923d629a 00000400 //X Button starts code
D3000000 023CDD40 //Set the Offset (I am using this with MKDS as a test)
FA000000 00000400 //Memory Dump (1024 Bytes)
d2000000 00000000

Memory Dumper -> GBA SRAM *UNTESTED*
023FE4D8 0A000003 ; Kenobi's "DB Code Type Fix"
94000130 FCFF0000 ; L+R to begin dumping
D3000000 02000000 ; Set dump source; Address to dump from (0x02000000)
C0000000 00010000 ; GBA SRAM is only 64KB is size (dump 64KB at a time)
DB000000 00000000 ; Load byte from dump source
D8000000 08000000 ; Store byte to GBA SRAM and increment dump source (***SEE NOTE***)
D2000000 00000000 ; End


***NOTE***
The second half of this line *MIGHT* have to be changed if you change the dump source. (Which you probably will!) To determine how to change it, subtract your dump source from 0x0A000000 to get the new 'store offset' for this line. Mathematically:
n = (0x0A000000 - source)

On the other hand, it might be safe to leave this line alone if your dump source is a multiple of 64KB. (Meaning the last four digits of the address are '0000' -- Which it should be!) I am currently unable to test this code, because I do not own an Action Replay. But experiment with this code, and let me know what works and what doesn't.

94000130 FDFF0000  //L Button activator
D3000000 02000000 //Load an offset
C0000000 00010000 //Should dump 128K since GBA Flashcart bus is 16 bits wide
DA000000 00000000 //Copy half-word into memory
D7000000 06000000 //Store half-word at 0x08000000
D2000000 00000000 //Loop

   mov r1, #0x08000000 @m3 unlock
   ldrh r0, [r1, #0x0]
   mov r2, #0x08E00000
   ldrh r0, [r2, #0x2]
   ldrh r0, [r1, #0xE]      
   mov r2, #0x08800000
   add r0, r2, #0x1F00
   ldrh r0, [r0, #0xFC]
   add r0, r1, #0x1000
   ldrh r0, [r0, #0x4A]
   add r0, r2, #0x600
   ldrh r0, [r0, #0x12]
   ldrh r0, [r1]
   add r0, r2, #0x1b00
   ldrh r0, [r0, #0x66]
   ldrh r0, [r2, #0x4]
   ldrh r0, [r1]   
   ldr r2, =0x09FFEFFE @ enable write
   ldr r0, =0xAA55
   strh r0, [r2]

inline void XROMUnlockSequence() {
   *(vu16*)0x08001554 = 0xAAAA;
   *(vu16*)0x08000AAA = 0x5555;
}

void xrom_erase(u32 addr, u32 size) {
   int i = 0;

   while (i < size) {
      *(vu16*)0x09000000 = 0x8000;
      *(vu16*)0x09000000 = (0x8400 | (addr >> 17));

      XROMUnlockSequence();
      *(vu16*)0x08001554 = 0x8080;

      XROMUnlockSequence();
      *(vu16*)0x08000000 = 0x3030;

      while ((*(vu16*)0x08000000 & 0x8080) != 0x8080);

      i += (128 * 1024);
      addr += (128 * 1024);
   }
}

void xrom_write(u32 addr, u16 *buffer, u32 size) {
   int i = 0, j;

   while (i < size) {
      *(vu16*)0x09000000 = 0x8000;
      *(vu16*)0x09000000 = (0x8400 | (addr >> 17));

      XROMUnlockSequence();
      *(vu16*)(0x08000000 | (addr & 0x0001FFFF)) = 0x2525;

      *(vu16*)(0x08000000 | (addr & 0x0001FFFF)) = 0x1F1F;

      for (j = 0; j < 32; j++) {
         *(vu16*)(0x08000000 | (addr & 0x0001FFFF) | (j << 1)) = buffer[(i >> 1) + j];
      }

      *(vu16*)0x08000000 = 0x2929;

      while ((*(vu16*)(0x08000000 | (addr & 0x0001FFFF) | (31 << 1)) & 0x8080) != (buffer[(i >> 1) + 31] & 0x8080));

      i += 64;
      addr += 64;
   }
}