diff --git a/core/cpu.c b/core/cpu.c
index 068b68f..4f334f5 100644
--- a/core/cpu.c
+++ b/core/cpu.c
@@ -703,7 +703,7 @@ static int cpuBSArithmetic(VB *sim) {
         case 2:
 
             /* Read the high source word */
-            if (cpuRead(sim, bsSrcAddr + 1, VB_S32, (int32_t *) &src)) {
+            if (cpuRead(sim, bsSrcAddr + 4, VB_S32, (int32_t *) &src)) {
                 sim->cpu.step = 2;
                 return 1;
             }
@@ -713,13 +713,14 @@ static int cpuBSArithmetic(VB *sim) {
         case 3:
 
             /* Read the destination word */
-            if (cpuRead(sim, bsDestAddr, VB_S32, (int32_t *) &dest)) {
+            if (cpuRead(sim, bsDestAddr, VB_S32, (int32_t *) &auxBS.dest)) {
                 sim->cpu.step = 3;
                 return 1;
             }
 
             /* Perform the ALU operation */
-            src = bsSrcBit < bsDestBit ?
+            dest = auxBS.dest;
+            src  = bsSrcBit < bsDestBit ?
                 auxBS.src << (bsDestBit - bsSrcBit ) :
                 auxBS.src >> (bsSrcBit  - bsDestBit)
             ;
@@ -755,18 +756,19 @@ static int cpuBSArithmetic(VB *sim) {
                 src = bsLength;
 
             /* Advance to next output word */
-            if (bsSrcBit  + src >= 32)
-                bsSrcAddr  += 4;
+            if (bsSrcBit + src >= 32) {
+                bsSrcAddr    += 4;
+                auxBS.src   >>= 32;
+                sim->cpu.step = 2; /* Read high source word */
+            } else sim->cpu.step = 3; /* Read destination word */
             if (bsDestBit + src >= 32)
                 bsDestAddr += 4;
             bsSrcBit    = (bsSrcBit  + src) & 31;
             bsDestBit   = (bsDestBit + src) & 31;
             bsLength   -= src;
-            auxBS.src >>= 32;
 
             /* Update state */
             sim->cpu.clocks += cpuClocks(12); /* TODO: Research */
-            sim->cpu.step    = 2; /* Read high source word */
     }
 
     /* Exit condition */