU.DMAStartTime[proc][3] == 0)		// Start Immediately
 					MMU_doDMA(proc, 3);
 				#ifdef LOG_DMA2
 				else
@@ -2091,6 +2091,15 @@ void FASTCALL MMU_write32(u32 proc, u32 adr, u32 val)
 				GPU_setBLDCNT	(SubScreen.gpu,val&0xffff);
 				GPU_setBLDALPHA (SubScreen.gpu,val>>16);
 				break;
+
+			case REG_DISPA_DISPMMEMFIFO:
+			{
+				// NOTE: right now, the capture unit is not taken into account,
+				//       I don't know is it should be handled here or 
+			
+				FIFOAdd(MMU.fifos + MAIN_MEMORY_DISP_FIFO, val);
+				break;
+			}
 			//case 0x21FDFF0 :  if(val==0) execute = FALSE;
 			//case 0x21FDFB0 :  if(val==0) execute = FALSE;
 			default :
@@ -2122,8 +2131,18 @@ void FASTCALL MMU_doDMA(u32 proc, u32 num)
 	 return;
 	}
 	
-		/* word count */
-        taille = (MMU.DMACrt[proc][num]&0xFFFF);
+
+	/* word count */
+    taille = (MMU.DMACrt[proc][num]&0xFFFF);
+
+	// If we are in "Main memory display" mode just copy an entire 
+	// screen (256x192 pixels). 
+	//    Reference:  http://nocash.emubase.de/gbatek.htm#dsvideocaptureandmainmemorydisplaymode
+	//       (under DISP_MMEM_FIFO)
+	if ((MMU.DMAStartTime[proc][num]==4) &&		// Must be in main memory display mode
+		(taille==4) &&							// Word must be 4
+		(((MMU.DMACrt[proc][num]>>26)&1) == 1))	// Transfer mode must be 32bit wide
+		taille = 256*192/2;
 	
 	if(MMU.DMAStartTime[proc][num] == 5) taille *= 0x80;
 	
diff --git a/desmume/src/MMU.h b/desmume/src/MMU.h
index f6a022a82..0d1e3c730 100644
--- a/desmume/src/MMU.h
+++ b/desmume/src/MMU.h
@@ -42,6 +42,7 @@ extern char szRomBaseName[512];
 #define ROM_8(m, a)  (((u8*)(m))[(a)])
  
 #define IPCFIFO  0
+#define MAIN_MEMORY_DISP_FIFO 2
  
 typedef struct {
         //ARM7 mem
diff --git a/desmume/src/NDSSystem.h b/desmume/src/NDSSystem.h
index d698ce5db..94db14615 100644
--- a/desmume/src/NDSSystem.h
+++ b/desmume/src/NDSSystem.h
@@ -265,6 +265,49 @@ int NDS_LoadFirmware(const char *filename);
                                 MMU_doDMA(0, 2);
                            if(MMU.DMAStartTime[0][3] == 3)
                                 MMU_doDMA(0, 3);
+
+						   // Main memory display
+						   if(MMU.DMAStartTime[0][0] == 4)
+						   {
+                                MMU_doDMA(0, 0);
+								MMU.DMAStartTime[0][0] = 0;
+						   }
+                           if(MMU.DMAStartTime[0][1] == 4)
+						   {
+                                MMU_doDMA(0, 1);
+								MMU.DMAStartTime[0][1] = 0;
+						   }
+                           if(MMU.DMAStartTime[0][2] == 4)
+						   {
+                                MMU_doDMA(0, 2);
+								MMU.DMAStartTime[0][2] = 0;
+						   }
+                           if(MMU.DMAStartTime[0][3] == 4)
+						   {
+                                MMU_doDMA(0, 3);
+								MMU.DMAStartTime[0][3] = 0;
+						   }
+
+						   if(MMU.DMAStartTime[1][0] == 4)
+						   {
+                                MMU_doDMA(1, 0);
+								MMU.DMAStartTime[1][0] = 0;
+						   }
+                           if(MMU.DMAStartTime[1][1] == 4)
+						   {
+                                MMU_doDMA(1, 1);
+								MMU.DMAStartTime[0][1] = 0;
+						   }
+                           if(MMU.DMAStartTime[1][2] == 4)
+						   {
+                                MMU_doDMA(1, 2);
+								MMU.DMAStartTime[1][2] = 0;
+						   }
+                           if(MMU.DMAStartTime[1][3] == 4)
+						   {
+                                MMU_doDMA(1, 3);
+								MMU.DMAStartTime[1][3] = 0;
+						   }
                                 
                            nds.lignerendu = FALSE;
                            if(nds.VCount==193)