Merge branch 'bjorn/erts/clang-opt'

* bjorn/erts/clang-opt: (27 commits)
  Improve unpacking performance on x86_64
  Slightly tweak the peformance for get_list
  Speed up list matching
  Eliminate the variable temp_bits at the top scope of process_main()
  Eliminate prefetch for conditional instructions
  Teach beam_makeops to pack operands for move3 and move_window
  Ensure that the move_call_ext_{last,only} instructions are used
  beam_makeops: Eliminate unnecessary masking when packing 3 operands
  Use a cheaper tag scheme for 'd' operands
  Introduce swap_temp/3 and swap/2
  Introduce specialized versions of move2
  Add back frequently used x(0) instructions
  Rewrite the hipe_mode_switch instructions
  Remove the last use of tmp_arg1
  Eliminate use of tmp_arg1 and tmp_arg2 in bit syntax
  Remove the i_fetch instruction
  Eliminate use of i_fetch for bit syntax instructions
  Eliminate the use of i_fetch for BIF instructions
  Eliminate the use of i_fetch for relational operators
  Eliminate the use of i_fetch in arithmetic instructions
  ...

9 files changed, 1207 insertions, 1653 deletions

diff --git a/erts/emulator/beam/beam_debug.c b/erts/emulator/beam/beam_debug.c
index 8a35ad17c6..ea1f2cd012 100644
--- a/erts/emulator/beam/beam_debug.c
+++ b/erts/emulator/beam/beam_debug.c
@@ -432,39 +432,33 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
     while (*sign) {
 	switch (*sign) {
 	case 'r':		/* x(0) */
-	    erts_print(to, to_arg, "x(0)");
+	    erts_print(to, to_arg, "r(0)");
 	    break;
 	case 'x':		/* x(N) */
-	    if (reg_index(ap[0]) == 0) {
-		erts_print(to, to_arg, "x[0]");
-	    } else {
-		erts_print(to, to_arg, "x(%d)", reg_index(ap[0]));
+	    {
+		Uint n = ap[0] / sizeof(Eterm);
+		erts_print(to, to_arg, "x(%d)", n);
+		ap++;
 	    }
-	    ap++;
 	    break;
 	case 'y':		/* y(N) */
-	    erts_print(to, to_arg, "y(%d)", reg_index(ap[0]) - CP_SIZE);
-	    ap++;
+	    {
+		Uint n = ap[0] / sizeof(Eterm) - CP_SIZE;
+		erts_print(to, to_arg, "y(%d)", n);
+		ap++;
+	    }
 	    break;
 	case 'n':		/* Nil */
 	    erts_print(to, to_arg, "[]");
 	    break;
 	case 's':		/* Any source (tagged constant or register) */
-	    tag = beam_reg_tag(*ap);
-	    if (tag == X_REG_DEF) {
-		if (reg_index(*ap) == 0) {
-		    erts_print(to, to_arg, "x[0]");
-		} else {
-		    erts_print(to, to_arg, "x(%d)", reg_index(*ap));
-		}
+	    tag = loader_tag(*ap);
+	    if (tag == LOADER_X_REG) {
+		erts_print(to, to_arg, "x(%d)", loader_x_reg_index(*ap));
 		ap++;
 		break;
-	    } else if (tag == Y_REG_DEF) {
-		erts_print(to, to_arg, "y(%d)", reg_index(*ap) - CP_SIZE);
-		ap++;
-		break;
-	    } else if (tag == R_REG_DEF) {
-		erts_print(to, to_arg, "x(0)");
+	    } else if (tag == LOADER_Y_REG) {
+		erts_print(to, to_arg, "y(%d)", loader_y_reg_index(*ap) - CP_SIZE);
 		ap++;
 		break;
 	    }
@@ -481,20 +475,12 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
 	    ap++;
 	    break;
 	case 'd':		/* Destination (x(0), x(N), y(N)) */
-	    switch (beam_reg_tag(*ap)) {
-	    case X_REG_DEF:
-		if (reg_index(*ap) == 0) {
-		    erts_print(to, to_arg, "x[0]");
-		} else {
-		    erts_print(to, to_arg, "x(%d)", reg_index(*ap));
-		}
-		break;
-	    case Y_REG_DEF:
-		erts_print(to, to_arg, "y(%d)", reg_index(*ap) - CP_SIZE);
-		break;
-	    case R_REG_DEF:
-		erts_print(to, to_arg, "x(0)");
-		break;
+	    if (*ap & 1) {
+		erts_print(to, to_arg, "y(%d)",
+			   *ap / sizeof(Eterm) - CP_SIZE);
+	    } else {
+		erts_print(to, to_arg, "x(%d)",
+			   *ap / sizeof(Eterm));
 	    }
 	    ap++;
 	    break;
@@ -561,7 +547,7 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
 	    ap++;
 	    break;
 	case 'l':		/* fr(N) */
-	    erts_print(to, to_arg, "fr(%d)", reg_index(ap[0]));
+	    erts_print(to, to_arg, "fr(%d)", loader_reg_index(ap[0]));
 	    ap++;
 	    break;
 	default:
@@ -580,7 +566,6 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
     unpacked = ap;
     ap = addr + size;
     switch (op) {
-    case op_i_select_val_lins_rfI:
     case op_i_select_val_lins_xfI:
     case op_i_select_val_lins_yfI:
 	{
@@ -600,7 +585,6 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
 	    }
 	}
 	break;
-    case op_i_select_val_bins_rfI:
     case op_i_select_val_bins_xfI:
     case op_i_select_val_bins_yfI:
 	{
@@ -614,7 +598,6 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
 	    }
 	}
 	break;
-    case op_i_select_tuple_arity_rfI:
     case op_i_select_tuple_arity_xfI:
     case op_i_select_tuple_arity_yfI:
         {
@@ -639,7 +622,6 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
             }
         }
         break;
-    case op_i_jump_on_val_rfII:
     case op_i_jump_on_val_xfII:
     case op_i_jump_on_val_yfII:
 	{
@@ -651,7 +633,6 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
 	    }
 	}
 	break;
-    case op_i_jump_on_val_zero_rfI:
     case op_i_jump_on_val_zero_xfI:
     case op_i_jump_on_val_zero_yfI:
 	{
@@ -663,7 +644,6 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
 	    }
 	}
 	break;
-    case op_i_put_tuple_rI:
     case op_i_put_tuple_xI:
     case op_i_put_tuple_yI:
     case op_new_map_dII:
@@ -673,20 +653,16 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
 	    int n = unpacked[-1];
 
 	    while (n > 0) {
-		if (!is_header(ap[0])) {
+		switch (loader_tag(ap[0])) {
+		case LOADER_X_REG:
+		    erts_print(to, to_arg, " x(%d)", loader_x_reg_index(ap[0]));
+		    break;
+		case LOADER_Y_REG:
+		    erts_print(to, to_arg, " x(%d)", loader_y_reg_index(ap[0]));
+		    break;
+		default:
 		    erts_print(to, to_arg, " %T", (Eterm) ap[0]);
-		} else {
-		    switch ((ap[0] >> 2) & 0x03) {
-		    case R_REG_DEF:
-			erts_print(to, to_arg, " x(0)");
-			break;
-		    case X_REG_DEF:
-			erts_print(to, to_arg, " x(%d)", ap[0] >> 4);
-			break;
-		    case Y_REG_DEF:
-			erts_print(to, to_arg, " y(%d)", ap[0] >> 4);
-			break;
-		    }
+		    break;
 		}
 		ap++, size++, n--;
 	    }
@@ -699,18 +675,16 @@ print_op(int to, void *to_arg, int op, int size, BeamInstr* addr)
 	    while (n > 0) {
 		if (n % 3 == 1) {
 		    erts_print(to, to_arg, " %X", ap[0]);
-		} else if (!is_header(ap[0])) {
-		    erts_print(to, to_arg, " %T", (Eterm) ap[0]);
 		} else {
-		    switch ((ap[0] >> 2) & 0x03) {
-		    case R_REG_DEF:
-			erts_print(to, to_arg, " x(0)");
+		    switch (loader_tag(ap[0])) {
+		    case LOADER_X_REG:
+			erts_print(to, to_arg, " x(%d)", loader_x_reg_index(ap[0]));
 			break;
-		    case X_REG_DEF:
-			erts_print(to, to_arg, " x(%d)", ap[0] >> 4);
+		    case LOADER_Y_REG:
+			erts_print(to, to_arg, " y(%d)", loader_y_reg_index(ap[0]));
 			break;
-		    case Y_REG_DEF:
-			erts_print(to, to_arg, " y(%d)", ap[0] >> 4);
+		    default:
+			erts_print(to, to_arg, " %T", (Eterm) ap[0]);
 			break;
 		    }
 		}
diff --git a/erts/emulator/beam/beam_emu.c b/erts/emulator/beam/beam_emu.c
index 806b31654c..a296713a0a 100644
--- a/erts/emulator/beam/beam_emu.c
+++ b/erts/emulator/beam/beam_emu.c
@@ -99,10 +99,7 @@ do {									\
 do {                                     \
   int i_;                                \
   int Arity_ = PC[-1];                   \
-  if (Arity_ > 0) {                      \
-	CHECK_TERM(r(0));                \
-  }                                      \
-  for (i_ = 1; i_ < Arity_; i_++) {      \
+  for (i_ = 0; i_ < Arity_; i_++) {      \
 	CHECK_TERM(x(i_));               \
   }                                      \
 } while (0)
@@ -151,25 +148,21 @@ do {                                     \
    ASSERT(VALID_INSTR(* (Eterm *)(ip))); \
    I = (ip)
 
-#define FetchArgs(S1, S2) tmp_arg1 = (S1); tmp_arg2 = (S2)
+/*
+ * Register target (X or Y register).
+ */
+#define REG_TARGET(Target) (*(((Target) & 1) ? &yb(Target-1) : &xb(Target)))
 
 /*
  * Store a result into a register given a destination descriptor.
  */
 
-#define StoreResult(Result, DestDesc)               \
-  do {                                              \
-    Eterm stb_reg;                                  \
-    stb_reg = (DestDesc);                           \
-    CHECK_TERM(Result);                             \
-    switch (beam_reg_tag(stb_reg)) {                \
-    case R_REG_DEF:                                 \
-      r(0) = (Result); break;                       \
-    case X_REG_DEF:                                 \
-      xb(x_reg_offset(stb_reg)) = (Result); break;  \
-    default:                                        \
-      yb(y_reg_offset(stb_reg)) = (Result); break;  \
-    }                                               \
+#define StoreResult(Result, DestDesc)		\
+  do {						\
+    Eterm stb_reg;				\
+    stb_reg = (DestDesc);			\
+    CHECK_TERM(Result);				\
+    REG_TARGET(stb_reg) = (Result);		\
   } while (0)
 
 #define StoreSimpleDest(Src, Dest) Dest = (Src)
@@ -180,22 +173,16 @@ do {                                     \
  * be just before the next instruction.
  */
  
-#define StoreBifResult(Dst, Result)                          \
-  do {                                                       \
-    BeamInstr* stb_next;                                         \
-    Eterm stb_reg;                                           \
-    stb_reg = Arg(Dst);                                      \
-    I += (Dst) + 2;                                          \
-    stb_next = (BeamInstr *) *I;                                 \
-    CHECK_TERM(Result);                                      \
-    switch (beam_reg_tag(stb_reg)) {                         \
-    case R_REG_DEF:                                          \
-      r(0) = (Result); Goto(stb_next);                       \
-    case X_REG_DEF:                                          \
-      xb(x_reg_offset(stb_reg)) = (Result); Goto(stb_next);  \
-    default:                                                 \
-      yb(y_reg_offset(stb_reg)) = (Result); Goto(stb_next);  \
-    }                                                        \
+#define StoreBifResult(Dst, Result)		\
+  do {						\
+    BeamInstr* stb_next;			\
+    Eterm stb_reg;				\
+    stb_reg = Arg(Dst);				\
+    I += (Dst) + 2;				\
+    stb_next = (BeamInstr *) *I;		\
+    CHECK_TERM(Result);				\
+    REG_TARGET(stb_reg) = (Result);		\
+    Goto(stb_next);				\
   } while (0)
 
 #define ClauseFail() goto jump_f
@@ -293,7 +280,7 @@ void** beam_ops;
 #define Ib(N) (N)
 #define x(N) reg[N]
 #define y(N) E[N]
-#define r(N) x##N
+#define r(N) x(N)
 
 /*
  * Makes sure that there are StackNeed + HeapNeed + 1 words available
@@ -309,12 +296,10 @@ void** beam_ops;
      needed = (StackNeed) + 1; \
      if (E - HTOP < (needed + (HeapNeed))) { \
            SWAPOUT; \
-           reg[0] = r(0); \
            PROCESS_MAIN_CHK_LOCKS(c_p); \
            FCALLS -= erts_garbage_collect(c_p, needed + (HeapNeed), reg, (M)); \
            ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p); \
            PROCESS_MAIN_CHK_LOCKS(c_p); \
-           r(0) = reg[0]; \
            SWAPIN; \
      } \
      E -= needed; \
@@ -363,12 +348,10 @@ void** beam_ops;
     unsigned need = (Nh);                                       		\
     if ((E - HTOP < need) || (MSO(c_p).overhead + (VNh) >= BIN_VHEAP_SZ(c_p))) {\
        SWAPOUT;                                                 		\
-       reg[0] = r(0);                                           		\
        PROCESS_MAIN_CHK_LOCKS(c_p);                             		\
        FCALLS -= erts_garbage_collect(c_p, need, reg, (Live));  		\
        ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);					\
        PROCESS_MAIN_CHK_LOCKS(c_p);                             		\
-       r(0) = reg[0];                                           		\
        SWAPIN;                                                  		\
     }                                                           		\
     HEAP_SPACE_VERIFIED(need);                                                  \
@@ -386,12 +369,10 @@ void** beam_ops;
     unsigned need = (Nh);                                       \
     if (E - HTOP < need) {                                      \
        SWAPOUT;                                                 \
-       reg[0] = r(0);                                           \
        PROCESS_MAIN_CHK_LOCKS(c_p);                             \
        FCALLS -= erts_garbage_collect(c_p, need, reg, (Live));  \
        ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);			\
        PROCESS_MAIN_CHK_LOCKS(c_p);                             \
-       r(0) = reg[0];                                           \
        SWAPIN;                                                  \
     }                                                           \
     HEAP_SPACE_VERIFIED(need);                             \
@@ -408,15 +389,11 @@ void** beam_ops;
     unsigned need = (Nh);						\
     if (E - HTOP < need) {						\
        SWAPOUT;								\
-       reg[0] = r(0);							\
        reg[Live] = Extra;						\
        PROCESS_MAIN_CHK_LOCKS(c_p);					\
        FCALLS -= erts_garbage_collect(c_p, need, reg, (Live)+1);	\
        ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);				\
        PROCESS_MAIN_CHK_LOCKS(c_p);					\
-       if (Live > 0) {							\
-	   r(0) = reg[0];						\
-       }								\
        Extra = reg[Live];						\
        SWAPIN;								\
     }									\
@@ -439,7 +416,6 @@ void** beam_ops;
 #define MakeFun(FunP, NumFree)					\
   do {								\
      SWAPOUT;							\
-     reg[0] = r(0);						\
      r(0) = new_fun(c_p, reg, (ErlFunEntry *) FunP, NumFree);	\
      SWAPIN;							\
   } while (0)
@@ -531,15 +507,19 @@ void** beam_ops;
     ASSERT(VALID_INSTR(Dst));  \
     Goto(Dst)
 
-#define GetR(pos, tr) \
-   do { \
-     tr = Arg(pos); \
-     switch (beam_reg_tag(tr)) { \
-     case R_REG_DEF: tr = r(0); break; \
-     case X_REG_DEF: tr = xb(x_reg_offset(tr)); break; \
-     case Y_REG_DEF: ASSERT(y_reg_offset(tr) >= 1); tr = yb(y_reg_offset(tr)); break; \
-     } \
-     CHECK_TERM(tr); \
+#define GetR(pos, tr)				\
+   do {						\
+     tr = Arg(pos);				\
+     switch (loader_tag(tr)) {			\
+     case LOADER_X_REG:				\
+        tr = x(loader_x_reg_index(tr));		\
+        break;					\
+     case LOADER_Y_REG:				\
+        ASSERT(loader_y_reg_index(tr) >= 1);	\
+        tr = y(loader_y_reg_index(tr));		\
+        break;					\
+     }						\
+     CHECK_TERM(tr);				\
    } while (0)
 
 #define GetArg1(N, Dst) GetR((N), Dst)
@@ -557,24 +537,93 @@ void** beam_ops;
    HTOP += 2;                      \
   } while (0)
 
+#define Swap(R1, R2)				\
+  do {						\
+    Eterm V = R1;				\
+    R1 = R2;					\
+    R2 = V;					\
+  } while (0)
+
+#define SwapTemp(R1, R2, Tmp)			\
+  do {						\
+    Eterm V = R1;				\
+    R1 = R2;					\
+    R2 = Tmp = V;				\
+  } while (0)
+
 #define Move(Src, Dst, Store)      \
    do {                            \
        Eterm term = (Src);         \
        Store(term, Dst);           \
    } while (0)
 
-#define Move2(S1, D1, S2, D2) D1 = (S1); D2 = (S2)
+#define Move2Par(S1, D1, S2, D2)		\
+  do {						\
+      Eterm V1, V2;				\
+      V1 = (S1); V2 = (S2); D1 = V1; D2 = V2;	\
+  } while (0)
+
+#define MoveShift(Src, SD, D)			\
+  do {						\
+    Eterm V;					\
+    V = Src; D = SD; SD = V;			\
+  } while (0)
+
+#define MoveDup(Src, D1, D2)			\
+  do {						\
+    D1 = D2 = (Src);				\
+  } while (0)
+
 #define Move3(S1, D1, S2, D2, S3, D3) D1 = (S1); D2 = (S2); D3 = (S3)
 
-#define MoveGenDest(src, dstp) \
-   if ((dstp) == NULL) { r(0) = (src); } else { *(dstp) = src; }
+#define MoveWindow3(S1, S2, S3, D)		\
+  do {						\
+      Eterm xt0, xt1, xt2;			\
+      Eterm *y = &D;				\
+      xt0  = S1;				\
+      xt1  = S2;				\
+      xt2  = S3;				\
+      y[0] = xt0;				\
+      y[1] = xt1;				\
+      y[2] = xt2;				\
+ } while (0)
+
+#define MoveWindow4(S1, S2, S3, S4, D)		\
+  do {						\
+      Eterm xt0, xt1, xt2, xt3;			\
+      Eterm *y = &D;				\
+      xt0  = S1;				\
+      xt1  = S2;				\
+      xt2  = S3;				\
+      xt3  = S4;				\
+      y[0] = xt0;				\
+      y[1] = xt1;				\
+      y[2] = xt2;				\
+      y[3] = xt3;				\
+ } while (0)
 
-#define MoveReturn(Src, Dest)       \
-    (Dest) = (Src);                 \
-    I = c_p->cp;                    \
-    ASSERT(VALID_INSTR(*c_p->cp));  \
-    c_p->cp = 0;                    \
-    CHECK_TERM(r(0));               \
+#define MoveWindow5(S1, S2, S3, S4, S5, D)	\
+  do {						\
+      Eterm xt0, xt1, xt2, xt3, xt4;		\
+      Eterm *y = &D;				\
+      xt0  = S1;				\
+      xt1  = S2;				\
+      xt2  = S3;				\
+      xt3  = S4;				\
+      xt4  = S5;				\
+      y[0] = xt0;				\
+      y[1] = xt1;				\
+      y[2] = xt2;				\
+      y[3] = xt3;				\
+      y[4] = xt4;				\
+ } while (0)
+
+#define MoveReturn(Src)				\
+    x(0) = (Src);				\
+    I = c_p->cp;				\
+    ASSERT(VALID_INSTR(*c_p->cp));		\
+    c_p->cp = 0;				\
+    CHECK_TERM(r(0));				\
     Goto(*I)
 
 #define DeallocateReturn(Deallocate)       \
@@ -586,26 +635,26 @@ void** beam_ops;
     Goto(*I);                              \
   } while (0)
 
-#define MoveDeallocateReturn(Src, Dest, Deallocate)  \
-    (Dest) = (Src);                                  \
+#define MoveDeallocateReturn(Src, Deallocate)	\
+    x(0) = (Src);				\
     DeallocateReturn(Deallocate)
 
-#define MoveCall(Src, Dest, CallDest, Size)	\
-    (Dest) = (Src);				\
+#define MoveCall(Src, CallDest, Size)		\
+    x(0) = (Src);				\
     SET_CP(c_p, I+Size+1);			\
-    SET_I((BeamInstr *) CallDest);			\
+    SET_I((BeamInstr *) CallDest);		\
     Dispatch();
 
-#define MoveCallLast(Src, Dest, CallDest, Deallocate)	\
-    (Dest) = (Src);					\
-    RESTORE_CP(E);					\
-    E = ADD_BYTE_OFFSET(E, (Deallocate));		\
-    SET_I((BeamInstr *) CallDest);				\
+#define MoveCallLast(Src, CallDest, Deallocate)	\
+    x(0) = (Src);				\
+    RESTORE_CP(E);				\
+    E = ADD_BYTE_OFFSET(E, (Deallocate));	\
+    SET_I((BeamInstr *) CallDest);		\
     Dispatch();
 
-#define MoveCallOnly(Src, Dest, CallDest)	\
-    (Dest) = (Src);				\
-    SET_I((BeamInstr *) CallDest);			\
+#define MoveCallOnly(Src, CallDest)		\
+    x(0) = (Src);				\
+    SET_I((BeamInstr *) CallDest);		\
     Dispatch();
 
 #define MoveJump(Src)				\
@@ -613,57 +662,56 @@ void** beam_ops;
      SET_I((BeamInstr *) Arg(0));		\
      Goto(*I);
 
-#define GetList(Src, H, T) do {			\
-   Eterm* tmp_ptr = list_val(Src);		\
-   H = CAR(tmp_ptr);				\
-   T = CDR(tmp_ptr); } while (0)
-
-#define GetTupleElement(Src, Element, Dest)				\
-  do {									\
-    tmp_arg1 = (Eterm) (((unsigned char *) tuple_val(Src)) + (Element));\
-    (Dest) = (*(Eterm *) tmp_arg1);	                		\
-  } while (0)
-
-#define ExtractNextElement(Dest)					  \
-    tmp_arg1 += sizeof(Eterm);						  \
-    (Dest) = (* (Eterm *) (((unsigned char *) tmp_arg1)))
-
-#define ExtractNextElement2(Dest)				\
-  do {								\
-    Eterm* ene_dstp = &(Dest);					\
-    ene_dstp[0] = ((Eterm *) tmp_arg1)[1];	\
-    ene_dstp[1] = ((Eterm *) tmp_arg1)[2];	\
-    tmp_arg1 += sizeof(Eterm) + sizeof(Eterm);			\
+#define GetList(Src, H, T)			\
+  do {						\
+    Eterm* tmp_ptr = list_val(Src);		\
+    Eterm hd, tl;				\
+    hd = CAR(tmp_ptr);				\
+    tl = CDR(tmp_ptr);				\
+    H = hd; T = tl;				\
   } while (0)
 
-#define ExtractNextElement3(Dest)		\
-  do {						\
-    Eterm* ene_dstp = &(Dest);			\
-    ene_dstp[0] = ((Eterm *) tmp_arg1)[1];	\
-    ene_dstp[1] = ((Eterm *) tmp_arg1)[2];	\
-    ene_dstp[2] = ((Eterm *) tmp_arg1)[3];	\
-    tmp_arg1 += 3*sizeof(Eterm);		\
+#define GetTupleElement(Src, Element, Dest)		\
+  do {							\
+    Eterm* src;						\
+    src = ADD_BYTE_OFFSET(tuple_val(Src), (Element));	\
+    (Dest) = *src;					\
   } while (0)
 
-#define ExtractNextElement4(Dest)		\
-  do {						\
-    Eterm* ene_dstp = &(Dest);			\
-    ene_dstp[0] = ((Eterm *) tmp_arg1)[1];	\
-    ene_dstp[1] = ((Eterm *) tmp_arg1)[2];	\
-    ene_dstp[2] = ((Eterm *) tmp_arg1)[3];	\
-    ene_dstp[3] = ((Eterm *) tmp_arg1)[4];	\
-    tmp_arg1 += 4*sizeof(Eterm);		\
+#define GetTupleElement2(Src, Element, Dest)		\
+  do {							\
+    Eterm* src;						\
+    Eterm* dst;						\
+    Eterm E1, E2;					\
+    src = ADD_BYTE_OFFSET(tuple_val(Src), (Element));	\
+    dst = &(Dest);					\
+    E1 = src[0];					\
+    E2 = src[1];					\
+    dst[0] = E1;					\
+    dst[1] = E2;					\
   } while (0)
 
-#define ExtractElement(Element, Dest)		\
-  do {						\
-     tmp_arg1 += (Element);			\
-     (Dest) = (* (Eterm *) tmp_arg1);		\
+#define GetTupleElement3(Src, Element, Dest)		\
+  do {							\
+    Eterm* src;						\
+    Eterm* dst;						\
+    Eterm E1, E2, E3;					\
+    src = ADD_BYTE_OFFSET(tuple_val(Src), (Element));	\
+    dst = &(Dest);					\
+    E1 = src[0];					\
+    E2 = src[1];					\
+    E3 = src[2];					\
+    dst[0] = E1;					\
+    dst[1] = E2;					\
+    dst[2] = E3;					\
   } while (0)
 
 #define EqualImmed(X, Y, Action) if (X != Y) { Action; }
 #define NotEqualImmed(X, Y, Action) if (X == Y) { Action; }
 #define EqualExact(X, Y, Action) if (!EQ(X,Y)) { Action; }
+#define NotEqualExact(X, Y, Action) if (EQ(X,Y)) { Action; }
+#define Equal(X, Y, Action) if (!CMP_EQ(X,Y)) { Action; }
+#define NotEqual(X, Y, Action) if (!CMP_NE(X,Y)) { Action; }
 #define IsLessThan(X, Y, Action) if (CMP_GE(X, Y)) { Action; }
 #define IsGreaterEqual(X, Y, Action) if (CMP_LT(X, Y)) { Action; }
 
@@ -689,17 +737,26 @@ void** beam_ops;
     if (is_not_list(Src)) { Fail; }                     \
     A(Need, Alive)
 
-#define IsNonemptyListTestHeap(Src, Need, Alive, Fail)  \
-    if (is_not_list(Src)) { Fail; }                     \
+#define IsNonemptyListTestHeap(Need, Alive, Fail)  \
+    if (is_not_list(x(0))) { Fail; }			\
     TestHeap(Need, Alive)
 
+#define IsNonemptyListGetList(Src, H, T, Fail)	\
+    if (is_not_list(Src)) {			\
+         Fail;					\
+    } else {					\
+       Eterm* tmp_ptr = list_val(Src);		\
+       Eterm hd, tl;				\
+       hd = CAR(tmp_ptr);			\
+       tl = CDR(tmp_ptr);			\
+       H = hd; T = tl;				\
+    }
+
 #define IsTuple(X, Action) if (is_not_tuple(X)) Action
 
-#define IsArity(Pointer, Arity, Fail)					  \
-    if (*(Eterm *)							  \
-	(tmp_arg1 = (Eterm) (tuple_val(Pointer))) != (Arity))             \
-    { 									  \
-        Fail; 								  \
+#define IsArity(Pointer, Arity, Fail)		\
+    if (*tuple_val(Pointer) != (Arity)) {	\
+        Fail;					\
     }
 
 #define IsMap(Src, Fail) if (!is_map(Src)) { Fail; }
@@ -736,14 +793,21 @@ void** beam_ops;
      }							\
   } while (0)
 
-#define IsTupleOfArity(Src, Arity, Fail)				      \
-  do {									      \
-    if (is_not_tuple(Src) || 						      \
-	*(Eterm *)							      \
-	(tmp_arg1 = (Eterm) (tuple_val(Src))) != Arity) {                     \
-        Fail;								      \
-    }									      \
+#ifdef DEBUG
+#define IsTupleOfArity(Src, Arityval, Fail)			\
+  do {								\
+    if (!(is_tuple(Src) && *tuple_val(Src) == Arityval)) {	\
+        Fail;							\
+    }								\
+  } while (0)
+#else
+#define IsTupleOfArity(Src, Arityval, Fail)			\
+  do {								\
+    if (!(is_boxed(Src) && *tuple_val(Src) == Arityval)) {	\
+        Fail;							\
+    }								\
   } while (0)
+#endif
 
 #define IsBoolean(X, Fail) if ((X) != am_true && (X) != am_false) { Fail; }
 
@@ -770,6 +834,7 @@ void** beam_ops;
 #define BsGetFieldSize(Bits, Unit, Fail, Target)	\
    do {							\
       Sint _signed_size; Uint _uint_size;		\
+      Uint temp_bits;					\
       if (is_small(Bits)) {				\
         _signed_size = signed_val(Bits);		\
          if (_signed_size < 0) { Fail; }		\
@@ -784,6 +849,7 @@ void** beam_ops;
 #define BsGetUncheckedFieldSize(Bits, Unit, Fail, Target)	\
    do {								\
       Sint _signed_size; Uint _uint_size;			\
+      Uint temp_bits;						\
       if (is_small(Bits)) {					\
         _signed_size = signed_val(Bits);			\
          if (_signed_size < 0) { Fail; }			\
@@ -995,23 +1061,17 @@ init_emulator(void)
  */
 
 #if defined(__GNUC__) && defined(sparc) && !defined(DEBUG)
-#  define REG_x0 asm("%l0")
 #  define REG_xregs asm("%l1")
 #  define REG_htop asm("%l2")
 #  define REG_stop asm("%l3")
 #  define REG_I asm("%l4")
 #  define REG_fcalls asm("%l5")
-#  define REG_tmp_arg1 asm("%l6")
-#  define REG_tmp_arg2 asm("%l7")
 #else
-#  define REG_x0
 #  define REG_xregs
 #  define REG_htop
 #  define REG_stop
 #  define REG_I
 #  define REG_fcalls
-#  define REG_tmp_arg1
-#  define REG_tmp_arg2
 #endif
 
 #ifdef USE_VM_PROBES
@@ -1129,11 +1189,6 @@ void process_main(void)
     ERTS_DECLARE_DUMMY(Eterm pid);
 #endif
 
-    /*
-     * X register zero; also called r(0)
-     */
-    register Eterm x0 REG_x0 = NIL;
-
     /* Pointer to X registers: x(1)..x(N); reg[0] is used when doing GC,
      * in all other cases x0 is used.
      */
@@ -1161,13 +1216,6 @@ void process_main(void)
     register Sint FCALLS REG_fcalls = 0;
 
     /*
-     * Temporaries used for picking up arguments for instructions.
-     */
-    register Eterm tmp_arg1 REG_tmp_arg1 = NIL;
-    register Eterm tmp_arg2 REG_tmp_arg2 = NIL;
-    Eterm tmp_big[2];
-
-    /*
      * X registers and floating point registers are located in
      * scheduler specific data.
      */
@@ -1178,8 +1226,6 @@ void process_main(void)
      */
     int neg_o_reds = 0;
 
-    Eterm (*arith_func)(Process* p, Eterm* reg, Uint live);
-
 #ifdef ERTS_OPCODE_COUNTER_SUPPORT
     static void* counting_opcodes[] = { DEFINE_COUNTING_OPCODES };
 #else
@@ -1190,8 +1236,6 @@ void process_main(void)
 #endif
 #endif
 
-    Uint temp_bits; /* Temporary used by BsSkipBits2 & BsGetInteger2 */
-
     Eterm pt_arity;		/* Used by do_put_tuple */
 
     Uint64 start_time = 0;          /* Monitor long schedule */
@@ -1265,7 +1309,7 @@ void process_main(void)
 	int i;
 
 	argp = c_p->arg_reg;
-	for (i = c_p->arity - 1; i > 0; i--) {
+	for (i = c_p->arity - 1; i >= 0; i--) {
 	    reg[i] = argp[i];
 	    CHECK_TERM(reg[i]);
 	}
@@ -1289,12 +1333,6 @@ void process_main(void)
 	}
 
 	next = (BeamInstr *) *I;
-	r(0) = c_p->arg_reg[0];
-#ifdef HARDDEBUG
-	if (c_p->arity > 0) {
-	    CHECK_TERM(r(0));
-	}
-#endif
 	SWAPIN;
 	ASSERT(VALID_INSTR(next));
 
@@ -1333,26 +1371,24 @@ void process_main(void)
 #endif
 #include "beam_hot.h"
 
-#define STORE_ARITH_RESULT(res) StoreBifResult(2, (res));
-#define ARITH_FUNC(name) erts_gc_##name
-
 	{
 	    Eterm increment_reg_val;
 	    Eterm increment_val;
 	    Uint live;
 	    Eterm result;
 
-	OpCase(i_increment_yIId):
-	    increment_reg_val = yb(Arg(0));
+	OpCase(i_increment_rIId):
+	    increment_reg_val = x(0);
+	    I--;
 	    goto do_increment;
 
 	OpCase(i_increment_xIId):
 	    increment_reg_val = xb(Arg(0));
 	    goto do_increment;
 
-	OpCase(i_increment_rIId):
-	    increment_reg_val = r(0);
-	    I--;
+	OpCase(i_increment_yIId):
+	    increment_reg_val = yb(Arg(0));
+	    goto do_increment;
 
 	do_increment:
 	    increment_val = Arg(1);
@@ -1361,229 +1397,133 @@ void process_main(void)
 		ASSERT(MY_IS_SSMALL(i) == IS_SSMALL(i));
 		if (MY_IS_SSMALL(i)) {
 		    result = make_small(i);
-		store_result:
 		    StoreBifResult(3, result);
 		}
 	    }
 
 	    live = Arg(2);
 	    SWAPOUT;
-	    reg[0] = r(0);
 	    reg[live] = increment_reg_val;
 	    reg[live+1] = make_small(increment_val);
 	    result = erts_gc_mixed_plus(c_p, reg, live);
-	    r(0) = reg[0];
 	    SWAPIN;
 	    ERTS_HOLE_CHECK(c_p);
 	    if (is_value(result)) {
-		goto store_result;
+		StoreBifResult(3, result);
 	    }
 	    ASSERT(c_p->freason != BADMATCH || is_value(c_p->fvalue));
 	    goto find_func_info;
 	}
 
-#define DO_BIG_ARITH(Func,Arg1,Arg2)     \
-    do {                                 \
-        Uint live = Arg(1);              \
-        SWAPOUT;                         \
-        reg[0] = r(0);                   \
-        reg[live] = (Arg1);              \
-        reg[live+1] = (Arg2);            \
-        result = (Func)(c_p, reg, live); \
-        r(0) = reg[0];                   \
-        SWAPIN;                          \
-        ERTS_HOLE_CHECK(c_p);            \
-        if (is_value(result)) {          \
-            StoreBifResult(4,result);    \
-        }                                \
-        goto lb_Cl_error;                \
-    } while(0)
+#define DO_OUTLINED_ARITH_2(name, Op1, Op2)	\
+ do {						\
+     Eterm result;				\
+     Uint live = Arg(1);			\
+						\
+     SWAPOUT;					\
+     reg[live] = Op1;				\
+     reg[live+1] = Op2;				\
+     result = erts_gc_##name(c_p, reg, live);	\
+     SWAPIN;					\
+     ERTS_HOLE_CHECK(c_p);			\
+     if (is_value(result)) {			\
+	 StoreBifResult(4, result);		\
+     }						\
+     goto lb_Cl_error;				\
+ } while (0)
 
- OpCase(i_plus_jIxxd):
  {
+     Eterm PlusOp1, PlusOp2;
      Eterm result;
 
-     if (is_both_small(xb(Arg(2)), xb(Arg(3)))) {
-	 Sint i = signed_val(xb(Arg(2))) + signed_val(xb(Arg(3)));
+ OpCase(i_plus_jIxxd):
+     PlusOp1 = xb(Arg(2));
+     PlusOp2 = xb(Arg(3));
+     goto do_plus;
+
+ OpCase(i_plus_jIxyd):
+     PlusOp1 = xb(Arg(2));
+     PlusOp2 = yb(Arg(3));
+     goto do_plus;
+
+ OpCase(i_plus_jIssd):
+     GetArg2(2, PlusOp1, PlusOp2);
+     goto do_plus;
+
+ do_plus:
+     if (is_both_small(PlusOp1, PlusOp2)) {
+	 Sint i = signed_val(PlusOp1) + signed_val(PlusOp2);
 	 ASSERT(MY_IS_SSMALL(i) == IS_SSMALL(i));
 	 if (MY_IS_SSMALL(i)) {
 	     result = make_small(i);
              StoreBifResult(4, result);
 	 }
      }
-     DO_BIG_ARITH(ARITH_FUNC(mixed_plus), xb(Arg(2)), xb(Arg(3)));
+     DO_OUTLINED_ARITH_2(mixed_plus, PlusOp1, PlusOp2);
  }
 
- OpCase(i_plus_jId):
  {
+     Eterm MinusOp1, MinusOp2;
      Eterm result;
 
-     if (is_both_small(tmp_arg1, tmp_arg2)) {
-	 Sint i = signed_val(tmp_arg1) + signed_val(tmp_arg2);
-	 ASSERT(MY_IS_SSMALL(i) == IS_SSMALL(i));
-	 if (MY_IS_SSMALL(i)) {
-	     result = make_small(i);
-	     STORE_ARITH_RESULT(result);
-	 }
-     }
-     arith_func = ARITH_FUNC(mixed_plus);
-     goto do_big_arith2;
- }
-
  OpCase(i_minus_jIxxd):
- {
-     Eterm result;
-
-     if (is_both_small(xb(Arg(2)), xb(Arg(3)))) {
-	 Sint i = signed_val(xb(Arg(2))) - signed_val(xb(Arg(3)));
-	 ASSERT(MY_IS_SSMALL(i) == IS_SSMALL(i));
-	 if (MY_IS_SSMALL(i)) {
-	     result = make_small(i);
-             StoreBifResult(4, result);
-	 }
-     }
-     DO_BIG_ARITH(ARITH_FUNC(mixed_minus), xb(Arg(2)), xb(Arg(3)));
- }
+     MinusOp1 = xb(Arg(2));
+     MinusOp2 = xb(Arg(3));
+     goto do_minus;
 
- OpCase(i_minus_jId):
- {
-     Eterm result;
+ OpCase(i_minus_jIssd):
+     GetArg2(2, MinusOp1, MinusOp2);
+     goto do_minus;
 
-     if (is_both_small(tmp_arg1, tmp_arg2)) {
-	 Sint i = signed_val(tmp_arg1) - signed_val(tmp_arg2);
+ do_minus:
+     if (is_both_small(MinusOp1, MinusOp2)) {
+	 Sint i = signed_val(MinusOp1) - signed_val(MinusOp2);
 	 ASSERT(MY_IS_SSMALL(i) == IS_SSMALL(i));
 	 if (MY_IS_SSMALL(i)) {
 	     result = make_small(i);
-	     STORE_ARITH_RESULT(result);
+             StoreBifResult(4, result);
 	 }
      }
-     arith_func = ARITH_FUNC(mixed_minus);
-     goto do_big_arith2;
+     DO_OUTLINED_ARITH_2(mixed_minus, MinusOp1, MinusOp2);
  }
 
- OpCase(i_is_lt_f):
-    if (CMP_GE(tmp_arg1, tmp_arg2)) {
-	ClauseFail();
-    }
-    Next(1);
-
- OpCase(i_is_ge_f):
-    if (CMP_LT(tmp_arg1, tmp_arg2)) {
-	ClauseFail();
-    }
-    Next(1);
-
- OpCase(i_is_eq_f):
-    if (CMP_NE(tmp_arg1, tmp_arg2)) {
-	ClauseFail();
-    }
-    Next(1);
-
- OpCase(i_is_ne_f):
-    if (CMP_EQ(tmp_arg1, tmp_arg2)) {
-	ClauseFail();
-    }
-    Next(1);
-
- OpCase(i_is_eq_exact_f):
-    if (!EQ(tmp_arg1, tmp_arg2)) {
-	ClauseFail();
-    }
-    Next(1);
-
     {
 	Eterm is_eq_exact_lit_val;
 
-    OpCase(i_is_eq_exact_literal_xfc):
-	is_eq_exact_lit_val = xb(Arg(0));
-	I++;
+    OpCase(i_is_eq_exact_literal_fxc):
+	is_eq_exact_lit_val = xb(Arg(1));
 	goto do_is_eq_exact_literal;
 
-    OpCase(i_is_eq_exact_literal_yfc):
-	is_eq_exact_lit_val = yb(Arg(0));
-	I++;
+    OpCase(i_is_eq_exact_literal_fyc):
+	is_eq_exact_lit_val = yb(Arg(1));
 	goto do_is_eq_exact_literal;
 
-    OpCase(i_is_eq_exact_literal_rfc):
-	is_eq_exact_lit_val = r(0);
-
     do_is_eq_exact_literal:
-	if (!eq(Arg(1), is_eq_exact_lit_val)) {
+	if (!eq(Arg(2), is_eq_exact_lit_val)) {
 	    ClauseFail();
 	}
-	Next(2);
+	Next(3);
     }
 
     {
 	Eterm is_ne_exact_lit_val;
 
-    OpCase(i_is_ne_exact_literal_xfc):
-	is_ne_exact_lit_val = xb(Arg(0));
-	I++;
+    OpCase(i_is_ne_exact_literal_fxc):
+	is_ne_exact_lit_val = xb(Arg(1));
 	goto do_is_ne_exact_literal;
 
-    OpCase(i_is_ne_exact_literal_yfc):
-	is_ne_exact_lit_val = yb(Arg(0));
-	I++;
+    OpCase(i_is_ne_exact_literal_fyc):
+	is_ne_exact_lit_val = yb(Arg(1));
 	goto do_is_ne_exact_literal;
 
-    OpCase(i_is_ne_exact_literal_rfc):
-	is_ne_exact_lit_val = r(0);
-
     do_is_ne_exact_literal:
-	if (eq(Arg(1), is_ne_exact_lit_val)) {
+	if (eq(Arg(2), is_ne_exact_lit_val)) {
 	    ClauseFail();
 	}
-	Next(2);
+	Next(3);
     }
 
- OpCase(move_window3_xxxy): {
-     BeamInstr *next;
-     Eterm xt0, xt1, xt2;
-     Eterm *y = (Eterm *)(((unsigned char *)E) + (Arg(3)));
-     PreFetch(4, next);
-     xt0  = xb(Arg(0));
-     xt1  = xb(Arg(1));
-     xt2  = xb(Arg(2));
-     y[0] = xt0;
-     y[1] = xt1;
-     y[2] = xt2;
-     NextPF(4, next);
- }
- OpCase(move_window4_xxxxy): {
-     BeamInstr *next;
-     Eterm xt0, xt1, xt2, xt3;
-     Eterm *y = (Eterm *)(((unsigned char *)E) + (Arg(4)));
-     PreFetch(5, next);
-     xt0  = xb(Arg(0));
-     xt1  = xb(Arg(1));
-     xt2  = xb(Arg(2));
-     xt3  = xb(Arg(3));
-     y[0] = xt0;
-     y[1] = xt1;
-     y[2] = xt2;
-     y[3] = xt3;
-     NextPF(5, next);
- }
- OpCase(move_window5_xxxxxy): {
-     BeamInstr *next;
-     Eterm xt0, xt1, xt2, xt3, xt4;
-     Eterm *y = (Eterm *)(((unsigned char *)E) + (Arg(5)));
-     PreFetch(6, next);
-     xt0  = xb(Arg(0));
-     xt1  = xb(Arg(1));
-     xt2  = xb(Arg(2));
-     xt3  = xb(Arg(3));
-     xt4  = xb(Arg(4));
-     y[0] = xt0;
-     y[1] = xt1;
-     y[2] = xt2;
-     y[3] = xt3;
-     y[4] = xt4;
-     NextPF(6, next);
- }
-
- OpCase(i_move_call_only_fcr): {
+ OpCase(i_move_call_only_fc): {
      r(0) = Arg(1);
  }
  /* FALL THROUGH */
@@ -1593,7 +1533,7 @@ void process_main(void)
      Dispatch();
  }
 
- OpCase(i_move_call_last_fPcr): {
+ OpCase(i_move_call_last_fPc): {
      r(0) = Arg(2);
  }
  /* FALL THROUGH */
@@ -1605,7 +1545,7 @@ void process_main(void)
      Dispatch();
  }
 
- OpCase(i_move_call_crf): {
+ OpCase(i_move_call_cf): {
      r(0) = Arg(0);
      I++;
  }
@@ -1617,7 +1557,7 @@ void process_main(void)
      Dispatch();
  }
 
- OpCase(i_move_call_ext_last_ePcr): {
+ OpCase(i_move_call_ext_last_ePc): {
      r(0) = Arg(2);
  }
  /* FALL THROUGH */
@@ -1633,7 +1573,7 @@ void process_main(void)
     DTRACE_GLOBAL_CALL_FROM_EXPORT(c_p, Arg(0));
     Dispatchx();
 
- OpCase(i_move_call_ext_cre): {
+ OpCase(i_move_call_ext_ce): {
      r(0) = Arg(0);
      I++;
  }
@@ -1643,7 +1583,7 @@ void process_main(void)
     DTRACE_GLOBAL_CALL_FROM_EXPORT(c_p, Arg(0));
     Dispatchx();
 
- OpCase(i_move_call_ext_only_ecr): {
+ OpCase(i_move_call_ext_only_ec): {
      r(0) = Arg(1);
  }
  /* FALL THROUGH */
@@ -1709,7 +1649,6 @@ void process_main(void)
 
      PRE_BIF_SWAPOUT(c_p);
      c_p->fcalls = FCALLS - 1;
-     reg[0] = r(0);
      result = erl_send(c_p, r(0), x(1));
      PreFetch(0, next);
      ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
@@ -1717,7 +1656,6 @@ void process_main(void)
      PROCESS_MAIN_CHK_LOCKS(c_p);
      if (c_p->mbuf || MSO(c_p).overhead >= BIN_VHEAP_SZ(c_p)) {
 	 result = erts_gc_after_bif_call(c_p, result, reg, 2);
-	 r(0) = reg[0];
 	 E = c_p->stop;
      }
      HTOP = HEAP_TOP(c_p);
@@ -1730,7 +1668,6 @@ void process_main(void)
 	 SET_CP(c_p, I+1);
 	 SET_I(c_p->i);
 	 SWAPIN;
-	 r(0) = reg[0];
 	 Dispatch();
      }
      goto find_func_info;
@@ -1740,29 +1677,23 @@ void process_main(void)
 	Eterm element_index;
 	Eterm element_tuple;
 
-    OpCase(i_element_xjsd):
-	element_tuple = xb(Arg(0));
-	I++;
+    OpCase(i_element_jxsd):
+	element_tuple = xb(Arg(1));
 	goto do_element;
 
-    OpCase(i_element_yjsd):
-	element_tuple = yb(Arg(0));
-	I++;
+    OpCase(i_element_jysd):
+	element_tuple = yb(Arg(1));
 	goto do_element;
 
-    OpCase(i_element_rjsd):
-	element_tuple = r(0);
-	/* Fall through */
-
     do_element:
-	GetArg1(1, element_index);
+	GetArg1(2, element_index);
 	if (is_small(element_index) && is_tuple(element_tuple)) {
 	    Eterm* tp = tuple_val(element_tuple);
 
 	    if ((signed_val(element_index) >= 1) &&
 		(signed_val(element_index) <= arityval(*tp))) {
 		Eterm result = tp[signed_val(element_index)];
-		StoreBifResult(2, result);
+		StoreBifResult(3, result);
 	    }
 	}
     }
@@ -1776,29 +1707,24 @@ void process_main(void)
     {
 	Eterm fast_element_tuple;
 
-    OpCase(i_fast_element_rjId):
-	fast_element_tuple = r(0);
+    OpCase(i_fast_element_jxId):
+     fast_element_tuple = xb(Arg(1));
+     goto do_fast_element;
+
+    OpCase(i_fast_element_jyId):
+     fast_element_tuple = yb(Arg(1));
+     goto do_fast_element;
 
     do_fast_element:
 	if (is_tuple(fast_element_tuple)) {
 	    Eterm* tp = tuple_val(fast_element_tuple);
-	    Eterm pos = Arg(1);	/* Untagged integer >= 1 */
+	    Eterm pos = Arg(2);	/* Untagged integer >= 1 */
 	    if (pos <= arityval(*tp)) {
 		Eterm result = tp[pos];
-		StoreBifResult(2, result);
+		StoreBifResult(3, result);
 	    }
 	}
      goto badarg;
-
-    OpCase(i_fast_element_xjId):
-     fast_element_tuple = xb(Arg(0));
-     I++;
-     goto do_fast_element;
-
-    OpCase(i_fast_element_yjId):
-     fast_element_tuple = yb(Arg(0));
-     I++;
-     goto do_fast_element;
  }
 
  OpCase(catch_yf):
@@ -1904,7 +1830,7 @@ void process_main(void)
      * Pick up the next message and place it in x(0).
      * If no message, jump to a wait or wait_timeout instruction.
      */
- OpCase(i_loop_rec_fr):
+ OpCase(i_loop_rec_f):
  {
      BeamInstr *next;
      ErlMessage* msgp;
@@ -1938,13 +1864,11 @@ void process_main(void)
      ErtsMoveMsgAttachmentIntoProc(msgp, c_p, E, HTOP, FCALLS,
 				   {
 				       SWAPOUT;
-				       reg[0] = r(0);
 				       PROCESS_MAIN_CHK_LOCKS(c_p);
 				   },
 				   {
 				       ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
 				       PROCESS_MAIN_CHK_LOCKS(c_p);
-				       r(0) = reg[0];
 				       SWAPIN;
 				   });
      if (is_non_value(ERL_MESSAGE_TERM(msgp))) {
@@ -2218,10 +2142,6 @@ void process_main(void)
      select_val2 = xb(Arg(0));
      goto do_select_tuple_arity2;
 
- OpCase(i_select_tuple_arity2_rfAAff):
-     select_val2 = r(0);
-     I--;
-
  do_select_tuple_arity2:
      if (is_not_tuple(select_val2)) {
 	 goto select_val2_fail;
@@ -2237,10 +2157,6 @@ void process_main(void)
      select_val2 = xb(Arg(0));
      goto do_select_val2;
 
- OpCase(i_select_val2_rfccff):
-     select_val2 = r(0);
-     I--;
-
  do_select_val2:
      if (select_val2 == Arg(2)) {
 	 I += 3;
@@ -2264,10 +2180,6 @@ void process_main(void)
      select_val = yb(Arg(0));
      goto do_select_tuple_arity;
 
- OpCase(i_select_tuple_arity_rfI):
-     select_val = r(0);
-     I--;
-
  do_select_tuple_arity:
      if (is_tuple(select_val)) {
 	 select_val = *tuple_val(select_val);
@@ -2284,10 +2196,6 @@ void process_main(void)
      select_val = yb(Arg(0));
      goto do_linear_search;
 
- OpCase(i_select_val_lins_rfI):
-     select_val = r(0);
-     I--;
-
  do_linear_search: {
      BeamInstr *vs = &Arg(3);
      int ix = 0;
@@ -2314,10 +2222,6 @@ void process_main(void)
      select_val = yb(Arg(0));
      goto do_binary_search;
      
- OpCase(i_select_val_bins_rfI):
-     select_val = r(0);
-     I--;
-
  do_binary_search:
  {
      struct Pairs {
@@ -2378,10 +2282,6 @@ void process_main(void)
      jump_on_val_zero_index = xb(Arg(0));
      goto do_jump_on_val_zero_index;
 
- OpCase(i_jump_on_val_zero_rfI):
-     jump_on_val_zero_index = r(0);
-     I--;
-
  do_jump_on_val_zero_index:
      if (is_small(jump_on_val_zero_index)) {
 	 jump_on_val_zero_index = signed_val(jump_on_val_zero_index);
@@ -2406,10 +2306,6 @@ void process_main(void)
      jump_on_val_index = xb(Arg(0));
      goto do_jump_on_val_index;
 
- OpCase(i_jump_on_val_rfII):
-     jump_on_val_index = r(0);
-     I--;
-
  do_jump_on_val_index:
      if (is_small(jump_on_val_index)) {
 	 jump_on_val_index = (Uint) (signed_val(jump_on_val_index) - Arg(3));
@@ -2429,15 +2325,12 @@ void process_main(void)
 
      do {
 	 Eterm term = *I++;
-	 switch (term & _TAG_IMMED1_MASK) {
-	 case (R_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER:
-	     *hp++ = r(0);
+	 switch (loader_tag(term)) {
+	 case LOADER_X_REG:
+	     *hp++ = x(loader_x_reg_index(term));
 	     break;
-	 case (X_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER:
-	     *hp++ = x(term >> _TAG_IMMED1_SIZE);
-	     break;
-	 case (Y_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER:
-	     *hp++ = y(term >> _TAG_IMMED1_SIZE);
+	 case LOADER_Y_REG:
+	     *hp++ = y(loader_y_reg_index(term));
 	     break;
 	 default:
 	     *hp++ = term;
@@ -2451,31 +2344,26 @@ void process_main(void)
  OpCase(new_map_dII): {
      Eterm res;
 
-     x(0) = r(0);
      SWAPOUT;
      res = new_map(c_p, reg, I-1);
      SWAPIN;
-     r(0) = x(0);
      StoreResult(res, Arg(0));
      Next(3+Arg(2));
  }
 
-#define PUT_TERM_REG(term, desc)				\
-do {								\
-    switch ((desc) & _TAG_IMMED1_MASK) {			\
-    case (R_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER:	\
-	r(0) = (term);						\
-	break;							\
-    case (X_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER:	\
-	x((desc) >> _TAG_IMMED1_SIZE) = (term);			\
-	break;							\
-    case (Y_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER:	\
-	y((desc) >> _TAG_IMMED1_SIZE) = (term);			\
-	break;							\
-    default:							\
-	ASSERT(0);						\
-	break;							\
-    }								\
+#define PUT_TERM_REG(term, desc)		\
+do {						\
+    switch (loader_tag(desc)) {			\
+    case LOADER_X_REG:				\
+	x(loader_x_reg_index(desc)) = (term);	\
+	break;					\
+    case LOADER_Y_REG:				\
+	y(loader_y_reg_index(desc)) = (term);	\
+	break;					\
+    default:					\
+	ASSERT(0);				\
+	break;					\
+    }						\
 } while(0)
 
  OpCase(i_get_map_elements_fsI): {
@@ -2546,12 +2434,10 @@ do {								\
      Eterm map;
 
      GetArg1(1, map);
-     x(0) = r(0);
      SWAPOUT;
      res = update_map_assoc(c_p, reg, map, I);
      SWAPIN;
      if (is_value(res)) {
-	 r(0) = x(0);
 	 StoreResult(res, Arg(2));
 	 Next(5+Arg(4));
      } else {
@@ -2570,12 +2456,10 @@ do {								\
      Eterm map;
 
      GetArg1(1, map);
-     x(0) = r(0);
      SWAPOUT;
      res = update_map_exact(c_p, reg, map, I);
      SWAPIN;
      if (is_value(res)) {
-	 r(0) = x(0);
 	 StoreResult(res, Arg(2));
 	 Next(5+Arg(4));
      } else {
@@ -2657,13 +2541,10 @@ do {								\
     {
 	typedef Eterm (*GcBifFunction)(Process*, Eterm*, Uint);
 	GcBifFunction bf;
-	Eterm arg;
 	Eterm result;
 	Uint live = (Uint) Arg(3);
 
-	GetArg1(2, arg);
-	reg[0] = r(0);
-	reg[live] = arg;
+	GetArg1(2, x(live));
 	bf = (GcBifFunction) Arg(1);
 	c_p->fcalls = FCALLS;
 	SWAPOUT;
@@ -2674,7 +2555,6 @@ do {								\
 	ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
 	PROCESS_MAIN_CHK_LOCKS(c_p);
 	SWAPIN;
-	r(0) = reg[0];
 	ERTS_HOLE_CHECK(c_p);
 	FCALLS = c_p->fcalls;
 	if (is_value(result)) {
@@ -2684,12 +2564,12 @@ do {								\
 	    SET_I((BeamInstr *) Arg(0));
 	    Goto(*I);
 	}
-	reg[0] = arg;
+	x(0) = x(live);
 	I = handle_error(c_p, I, reg, translate_gc_bif((void *) bf));
 	goto post_error_handling;
     }
 
- OpCase(i_gc_bif2_jIId): /* Note, one less parameter than the i_gc_bif1
+ OpCase(i_gc_bif2_jIIssd): /* Note, one less parameter than the i_gc_bif1
 			    and i_gc_bif3 */
     {
 	typedef Eterm (*GcBifFunction)(Process*, Eterm*, Uint);
@@ -2697,9 +2577,13 @@ do {								\
 	Eterm result;
 	Uint live = (Uint) Arg(2);
 
-	reg[0] = r(0);
-	reg[live++] = tmp_arg1;
-	reg[live] = tmp_arg2;
+	GetArg2(3, x(live), x(live+1));
+	/*
+	 * XXX This calling convention does not make sense. 'live'
+	 * should point out the first argument, not the second
+	 * (i.e. 'live' should not be incremented below).
+	 */
+	live++;
 	bf = (GcBifFunction) Arg(1);
 	c_p->fcalls = FCALLS;
 	SWAPOUT;
@@ -2710,35 +2594,37 @@ do {								\
 	ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
 	PROCESS_MAIN_CHK_LOCKS(c_p);
 	SWAPIN;
-	r(0) = reg[0];
 	ERTS_HOLE_CHECK(c_p);
 	FCALLS = c_p->fcalls;
 	if (is_value(result)) {
-	    StoreBifResult(3, result);
+	    StoreBifResult(5, result);
 	}
 	if (Arg(0) != 0) {
 	    SET_I((BeamInstr *) Arg(0));
 	    Goto(*I);
 	}
-	reg[0] = tmp_arg1;
-	reg[1] = tmp_arg2;
+	live--;
+	x(0) = x(live);
+	x(1) = x(live+1);
 	I = handle_error(c_p, I, reg, translate_gc_bif((void *) bf));
 	goto post_error_handling;
     }
 
- OpCase(i_gc_bif3_jIsId):
+ OpCase(i_gc_bif3_jIIssd):
     {
 	typedef Eterm (*GcBifFunction)(Process*, Eterm*, Uint);
 	GcBifFunction bf;
-	Eterm arg;
 	Eterm result;
-	Uint live = (Uint) Arg(3);
+	Uint live = (Uint) Arg(2);
 
-	GetArg1(2, arg);
-	reg[0] = r(0);
-	reg[live++] = arg;
-	reg[live++] = tmp_arg1;
-	reg[live] = tmp_arg2;
+	x(live) = x(SCRATCH_X_REG);
+	GetArg2(3, x(live+1), x(live+2));
+	/*
+	 * XXX This calling convention does not make sense. 'live'
+	 * should point out the first argument, not the third
+	 * (i.e. 'live' should not be incremented below).
+	 */
+	live += 2;
 	bf = (GcBifFunction) Arg(1);
 	c_p->fcalls = FCALLS;
 	SWAPOUT;
@@ -2749,19 +2635,19 @@ do {								\
 	ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
 	PROCESS_MAIN_CHK_LOCKS(c_p);
 	SWAPIN;
-	r(0) = reg[0];
 	ERTS_HOLE_CHECK(c_p);
 	FCALLS = c_p->fcalls;
 	if (is_value(result)) {
-	    StoreBifResult(4, result);
+	    StoreBifResult(5, result);
 	}
 	if (Arg(0) != 0) {
 	    SET_I((BeamInstr *) Arg(0));
 	    Goto(*I);
 	}
-	reg[0] = arg;
-	reg[1] = tmp_arg1;
-	reg[2] = tmp_arg2;
+	live -= 2;
+	x(0) = x(live);
+	x(1) = x(live+1);
+	x(2) = x(live+2);
 	I = handle_error(c_p, I, reg, translate_gc_bif((void *) bf));
 	goto post_error_handling;
     }
@@ -2769,12 +2655,13 @@ do {								\
  /*
   * Guards bifs and, or, xor in guards.
   */
- OpCase(i_bif2_fbd):
+ OpCase(i_bif2_fbssd):
     {
-	Eterm tmp_reg[2] = {tmp_arg1, tmp_arg2};
+	Eterm tmp_reg[2];
 	Eterm (*bf)(Process*, Eterm*);
 	Eterm result;
 
+	GetArg2(2, tmp_reg[0], tmp_reg[1]);
 	bf = (BifFunction) Arg(1);
 	c_p->fcalls = FCALLS;
 	PROCESS_MAIN_CHK_LOCKS(c_p);
@@ -2786,7 +2673,7 @@ do {								\
 	ERTS_HOLE_CHECK(c_p);
 	FCALLS = c_p->fcalls;
 	if (is_value(result)) {
-	    StoreBifResult(2, result);
+	    StoreBifResult(4, result);
 	}
 	SET_I((BeamInstr *) Arg(0));
 	Goto(*I);
@@ -2795,12 +2682,13 @@ do {								\
  /*
   * Guards bifs and, or, xor, relational operators in body.
   */
- OpCase(i_bif2_body_bd):
+ OpCase(i_bif2_body_bssd):
     {
-	Eterm tmp_reg[2] = {tmp_arg1, tmp_arg2};
+	Eterm tmp_reg[2];
 	Eterm (*bf)(Process*, Eterm*);
 	Eterm result;
 
+	GetArg2(1, tmp_reg[0], tmp_reg[1]);
 	bf = (BifFunction) Arg(0);
 	PROCESS_MAIN_CHK_LOCKS(c_p);
 	ASSERT(!ERTS_PROC_IS_EXITING(c_p));
@@ -2811,10 +2699,10 @@ do {								\
 	ERTS_HOLE_CHECK(c_p);
 	if (is_value(result)) {
 	    ASSERT(!is_CP(result));
-	    StoreBifResult(1, result);
+	    StoreBifResult(3, result);
 	}
-	reg[0] = tmp_arg1;
-	reg[1] = tmp_arg2;
+	reg[0] = tmp_reg[0];
+	reg[1] = tmp_reg[1];
 	SWAPOUT;
 	I = handle_error(c_p, I, reg, bf);
 	goto post_error_handling;
@@ -2838,7 +2726,6 @@ do {								\
 	PreFetch(1, next);
 	ASSERT(!ERTS_PROC_IS_EXITING(c_p));
 	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	reg[0] = r(0);
 	result = (*bf)(c_p, reg, I);
 	ASSERT(!ERTS_PROC_IS_EXITING(c_p) || is_non_value(result));
 	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
@@ -2860,7 +2747,6 @@ do {								\
 	    SET_CP(c_p, I+2);
 	    SET_I(c_p->i);
 	    SWAPIN;
-	    r(0) = reg[0];
 	    Dispatch();
 	}
 
@@ -2876,111 +2762,81 @@ do {								\
   * Arithmetic operations.
   */
 
- OpCase(i_times_jId):
+ OpCase(i_times_jIssd):
  {
-     arith_func = ARITH_FUNC(mixed_times);
-     goto do_big_arith2;
+     Eterm Op1, Op2;
+     GetArg2(2, Op1, Op2);
+     DO_OUTLINED_ARITH_2(mixed_times, Op1, Op2);
  }
 
- OpCase(i_m_div_jId):
+ OpCase(i_m_div_jIssd):
  {
-     arith_func = ARITH_FUNC(mixed_div);
-     goto do_big_arith2;
+     Eterm Op1, Op2;
+     GetArg2(2, Op1, Op2);
+     DO_OUTLINED_ARITH_2(mixed_div, Op1, Op2);
  }
 
- OpCase(i_int_div_jId):
+ OpCase(i_int_div_jIssd):
  {
-     Eterm result;
+     Eterm Op1, Op2;
 
-     if (tmp_arg2 == SMALL_ZERO) {
+     GetArg2(2, Op1, Op2);
+     if (Op2 == SMALL_ZERO) {
 	 goto badarith;
-     } else if (is_both_small(tmp_arg1, tmp_arg2)) {
-	 Sint ires = signed_val(tmp_arg1) / signed_val(tmp_arg2);
+     } else if (is_both_small(Op1, Op2)) {
+	 Sint ires = signed_val(Op1) / signed_val(Op2);
 	 if (MY_IS_SSMALL(ires)) {
-	     result = make_small(ires);
-	     STORE_ARITH_RESULT(result);
+	     Eterm result = make_small(ires);
+	     StoreBifResult(4, result);
 	 }
      }
-     arith_func = ARITH_FUNC(int_div);
-     goto do_big_arith2;
+     DO_OUTLINED_ARITH_2(int_div, Op1, Op2);
  }
 
- OpCase(i_rem_jIxxd):
  {
-     Eterm result;
+     Eterm RemOp1, RemOp2;
 
-     if (xb(Arg(3)) == SMALL_ZERO) {
-	 goto badarith;
-     } else if (is_both_small(xb(Arg(2)), xb(Arg(3)))) {
-	 result = make_small(signed_val(xb(Arg(2))) % signed_val(xb(Arg(3))));
+ OpCase(i_rem_jIxxd):
+     RemOp1 = xb(Arg(2));
+     RemOp2 = xb(Arg(3));
+     goto do_rem;
+
+ OpCase(i_rem_jIssd):
+     GetArg2(2, RemOp1, RemOp2);
+     goto do_rem;
+
+ do_rem:
+     if (RemOp2 == SMALL_ZERO) {
+         goto badarith;
+     } else if (is_both_small(RemOp1, RemOp2)) {
+         Eterm result = make_small(signed_val(RemOp1) % signed_val(RemOp2));
          StoreBifResult(4, result);
+     } else {
+	 DO_OUTLINED_ARITH_2(int_rem, RemOp1, RemOp2);
      }
-     DO_BIG_ARITH(ARITH_FUNC(int_rem),xb(Arg(2)),xb(Arg(3)));
  }
 
- OpCase(i_rem_jId):
  {
-     Eterm result;
-
-     if (tmp_arg2 == SMALL_ZERO) {
-	 goto badarith;
-     } else if (is_both_small(tmp_arg1, tmp_arg2)) {
-	 result = make_small(signed_val(tmp_arg1) % signed_val(tmp_arg2));
-	 STORE_ARITH_RESULT(result);
-     } else {
-	 arith_func = ARITH_FUNC(int_rem);
-	 goto do_big_arith2;
-     }
- }
+     Eterm BandOp1, BandOp2;
 
  OpCase(i_band_jIxcd):
- {
-     Eterm result;
+     BandOp1 = xb(Arg(2));
+     BandOp2 = Arg(3);
+     goto do_band;
+
+ OpCase(i_band_jIssd):
+     GetArg2(2, BandOp1, BandOp2);
+     goto do_band;
 
-     if (is_both_small(xb(Arg(2)), Arg(3))) {
+ do_band:
+     if (is_both_small(BandOp1, BandOp2)) {
          /*
           * No need to untag -- TAG & TAG == TAG.
           */
-         result = xb(Arg(2)) & Arg(3);
+         Eterm result = BandOp1 & BandOp2;
          StoreBifResult(4, result);
      }
-     DO_BIG_ARITH(ARITH_FUNC(band),xb(Arg(2)),Arg(3));
- }
-
- OpCase(i_band_jId):
- {
-     Eterm result;
-
-     if (is_both_small(tmp_arg1, tmp_arg2)) {
-	 /*
-	  * No need to untag -- TAG & TAG == TAG.
-	  */
-	 result = tmp_arg1 & tmp_arg2;
-	 STORE_ARITH_RESULT(result);
-     }
-     arith_func = ARITH_FUNC(band);
-     goto do_big_arith2;
- }
-
-#undef DO_BIG_ARITH
-
- do_big_arith2:
- {
-     Eterm result;
-     Uint live = Arg(1);
-
-     SWAPOUT;
-     reg[0] = r(0);
-     reg[live] = tmp_arg1;
-     reg[live+1] = tmp_arg2;
-     result = arith_func(c_p, reg, live);
-     r(0) = reg[0];
-     SWAPIN;
-     ERTS_HOLE_CHECK(c_p);
-     if (is_value(result)) {
-	 STORE_ARITH_RESULT(result);
-     }
-     goto lb_Cl_error;
+     DO_OUTLINED_ARITH_2(band, BandOp1, BandOp2);
  }
 
  /*
@@ -3001,97 +2857,102 @@ do {								\
      goto find_func_info;
  }
 
- OpCase(i_bor_jId):
+ OpCase(i_bor_jIssd):
  {
-     Eterm result;
+     Eterm Op1, Op2;
 
-     if (is_both_small(tmp_arg1, tmp_arg2)) {
+     GetArg2(2, Op1, Op2);
+     if (is_both_small(Op1, Op2)) {
 	 /*
 	  * No need to untag -- TAG | TAG == TAG.
 	  */
-	 result = tmp_arg1 | tmp_arg2;
-	 STORE_ARITH_RESULT(result);
+	 Eterm result = Op1 | Op2;
+	 StoreBifResult(4, result);
      }
-     arith_func = ARITH_FUNC(bor);
-     goto do_big_arith2;
+     DO_OUTLINED_ARITH_2(bor, Op1, Op2);
  }
 
- OpCase(i_bxor_jId):
+ OpCase(i_bxor_jIssd):
  {
-     Eterm result;
+     Eterm Op1, Op2;
 
-     if (is_both_small(tmp_arg1, tmp_arg2)) {
+     GetArg2(2, Op1, Op2);
+     if (is_both_small(Op1, Op2)) {
 	 /*
 	  * We could extract the tag from one argument, but a tag extraction
 	  * could mean a shift.  Therefore, play it safe here.
 	  */
-	 result = make_small(signed_val(tmp_arg1) ^ signed_val(tmp_arg2));
-	 STORE_ARITH_RESULT(result);
+	 Eterm result = make_small(signed_val(Op1) ^ signed_val(Op2));
+	 StoreBifResult(4, result);
      }
-     arith_func = ARITH_FUNC(bxor);
-     goto do_big_arith2;
+     DO_OUTLINED_ARITH_2(bxor, Op1, Op2);
  }
 
  {
+     Eterm Op1, Op2;
      Sint i;
      Sint ires;
      Eterm* bigp;
+     Eterm tmp_big[2];
 
-     OpCase(i_bsr_jId):
-	 if (is_small(tmp_arg2)) {
-	     i = -signed_val(tmp_arg2);
-	     if (is_small(tmp_arg1)) {
+     OpCase(i_bsr_jIssd):
+         GetArg2(2, Op1, Op2);
+	 if (is_small(Op2)) {
+	     i = -signed_val(Op2);
+	     if (is_small(Op1)) {
 		 goto small_shift;
-	     } else if (is_big(tmp_arg1)) {
+	     } else if (is_big(Op1)) {
 		 if (i == 0) {
-		     StoreBifResult(2, tmp_arg1);
+		     StoreBifResult(4, Op1);
 		 }
 		 goto big_shift;
 	     }
-	 } else if (is_big(tmp_arg2)) {
+	 } else if (is_big(Op2)) {
 	     /*
 	      * N bsr NegativeBigNum == N bsl MAX_SMALL
 	      * N bsr PositiveBigNum == N bsl MIN_SMALL
 	      */
-	     tmp_arg2 = make_small(bignum_header_is_neg(*big_val(tmp_arg2)) ?
+	     Op2 = make_small(bignum_header_is_neg(*big_val(Op2)) ?
 				   MAX_SMALL : MIN_SMALL);
 	     goto do_bsl;
 	}
      goto badarith;
      
-     OpCase(i_bsl_jId):
+     OpCase(i_bsl_jIssd):
+         GetArg2(2, Op1, Op2);
+
  do_bsl:
-	 if (is_small(tmp_arg2)) {
-	     i = signed_val(tmp_arg2);
+	 if (is_small(Op2)) {
+	     i = signed_val(Op2);
 
-	     if (is_small(tmp_arg1)) {
+	     if (is_small(Op1)) {
 	     small_shift:
-		 ires = signed_val(tmp_arg1);
+		 ires = signed_val(Op1);
 	     
 		 if (i == 0 || ires == 0) {
-		     StoreBifResult(2, tmp_arg1);
+		     StoreBifResult(4, Op1);
 		 } else if (i < 0)  { /* Right shift */
 		     i = -i;
 		     if (i >= SMALL_BITS-1) {
-			 tmp_arg1 = (ires < 0) ? SMALL_MINUS_ONE : SMALL_ZERO;
+			 Op1 = (ires < 0) ? SMALL_MINUS_ONE : SMALL_ZERO;
 		     } else {
-			 tmp_arg1 = make_small(ires >> i);
+			 Op1 = make_small(ires >> i);
 		     }
-		     StoreBifResult(2, tmp_arg1);
+		     StoreBifResult(4, Op1);
 		 } else if (i < SMALL_BITS-1) { /* Left shift */
 		     if ((ires > 0 && ((~(Uint)0 << ((SMALL_BITS-1)-i)) & ires) == 0) ||
 			 ((~(Uint)0 << ((SMALL_BITS-1)-i)) & ~ires) == 0) {
-			 tmp_arg1 = make_small(ires << i);
-			 StoreBifResult(2, tmp_arg1);
+			 Op1 = make_small(ires << i);
+			 StoreBifResult(4, Op1);
 		     }
 		 }
-		 tmp_arg1 = small_to_big(ires, tmp_big);
+		 Op1 = small_to_big(ires, tmp_big);
 
 	     big_shift:
 		 if (i > 0) {	/* Left shift. */
-		     ires = big_size(tmp_arg1) + (i / D_EXP);
+		     ires = big_size(Op1) + (i / D_EXP);
 		 } else {	/* Right shift. */
-		     ires = big_size(tmp_arg1);
+		     ires = big_size(Op1);
 		     if (ires <= (-i / D_EXP))
 			 ires = 3; /* ??? */
 		     else
@@ -3109,14 +2970,14 @@ do {								\
 			 c_p->freason = SYSTEM_LIMIT;
 			 goto lb_Cl_error;
 		     }
-		     TestHeapPreserve(ires+1, Arg(1), tmp_arg1);
+		     TestHeapPreserve(ires+1, Arg(1), Op1);
 		     bigp = HTOP;
-		     tmp_arg1 = big_lshift(tmp_arg1, i, bigp);
-		     if (is_big(tmp_arg1)) {
+		     Op1 = big_lshift(Op1, i, bigp);
+		     if (is_big(Op1)) {
 			 HTOP += bignum_header_arity(*HTOP) + 1;
 		     }
 		     HEAP_SPACE_VERIFIED(0);
-		     if (is_nil(tmp_arg1)) {
+		     if (is_nil(Op1)) {
 			 /*
 			  * This result must have been only slight larger
 			  * than allowed since it wasn't caught by the
@@ -3126,25 +2987,25 @@ do {								\
 			 goto lb_Cl_error;
 		     }
 		     ERTS_HOLE_CHECK(c_p);
-		     StoreBifResult(2, tmp_arg1);
+		     StoreBifResult(4, Op1);
 		 }
-	     } else if (is_big(tmp_arg1)) {
+	     } else if (is_big(Op1)) {
 		 if (i == 0) {
-		     StoreBifResult(2, tmp_arg1);
+		     StoreBifResult(4, Op1);
 		 }
 		 goto big_shift;
 	     }
-	 } else if (is_big(tmp_arg2)) {
-	     if (bignum_header_is_neg(*big_val(tmp_arg2))) {
+	 } else if (is_big(Op2)) {
+	     if (bignum_header_is_neg(*big_val(Op2))) {
 		 /*
 		  * N bsl NegativeBigNum is either 0 or -1, depending on
 		  * the sign of N. Since we don't believe this case
 		  * is common, do the calculation with the minimum
 		  * amount of code.
 		  */
-		 tmp_arg2 = make_small(MIN_SMALL);
+		 Op2 = make_small(MIN_SMALL);
 		 goto do_bsl;
-	     } else if (is_small(tmp_arg1) || is_big(tmp_arg1)) {
+	     } else if (is_small(Op1) || is_big(Op1)) {
 		 /*
 		  * N bsl PositiveBigNum is too large to represent.
 		  */
@@ -3169,10 +3030,8 @@ do {								\
      } else {
 	 Uint live = Arg(2);
 	 SWAPOUT;
-	 reg[0] = r(0);
 	 reg[live] = bnot_val;
 	 bnot_val = erts_gc_bnot(c_p, reg, live);
-	 r(0) = reg[0];
 	 SWAPIN;
 	 ERTS_HOLE_CHECK(c_p);
 	 if (is_nil(bnot_val)) {
@@ -3192,7 +3051,6 @@ do {								\
      next = apply(c_p, r(0), x(1), x(2), reg);
      SWAPIN;
      if (next != NULL) {
-	 r(0) = reg[0];
 	 SET_CP(c_p, I+1);
 	 SET_I(next);
 	 Dispatch();
@@ -3207,7 +3065,6 @@ do {								\
      next = apply(c_p, r(0), x(1), x(2), reg);
      SWAPIN;
      if (next != NULL) {
-	 r(0) = reg[0];
 	 SET_CP(c_p, (BeamInstr *) E[0]);
 	 E = ADD_BYTE_OFFSET(E, Arg(0));
 	 SET_I(next);
@@ -3223,7 +3080,6 @@ do {								\
      next = apply(c_p, r(0), x(1), x(2), reg);
      SWAPIN;
      if (next != NULL) {
-	 r(0) = reg[0];
 	 SET_I(next);
 	 Dispatch();
      }
@@ -3234,12 +3090,10 @@ do {								\
  OpCase(apply_I): {
      BeamInstr *next;
 
-     reg[0] = r(0);
      SWAPOUT;
      next = fixed_apply(c_p, reg, Arg(0));
      SWAPIN;
      if (next != NULL) {
-	 r(0) = reg[0];
 	 SET_CP(c_p, I+2);
 	 SET_I(next);
 	 Dispatch();
@@ -3251,12 +3105,10 @@ do {								\
  OpCase(apply_last_IP): {
      BeamInstr *next;
 
-     reg[0] = r(0);
      SWAPOUT;
      next = fixed_apply(c_p, reg, Arg(0));
      SWAPIN;
      if (next != NULL) {
-	 r(0) = reg[0];
 	 SET_CP(c_p, (BeamInstr *) E[0]);
 	 E = ADD_BYTE_OFFSET(E, Arg(1));
 	 SET_I(next);
@@ -3273,7 +3125,6 @@ do {								\
      next = apply_fun(c_p, r(0), x(1), reg);
      SWAPIN;
      if (next != NULL) {
-	 r(0) = reg[0];
 	 SET_CP(c_p, I+1);
 	 SET_I(next);
 	 Dispatchfun();
@@ -3288,7 +3139,6 @@ do {								\
      next = apply_fun(c_p, r(0), x(1), reg);
      SWAPIN;
      if (next != NULL) {
-	 r(0) = reg[0];
 	 SET_CP(c_p, (BeamInstr *) E[0]);
 	 E = ADD_BYTE_OFFSET(E, Arg(0));
 	 SET_I(next);
@@ -3304,7 +3154,6 @@ do {								\
      next = apply_fun(c_p, r(0), x(1), reg);
      SWAPIN;
      if (next != NULL) {
-	 r(0) = reg[0];
 	 SET_I(next);
 	 Dispatchfun();
      }
@@ -3315,12 +3164,9 @@ do {								\
      BeamInstr *next;
 
      SWAPOUT;
-     reg[0] = r(0);
-
      next = call_fun(c_p, Arg(0), reg, THE_NON_VALUE);
      SWAPIN;
      if (next != NULL) {
-	 r(0) = reg[0];
 	 SET_CP(c_p, I+2);
 	 SET_I(next);
 	 Dispatchfun();
@@ -3332,11 +3178,9 @@ do {								\
      BeamInstr *next;
 
      SWAPOUT;
-     reg[0] = r(0);
      next = call_fun(c_p, Arg(0), reg, THE_NON_VALUE);
      SWAPIN;
      if (next != NULL) {
-	r(0) = reg[0];
 	SET_CP(c_p, (BeamInstr *) E[0]);
 	E = ADD_BYTE_OFFSET(E, Arg(1));
 	SET_I(next);
@@ -3421,10 +3265,9 @@ do {								\
       */
 
      argp = c_p->arg_reg;
-     for (i = c_p->arity - 1; i > 0; i--) {
+     for (i = c_p->arity - 1; i >= 0; i--) {
 	 argp[i] = reg[i];
      }
-     c_p->arg_reg[0] = r(0);
      SWAPOUT;
      c_p->i = I;
      goto do_schedule1;
@@ -3437,19 +3280,14 @@ do {								\
      Eterm* p;
      
      PreFetch(3, next);
-     GetArg2(0, element, tuple);
+     GetArg1(0, element);
+     tuple = REG_TARGET(Arg(1));
      ASSERT(is_tuple(tuple));
      p = (Eterm *) ((unsigned char *) tuple_val(tuple) + Arg(2));
      *p = element;
      NextPF(3, next);
  }
 
- OpCase(i_is_ne_exact_f):
-    if (EQ(tmp_arg1, tmp_arg2)) {
-	ClauseFail();
-    }
-    Next(1);
-
  OpCase(normal_exit): {
      SWAPOUT;
      c_p->freason = EXC_NORMAL;
@@ -3516,25 +3354,14 @@ do {								\
     {
 	Eterm badmatch_val;
 
-    OpCase(badmatch_y):
-	badmatch_val = yb(Arg(0));
-	goto do_badmatch;
-
     OpCase(badmatch_x):
 	badmatch_val = xb(Arg(0));
-	goto do_badmatch;
-
-    OpCase(badmatch_r):
-	badmatch_val = r(0);
-
-    do_badmatch:
 	c_p->fvalue = badmatch_val;
 	c_p->freason = BADMATCH;
     }
  /* Fall through here */
 
  find_func_info: {
-     reg[0] = r(0);
      SWAPOUT;
      I = handle_error(c_p, I, reg, NULL);
      goto post_error_handling;
@@ -3552,9 +3379,7 @@ do {								\
      * code[4]: Not used
      */
     SWAPOUT;
-    reg[0] = r(0);
     I = call_error_handler(c_p, I-3, reg, am_undefined_function);
-    r(0) = reg[0];
     SWAPIN;
     if (I) {
 	Goto(*I);
@@ -3563,14 +3388,12 @@ do {								\
  /* Fall through */
  OpCase(error_action_code): {
     handle_error:
-     reg[0] = r(0);
      SWAPOUT;
      I = handle_error(c_p, NULL, reg, NULL);
  post_error_handling:
      if (I == 0) {
 	 goto do_schedule;
      } else {
-	 r(0) = reg[0];
 	 ASSERT(!is_value(r(0)));
 	 if (c_p->mbuf) {
 	     erts_garbage_collect(c_p, 0, reg+1, 3);
@@ -3613,7 +3436,6 @@ do {								\
 		NifF* fp = vbf = (NifF*) I[1];
 		struct enif_environment_t env;
 		erts_pre_nif(&env, c_p, (struct erl_module_nif*)I[2]);
-		reg[0] = r(0);
 		nif_bif_result = (*fp)(&env, bif_nif_arity, reg);
 		if (env.exception_thrown)
 		    nif_bif_result = THE_NON_VALUE;
@@ -3652,7 +3474,7 @@ do {								\
 	    bif_nif_arity = I[-1];
             ASSERT(bif_nif_arity <= 4);
 	    ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
-	    reg[0] = r(0);
+	    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
 	    {
 		Eterm (*bf)(Process*, Eterm*, BeamInstr*) = vbf;
 		ASSERT(!ERTS_PROC_IS_EXITING(c_p));
@@ -3682,7 +3504,6 @@ do {								\
 		Goto(*I);
 	    } else if (c_p->freason == TRAP) {
 		SET_I(c_p->i);
-		r(0) = reg[0];
 		if (c_p->flags & F_HIBERNATE_SCHED) {
 		    c_p->flags &= ~F_HIBERNATE_SCHED;
 		    goto do_schedule;
@@ -3709,16 +3530,6 @@ do {								\
 
     OpCase(case_end_x):
 	case_end_val = xb(Arg(0));
-	goto do_case_end;
-
-    OpCase(case_end_y):
-	case_end_val = yb(Arg(0));
-	goto do_case_end;
-
-    OpCase(case_end_r):
-	case_end_val = r(0);
-
-    do_case_end:
 	c_p->fvalue = case_end_val;
 	c_p->freason = EXC_CASE_CLAUSE;
 	goto find_func_info;
@@ -3766,19 +3577,13 @@ do {								\
 	 goto do_bs_init_bits_known;
      }
 
-     OpCase(i_bs_init_bits_fail_heap_IjId): {
-	 /* tmp_arg1 was fetched by an i_fetch instruction */
-	 num_bits_term = tmp_arg1;
-	 alloc = Arg(0);
-	 I++;
+     OpCase(i_bs_init_bits_fail_heap_sIjId): {
+	 GetArg1(0, num_bits_term);
+	 alloc = Arg(1);
+	 I += 2;
 	 goto do_bs_init_bits;
      }
 
-     OpCase(i_bs_init_bits_fail_rjId): {
-	 num_bits_term = r(0);
-	 alloc = 0;
-	 goto do_bs_init_bits;
-     }
      OpCase(i_bs_init_bits_fail_yjId): {
 	 num_bits_term = yb(Arg(0));
 	 I++;
@@ -3898,52 +3703,48 @@ do {								\
  }
 
  {
-     OpCase(i_bs_init_fail_heap_IjId): {
-	 /* tmp_arg1 was fetched by an i_fetch instruction */
-	 tmp_arg2 = Arg(0);
-	 I++;
-	 goto do_bs_init;
-     }
+     Eterm BsOp1, BsOp2;
 
-     OpCase(i_bs_init_fail_rjId): {
-	 tmp_arg1 = r(0);
-	 tmp_arg2 = 0;
+     OpCase(i_bs_init_fail_heap_sIjId): {
+	 GetArg1(0, BsOp1);
+	 BsOp2 = Arg(1);
+	 I += 2;
 	 goto do_bs_init;
      }
 
      OpCase(i_bs_init_fail_yjId): {
-	 tmp_arg1 = yb(Arg(0));
-	 tmp_arg2 = 0;
+	 BsOp1 = yb(Arg(0));
+	 BsOp2 = 0;
 	 I++;
 	 goto do_bs_init;
      }
 
      OpCase(i_bs_init_fail_xjId): {
-	 tmp_arg1 = xb(Arg(0));
-	 tmp_arg2 = 0;
+	 BsOp1 = xb(Arg(0));
+	 BsOp2 = 0;
 	 I++;
      }
 	 /* FALL THROUGH */
      do_bs_init:
-         if (is_small(tmp_arg1)) {
-	     Sint size = signed_val(tmp_arg1);
+         if (is_small(BsOp1)) {
+	     Sint size = signed_val(BsOp1);
 	     if (size < 0) {
 		 goto badarg;
 	     }
-	     tmp_arg1 = (Eterm) size;
+	     BsOp1 = (Eterm) size;
 	 } else {
 	     Uint bytes;
 
-	     if (!term_to_Uint(tmp_arg1, &bytes)) {
+	     if (!term_to_Uint(BsOp1, &bytes)) {
 		 c_p->freason = bytes;
 		 goto lb_Cl_error;
 	     }
 	     if ((bytes >> (8*sizeof(Uint)-3)) != 0) {
 		 goto system_limit;
 	     }
-	     tmp_arg1 = (Eterm) bytes;
+	     BsOp1 = (Eterm) bytes;
 	 }
-	 if (tmp_arg1 <= ERL_ONHEAP_BIN_LIMIT) {
+	 if (BsOp1 <= ERL_ONHEAP_BIN_LIMIT) {
 	     goto do_heap_bin_alloc;
 	 } else {
 	     goto do_proc_bin_alloc;
@@ -3951,15 +3752,15 @@ do {								\
 
 
      OpCase(i_bs_init_heap_IIId): {
-	 tmp_arg1 = Arg(0);
-	 tmp_arg2 = Arg(1);
+	 BsOp1 = Arg(0);
+	 BsOp2 = Arg(1);
 	 I++;
 	 goto do_proc_bin_alloc;
      }
 
      OpCase(i_bs_init_IId): {
-	 tmp_arg1 = Arg(0);
-	 tmp_arg2 = 0;
+	 BsOp1 = Arg(0);
+	 BsOp2 = 0;
      }
      /* FALL THROUGH */
      do_proc_bin_alloc: {
@@ -3968,13 +3769,13 @@ do {								\
 
 	 erts_bin_offset = 0;
 	 erts_writable_bin = 0;
-	 TestBinVHeap(tmp_arg1 / sizeof(Eterm), 
-	 	      tmp_arg2 + PROC_BIN_SIZE + ERL_SUB_BIN_SIZE, Arg(1));
+	 TestBinVHeap(BsOp1 / sizeof(Eterm),
+		      BsOp2 + PROC_BIN_SIZE + ERL_SUB_BIN_SIZE, Arg(1));
 
 	 /*
 	  * Allocate the binary struct itself.
 	  */
-	 bptr = erts_bin_nrml_alloc(tmp_arg1);
+	 bptr = erts_bin_nrml_alloc(BsOp1);
 	 erts_refc_init(&bptr->refc, 1);
 	 erts_current_bin = (byte *) bptr->orig_bytes;
 
@@ -3984,28 +3785,28 @@ do {								\
 	 pb = (ProcBin *) HTOP;
 	 HTOP += PROC_BIN_SIZE;
 	 pb->thing_word = HEADER_PROC_BIN;
-	 pb->size = tmp_arg1;
+	 pb->size = BsOp1;
 	 pb->next = MSO(c_p).first;
 	 MSO(c_p).first = (struct erl_off_heap_header*) pb;
 	 pb->val = bptr;
 	 pb->bytes = (byte*) bptr->orig_bytes;
 	 pb->flags = 0;
 	 
-	 OH_OVERHEAD(&(MSO(c_p)), tmp_arg1 / sizeof(Eterm));
+	 OH_OVERHEAD(&(MSO(c_p)), BsOp1 / sizeof(Eterm));
 
 	 StoreBifResult(2, make_binary(pb));
      }
 
      OpCase(i_bs_init_heap_bin_heap_IIId): {
-	 tmp_arg1 = Arg(0);
-	 tmp_arg2 = Arg(1);
+	 BsOp1 = Arg(0);
+	 BsOp2 = Arg(1);
 	 I++;
 	 goto do_heap_bin_alloc;
      }
 
      OpCase(i_bs_init_heap_bin_IId): {
-	 tmp_arg1 = Arg(0);
-	 tmp_arg2 = 0;
+	 BsOp1 = Arg(0);
+	 BsOp2 = 0;
      }
      /* Fall through */
      do_heap_bin_alloc:
@@ -4013,33 +3814,36 @@ do {								\
 	     ErlHeapBin* hb;
 	     Uint bin_need;
 
-	     bin_need = heap_bin_size(tmp_arg1);
+	     bin_need = heap_bin_size(BsOp1);
 	     erts_bin_offset = 0;
 	     erts_writable_bin = 0;
-	     TestHeap(bin_need+tmp_arg2+ERL_SUB_BIN_SIZE, Arg(1));
+	     TestHeap(bin_need+BsOp2+ERL_SUB_BIN_SIZE, Arg(1));
 	     hb = (ErlHeapBin *) HTOP;
 	     HTOP += bin_need;
-	     hb->thing_word = header_heap_bin(tmp_arg1);
-	     hb->size = tmp_arg1;
+	     hb->thing_word = header_heap_bin(BsOp1);
+	     hb->size = BsOp1;
 	     erts_current_bin = (byte *) hb->data;
-	     tmp_arg1 = make_binary(hb);
-	     StoreBifResult(2, tmp_arg1);
+	     BsOp1 = make_binary(hb);
+	     StoreBifResult(2, BsOp1);
 	 }
  }
 
- OpCase(i_bs_add_jId): {
-     Uint Unit = Arg(1);
-     if (is_both_small(tmp_arg1, tmp_arg2)) {
-	 Sint Arg1 = signed_val(tmp_arg1);
-	 Sint Arg2 = signed_val(tmp_arg2);
+ OpCase(bs_add_jssId): {
+     Eterm Op1, Op2;
+     Uint Unit = Arg(3);
+
+     GetArg2(1, Op1, Op2);
+     if (is_both_small(Op1, Op2)) {
+	 Sint Arg1 = signed_val(Op1);
+	 Sint Arg2 = signed_val(Op2);
 
 	 if (Arg1 >= 0 && Arg2 >= 0) {
-	     BsSafeMul(Arg2, Unit, goto system_limit, tmp_arg1);
-	     tmp_arg1 += Arg1;
+	     BsSafeMul(Arg2, Unit, goto system_limit, Op1);
+	     Op1 += Arg1;
 
 	 store_bs_add_result:
-	     if (MY_IS_SSMALL((Sint) tmp_arg1)) {
-		 tmp_arg1 = make_small(tmp_arg1);
+	     if (MY_IS_SSMALL((Sint) Op1)) {
+		 Op1 = make_small(Op1);
 	     } else {
 		 /*
 		  * May generate a heap fragment, but in this
@@ -4051,10 +3855,10 @@ do {								\
 		  * references (such as the heap).
 		  */
 		 SWAPOUT;
-		 tmp_arg1 = erts_make_integer(tmp_arg1, c_p);
+		 Op1 = erts_make_integer(Op1, c_p);
 		 HTOP = HEAP_TOP(c_p);
 	     }
-	     StoreBifResult(2, tmp_arg1);
+	     StoreBifResult(4, Op1);
 	 }
 	 goto badarg;
      } else {
@@ -4077,16 +3881,16 @@ do {								\
 	  * an Uint, the reason is SYSTEM_LIMIT.
 	  */
 
-	 if (!term_to_Uint(tmp_arg1, &a)) {
+	 if (!term_to_Uint(Op1, &a)) {
 	     if (a == BADARG) {
 		 goto badarg;
 	     }
-	     if (!term_to_Uint(tmp_arg2, &b)) {
+	     if (!term_to_Uint(Op2, &b)) {
 		 c_p->freason = b;
 		 goto lb_Cl_error;
 	     }
 	     goto system_limit;
-	 } else if (!term_to_Uint(tmp_arg2, &b)) {
+	 } else if (!term_to_Uint(Op2, &b)) {
 	     c_p->freason = b;
 	     goto lb_Cl_error;
 	 }
@@ -4096,8 +3900,8 @@ do {								\
 	  */
 	 
 	 BsSafeMul(b, Unit, goto system_limit, c);
-	 tmp_arg1 = a + c;
-	 if (tmp_arg1 < a) {
+	 Op1 = a + c;
+	 if (Op1 < a) {
 	     /*
 	      * If the result is less than one of the
 	      * arguments, there must have been an overflow.
@@ -4119,48 +3923,43 @@ do {								\
     }
 
  /*
-  * tmp_arg1 = Number of bytes to build
-  * tmp_arg2 = Source binary
-  * Operands: Fail ExtraHeap Live Unit Dst
+  * x(SCRATCH_X_REG);
+  * Operands: Fail ExtraHeap Live Unit Size Dst
   */
 
- OpCase(i_bs_append_jIIId): {
+ OpCase(i_bs_append_jIIIsd): {
      Uint live = Arg(2);
      Uint res;
+     Eterm Size;
 
+     GetArg1(4, Size);
      SWAPOUT;
-     reg[0] = r(0);
-     reg[live] = tmp_arg2;
-     res = erts_bs_append(c_p, reg, live, tmp_arg1, Arg(1), Arg(3));
-     r(0) = reg[0];
+     reg[live] = x(SCRATCH_X_REG);
+     res = erts_bs_append(c_p, reg, live, Size, Arg(1), Arg(3));
      SWAPIN;
      if (is_non_value(res)) {
 	 /* c_p->freason is already set (may be either BADARG or SYSTEM_LIMIT). */
 	 goto lb_Cl_error;
      }
-     StoreBifResult(4, res);
+     StoreBifResult(5, res);
  }
 
  /*
-  * tmp_arg1 = Number of bytes to build
-  * tmp_arg2 = Source binary
-  * Operands: Fail Unit Dst
+  * Operands: Fail Size Src Unit Dst
   */
- OpCase(i_bs_private_append_jId): {
+ OpCase(i_bs_private_append_jIssd): {
      Eterm res;
+     Eterm Size, Src;
 
-     res = erts_bs_private_append(c_p, tmp_arg2, tmp_arg1, Arg(1));
+     GetArg2(2, Size, Src);
+     res = erts_bs_private_append(c_p, Src, Size, Arg(1));
      if (is_non_value(res)) {
 	 /* c_p->freason is already set (may be either BADARG or SYSTEM_LIMIT). */
 	 goto lb_Cl_error;
      }
-     StoreBifResult(2, res);
+     StoreBifResult(4, res);
  }
 
- /*
-  * tmp_arg1 = Initial size of writable binary
-  * Operands: Live Dst
-  */
  OpCase(bs_init_writable): {
      SWAPOUT;
      r(0) = erts_bs_init_writable(c_p, r(0));
@@ -4255,26 +4054,29 @@ do {								\
 
  /*
   * Only used for validating a value matched out. 
-  *
-  * tmp_arg1 = Integer to validate
-  * tmp_arg2 = Match context
   */
- OpCase(i_bs_validate_unicode_retract_j): {
-     /*
-      * There is no need to untag the integer, but it IS necessary
-      * to make sure it is small (a bignum pointer could fall in
-      * the valid range).
-      */
-     if (is_not_small(tmp_arg1) || tmp_arg1 > make_small(0x10FFFFUL) ||
-	 (make_small(0xD800UL) <= tmp_arg1 &&
-	  tmp_arg1 <= make_small(0xDFFFUL))) {
-	 ErlBinMatchBuffer *mb = ms_matchbuffer(tmp_arg2);
+ OpCase(i_bs_validate_unicode_retract_jss): {
+	Eterm i; 		/* Integer to validate */
 
-	 mb->offset -= 32;
-	 goto badarg;
-     }
-     Next(1);
- }
+	/*
+	 * There is no need to untag the integer, but it IS necessary
+	 * to make sure it is small (a bignum pointer could fall in
+	 * the valid range).
+	 */
+
+	GetArg1(1, i);
+	if (is_not_small(i) || i > make_small(0x10FFFFUL) ||
+	    (make_small(0xD800UL) <= i && i <= make_small(0xDFFFUL))) {
+	    Eterm ms;		/* Match context */
+	    ErlBinMatchBuffer* mb;
+
+	    GetArg1(2, ms);
+	    mb = ms_matchbuffer(ms);
+	    mb->offset -= 32;
+	    goto badarg;
+	}
+	Next(3);
+    }
 
  /*
   * Matching of binaries.
@@ -4286,9 +4088,6 @@ do {								\
      Uint slots;
      Eterm context;
 
-     OpCase(i_bs_start_match2_rfIId): {
-	 context = r(0);
-
      do_start_match:
 	 slots = Arg(2);
 	 if (!is_boxed(context)) {
@@ -4335,7 +4134,7 @@ do {								\
 	     ClauseFail();
 	 }
 	 NextPF(4, next);
-     }
+
      OpCase(i_bs_start_match2_xfIId): {
 	 context = xb(Arg(0));
 	 I++;
@@ -4348,18 +4147,6 @@ do {								\
      }
  }
 
- OpCase(bs_test_zero_tail2_fr): {
-     BeamInstr *next;
-     ErlBinMatchBuffer *_mb;
-     
-     PreFetch(1, next);
-     _mb = (ErlBinMatchBuffer*) ms_matchbuffer(r(0));
-     if (_mb->size != _mb->offset) {
-	 ClauseFail();
-     }
-     NextPF(1, next);
- }
-
  OpCase(bs_test_zero_tail2_fx): {
      BeamInstr *next;
      ErlBinMatchBuffer *_mb;
@@ -4372,16 +4159,6 @@ do {								\
      NextPF(2, next);
  }
 
- OpCase(bs_test_tail_imm2_frI): {
-     BeamInstr *next;
-     ErlBinMatchBuffer *_mb;
-     PreFetch(2, next);
-     _mb = ms_matchbuffer(r(0));
-     if (_mb->size - _mb->offset != Arg(1)) {
-	 ClauseFail();
-     }
-     NextPF(2, next);
- }
  OpCase(bs_test_tail_imm2_fxI): {
      BeamInstr *next;
      ErlBinMatchBuffer *_mb;
@@ -4393,16 +4170,6 @@ do {								\
      NextPF(3, next);
  }
 
- OpCase(bs_test_unit_frI): {
-     BeamInstr *next;
-     ErlBinMatchBuffer *_mb;
-     PreFetch(2, next);
-     _mb = ms_matchbuffer(r(0));
-     if ((_mb->size - _mb->offset) % Arg(1)) {
-	 ClauseFail();
-     }
-     NextPF(2, next);
- }
  OpCase(bs_test_unit_fxI): {
      BeamInstr *next;
      ErlBinMatchBuffer *_mb;
@@ -4414,16 +4181,6 @@ do {								\
      NextPF(3, next);
  }
 
- OpCase(bs_test_unit8_fr): {
-     BeamInstr *next;
-     ErlBinMatchBuffer *_mb;
-     PreFetch(1, next);
-     _mb = ms_matchbuffer(r(0));
-     if ((_mb->size - _mb->offset) & 7) {
-	 ClauseFail();
-     }
-     NextPF(1, next);
- }
  OpCase(bs_test_unit8_fx): {
      BeamInstr *next;
      ErlBinMatchBuffer *_mb;
@@ -4438,19 +4195,11 @@ do {								\
  {
      Eterm bs_get_integer8_context;
 
- OpCase(i_bs_get_integer_8_rfd): {
-	 bs_get_integer8_context = r(0);
-	 goto do_bs_get_integer_8;
-     }
-
  OpCase(i_bs_get_integer_8_xfd): {
-	 bs_get_integer8_context = xb(Arg(0));
-	 I++;
-     }
-
- do_bs_get_integer_8: {
 	 ErlBinMatchBuffer *_mb;
 	 Eterm _result;
+	 bs_get_integer8_context = xb(Arg(0));
+	 I++;
 	 _mb = ms_matchbuffer(bs_get_integer8_context);
 	 if (_mb->size - _mb->offset < 8) {
 	     ClauseFail();
@@ -4468,15 +4217,10 @@ do {								\
  {
      Eterm bs_get_integer_16_context;
 
- OpCase(i_bs_get_integer_16_rfd):
-     bs_get_integer_16_context = r(0);
-     goto do_bs_get_integer_16;
-
  OpCase(i_bs_get_integer_16_xfd):
      bs_get_integer_16_context = xb(Arg(0));
      I++;
 
- do_bs_get_integer_16:
      {
 	 ErlBinMatchBuffer *_mb;
 	 Eterm _result;
@@ -4497,17 +4241,10 @@ do {								\
  {
      Eterm bs_get_integer_32_context;
 
- OpCase(i_bs_get_integer_32_rfId):
-     bs_get_integer_32_context = r(0);
-     goto do_bs_get_integer_32;
-
-     
  OpCase(i_bs_get_integer_32_xfId):
      bs_get_integer_32_context = xb(Arg(0));
      I++;
 
-
- do_bs_get_integer_32:
      {
 	 ErlBinMatchBuffer *_mb;
 	 Uint32 _integer;
@@ -4536,103 +4273,82 @@ do {								\
      }
  }
 
- /* Operands: Size Live Fail Flags Dst */
- OpCase(i_bs_get_integer_imm_rIIfId): {
-     tmp_arg1 = r(0);
-     /* Operands: Size Live Fail Flags Dst */
-     goto do_bs_get_integer_imm_test_heap;
- }
+ {
+     Eterm Ms, Sz;
 
- /* Operands: x(Reg) Size Live Fail Flags Dst */
+     /* Operands: x(Reg) Size Live Fail Flags Dst */
  OpCase(i_bs_get_integer_imm_xIIfId): {
-     tmp_arg1 = xb(Arg(0));
-     I++;
-     /* Operands: Size Live Fail Flags Dst */
-     goto do_bs_get_integer_imm_test_heap;
- }
-
- /*
-  * tmp_arg1 = match context
-  * Operands: Size Live Fail Flags Dst
-  */
- do_bs_get_integer_imm_test_heap: {
-     Uint wordsneeded;
-     tmp_arg2 = Arg(0);
-     wordsneeded = 1+WSIZE(NBYTES(tmp_arg2));
-     TestHeapPreserve(wordsneeded, Arg(1), tmp_arg1);
-     I += 2;
-     /* Operands: Fail Flags Dst */
-     goto do_bs_get_integer_imm;
- }
-
- /* Operands: Size Fail Flags Dst */
- OpCase(i_bs_get_integer_small_imm_rIfId): {
-     tmp_arg1 = r(0);
-     tmp_arg2 = Arg(0);
-     I++;
-     /* Operands: Fail Flags Dst */
-     goto do_bs_get_integer_imm;
- }
+	 Uint wordsneeded;
+	 Ms = xb(Arg(0));
+	 Sz = Arg(1);
+	 wordsneeded = 1+WSIZE(NBYTES(Sz));
+	 TestHeapPreserve(wordsneeded, Arg(2), Ms);
+	 I += 3;
+	 /* Operands: Fail Flags Dst */
+	 goto do_bs_get_integer_imm;
+     }
 
- /* Operands: x(Reg) Size Fail Flags Dst */
+     /* Operands: x(Reg) Size Fail Flags Dst */
  OpCase(i_bs_get_integer_small_imm_xIfId): {
-     tmp_arg1 = xb(Arg(0));
-     tmp_arg2 = Arg(1);
-     I += 2;
-     /* Operands: Fail Flags Dst */
-     goto do_bs_get_integer_imm;
- }
+	 Ms = xb(Arg(0));
+	 Sz = Arg(1);
+	 I += 2;
+	 /* Operands: Fail Flags Dst */
+	 goto do_bs_get_integer_imm;
+     }
  
- /*
-  * tmp_arg1 = match context
-  * tmp_arg2 = size of field
-  * Operands: Fail Flags Dst
-  */
+     /*
+      * Ms = match context
+      * Sz = size of field
+      * Operands: Fail Flags Dst
+      */
  do_bs_get_integer_imm: {
-     ErlBinMatchBuffer* mb;
-     Eterm result;
+	 ErlBinMatchBuffer* mb;
+	 Eterm result;
 
-     mb = ms_matchbuffer(tmp_arg1);
-     LIGHT_SWAPOUT;
-     result = erts_bs_get_integer_2(c_p, tmp_arg2, Arg(1), mb);
-     LIGHT_SWAPIN;
-     HEAP_SPACE_VERIFIED(0);
-     if (is_non_value(result)) {
-	 ClauseFail();
+	 mb = ms_matchbuffer(Ms);
+	 LIGHT_SWAPOUT;
+	 result = erts_bs_get_integer_2(c_p, Sz, Arg(1), mb);
+	 LIGHT_SWAPIN;
+	 HEAP_SPACE_VERIFIED(0);
+	 if (is_non_value(result)) {
+	     ClauseFail();
+	 }
+	 StoreBifResult(2, result);
      }
-     StoreBifResult(2, result);
  }
 
  /*
-  * tmp_arg1 = Match context
-  * tmp_arg2 = Size field
-  * Operands: Fail Live FlagsAndUnit Dst
+  * Operands: Fail Live FlagsAndUnit Ms Sz Dst
   */
- OpCase(i_bs_get_integer_fIId): {
+ OpCase(i_bs_get_integer_fIIssd): {
      Uint flags;
      Uint size;
+     Eterm Ms;
+     Eterm Sz;
      ErlBinMatchBuffer* mb;
      Eterm result;
 
      flags = Arg(2);
-     BsGetFieldSize(tmp_arg2, (flags >> 3), ClauseFail(), size);
+     GetArg2(3, Ms, Sz);
+     BsGetFieldSize(Sz, (flags >> 3), ClauseFail(), size);
      if (size >= SMALL_BITS) {
 	 Uint wordsneeded;
-	 /* check bits size before potential gc.
+	 /* Check bits size before potential gc.
 	  * We do not want a gc and then realize we don't need
-	  * the allocated space (i.e. if the op fails)
+	  * the allocated space (i.e. if the op fails).
 	  *
-	  * remember to reacquire the matchbuffer after gc.
+	  * Remember to re-acquire the matchbuffer after gc.
 	  */
 
-	 mb = ms_matchbuffer(tmp_arg1);
+	 mb = ms_matchbuffer(Ms);
 	 if (mb->size - mb->offset < size) {
 	     ClauseFail();
 	 }
 	 wordsneeded = 1+WSIZE(NBYTES((Uint) size));
-	 TestHeapPreserve(wordsneeded, Arg(1), tmp_arg1);
+	 TestHeapPreserve(wordsneeded, Arg(1), Ms);
      }
-     mb = ms_matchbuffer(tmp_arg1);
+     mb = ms_matchbuffer(Ms);
      LIGHT_SWAPOUT;
      result = erts_bs_get_integer_2(c_p, size, flags, mb);
      LIGHT_SWAPIN;
@@ -4640,18 +4356,13 @@ do {								\
      if (is_non_value(result)) {
 	 ClauseFail();
      }
-     StoreBifResult(3, result);
+     StoreBifResult(5, result);
  }
 
  {
      Eterm get_utf8_context;
 
      /* Operands: MatchContext Fail Dst */
- OpCase(i_bs_get_utf8_rfd): {
-	 get_utf8_context = r(0);
-	 goto do_bs_get_utf8;
-     }
-
  OpCase(i_bs_get_utf8_xfd): {
 	 get_utf8_context = xb(Arg(0));
 	 I++;
@@ -4662,7 +4373,7 @@ do {								\
       * Operands: Fail Dst
       */
 
- do_bs_get_utf8: {
+     {
 	 Eterm result = erts_bs_get_utf8(ms_matchbuffer(get_utf8_context));
 	 if (is_non_value(result)) {
 	     ClauseFail();
@@ -4675,12 +4386,7 @@ do {								\
      Eterm get_utf16_context;
 
      /* Operands: MatchContext Fail Flags Dst */
- OpCase(i_bs_get_utf16_rfId): {
-	 get_utf16_context = r(0);
-	 goto do_bs_get_utf16;
-     }
-
- OpCase(i_bs_get_utf16_xfId): {
+     OpCase(i_bs_get_utf16_xfId): {
 	 get_utf16_context = xb(Arg(0));
 	 I++;
      }
@@ -4689,7 +4395,7 @@ do {								\
       * get_utf16_context = match_context
       * Operands: Fail Flags Dst
       */
- do_bs_get_utf16: {
+     {
 	 Eterm result = erts_bs_get_utf16(ms_matchbuffer(get_utf16_context),
 					  Arg(1));
 	 if (is_non_value(result)) {
@@ -4708,26 +4414,10 @@ do {								\
      Uint orig;
      Uint hole_size;
 
-     OpCase(bs_context_to_binary_r): {
-	 context_to_binary_context = x0;
-	 I -= 2;
-	 goto do_context_to_binary;
-     }
-
-     /* Unfortunately, inlining can generate this instruction. */
-     OpCase(bs_context_to_binary_y): {
-	 context_to_binary_context = yb(Arg(0));
-	 goto do_context_to_binary0;
-     }
-
-     OpCase(bs_context_to_binary_x): {
+     OpCase(bs_context_to_binary_x):
 	 context_to_binary_context = xb(Arg(0));
-     
-     do_context_to_binary0:
 	 I--;
-     }
 
- do_context_to_binary:
      if (is_boxed(context_to_binary_context) &&
 	 header_is_bin_matchstate(*boxed_val(context_to_binary_context))) {
 	 ErlBinMatchState* ms;
@@ -4739,17 +4429,11 @@ do {								\
      }
      Next(2);
 
-     OpCase(i_bs_get_binary_all_reuse_rfI): {
-	 context_to_binary_context = x0;
-	 goto do_bs_get_binary_all_reuse;
-     }
-
      OpCase(i_bs_get_binary_all_reuse_xfI): {
 	 context_to_binary_context = xb(Arg(0));
 	 I++;
      }
 
- do_bs_get_binary_all_reuse:
      mb = ms_matchbuffer(context_to_binary_context);
      size = mb->size - mb->offset;
      if (size % Arg(1) != 0) {
@@ -4777,16 +4461,11 @@ do {								\
  {
      Eterm match_string_context;
 
-     OpCase(i_bs_match_string_rfII): {
-	 match_string_context = r(0);
-	 goto do_bs_match_string;
-     }
      OpCase(i_bs_match_string_xfII): {
 	 match_string_context = xb(Arg(0));
 	 I++;
      }
 
- do_bs_match_string:
      {
 	 BeamInstr *next;
 	 byte* bytes;
@@ -4814,14 +4493,6 @@ do {								\
      }
  }
 
- OpCase(i_bs_save2_rI): {
-     BeamInstr *next;
-     ErlBinMatchState *_ms;
-     PreFetch(1, next);
-     _ms = (ErlBinMatchState*) boxed_val((Eterm) r(0));
-     _ms->save_offset[Arg(0)] = _ms->mb.offset;
-     NextPF(1, next);
- }
  OpCase(i_bs_save2_xI): {
      BeamInstr *next;
      ErlBinMatchState *_ms;
@@ -4831,14 +4502,6 @@ do {								\
      NextPF(2, next);
  }
 
- OpCase(i_bs_restore2_rI): {
-     BeamInstr *next;
-     ErlBinMatchState *_ms;
-     PreFetch(1, next);
-     _ms = (ErlBinMatchState*) boxed_val((Eterm) r(0));
-     _ms->mb.offset = _ms->save_offset[Arg(0)];
-     NextPF(1, next);
- }
  OpCase(i_bs_restore2_xI): {
      BeamInstr *next;
      ErlBinMatchState *_ms;
@@ -4886,9 +4549,7 @@ do {								\
      BeamInstr real_I;
      ASSERT(I[-5] == (BeamInstr) BeamOp(op_i_func_info_IaaI));
      SWAPOUT;
-     reg[0] = r(0);
      real_I = erts_generic_breakpoint(c_p, I, reg);
-     r(0) = reg[0];
      SWAPIN;
      ASSERT(VALID_INSTR(real_I));
      Goto(real_I);
@@ -4948,7 +4609,7 @@ do {								\
      BeamInstr *next;
 
      PreFetch(2, next);
-     GetR(0, targ1);
+     targ1 = REG_TARGET(Arg(0));
      /* Arg(0) == HEADER_FLONUM */
      GET_DOUBLE(targ1, *(FloatDef*)ADD_BYTE_OFFSET(freg, fr));
      NextPF(2, next);
@@ -4968,7 +4629,7 @@ do {								\
      Eterm fr = Arg(1);
      BeamInstr *next;
 
-     GetR(0, targ1);
+     targ1 = REG_TARGET(Arg(0));
      PreFetch(2, next);
      if (is_small(targ1)) {
 	 fb(fr) = (double) signed_val(targ1);
@@ -5064,7 +4725,12 @@ do {								\
 
 #ifdef HIPE
  {
-     unsigned cmd;
+#define HIPE_MODE_SWITCH(Cmd)			\
+     SWAPOUT;					\
+     c_p->fcalls = FCALLS;			\
+     c_p->def_arg_reg[4] = -neg_o_reds;		\
+     c_p = hipe_mode_switch(c_p, Cmd, reg);     \
+     goto L_post_hipe_mode_switch
 
      OpCase(hipe_trap_call): {
 	 /*
@@ -5078,52 +4744,45 @@ do {								\
 	  */
 	 ASSERT(I[-5] == (Uint) OpCode(i_func_info_IaaI));
 	 c_p->hipe.u.ncallee = (void(*)(void)) I[-4];
-	 cmd = HIPE_MODE_SWITCH_CMD_CALL | (I[-1] << 8);
 	 ++hipe_trap_count;
-	 goto L_hipe_mode_switch;
+	 HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_CALL | (I[-1] << 8));
      }
      OpCase(hipe_trap_call_closure): {
        ASSERT(I[-5] == (Uint) OpCode(i_func_info_IaaI));
        c_p->hipe.u.ncallee = (void(*)(void)) I[-4];
-       cmd = HIPE_MODE_SWITCH_CMD_CALL_CLOSURE | (I[-1] << 8);
        ++hipe_trap_count;
-       goto L_hipe_mode_switch;
+       HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_CALL_CLOSURE | (I[-1] << 8));
      }
      OpCase(hipe_trap_return): {
-	 cmd = HIPE_MODE_SWITCH_CMD_RETURN;
-	 goto L_hipe_mode_switch;
+	 HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_RETURN);
      }
      OpCase(hipe_trap_throw): {
-	 cmd = HIPE_MODE_SWITCH_CMD_THROW;
-	 goto L_hipe_mode_switch;
+	 HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_THROW);
      }
      OpCase(hipe_trap_resume): {
-	 cmd = HIPE_MODE_SWITCH_CMD_RESUME;
-	 goto L_hipe_mode_switch;
+	 HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_RESUME);
      }
- L_hipe_mode_switch:
-     /* XXX: this abuse of def_arg_reg[] is horrid! */
-     SWAPOUT;
-     c_p->fcalls = FCALLS;
-     c_p->def_arg_reg[4] = -neg_o_reds;
-     reg[0] = r(0);
-     c_p = hipe_mode_switch(c_p, cmd, reg);
+#undef HIPE_MODE_SWITCH
+
+ L_post_hipe_mode_switch:
      reg = ERTS_PROC_GET_SCHDATA(c_p)->x_reg_array;
      freg = ERTS_PROC_GET_SCHDATA(c_p)->f_reg_array;
      ERL_BITS_RELOAD_STATEP(c_p);
+     /* XXX: this abuse of def_arg_reg[] is horrid! */
      neg_o_reds = -c_p->def_arg_reg[4];
      FCALLS = c_p->fcalls;
      SWAPIN;
      switch( c_p->def_arg_reg[3] ) {
        case HIPE_MODE_SWITCH_RES_RETURN:
 	 ASSERT(is_value(reg[0]));
-	 MoveReturn(reg[0], r(0));
+	 SET_I(c_p->cp);
+	 c_p->cp = 0;
+	 Goto(*I);
        case HIPE_MODE_SWITCH_RES_CALL_EXPORTED:
 	 c_p->i = c_p->hipe.u.callee_exp->addressv[erts_active_code_ix()];
 	 /*fall through*/
        case HIPE_MODE_SWITCH_RES_CALL_BEAM:
 	 SET_I(c_p->i);
-	 r(0) = reg[0];
 	 Dispatch();
        case HIPE_MODE_SWITCH_RES_CALL_CLOSURE:
 	 /* This can be used to call any function value, but currently it's
@@ -5134,7 +4793,6 @@ do {								\
 	     next = call_fun(c_p, c_p->arity - 1, reg, THE_NON_VALUE);
 	     SWAPIN;
 	     if (next != NULL) {
-		 r(0) = reg[0];
 		 SET_I(next);
 		 Dispatchfun();
 	     }
@@ -5194,9 +4852,7 @@ do {								\
 
  OpCase(i_debug_breakpoint): {
      SWAPOUT;
-     reg[0] = r(0);
      I = call_error_handler(c_p, I-3, reg, am_breakpoint);
-     r(0) = reg[0];
      SWAPIN;
      if (I) {
 	 Goto(*I);
@@ -6587,23 +6243,20 @@ static Eterm get_map_element_hash(Eterm map, Eterm key, Uint32 hx)
     return vs ? *vs : THE_NON_VALUE;
 }
 
-#define GET_TERM(term, dest)					\
-do {								\
-    Eterm src = (Eterm)(term);					\
-    switch (src & _TAG_IMMED1_MASK) {				\
-    case (R_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER:	\
-	dest = x(0);						\
-	break;							\
-    case (X_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER:	\
-	dest = x(src >> _TAG_IMMED1_SIZE);			\
-	break;							\
-    case (Y_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER:	\
-	dest = y(src >> _TAG_IMMED1_SIZE);			\
-	break;							\
-    default:							\
-	dest = src;						\
-	break;							\
-    }								\
+#define GET_TERM(term, dest)			\
+do {						\
+    Eterm src = (Eterm)(term);			\
+    switch (loader_tag(src)) {			\
+    case LOADER_X_REG:				\
+        dest = x(loader_x_reg_index(src));	\
+	break;					\
+    case LOADER_Y_REG:				\
+        dest = y(loader_y_reg_index(src));	\
+	break;					\
+    default:					\
+	dest = src;				\
+	break;					\
+    }						\
 } while(0)
 
 
diff --git a/erts/emulator/beam/beam_load.c b/erts/emulator/beam/beam_load.c
index 10baab506a..a9d47eb7b0 100644
--- a/erts/emulator/beam/beam_load.c
+++ b/erts/emulator/beam/beam_load.c
@@ -1847,9 +1847,7 @@ load_code(LoaderState* stp)
 	    case TAG_o:
 		break;
 	    case TAG_x:
-		if (last_op->a[arg].val == 0) {
-		    last_op->a[arg].type = TAG_r;
-		} else if (last_op->a[arg].val >= MAX_REG) {
+		if (last_op->a[arg].val >= MAX_REG) {
 		    LoadError1(stp, "invalid x register number: %u",
 			       last_op->a[arg].val);
 		}
@@ -2055,7 +2053,42 @@ load_code(LoaderState* stp)
 		if (((opc[specific].mask[0] & mask[0]) == mask[0]) &&
 		    ((opc[specific].mask[1] & mask[1]) == mask[1]) &&
 		    ((opc[specific].mask[2] & mask[2]) == mask[2])) {
-		    break;
+
+		    if (!opc[specific].involves_r) {
+			break;	/* No complications - match */
+		    }
+
+		    /*
+		     * The specific operation uses the 'r' operand,
+		     * which is shorthand for x(0). Now things
+		     * get complicated. First we must check whether
+		     * all operands that should be of type 'r' use
+		     * x(0) (as opposed to some other X register).
+		     */
+		    for (arg = 0; arg < arity; arg++) {
+			if (opc[specific].involves_r & (1 << arg) &&
+			    tmp_op->a[arg].type == TAG_x) {
+			    if (tmp_op->a[arg].val != 0) {
+				break; /* Other X register than 0 */
+			    }
+			}
+		    }
+
+		    if (arg == arity) {
+			/*
+			 * All 'r' operands use x(0) in the generic
+			 * operation. That means a match. Now we
+			 * will need to rewrite the generic instruction
+			 * to actually use 'r' instead of 'x(0)'.
+			 */
+			for (arg = 0; arg < arity; arg++) {
+			    if (opc[specific].involves_r & (1 << arg) &&
+				tmp_op->a[arg].type == TAG_x) {
+				tmp_op->a[arg].type = TAG_r;
+			    }
+			}
+			break;	/* Match */
+		    }
 		}
 		specific++;
 	    }
@@ -2166,14 +2199,11 @@ load_code(LoaderState* stp)
 		break;
 	    case 's':	/* Any source (tagged constant or register) */
 		switch (tag) {
-		case TAG_r:
-		    code[ci++] = make_rreg();
-		    break;
 		case TAG_x:
-		    code[ci++] = make_xreg(tmp_op->a[arg].val);
+		    code[ci++] = make_loader_x_reg(tmp_op->a[arg].val);
 		    break;
 		case TAG_y:
-		    code[ci++] = make_yreg(tmp_op->a[arg].val);
+		    code[ci++] = make_loader_y_reg(tmp_op->a[arg].val);
 		    break;
 		case TAG_i:
 		    code[ci++] = (BeamInstr) make_small((Uint)tmp_op->a[arg].val);
@@ -2184,6 +2214,10 @@ load_code(LoaderState* stp)
 		case TAG_n:
 		    code[ci++] = NIL;
 		    break;
+		case TAG_q:
+		    new_literal_patch(stp, ci);
+		    code[ci++] = tmp_op->a[arg].val;
+		    break;
 		default:
 		    LoadError1(stp, "bad tag %d for general source",
 			       tmp_op->a[arg].type);
@@ -2192,14 +2226,11 @@ load_code(LoaderState* stp)
 		break;
 	    case 'd':	/* Destination (x(0), x(N), y(N) */
 		switch (tag) {
-		case TAG_r:
-		    code[ci++] = make_rreg();
-		    break;
 		case TAG_x:
-		    code[ci++] = make_xreg(tmp_op->a[arg].val);
+		    code[ci++] = tmp_op->a[arg].val * sizeof(Eterm);
 		    break;
 		case TAG_y:
-		    code[ci++] = make_yreg(tmp_op->a[arg].val);
+		    code[ci++] = tmp_op->a[arg].val * sizeof(Eterm) + 1;
 		    break;
 		default:
 		    LoadError1(stp, "bad tag %d for destination",
@@ -2356,20 +2387,13 @@ load_code(LoaderState* stp)
 		stp->labels[tmp_op->a[arg].val].patches = ci;
 		ci++;
 		break;
-	    case TAG_r:
-		CodeNeed(1);
-		code[ci++] = (R_REG_DEF << _TAG_PRIMARY_SIZE) |
-		    TAG_PRIMARY_HEADER;
-		break;
 	    case TAG_x:
 		CodeNeed(1);
-		code[ci++] = (tmp_op->a[arg].val << _TAG_IMMED1_SIZE) |
-		    (X_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER;
+		code[ci++] = make_loader_x_reg(tmp_op->a[arg].val);
 		break;
 	    case TAG_y:
 		CodeNeed(1);
-		code[ci++] = (tmp_op->a[arg].val << _TAG_IMMED1_SIZE) |
-		    (Y_REG_DEF << _TAG_PRIMARY_SIZE) | TAG_PRIMARY_HEADER;
+		code[ci++] = make_loader_y_reg(tmp_op->a[arg].val);
 		break;
 	    case TAG_n:
 		CodeNeed(1);
@@ -2457,7 +2481,6 @@ load_code(LoaderState* stp)
 	    stp->on_load = ci;
 	    break;
 	case op_bs_put_string_II:
-	case op_i_bs_match_string_rfII:
 	case op_i_bs_match_string_xfII:
 	    new_string_patch(stp, ci-1);
 	    break;
@@ -2679,6 +2702,12 @@ same_label(LoaderState* stp, GenOpArg Target, GenOpArg Label)
 	Target.val == Label.val;
 }
 
+static int
+is_killed(LoaderState* stp, GenOpArg Reg, GenOpArg Live)
+{
+    return Reg.type == TAG_x && Live.type == TAG_u &&
+	Live.val <= Reg.val;
+}
 
 /*
  * Generate an instruction for element/2.
@@ -2695,17 +2724,17 @@ gen_element(LoaderState* stp, GenOpArg Fail, GenOpArg Index,
     op->next = NULL;
 
     if (Index.type == TAG_i && Index.val > 0 &&
-	(Tuple.type == TAG_r || Tuple.type == TAG_x || Tuple.type == TAG_y)) {
+	(Tuple.type == TAG_x || Tuple.type == TAG_y)) {
 	op->op = genop_i_fast_element_4;
-	op->a[0] = Tuple;
-	op->a[1] = Fail;
+	op->a[0] = Fail;
+	op->a[1] = Tuple;
 	op->a[2].type = TAG_u;
 	op->a[2].val = Index.val;
 	op->a[3] = Dst;
     } else {
 	op->op = genop_i_element_4;
-	op->a[0] = Tuple;
-	op->a[1] = Fail;
+	op->a[0] = Fail;
+	op->a[1] = Tuple;
 	op->a[2] = Index;
 	op->a[3] = Dst;
     }
@@ -2831,23 +2860,16 @@ gen_get_integer2(LoaderState* stp, GenOpArg Fail, GenOpArg Ms, GenOpArg Live,
 	    goto generic;
 	}
     } else {
-	GenOp* op2;
-	NEW_GENOP(stp, op2);
-	
-	op->op = genop_i_fetch_2;
-	op->arity = 2;
-	op->a[0] = Ms;
-	op->a[1] = Size;
-	op->next = op2;
-
-	op2->op = genop_i_bs_get_integer_4;
-	op2->arity = 4;
-	op2->a[0] = Fail;
-	op2->a[1] = Live;
-	op2->a[2].type = TAG_u;
-	op2->a[2].val = (Unit.val << 3) | Flags.val;
-	op2->a[3] = Dst;
-	op2->next = NULL;
+	op->op = genop_i_bs_get_integer_6;
+	op->arity = 6;
+	op->a[0] = Fail;
+	op->a[1] = Live;
+	op->a[2].type = TAG_u;
+	op->a[2].val = (Unit.val << 3) | Flags.val;
+	op->a[3] = Ms;
+	op->a[4] = Size;
+	op->a[5] = Dst;
+	op->next = NULL;
 	return op;
     }
     op->next = NULL;
@@ -3898,9 +3920,7 @@ gen_make_fun2(LoaderState* stp, GenOpArg idx)
 /*
  * Rewrite gc_bifs with one parameter (the common case). Utilized
  * in ops.tab to rewrite instructions calling bif's in guards
- * to use a garbage collecting implementation. The instructions
- * are sometimes once again rewritten to handle literals (putting the
- * parameter in the mostly unused r[0] before the instruction is executed).
+ * to use a garbage collecting implementation.
  */
 static GenOp*
 gen_guard_bif1(LoaderState* stp, GenOpArg Fail, GenOpArg Live, GenOpArg Bif,
@@ -3955,10 +3975,6 @@ gen_guard_bif1(LoaderState* stp, GenOpArg Fail, GenOpArg Live, GenOpArg Bif,
 
 /*
  * This is used by the ops.tab rule that rewrites gc_bifs with two parameters.
- * The instruction returned is then again rewritten to an i_load instruction
- * followed by i_gc_bif2_jIId, to handle literals properly.
- * As opposed to the i_gc_bif1_jIsId, the instruction  i_gc_bif2_jIId is
- * always rewritten, regardless of if there actually are any literals.
  */
 static GenOp*
 gen_guard_bif2(LoaderState* stp, GenOpArg Fail, GenOpArg Live, GenOpArg Bif,
@@ -3985,23 +4001,19 @@ gen_guard_bif2(LoaderState* stp, GenOpArg Fail, GenOpArg Live, GenOpArg Bif,
 	op->a[2].val = stp->import[Bif.val].arity;
 	return op;
     }
-    op->op = genop_ii_gc_bif2_6;
+    op->op = genop_i_gc_bif2_6;
     op->arity = 6;
     op->a[0] = Fail;
     op->a[1].type = TAG_u;
-    op->a[2] = S1;
-    op->a[3] = S2;
-    op->a[4] = Live;
+    op->a[2] = Live;
+    op->a[3] = S1;
+    op->a[4] = S2;
     op->a[5] = Dst;
     return op;
 }
 
 /*
  * This is used by the ops.tab rule that rewrites gc_bifs with three parameters.
- * The instruction returned is then again rewritten to a move instruction that
- * uses r[0] for temp storage, followed by an i_load instruction,
- * followed by i_gc_bif3_jIsId, to handle literals properly. Rewriting
- * always occur, as with the gc_bif2 counterpart.
  */
 static GenOp*
 gen_guard_bif3(LoaderState* stp, GenOpArg Fail, GenOpArg Live, GenOpArg Bif,
@@ -4032,10 +4044,10 @@ gen_guard_bif3(LoaderState* stp, GenOpArg Fail, GenOpArg Live, GenOpArg Bif,
     op->arity = 7;
     op->a[0] = Fail;
     op->a[1].type = TAG_u;
-    op->a[2] = S1;
-    op->a[3] = S2;
-    op->a[4] = S3;
-    op->a[5] = Live;
+    op->a[2] = Live;
+    op->a[3] = S1;
+    op->a[4] = S2;
+    op->a[5] = S3;
     op->a[6] = Dst;
     op->next = NULL;
     return op;
@@ -4322,7 +4334,7 @@ gen_has_map_fields(LoaderState* stp, GenOpArg Fail, GenOpArg Src,
     for (i = 0; i < n; i++) {
 	op->a[3+2*i] = Rest[i];
 	op->a[3+2*i+1].type = TAG_x;
-	op->a[3+2*i+1].val = 0;	/* x(0); normally not used */
+	op->a[3+2*i+1].val = SCRATCH_X_REG; /* Ignore result */
     }
     return op;
 }
@@ -4800,7 +4812,8 @@ transform_engine(LoaderState* st)
 	    if (var[i].type != instr->a[ap].type)
 		goto restart;
 	    switch (var[i].type) {
-	    case TAG_r: case TAG_n: break;
+	    case TAG_n:
+		break;
 	    default:
 		if (var[i].val != instr->a[ap].val)
 		    goto restart;
@@ -5867,7 +5880,7 @@ make_stub(BeamInstr* fp, Eterm mod, Eterm func, Uint arity, Uint native, BeamIns
     fp[4] = arity;
 #ifdef HIPE
     if (native) {
-	fp[5] = BeamOpCode(op_move_return_nr);
+	fp[5] = BeamOpCode(op_move_return_n);
 	hipe_mfa_save_orig_beam_op(mod, func, arity, fp+5);
     }
 #endif
@@ -6303,7 +6316,7 @@ erts_make_stub_module(Process* p, Eterm Mod, Eterm Beam, Eterm Info)
 #ifdef HIPE
 	op = (Eterm) BeamOpCode(op_hipe_trap_call); /* Might be changed later. */
 #else
-	op = (Eterm) BeamOpCode(op_move_return_nr);
+	op = (Eterm) BeamOpCode(op_move_return_n);
 #endif
 	fp = make_stub(fp, Mod, func, arity, (Uint)native_address, op);
     }
diff --git a/erts/emulator/beam/beam_load.h b/erts/emulator/beam/beam_load.h
index d5af634fad..eedb5ee4cd 100644
--- a/erts/emulator/beam/beam_load.h
+++ b/erts/emulator/beam/beam_load.h
@@ -52,6 +52,7 @@ extern BeamInstr* em_call_error_handler;
 extern BeamInstr* em_apply_bif;
 extern BeamInstr* em_call_nif;
 
+
 /*
  * The following variables keep a sorted list of address ranges for
  * each module.  It allows us to quickly find a function given an
diff --git a/erts/emulator/beam/erl_term.c b/erts/emulator/beam/erl_term.c
index 89459fb278..3a5fbcc284 100644
--- a/erts/emulator/beam/erl_term.c
+++ b/erts/emulator/beam/erl_term.c
@@ -173,9 +173,7 @@ ET_DEFINE_CHECKED(Uint,external_thing_data_words,ExternalThing*,is_thing_ptr);
 ET_DEFINE_CHECKED(Eterm,make_cp,UWord *,_is_taggable_pointer);
 ET_DEFINE_CHECKED(UWord *,cp_val,Eterm,is_CP);
 ET_DEFINE_CHECKED(Uint,catch_val,Eterm,is_catch);
-ET_DEFINE_CHECKED(Uint,x_reg_offset,Uint,_is_xreg);
-ET_DEFINE_CHECKED(Uint,y_reg_offset,Uint,_is_yreg);
-ET_DEFINE_CHECKED(Uint,x_reg_index,Uint,_is_xreg);
-ET_DEFINE_CHECKED(Uint,y_reg_index,Uint,_is_yreg);
+ET_DEFINE_CHECKED(Uint,loader_x_reg_index,Uint,_is_loader_x_reg);
+ET_DEFINE_CHECKED(Uint,loader_y_reg_index,Uint,_is_loader_y_reg);
 
 #endif	/* ET_DEBUG */
diff --git a/erts/emulator/beam/erl_term.h b/erts/emulator/beam/erl_term.h
index b928e4ca1e..089eecf024 100644
--- a/erts/emulator/beam/erl_term.h
+++ b/erts/emulator/beam/erl_term.h
@@ -1027,44 +1027,40 @@ _ET_DECLARE_CHECKED(Uint,catch_val,Eterm)
 /*
  * Overloaded tags.
  *
- * SMALL = 15
- * ATOM/NIL=7
+ * In the loader, we want to tag a term in a way so that it can
+ * be any literal (atom/integer/float/tuple/list/binary) or a
+ * register.
  *
- * Note that the two least significant bits in SMALL/ATOM/NIL always are 3;
- * thus, we can distinguish register from literals by looking at only these
- * two bits.
+ * We can achive that by overloading the PID and PORT tags to
+ * mean X and Y registers. That works because there are no
+ * pid or port literals.
  */
 
-#define X_REG_DEF	0
-#define Y_REG_DEF	1
-#define R_REG_DEF	2
+#define _LOADER_TAG_XREG _TAG_IMMED1_PID
+#define _LOADER_TAG_YREG _TAG_IMMED1_PORT
+#define _LOADER_TAG_SIZE _TAG_IMMED1_SIZE
+#define _LOADER_MASK _TAG_IMMED1_MASK
 
-#define beam_reg_tag(x)	((x) & 3)
+#define LOADER_X_REG _LOADER_TAG_XREG
+#define LOADER_Y_REG _LOADER_TAG_YREG
 
-#define make_rreg()	R_REG_DEF
-#define make_xreg(ix)	(((ix) * sizeof(Eterm)) | X_REG_DEF)
-#define make_yreg(ix)	(((ix) * sizeof(Eterm)) | Y_REG_DEF)
+#define make_loader_x_reg(R) (((R) << _LOADER_TAG_SIZE) | _LOADER_TAG_XREG)
+#define make_loader_y_reg(R) (((R) << _LOADER_TAG_SIZE) | _LOADER_TAG_YREG)
 
-#define _is_xreg(x)	(beam_reg_tag(x) == X_REG_DEF)
-#define _is_yreg(x)	(beam_reg_tag(x) == Y_REG_DEF)
+#define loader_reg_index(R) ((R) >> _LOADER_TAG_SIZE)
 
-#define _unchecked_x_reg_offset(R)	((R) - X_REG_DEF)
-_ET_DECLARE_CHECKED(Uint,x_reg_offset,Uint)
-#define x_reg_offset(R)	_ET_APPLY(x_reg_offset,(R))
+#define loader_tag(T) ((T) & _LOADER_MASK)
 
-#define _unchecked_y_reg_offset(R)	((R) - Y_REG_DEF)
-_ET_DECLARE_CHECKED(Uint,y_reg_offset,Uint)
-#define y_reg_offset(R)	_ET_APPLY(y_reg_offset,(R))
+#define _is_loader_x_reg(x) (loader_tag(x) == _LOADER_TAG_XREG)
+#define _is_loader_y_reg(x) (loader_tag(x) == _LOADER_TAG_YREG)
 
-#define reg_index(R) ((R) / sizeof(Eterm))
+#define _unchecked_loader_x_reg_index(R) ((R) >> _LOADER_TAG_SIZE)
+_ET_DECLARE_CHECKED(Uint,loader_x_reg_index,Uint)
+#define loader_x_reg_index(R) _ET_APPLY(loader_x_reg_index,(R))
 
-#define _unchecked_x_reg_index(R)	((R) >> 2)
-_ET_DECLARE_CHECKED(Uint,x_reg_index,Uint)
-#define x_reg_index(R)	_ET_APPLY(x_reg_index,(R))
-
-#define _unchecked_y_reg_index(R)	((R) >> 2)
-_ET_DECLARE_CHECKED(Uint,y_reg_index,Uint)
-#define y_reg_index(R)	_ET_APPLY(y_reg_index,(R))
+#define _unchecked_loader_y_reg_index(R)  ((R) >> _LOADER_TAG_SIZE)
+_ET_DECLARE_CHECKED(Uint,loader_y_reg_index,Uint)
+#define loader_y_reg_index(R) _ET_APPLY(loader_y_reg_index,(R))
 
 /*
  * Backwards compatibility definitions:
diff --git a/erts/emulator/beam/erl_vm.h b/erts/emulator/beam/erl_vm.h
index cfc978b5a5..98f27a1725 100644
--- a/erts/emulator/beam/erl_vm.h
+++ b/erts/emulator/beam/erl_vm.h
@@ -141,6 +141,7 @@
 typedef struct op_entry {
    char* name;			/* Name of instruction. */
    Uint32 mask[3];		/* Signature mask. */
+   unsigned involves_r;		/* Needs special attention when matching. */
    int sz;			/* Number of loaded words. */
    char* pack;			/* Instructions for packing engine. */
    char* sign;			/* Signature string. */
diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 1d32e72247..46fefb88af 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -39,8 +39,8 @@ too_old_compiler | never() =>
 # necessary.) Since the instructions don't work correctly in R12B, simply
 # refuse to load the module.
 
-func_info M=a a==am_module_info A=u==0 | label L | move n r => too_old_compiler
-func_info M=a a==am_module_info A=u==1 | label L | move n r => too_old_compiler
+func_info M=a a==am_module_info A=u==0 | label L | move n x==0 => too_old_compiler
+func_info M=a a==am_module_info A=u==1 | label L | move n x==0 => too_old_compiler
 
 # The undocumented and unsupported guard BIF is_constant/1 was removed
 # in R13. The is_constant/2 operation is marked as obsolete in genop.tab,
@@ -76,17 +76,6 @@ return
 # with the following call instruction, we need to make sure that
 # there is no line/1 instruction between the move and the call.
 #
-
-move S r | line Loc | call_ext Ar Func => \
-     line Loc | move S r | call_ext Ar Func
-move S r | line Loc | call_ext_last Ar Func=u$is_bif D => \
-     line Loc | move S r | call_ext_last Ar Func D
-move S r | line Loc | call_ext_only Ar Func=u$is_bif => \
-     line Loc | move S r | call_ext_only Ar Func
-move S r | line Loc | call Ar Func => \
-     line Loc | move S r | call Ar Func
-
-#
 # A tail-recursive call to an external function (non-BIF) will
 # never be saved on the stack, so there is no reason to keep
 # the line instruction. (The compiler did not remove the line
@@ -94,10 +83,14 @@ move S r | line Loc | call Ar Func => \
 # BIFs and ordinary Erlang functions.)
 #
 
-line Loc | call_ext_last Ar Func=u$is_not_bif D => \
-     call_ext_last Ar Func D
-line Loc | call_ext_only Ar Func=u$is_not_bif => \
-     call_ext_only Ar Func
+move S X0=x==0 | line Loc | call_ext Ar Func => \
+     line Loc | move S X0 | call_ext Ar Func
+move S X0=x==0 | line Loc | call_ext_last Ar Func=u$is_not_bif D => \
+     move S X0 | call_ext_last Ar Func D
+move S X0=x==0 | line Loc | call_ext_only Ar Func=u$is_not_bif => \
+     move S X0 | call_ext_only Ar Func
+move S X0=x==0 | line Loc | call Ar Func => \
+     line Loc | move S X0 | call Ar Func
 
 line Loc | func_info M F A => func_info M F A | line Loc
 
@@ -167,31 +160,24 @@ is_tuple Fail=f S | select_tuple_arity S=d Fail=f Size=u Rest=* => \
 select_tuple_arity S=d Fail=f Size=u Rest=* => \
   gen_select_tuple_arity(S, Fail, Size, Rest)
 
-i_select_val_bins r f I
 i_select_val_bins x f I
 i_select_val_bins y f I
 
-i_select_val_lins r f I
 i_select_val_lins x f I
 i_select_val_lins y f I
 
-i_select_val2 r f c c f f
 i_select_val2 x f c c f f
 i_select_val2 y f c c f f
 
-i_select_tuple_arity r f I
 i_select_tuple_arity x f I
 i_select_tuple_arity y f I
 
-i_select_tuple_arity2 r f A A f f
 i_select_tuple_arity2 x f A A f f
 i_select_tuple_arity2 y f A A f f
 
-i_jump_on_val_zero r f I
 i_jump_on_val_zero x f I
 i_jump_on_val_zero y f I
 
-i_jump_on_val r f I I
 i_jump_on_val x f I I
 i_jump_on_val y f I I
 
@@ -203,30 +189,21 @@ is_ne_exact L1 S1 S2 | jump Fail | label L2 | same_label(L1, L2) => \
 %macro: get_list GetList -pack
 get_list x x x
 get_list x x y
-get_list x x r
 get_list x y x
 get_list x y y
-get_list x y r
-get_list x r x
-get_list x r y
 
 get_list y x x
 get_list y x y
-get_list y x r
 get_list y y x
 get_list y y y
-get_list y y r
-get_list y r x
-get_list y r y
 
+# The following get_list instructions using x(0) are frequently used.
 get_list r x x
+get_list r r y
+get_list x r x
 get_list r x y
-get_list r x r
-get_list r y x
-get_list r y y
 get_list r y r
-get_list r r x
-get_list r r y
+get_list r x r
 
 # Old-style catch.
 catch y f
@@ -237,33 +214,31 @@ try Y F => catch Y F
 try_case Y => try_end Y
 try_end y
 
-try_case_end Literal=q => move Literal x | try_case_end x
 try_case_end s
 
 # Destructive set tuple element
 
-set_tuple_element Lit=q Tuple Pos => move Lit x | set_tuple_element x Tuple Pos
 set_tuple_element s d P
 
 # Get tuple element
 
 %macro: i_get_tuple_element GetTupleElement -pack
 i_get_tuple_element x P x
-i_get_tuple_element r P x
 i_get_tuple_element y P x
-i_get_tuple_element x P r
-i_get_tuple_element y P r
 
 %cold
-i_get_tuple_element r P r
 i_get_tuple_element x P y
-i_get_tuple_element r P y
 i_get_tuple_element y P y
 %hot
 
+%macro: i_get_tuple_element2 GetTupleElement2 -pack
+i_get_tuple_element2 x P x
+
+%macro: i_get_tuple_element3 GetTupleElement3 -pack
+i_get_tuple_element3 x P x
+
 %macro: is_number IsNumber -fail_action
 %cold
-is_number f r
 is_number f x
 is_number f y
 %hot
@@ -273,16 +248,12 @@ is_number Fail Literal=q => move Literal x | is_number Fail x
 
 jump f
 
-case_end Literal=cq => move Literal x | case_end x
-badmatch Literal=cq => move Literal x | badmatch x
+case_end NotInX=cy => move NotInX x | case_end x
+badmatch NotInX=cy => move NotInX x | badmatch x
 
-case_end r
 case_end x
-case_end y
 
-badmatch r
 badmatch x
-badmatch y
 
 if_end
 raise s s
@@ -291,7 +262,7 @@ raise s s
 badarg j
 system_limit j
 
-move C=cxy r | jump Lbl => move_jump Lbl C
+move C=cxy x==0 | jump Lbl => move_jump Lbl C
 
 %macro: move_jump MoveJump -nonext
 move_jump f n
@@ -309,8 +280,6 @@ move_window/6
 
 # x -> y
 
-move S1=r S2=y | move X1=x Y1=y => move2 S1 S2 X1 Y1
-
 move X1=x Y1=y | move X2=x Y2=y | move X3=x Y3=y | succ(Y1,Y2) | succ(Y2,Y3) => \
     move_window X1 X2 X3 Y1 Y3
 
@@ -323,24 +292,76 @@ move_window X1=x X2=x X3=x X4=x Y1=y Y4=y | move X5=x Y5=y | succ(Y4,Y5) => \
 move_window X1=x X2=x X3=x Y1=y Y3=y => move_window3 X1 X2 X3 Y1
 move_window X1=x X2=x X3=x X4=x Y1=y Y4=y => move_window4 X1 X2 X3 X4 Y1
 
+%macro: move_window3 MoveWindow3 -pack
+%macro: move_window4 MoveWindow4 -pack
+%macro: move_window5 MoveWindow5 -pack
+
 move_window3 x x x y
 move_window4 x x x x y
 move_window5 x x x x x y
 
-move X1=x Y1=y | move X2=x Y2=y => move2 X1 Y1 X2 Y2
-move Y1=y X1=x | move Y2=y X2=x => move2 Y1 X1 Y2 X2
-move X1=x X2=x | move X3=x X4=x => move2 X1 X2 X3 X4
+# Swap registers.
+move R1=x Tmp=x | move R2=xy R1 | move Tmp R2 => swap_temp R1 R2 Tmp
+
+swap_temp R1 R2 Tmp | line Loc | apply Live | is_killed(Tmp, Live) => \
+  swap R1 R2 | line Loc | apply Live
+
+swap_temp R1 R2 Tmp | line Loc | call Live Addr | is_killed(Tmp, Live) => \
+  swap R1 R2 | line Loc | call Live Addr
+swap_temp R1 R2 Tmp | call_only Live Addr | \
+  is_killed(Tmp, Live) => swap R1 R2 | call_only Live Addr
+swap_temp R1 R2 Tmp | call_last Live Addr D | \
+  is_killed(Tmp, Live) => swap R1 R2 | call_last Live Addr D
 
-move S1=x S2=r | move S3=x S4=x => move2 S1 S2 S3 S4
-move S1=x S2=r | move X1=x Y1=y => move2 S1 S2 X1 Y1
-move S1=y S2=r | move X1=x Y1=y => move2 S1 S2 X1 Y1
+swap_temp R1 R2 Tmp | line Loc | call_ext Live Addr | is_killed(Tmp, Live) => \
+  swap R1 R2 | line Loc | call_ext Live Addr
+swap_temp R1 R2 Tmp | line Loc | call_ext_only Live Addr | \
+  is_killed(Tmp, Live) => swap R1 R2 | line Loc | call_ext_only Live Addr
+swap_temp R1 R2 Tmp | line Loc | call_ext_last Live Addr D | \
+  is_killed(Tmp, Live) => swap R1 R2 | line Loc | call_ext_last Live Addr D
 
-move Y1=y X1=x | move S1=r D1=x => move2 Y1 X1 S1 D1
-move S1=r D1=x | move Y1=y X1=x => move2 S1 D1 Y1 X1
+%macro: swap_temp SwapTemp -pack
+swap_temp x x x
+swap_temp x y x
 
-move2 X1=x Y1=y X2=x Y2=y | move X3=x Y3=y => move3 X1 Y1 X2 Y2 X3 Y3
-move2 Y1=y X1=x Y2=y X2=x | move Y3=y X3=x => move3 Y1 X1 Y2 X2 Y3 X3
-move2 X1=x X2=x X3=x X4=x | move X5=x X6=x => move3 X1 X2 X3 X4 X5 X6
+%macro: swap Swap -pack
+swap x x
+swap x y
+
+move Src=x D1=x | move Src=x D2=x => move_dup Src D1 D2
+move Src=x SD=x | move SD=x D=x   => move_dup Src SD D
+move Src=x D1=x | move Src=x D2=y => move_dup Src D1 D2
+move Src=y SD=x | move SD=x D=y   => move_dup Src SD D
+move Src=x SD=x | move SD=x D=y   => move_dup Src SD D
+move Src=y SD=x | move SD=x D=x   => move_dup Src SD D
+
+move SD=x D=x | move Src=xy SD=x => move_shift Src SD D
+move SD=y D=x | move Src=x  SD=y => move_shift Src SD D
+move SD=x D=y | move Src=x  SD=x => move_shift Src SD D
+
+# The transformations above guarantee that the source for
+# the second move is not the same as the destination for
+# the first move. That means that we can do the moves in
+# parallel (fetch both values, then store them) which could
+# be faster.
+
+move X1=x Y1=y | move X2=x Y2=y => move2_par X1 Y1 X2 Y2
+move Y1=y X1=x | move Y2=y X2=x => move2_par Y1 X1 Y2 X2
+
+move X1=x X2=x | move X3=x X4=x => move2_par X1 X2 X3 X4
+
+move X1=x X2=x | move X3=x Y1=y => move2_par X1 X2 X3 Y1
+
+move S1=x S2=x | move X1=x Y1=y => move2_par S1 S2 X1 Y1
+
+move S1=y S2=x | move X1=x Y1=y => move2_par S1 S2 X1 Y1
+
+move Y1=y X1=x | move S1=x D1=x => move2_par Y1 X1 S1 D1
+move S1=x D1=x | move Y1=y X1=x => move2_par S1 D1 Y1 X1
+
+move2_par X1=x Y1=y X2=x Y2=y | move X3=x Y3=y => move3 X1 Y1 X2 Y2 X3 Y3
+move2_par Y1=y X1=x Y2=y X2=x | move Y3=y X3=x => move3 Y1 X1 Y2 X2 Y3 X3
+move2_par X1=x X2=x X3=x X4=x | move X5=x X6=x => move3 X1 X2 X3 X4 X5 X6
 
 move C=aiq X=x==1 => move_x1 C
 move C=aiq X=x==2 => move_x2 C
@@ -348,21 +369,32 @@ move C=aiq X=x==2 => move_x2 C
 move_x1 c
 move_x2 c
 
-%macro: move2 Move2 -pack
-move2 x y x y
-move2 y x y x
-move2 x x x x
+%macro: move_shift MoveShift -pack
+move_shift x x x
+move_shift y x x
+move_shift x y x
+move_shift x x y
+
+%macro: move_dup MoveDup -pack
+move_dup x x x
+move_dup x x y
+move_dup y x x
+move_dup y x y
 
-move2 x r x x
+%macro: move2_par Move2Par -pack
 
-move2 x r x y
-move2 r y x y
-move2 y r x y
+move2_par x y x y
+move2_par y x y x
+move2_par x x x x
 
-move2 r x y x
-move2 y x r x
+move2_par x x x y
 
-%macro: move3 Move3
+move2_par y x x y
+
+move2_par x x y x
+move2_par y x x x
+
+%macro: move3 Move3 -pack
 move3 x y x y x y
 move3 y x y x y x
 move3 x x x x x x
@@ -375,20 +407,22 @@ move S=c D=y => move S x | move x D
 %macro:move Move -pack -gen_dest
 move x x
 move x y
-move x r
 move y x
-move y r
-move r x
-move r y
-move c r
 move c x
 move n x
-move n r
 move y y
 
+# The following move instructions using x(0) are frequently used.
+
+move x r
+move r x
+move y r
+move c r
+move r y
+
 # Receive operations.
 
-loop_rec Fail Src | smp_mark_target_label(Fail) => i_loop_rec Fail Src
+loop_rec Fail x==0 | smp_mark_target_label(Fail) => i_loop_rec Fail
 
 label L | wait_timeout Fail Src | smp_already_locked(L) => label L | i_wait_timeout_locked Fail Src
 wait_timeout Fail Src => i_wait_timeout Fail Src
@@ -403,7 +437,7 @@ label L | timeout | smp_already_locked(L) => label L | timeout_locked
 remove_message
 timeout
 timeout_locked
-i_loop_rec f r
+i_loop_rec f
 loop_rec_end f
 wait f
 wait_locked f
@@ -421,93 +455,57 @@ send
 # Optimized comparisons with one immediate/literal operand.
 #
 
-is_eq_exact Lbl R=rxy C=ian => i_is_eq_exact_immed Lbl R C
-is_eq_exact Lbl R=rxy C=q => i_is_eq_exact_literal R Lbl C
+is_eq_exact Lbl R=xy C=ian => i_is_eq_exact_immed Lbl R C
+is_eq_exact Lbl R=xy C=q => i_is_eq_exact_literal Lbl R C
 
-is_ne_exact Lbl R=rxy C=ian => i_is_ne_exact_immed Lbl R C
-is_ne_exact Lbl R=rxy C=q => i_is_ne_exact_literal R Lbl C
+is_ne_exact Lbl R=xy C=ian => i_is_ne_exact_immed Lbl R C
+is_ne_exact Lbl R=xy C=q => i_is_ne_exact_literal Lbl R C
 
 %macro: i_is_eq_exact_immed EqualImmed -fail_action
 i_is_eq_exact_immed f r c
 i_is_eq_exact_immed f x c
 i_is_eq_exact_immed f y c
 
-i_is_eq_exact_literal r f c
-i_is_eq_exact_literal x f c
-i_is_eq_exact_literal y f c
+i_is_eq_exact_literal f x c
+i_is_eq_exact_literal f y c
 
 %macro: i_is_ne_exact_immed NotEqualImmed -fail_action
-i_is_ne_exact_immed f r c
 i_is_ne_exact_immed f x c
 i_is_ne_exact_immed f y c
 
-i_is_ne_exact_literal r f c
-i_is_ne_exact_literal x f c
-i_is_ne_exact_literal y f c
-
-#
-# Common Compare Specializations
-# We don't do all of them since we want
-# to keep the instruction set small-ish
-#
-
-is_eq_exact Lbl S1=xy S2=r => is_eq_exact Lbl S2 S1
-is_eq_exact Lbl S1=rx S2=xy => i_is_eq_exact_spec Lbl S1 S2
-%macro: i_is_eq_exact_spec EqualExact -fail_action
-
-i_is_eq_exact_spec f x x
-i_is_eq_exact_spec f x y
-i_is_eq_exact_spec f r x
-i_is_eq_exact_spec f r y
-%cold
-i_is_eq_exact_spec f r r
-%hot
+i_is_ne_exact_literal f x c
+i_is_ne_exact_literal f y c
 
-is_lt Lbl S1=rxc S2=rxc  => i_is_lt_spec Lbl S1 S2
+is_eq_exact Lbl Y=y X=x => is_eq_exact Lbl X Y
+%macro: is_eq_exact EqualExact -fail_action -pack
+is_eq_exact f x x
+is_eq_exact f x y
+is_eq_exact f s s
 
-%macro: i_is_lt_spec IsLessThan -fail_action
-
-i_is_lt_spec f x x
-i_is_lt_spec f x r
-i_is_lt_spec f x c
-i_is_lt_spec f r x
-i_is_lt_spec f r c
-i_is_lt_spec f c x
-i_is_lt_spec f c r
+%macro: is_lt IsLessThan -fail_action
+is_lt f x x
+is_lt f x c
+is_lt f c x
 %cold
-i_is_lt_spec f r r
-i_is_lt_spec f c c
+is_lt f s s
 %hot
 
-is_ge Lbl S1=xc S2=xc => i_is_ge_spec Lbl S1 S2
-
-%macro: i_is_ge_spec IsGreaterEqual -fail_action
-
-i_is_ge_spec f x x
-i_is_ge_spec f x c
-i_is_ge_spec f c x
+%macro: is_ge IsGreaterEqual -fail_action
+is_ge f x x
+is_ge f x c
+is_ge f c x
 %cold
-i_is_ge_spec f c c
+is_ge f s s
 %hot
 
-#
-# All other comparisons.
-#
+%macro: is_ne_exact NotEqualExact -fail_action
+is_ne_exact f s s
 
-is_eq_exact Lbl S1 S2 => i_fetch S1 S2 | i_is_eq_exact Lbl
-is_ne_exact Lbl S1 S2 => i_fetch S1 S2 | i_is_ne_exact Lbl
+%macro: is_eq Equal -fail_action
+is_eq f s s
 
-is_lt Lbl S1 S2 => i_fetch S1 S2 | i_is_lt Lbl
-is_ge Lbl S1 S2 => i_fetch S1 S2 | i_is_ge Lbl
-is_eq Lbl S1 S2 => i_fetch S1 S2 | i_is_eq Lbl
-is_ne Lbl S1 S2 => i_fetch S1 S2 | i_is_ne Lbl
-
-i_is_eq_exact f
-i_is_ne_exact f
-i_is_lt f
-i_is_ge f
-i_is_eq f
-i_is_ne f
+%macro: is_ne NotEqual -fail_action
+is_ne f s s
 
 #
 # Putting things.
@@ -525,7 +523,6 @@ i_put_tuple Dst Arity Puts=* | put S => \
 i_put_tuple/2
 
 %macro:i_put_tuple PutTuple -pack -goto:do_put_tuple
-i_put_tuple r I
 i_put_tuple x I
 i_put_tuple y I
 
@@ -542,74 +539,36 @@ put_list x n x
 put_list y n x
 put_list x x x
 put_list y x x
-put_list x x r
-put_list y r r
 
 put_list y y x
 put_list x y x
-put_list r x x
-put_list r y x
-put_list r x r
-put_list y y r
-put_list y r x
-put_list r n x
 
-put_list x r x
-put_list x y r
-put_list y x r
 put_list y x x
 
-put_list x r r
-
 # put_list SrcReg Constant Dst
-put_list r c r
-put_list r c x
-put_list r c y
 
-put_list x c r
 put_list x c x
 put_list x c y
 
-put_list y c r
 put_list y c x
-put_list y c y
 
 # put_list Constant SrcReg Dst
-put_list c r r
-put_list c r x
-put_list c r y
 
-put_list c x r
 put_list c x x
-put_list c x y
-
-put_list c y r
 put_list c y x
-put_list c y y
 
-%cold
-put_list s s d
-%hot
+# The following put_list instructions using x(0) are frequently used.
 
-%macro: i_fetch FetchArgs -pack
-i_fetch c r
-i_fetch c x
-i_fetch c y
-i_fetch r c
-i_fetch r x
-i_fetch r y
-i_fetch x c
-i_fetch x r
-i_fetch x x
-i_fetch x y
-i_fetch y c
-i_fetch y r
-i_fetch y x
-i_fetch y y
+put_list y r r
+put_list x r r
+put_list r n r
+put_list r n x
+put_list r x x
+put_list r x r
+put_list x x r
 
 %cold
-i_fetch c c
-i_fetch s s
+put_list s s d
 %hot
 
 #
@@ -631,27 +590,27 @@ return_trace
 # Note: There is no 'move_return y r', since there never are any y registers
 # when we do move_return (if we have y registers, we must do move_deallocate_return).
 
-move S r | return => move_return S r
+move S x==0 | return => move_return S
 
 %macro: move_return MoveReturn -nonext
-move_return x r
-move_return c r
-move_return n r
+move_return x
+move_return c
+move_return n
 
-move S r | deallocate D | return => move_deallocate_return S r D
+move S x==0 | deallocate D | return => move_deallocate_return S D
 
 %macro: move_deallocate_return MoveDeallocateReturn -pack -nonext
-move_deallocate_return x r Q
-move_deallocate_return y r Q
-move_deallocate_return c r Q
-move_deallocate_return n r Q
+move_deallocate_return x Q
+move_deallocate_return y Q
+move_deallocate_return c Q
+move_deallocate_return n Q
 
 deallocate D | return => deallocate_return D
 
 %macro: deallocate_return DeallocateReturn -nonext
 deallocate_return Q
 
-test_heap Need u==1 | put_list Y=y r r => test_heap_1_put_list Need Y
+test_heap Need u==1 | put_list Y=y x==0 x==0 => test_heap_1_put_list Need Y
 
 %macro: test_heap_1_put_list TestHeapPutList -pack
 test_heap_1_put_list I y
@@ -660,18 +619,18 @@ test_heap_1_put_list I y
 
 is_tuple Fail Literal=q => move Literal x | is_tuple Fail x
 is_tuple Fail=f c => jump Fail
-is_tuple Fail=f S=rxy | test_arity Fail=f S=rxy Arity => is_tuple_of_arity Fail S Arity
+is_tuple Fail=f S=xy | test_arity Fail=f S=xy Arity => is_tuple_of_arity Fail S Arity
 
 %macro:is_tuple_of_arity IsTupleOfArity -fail_action
 
+is_tuple_of_arity f r A
 is_tuple_of_arity f x A
 is_tuple_of_arity f y A
-is_tuple_of_arity f r A
 
 %macro: is_tuple IsTuple -fail_action
+is_tuple f r
 is_tuple f x
 is_tuple f y
-is_tuple f r
 
 test_arity Fail Literal=q Arity => move Literal x | test_arity Fail x Arity
 test_arity Fail=f c Arity => jump Fail
@@ -679,95 +638,63 @@ test_arity Fail=f c Arity => jump Fail
 %macro: test_arity IsArity -fail_action
 test_arity f x A
 test_arity f y A
-test_arity f r A
-
-is_tuple_of_arity Fail=f Reg Arity | get_tuple_element Reg P=u==0 Dst=xy => \
-  is_tuple_of_arity Fail Reg Arity | extract_next_element Dst | original_reg Reg P
-
-test_arity Fail Reg Arity | get_tuple_element Reg P=u==0 Dst=xy => \
-  test_arity Fail Reg Arity | extract_next_element Dst | original_reg Reg P
-
-original_reg Reg P1 | get_tuple_element Reg P2 Dst=xy | succ(P1, P2) => \
-  extract_next_element Dst | original_reg Reg P2
-
-get_tuple_element Reg P Dst => i_get_tuple_element Reg P Dst | original_reg Reg P
-
-original_reg Reg Pos =>
-
-original_reg/2
-
-extract_next_element D1=xy | original_reg Reg P1 | get_tuple_element Reg P2 D2=xy | \
-succ(P1, P2) | succ(D1, D2) => \
-  extract_next_element2 D1 | original_reg Reg P2
-
-extract_next_element2 D1=xy | original_reg Reg P1 | get_tuple_element Reg P2 D2=xy | \
-succ(P1, P2) | succ2(D1, D2) => \
-  extract_next_element3 D1 | original_reg Reg P2
-
-#extract_next_element3 D1=xy | original_reg Reg P1 | get_tuple_element Reg P2 D2=xy | \
-#succ(P1, P2) | succ3(D1, D2) => \
-#  extract_next_element4 D1 | original_reg Reg P2
-
-%macro: extract_next_element ExtractNextElement -pack
-extract_next_element x
-extract_next_element y
 
-%macro: extract_next_element2 ExtractNextElement2 -pack
-extract_next_element2 x
-extract_next_element2 y
+get_tuple_element Reg=x P1 D1=x | get_tuple_element Reg=x P2 D2=x | \
+   get_tuple_element Reg=x P3 D3=x | \
+   succ(P1, P2) | succ(P2, P3) | \
+   succ(D1, D2) | succ(D2, D3) => i_get_tuple_element3 Reg P1 D1
 
-%macro: extract_next_element3 ExtractNextElement3 -pack
-extract_next_element3 x
-extract_next_element3 y
+get_tuple_element Reg=x P1 D1=x | get_tuple_element Reg=x P2 D2=x | \
+   succ(P1, P2) | succ(D1, D2) => i_get_tuple_element2 Reg P1 D1
 
-#%macro: extract_next_element4 ExtractNextElement4 -pack
-#extract_next_element4 x
-#extract_next_element4 y
+get_tuple_element Reg P Dst => i_get_tuple_element Reg P Dst
 
 is_integer Fail=f i =>
 is_integer Fail=f an => jump Fail
 is_integer Fail Literal=q => move Literal x | is_integer Fail x
 
-is_integer Fail=f S=rx | allocate Need Regs => is_integer_allocate Fail S Need Regs
+is_integer Fail=f S=x | allocate Need Regs => is_integer_allocate Fail S Need Regs
 
 %macro: is_integer_allocate IsIntegerAllocate -fail_action
 is_integer_allocate f x I I
-is_integer_allocate f r I I
 
 %macro: is_integer IsInteger -fail_action
 is_integer f x
 is_integer f y
-is_integer f r
 
 is_list Fail=f n =>
 is_list Fail Literal=q => move Literal x | is_list Fail x
 is_list Fail=f c => jump Fail
 %macro: is_list IsList -fail_action
-is_list f r
 is_list f x
 %cold
 is_list f y
 %hot
 
-is_nonempty_list Fail=f S=rx | allocate Need Rs => is_nonempty_list_allocate Fail S Need Rs
+is_nonempty_list Fail=f S=x | allocate Need Rs => is_nonempty_list_allocate Fail S Need Rs
 
 %macro:is_nonempty_list_allocate IsNonemptyListAllocate -fail_action -pack
-is_nonempty_list_allocate f x I t
 is_nonempty_list_allocate f r I t
+is_nonempty_list_allocate f x I t
 
-is_nonempty_list F=f r | test_heap I1 I2 => is_non_empty_list_test_heap F r I1 I2
+is_nonempty_list F=f x==0 | test_heap I1 I2 => is_non_empty_list_test_heap F I1 I2
 
 %macro: is_non_empty_list_test_heap IsNonemptyListTestHeap -fail_action -pack
-is_non_empty_list_test_heap f r I t
+is_non_empty_list_test_heap f I t
+
+is_nonempty_list Fail=f S=x | get_list S D1=x D2=x => \
+  is_nonempty_list_get_list Fail S D1 D2
+
+%macro: is_nonempty_list_get_list IsNonemptyListGetList -fail_action -pack
+is_nonempty_list_get_list f r x x
+is_nonempty_list_get_list f x x x
 
 %macro: is_nonempty_list IsNonemptyList -fail_action
 is_nonempty_list f x
 is_nonempty_list f y
-is_nonempty_list f r
 
 %macro: is_atom IsAtom -fail_action
 is_atom f x
-is_atom f r
 %cold
 is_atom f y
 %hot
@@ -775,7 +702,6 @@ is_atom Fail=f a =>
 is_atom Fail=f niq => jump Fail
 
 %macro: is_float IsFloat -fail_action
-is_float f r
 is_float f x
 %cold
 is_float f y
@@ -789,12 +715,10 @@ is_nil Fail=f qia => jump Fail
 %macro: is_nil IsNil -fail_action
 is_nil f x
 is_nil f y
-is_nil f r
 
 is_binary Fail Literal=q => move Literal x | is_binary Fail x
 is_binary Fail=f c => jump Fail
 %macro: is_binary IsBinary -fail_action
-is_binary f r
 is_binary f x
 %cold
 is_binary f y
@@ -806,7 +730,6 @@ is_bitstr Fail Term => is_bitstring Fail Term
 is_bitstring Fail Literal=q => move Literal x | is_bitstring Fail x
 is_bitstring Fail=f c => jump Fail
 %macro: is_bitstring IsBitstring -fail_action
-is_bitstring f r
 is_bitstring f x
 %cold
 is_bitstring f y
@@ -814,7 +737,6 @@ is_bitstring f y
 
 is_reference Fail=f cq => jump Fail
 %macro: is_reference IsRef -fail_action
-is_reference f r
 is_reference f x
 %cold
 is_reference f y
@@ -822,7 +744,6 @@ is_reference f y
 
 is_pid Fail=f cq => jump Fail
 %macro: is_pid IsPid -fail_action
-is_pid f r
 is_pid f x
 %cold
 is_pid f y
@@ -830,7 +751,6 @@ is_pid f y
 
 is_port Fail=f cq => jump Fail
 %macro: is_port IsPort -fail_action
-is_port f r
 is_port f x
 %cold
 is_port f y
@@ -842,14 +762,12 @@ is_boolean Fail=f ac => jump Fail
 
 %cold
 %macro: is_boolean IsBoolean -fail_action
-is_boolean f r
 is_boolean f x
 is_boolean f y
 %hot
 
 is_function2 Fail=f acq Arity => jump Fail
 is_function2 Fail=f Fun a => jump Fail
-is_function2 Fail Fun Literal=q => move Literal x | is_function2 Fail Fun x
 
 is_function2 f s s
 %macro: is_function2 IsFunction2 -fail_action
@@ -989,76 +907,76 @@ call_ext_only u==3 u$func:erlang:hibernate/3 => i_hibernate
 %unless USE_VM_PROBES
 
 call_ext Arity u$func:erlang:dt_get_tag/0 => \
-    move a=am_undefined r
+    move a=am_undefined x=0
 call_ext_last Arity u$func:erlang:dt_get_tag/0 D => \
-    move a=am_undefined r | deallocate D | return
+    move a=am_undefined x=0 | deallocate D | return
 call_ext_only Arity u$func:erlang:dt_get_tag/0 => \
-    move a=am_undefined r | return
-
-move Any r | call_ext Arity u$func:erlang:dt_put_tag/1 => \
-    move a=am_undefined r
-move Any r | call_ext_last Arity u$func:erlang:dt_put_tag/1 D => \
-    move a=am_undefined r | deallocate D | return
-move Any r | call_ext_only Arity u$func:erlang:dt_put_tag/1 => \
-    move a=am_undefined r | return
+    move a=am_undefined x=0 | return
+
+move Any x==0 | call_ext Arity u$func:erlang:dt_put_tag/1 => \
+    move a=am_undefined x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_put_tag/1 D => \
+    move a=am_undefined x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_put_tag/1 => \
+    move a=am_undefined x=0 | return
 call_ext Arity u$func:erlang:dt_put_tag/1 => \
-    move a=am_undefined r
+    move a=am_undefined x=0
 call_ext_last Arity u$func:erlang:dt_put_tag/1 D => \
-    move a=am_undefined r | deallocate D | return
+    move a=am_undefined x=0 | deallocate D | return
 call_ext_only Arity u$func:erlang:dt_put_tag/1 => \
-    move a=am_undefined r | return
+    move a=am_undefined x=0 | return
 
 call_ext Arity u$func:erlang:dt_get_tag_data/0 => \
-    move a=am_undefined r
+    move a=am_undefined x=0
 call_ext_last Arity u$func:erlang:dt_get_tag_data/0 D => \
-    move a=am_undefined r | deallocate D | return
+    move a=am_undefined x=0 | deallocate D | return
 call_ext_only Arity u$func:erlang:dt_get_tag_data/0 => \
-    move a=am_undefined r | return
-
-move Any r | call_ext Arity u$func:erlang:dt_spread_tag/1 => \
-    move a=am_true r
-move Any r | call_ext_last Arity u$func:erlang:dt_spread_tag/1 D => \
-    move a=am_true r | deallocate D | return
-move Any r | call_ext_only Arity u$func:erlang:dt_spread_tag/1 => \
-    move a=am_true r | return
+    move a=am_undefined x=0 | return
+
+move Any x==0 | call_ext Arity u$func:erlang:dt_spread_tag/1 => \
+    move a=am_true x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_spread_tag/1 D => \
+    move a=am_true x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_spread_tag/1 => \
+    move a=am_true x=0 | return
 call_ext Arity u$func:erlang:dt_spread_tag/1 => \
-    move a=am_true r
+    move a=am_true x=0
 call_ext_last Arity u$func:erlang:dt_spread_tag/1 D => \
-    move a=am_true r | deallocate D | return
+    move a=am_true x=0 | deallocate D | return
 call_ext_only Arity u$func:erlang:dt_spread_tag/1 => \
-    move a=am_true r | return
-
-move Any r | call_ext Arity u$func:erlang:dt_restore_tag/1 => \
-    move a=am_true r
-move Any r | call_ext_last Arity u$func:erlang:dt_restore_tag/1 D => \
-    move a=am_true r | deallocate D | return
-move Any r | call_ext_only Arity u$func:erlang:dt_restore_tag/1 => \
-    move a=am_true r | return
+    move a=am_true x=0 | return
+
+move Any x==0 | call_ext Arity u$func:erlang:dt_restore_tag/1 => \
+    move a=am_true x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_restore_tag/1 D => \
+    move a=am_true x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_restore_tag/1 => \
+    move a=am_true x=0 | return
 call_ext Arity u$func:erlang:dt_restore_tag/1 => \
-    move a=am_true r
+    move a=am_true x=0
 call_ext_last Arity u$func:erlang:dt_restore_tag/1 D => \
-    move a=am_true r | deallocate D | return
+    move a=am_true x=0 | deallocate D | return
 call_ext_only Arity u$func:erlang:dt_restore_tag/1 => \
-    move a=am_true r | return
-
-move Any r | call_ext Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \
-    move Any r
-move Any r | call_ext_last Arity u$func:erlang:dt_prepend_vm_tag_data/1 D => \
-    move Any r | deallocate D | return
-move Any r | call_ext_only Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \
-    move Any r | return
+    move a=am_true x=0 | return
+
+move Any x==0 | call_ext Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \
+    move Any x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_prepend_vm_tag_data/1 D => \
+    move Any x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \
+    move Any x=0 | return
 call_ext Arity u$func:erlang:dt_prepend_vm_tag_data/1 =>
 call_ext_last Arity u$func:erlang:dt_prepend_vm_tag_data/1 D => \
     deallocate D | return
 call_ext_only Arity u$func:erlang:dt_prepend_vm_tag_data/1 => \
     return
 
-move Any r | call_ext Arity u$func:erlang:dt_append_vm_tag_data/1 => \
-    move Any r
-move Any r | call_ext_last Arity u$func:erlang:dt_append_vm_tag_data/1 D => \
-    move Any r | deallocate D | return
-move Any r | call_ext_only Arity u$func:erlang:dt_append_vm_tag_data/1 => \
-    move Any r | return
+move Any x==0 | call_ext Arity u$func:erlang:dt_append_vm_tag_data/1 => \
+    move Any x=0
+move Any x==0 | call_ext_last Arity u$func:erlang:dt_append_vm_tag_data/1 D => \
+    move Any x=0 | deallocate D | return
+move Any x==0 | call_ext_only Arity u$func:erlang:dt_append_vm_tag_data/1 => \
+    move Any x=0 | return
 call_ext Arity u$func:erlang:dt_append_vm_tag_data/1 =>
 call_ext_last Arity u$func:erlang:dt_append_vm_tag_data/1 D => \
     deallocate D | return
@@ -1066,7 +984,7 @@ call_ext_only Arity u$func:erlang:dt_append_vm_tag_data/1 => \
     return
 
 # Can happen after one of the transformations above.
-move Discarded r | move Something r => move Something r
+move Discarded x==0 | move Something x==0 => move Something x=0
 
 %endif
 
@@ -1091,9 +1009,9 @@ call_ext_only Ar=u Bif=u$is_bif => \
 # with call instructions.
 #
 
-move S=c r | call_ext Ar=u Func=u$is_not_bif => i_move_call_ext S r Func
-move S=c r | call_ext_last Ar=u Func=u$is_not_bif D => i_move_call_ext_last Func D S r
-move S=c r | call_ext_only Ar=u Func=u$is_not_bif => i_move_call_ext_only Func S r 
+move S=c x==0 | call_ext Ar=u Func=u$is_not_bif => i_move_call_ext S Func
+move S=c x==0 | call_ext_last Ar=u Func=u$is_not_bif D => i_move_call_ext_last Func D S
+move S=c x==0 | call_ext_only Ar=u Func=u$is_not_bif => i_move_call_ext_only Func S
 
 call_ext Ar Func        => i_call_ext Func
 call_ext_last Ar Func D => i_call_ext_last Func D
@@ -1120,78 +1038,71 @@ bif0 u$bif:erlang:node/0 Dst=d => node Dst
 
 bif1 Fail Bif=u$bif:erlang:get/1 Src=s Dst=d => i_get Src Dst
 
-bif2 Jump=j u$bif:erlang:element/2 S1=s S2=rxy Dst=d => gen_element(Jump, S1, S2, Dst)
+bif2 Jump=j u$bif:erlang:element/2 S1=s S2=xy Dst=d => gen_element(Jump, S1, S2, Dst)
 
-bif1 Fail Bif Literal=q Dst => move Literal x | bif1 Fail Bif x Dst
 bif1 p Bif S1 Dst => bif1_body Bif S1 Dst
 
-bif1_body Bif Literal=q Dst => move Literal x | bif1_body Bif x Dst
-
-bif2 p Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2_body Bif Dst
-bif2 Fail Bif S1 S2 Dst => i_fetch S1 S2 | i_bif2 Fail Bif Dst
+bif2 p Bif S1 S2 Dst => i_bif2_body Bif S1 S2 Dst
+bif2 Fail Bif S1 S2 Dst => i_bif2 Fail Bif S1 S2 Dst
 
 i_get s d
 
 %macro: self Self
-self r
 self x
 self y
 
 %macro: node Node
-node r
 node x
 %cold
 node y
 %hot
 
-i_fast_element r j I d
-i_fast_element x j I d
-i_fast_element y j I d
+i_fast_element j x I d
+i_fast_element j y I d
 
-i_element r j s d
-i_element x j s d
-i_element y j s d
+i_element j x s d
+i_element j y s d
 
 bif1 f b s d
 bif1_body b s d
-i_bif2 f b d
-i_bif2_body b d
+i_bif2 f b s s d
+i_bif2_body b s s d
 
 #
 # Internal calls.
 #
 
-move S=c r | call Ar P=f => i_move_call S r P
-move S=s r | call Ar P=f => move_call S r P
+move S=c x==0 | call Ar P=f => i_move_call S P
+move S=s x==0 | call Ar P=f => move_call S P
 
-i_move_call c r f
+i_move_call c f
 
 %macro:move_call MoveCall -arg_f -size -nonext
-move_call/3
+move_call/2
 
-move_call x r f
-move_call y r f
+move_call x f
+move_call y f
 
-move S=c r | call_last Ar P=f D => i_move_call_last P D S r
-move S r | call_last Ar P=f D => move_call_last S r P D
+move S=c x==0 | call_last Ar P=f D => i_move_call_last P D S
+move S x==0 | call_last Ar P=f D => move_call_last S P D
 
-i_move_call_last f P c r
+i_move_call_last f P c
 
 %macro:move_call_last MoveCallLast -arg_f -nonext -pack
 
-move_call_last/4
-move_call_last x r f Q
-move_call_last y r f Q
+move_call_last/3
+move_call_last x f Q
+move_call_last y f Q
 
-move S=c r | call_only Ar P=f => i_move_call_only P S r
-move S=x r | call_only Ar P=f => move_call_only S r P
+move S=c x==0 | call_only Ar P=f => i_move_call_only P S
+move S=x x==0 | call_only Ar P=f => move_call_only S P
 
-i_move_call_only f c r
+i_move_call_only f c
 
 %macro:move_call_only MoveCallOnly -arg_f -nonext
-move_call_only/3
+move_call_only/2
 
-move_call_only x r f
+move_call_only x f
 
 call Ar Func        => i_call Func
 call_last Ar Func D => i_call_last Func D
@@ -1205,9 +1116,9 @@ i_call_ext e
 i_call_ext_last e P
 i_call_ext_only e
 
-i_move_call_ext c r e
-i_move_call_ext_last e P c r
-i_move_call_ext_only e c r
+i_move_call_ext c e
+i_move_call_ext_last e P c
+i_move_call_ext_only e c
 
 # Fun calls.
 
@@ -1227,7 +1138,6 @@ i_make_fun I t
 %macro: is_function IsFunction -fail_action
 is_function f x
 is_function f y
-is_function f r
 is_function Fail=f c => jump Fail
 
 func_info M F A => i_func_info u M F A
@@ -1239,131 +1149,105 @@ func_info M F A => i_func_info u M F A
 %cold
 bs_start_match2 Fail=f ica X Y D => jump Fail
 bs_start_match2 Fail Bin X Y D => i_bs_start_match2 Bin Fail X Y D
-i_bs_start_match2 r f I I d
 i_bs_start_match2 x f I I d
 i_bs_start_match2 y f I I d
 
 bs_save2 Reg Index => gen_bs_save(Reg, Index)
-i_bs_save2 r I
 i_bs_save2 x I
 
 bs_restore2 Reg Index => gen_bs_restore(Reg, Index)
-i_bs_restore2 r I
 i_bs_restore2 x I
 
 # Matching integers
 bs_match_string Fail Ms Bits Val => i_bs_match_string Ms Fail Bits Val
 
-i_bs_match_string r f I I
 i_bs_match_string x f I I
 
 # Fetching integers from binaries.
-bs_get_integer2 Fail=f Ms=rx Live=u Sz=sq Unit=u Flags=u Dst=d => \
+bs_get_integer2 Fail=f Ms=x Live=u Sz=sq Unit=u Flags=u Dst=d => \
 			gen_get_integer2(Fail, Ms, Live, Sz, Unit, Flags, Dst)
 
-i_bs_get_integer_small_imm r I f I d
 i_bs_get_integer_small_imm x I f I d
-i_bs_get_integer_imm r I I f I d
 i_bs_get_integer_imm x I I f I d
-i_bs_get_integer f I I d
-i_bs_get_integer_8 r f d
+i_bs_get_integer f I I s s d
 i_bs_get_integer_8 x f d
-i_bs_get_integer_16 r f d
 i_bs_get_integer_16 x f d
-i_bs_get_integer_32 r f I d
 i_bs_get_integer_32 x f I d
 
 # Fetching binaries from binaries.
-bs_get_binary2 Fail=f Ms=rx Live=u Sz=sq Unit=u Flags=u Dst=d => \
+bs_get_binary2 Fail=f Ms=x Live=u Sz=sq Unit=u Flags=u Dst=d => \
 			gen_get_binary2(Fail, Ms, Live, Sz, Unit, Flags, Dst)
 
 %macro: i_bs_get_binary_imm2 BsGetBinaryImm_2 -fail_action -gen_dest
 %macro: i_bs_get_binary2 BsGetBinary_2 -fail_action -gen_dest
 %macro: i_bs_get_binary_all2 BsGetBinaryAll_2 -fail_action -gen_dest
 
-i_bs_get_binary_imm2 f r I I I d
 i_bs_get_binary_imm2 f x I I I d
-i_bs_get_binary2 f r I s I d
 i_bs_get_binary2 f x I s I d
-i_bs_get_binary_all2 f r I I d
 i_bs_get_binary_all2 f x I I d
-i_bs_get_binary_all_reuse r f I
 i_bs_get_binary_all_reuse x f I
 
 # Fetching float from binaries.
-bs_get_float2 Fail=f Ms=rx Live=u Sz=s Unit=u Flags=u Dst=d => \
+bs_get_float2 Fail=f Ms=x Live=u Sz=s Unit=u Flags=u Dst=d => \
 		gen_get_float2(Fail, Ms, Live, Sz, Unit, Flags, Dst)
 
-bs_get_float2 Fail=f Ms=rx Live=u Sz=q Unit=u Flags=u Dst=d => jump Fail
+bs_get_float2 Fail=f Ms=x Live=u Sz=q Unit=u Flags=u Dst=d => jump Fail
 
 %macro: i_bs_get_float2 BsGetFloat2 -fail_action -gen_dest
-i_bs_get_float2 f r I s I d
 i_bs_get_float2 f x I s I d
 
 # Miscellanous
 
-bs_skip_bits2 Fail=f Ms=rx Sz=s Unit=u Flags=u => \
-			gen_skip_bits2(Fail, Ms, Sz, Unit, Flags)
-bs_skip_bits2 Fail=f Ms=rx Sz=q Unit=u Flags=u => \
+bs_skip_bits2 Fail=f Ms=x Sz=sq Unit=u Flags=u => \
 			gen_skip_bits2(Fail, Ms, Sz, Unit, Flags)
 
 %macro: i_bs_skip_bits_imm2 BsSkipBitsImm2 -fail_action
-i_bs_skip_bits_imm2 f r I
 i_bs_skip_bits_imm2 f x I
 
 %macro: i_bs_skip_bits2 BsSkipBits2 -fail_action
-i_bs_skip_bits2 f r x I
-i_bs_skip_bits2 f r y I
 i_bs_skip_bits2 f x x I
-i_bs_skip_bits2 f x r I
 i_bs_skip_bits2 f x y I
 
 %macro: i_bs_skip_bits_all2 BsSkipBitsAll2 -fail_action
-i_bs_skip_bits_all2 f r I
 i_bs_skip_bits_all2 f x I 
 
-bs_test_tail2 Fail=f Ms=rx Bits=u==0 => bs_test_zero_tail2 Fail Ms
-bs_test_tail2 Fail=f Ms=rx Bits=u => bs_test_tail_imm2 Fail Ms Bits
-bs_test_zero_tail2 f r
+bs_test_tail2 Fail=f Ms=x Bits=u==0 => bs_test_zero_tail2 Fail Ms
+bs_test_tail2 Fail=f Ms=x Bits=u => bs_test_tail_imm2 Fail Ms Bits
 bs_test_zero_tail2 f x
-bs_test_tail_imm2 f r I
 bs_test_tail_imm2 f x I
 
 bs_test_unit F Ms Unit=u==8 => bs_test_unit8 F Ms
-bs_test_unit f r I
 bs_test_unit f x I
-bs_test_unit8 f r
 bs_test_unit8 f x
 
-bs_context_to_binary r
+# An y register operand for bs_context_to_binary is rare,
+# but can happen because of inlining.
+
+bs_context_to_binary Y=y => move Y x | bs_context_to_binary x
+
 bs_context_to_binary x
-bs_context_to_binary y
 
 #
 # Utf8/utf16/utf32 support. (R12B-5)
 #
-bs_get_utf8 Fail=f Ms=rx u u Dst=d => i_bs_get_utf8 Ms Fail Dst
-i_bs_get_utf8 r f d
+bs_get_utf8 Fail=f Ms=x u u Dst=d => i_bs_get_utf8 Ms Fail Dst
 i_bs_get_utf8 x f d
 
-bs_skip_utf8 Fail=f Ms=rx u u => i_bs_get_utf8 Ms Fail x
+bs_skip_utf8 Fail=f Ms=x u u => i_bs_get_utf8 Ms Fail x
 
-bs_get_utf16 Fail=f Ms=rx u Flags=u Dst=d => i_bs_get_utf16 Ms Fail Flags Dst
-bs_skip_utf16 Fail=f Ms=rx u Flags=u => i_bs_get_utf16 Ms Fail Flags x
+bs_get_utf16 Fail=f Ms=x u Flags=u Dst=d => i_bs_get_utf16 Ms Fail Flags Dst
+bs_skip_utf16 Fail=f Ms=x u Flags=u => i_bs_get_utf16 Ms Fail Flags x
 
-i_bs_get_utf16 r f I d
 i_bs_get_utf16 x f I d
 
-bs_get_utf32 Fail=f Ms=rx Live=u Flags=u Dst=d => \
+bs_get_utf32 Fail=f Ms=x Live=u Flags=u Dst=d => \
 	bs_get_integer2 Fail Ms Live i=32 u=1 Flags Dst | \
-	i_fetch Dst Ms | \
-	i_bs_validate_unicode_retract Fail
-bs_skip_utf32 Fail=f Ms=rx Live=u Flags=u => \
+	i_bs_validate_unicode_retract Fail Dst Ms
+bs_skip_utf32 Fail=f Ms=x Live=u Flags=u => \
 	bs_get_integer2 Fail Ms Live i=32 u=1 Flags x | \
-	i_fetch x Ms | \
-	i_bs_validate_unicode_retract Fail
+	i_bs_validate_unicode_retract Fail x Ms
 
-i_bs_validate_unicode_retract j
+i_bs_validate_unicode_retract j s s
 %hot
 
 #
@@ -1385,13 +1269,12 @@ bs_init2 Fail Sz=u Words Regs Flags Dst => \
 bs_init2 Fail Sz Words=u==0 Regs Flags Dst => \
   i_bs_init_fail Sz Fail Regs Dst
 bs_init2 Fail Sz Words Regs Flags Dst => \
-  i_fetch Sz r | i_bs_init_fail_heap Words Fail Regs Dst
+  i_bs_init_fail_heap Sz Words Fail Regs Dst
 
-i_bs_init_fail r j I d
 i_bs_init_fail x j I d
 i_bs_init_fail y j I d
 
-i_bs_init_fail_heap I j I d
+i_bs_init_fail_heap s I j I d
 
 i_bs_init I I d
 i_bs_init_heap_bin I I d
@@ -1408,39 +1291,35 @@ bs_init_bits Fail Sz=u Words Regs Flags Dst =>  i_bs_init_bits_heap Sz Words Reg
 bs_init_bits Fail Sz Words=u==0 Regs Flags Dst => \
   i_bs_init_bits_fail Sz Fail Regs Dst
 bs_init_bits Fail Sz Words Regs Flags Dst => \
-  i_fetch Sz r | i_bs_init_bits_fail_heap Words Fail Regs Dst
+  i_bs_init_bits_fail_heap Sz Words Fail Regs Dst
 
-i_bs_init_bits_fail r j I d
 i_bs_init_bits_fail x j I d
 i_bs_init_bits_fail y j I d
 
-i_bs_init_bits_fail_heap I j I d
+i_bs_init_bits_fail_heap s I j I d
 
 i_bs_init_bits I I d
 i_bs_init_bits_heap I I I d
 
 bs_add Fail S1=i==0 S2 Unit=u==1 D => move S2 D
-bs_add Fail S1 S2 Unit D => i_fetch S1 S2 | i_bs_add Fail Unit D
 
-i_bs_add j I d
+bs_add j s s I d
 
 bs_append Fail Size Extra Live Unit Bin Flags Dst => \
-  i_fetch Size Bin | i_bs_append Fail Extra Live Unit Dst
+  move Bin x | i_bs_append Fail Extra Live Unit Size Dst
 
 bs_private_append Fail Size Unit Bin Flags Dst => \
-  i_fetch Size Bin | i_bs_private_append Fail Unit Dst
+  i_bs_private_append Fail Unit Size Bin Dst
 
 bs_init_writable
 
-i_bs_append j I I I d
-i_bs_private_append j I d
+i_bs_append j I I I s d
+i_bs_private_append j I s s d
 
 #
 # Storing integers into binaries.
 #
 
-bs_put_integer Fail=j Sz=s Unit=u Flags=u Literal=q => \
-	move Literal x | bs_put_integer Fail Sz Unit Flags x
 bs_put_integer Fail=j Sz=sq Unit=u Flags=u Src=s => \
 			gen_put_integer(Fail, Sz, Unit, Flags, Src)
 
@@ -1454,32 +1333,22 @@ i_new_bs_put_integer_imm j I I s
 # Utf8/utf16/utf32 support. (R12B-5)
 #
 
-bs_utf8_size Fail Literal=q Dst=d => \
-	move Literal x | bs_utf8_size Fail x Dst
 bs_utf8_size j Src=s Dst=d => i_bs_utf8_size Src Dst
 
 i_bs_utf8_size s d
 
-bs_utf16_size Fail Literal=q Dst=d => \
-	move Literal x | bs_utf16_size Fail x Dst
 bs_utf16_size j Src=s Dst=d => i_bs_utf16_size Src Dst
 
 i_bs_utf16_size s d
 
-bs_put_utf8 Fail=j Flags=u Literal=q => \
-	move Literal x | bs_put_utf8 Fail Flags x
 bs_put_utf8 Fail u Src=s => i_bs_put_utf8 Fail Src
 
 i_bs_put_utf8 j s
 
-bs_put_utf16 Fail=j Flags=u Literal=q => \
-	move Literal x | bs_put_utf16 Fail Flags x
 bs_put_utf16 Fail Flags=u Src=s => i_bs_put_utf16 Fail Flags Src
 
 i_bs_put_utf16 j I s
 
-bs_put_utf32 Fail=j Flags=u Literal=q => \
-	move Literal x | bs_put_utf32 Fail Flags x
 bs_put_utf32 Fail=j Flags=u Src=s => \
    i_bs_validate_unicode Fail Src | bs_put_integer Fail i=32 u=1 Flags Src
 
@@ -1490,9 +1359,6 @@ i_bs_validate_unicode j s
 #
 bs_put_float Fail Sz=q Unit Flags Val => badarg Fail
 
-bs_put_float Fail=j Sz Unit=u Flags=u Literal=q => \
-	move Literal x | bs_put_float Fail Sz Unit Flags x
-
 bs_put_float Fail=j Sz=s Unit=u Flags=u Src=s => \
 			gen_put_float(Fail, Sz, Unit, Flags, Src)
 
@@ -1506,8 +1372,6 @@ i_new_bs_put_float_imm j I I s
 # Storing binaries into binaries.
 #
 
-bs_put_binary Fail Sz Unit Flags Literal=q => \
- 	move Literal x | bs_put_binary Fail Sz Unit Flags x
 bs_put_binary Fail=j Sz=s Unit=u Flags=u Src=s => \
 			gen_put_binary(Fail, Sz, Unit, Flags, Src)
 
@@ -1600,7 +1464,6 @@ is_map Fail Lit=q | literal_is_map(Lit) =>
 is_map Fail cq => jump Fail
 
 %macro: is_map IsMap -fail_action
-is_map f r
 is_map f x
 is_map f y
 
@@ -1611,101 +1474,120 @@ has_map_fields Fail Src Size Rest=* => \
 
 ## Transform get_map_elements(s) #{ K1 := V1, K2 := V2 }
 
-get_map_elements Fail Src=rxy Size=u==2 Rest=* => \
+get_map_elements Fail Src=xy Size=u==2 Rest=* => \
     gen_get_map_element(Fail, Src, Size, Rest)
 get_map_elements Fail Src Size Rest=* | map_key_sort(Size, Rest) => \
    gen_get_map_elements(Fail, Src, Size, Rest)
 
 i_get_map_elements f s I
 
-i_get_map_element Fail Src=rxy Key=ry Dst => \
+i_get_map_element Fail Src=xy Key=y Dst => \
     move Key x | i_get_map_element Fail Src x Dst
 
 %macro: i_get_map_element_hash GetMapElementHash -fail_action
-i_get_map_element_hash f r c I r
-i_get_map_element_hash f x c I r
-i_get_map_element_hash f y c I r
-i_get_map_element_hash f r c I x
 i_get_map_element_hash f x c I x
 i_get_map_element_hash f y c I x
-i_get_map_element_hash f r c I y
 i_get_map_element_hash f x c I y
 i_get_map_element_hash f y c I y
 
 %macro: i_get_map_element GetMapElement -fail_action
-i_get_map_element f r x r
-i_get_map_element f x x r
-i_get_map_element f y x r
-i_get_map_element f r x x
 i_get_map_element f x x x
 i_get_map_element f y x x
-i_get_map_element f r x y
 i_get_map_element f x x y
 i_get_map_element f y x y
 
 #
+# Convert the plus operations to a generic plus instruction.
+#
+gen_plus/5
+gen_minus/5
+
+gc_bif1 Fail Live u$bif:erlang:splus/1 Src Dst => \
+   gen_plus Fail Live Src i Dst
+gc_bif2 Fail Live u$bif:erlang:splus/2 S1 S2 Dst => \
+   gen_plus Fail Live S1 S2 Dst
+
+gc_bif1 Fail Live u$bif:erlang:sminus/1 Src Dst => \
+   gen_minus Fail Live i Src Dst
+gc_bif2 Fail Live u$bif:erlang:sminus/2 S1 S2 Dst => \
+   gen_minus Fail Live S1 S2 Dst
+
+#
 # Optimize addition and subtraction of small literals using
 # the i_increment/4 instruction (in bodies, not in guards).
 #
 
-gc_bif2 p Live u$bif:erlang:splus/2 Int=i Reg=d Dst => \
+gen_plus p Live Int=i Reg=d Dst => \
 	gen_increment(Reg, Int, Live, Dst)
-gc_bif2 p Live u$bif:erlang:splus/2 Reg=d Int=i Dst => \
+gen_plus p Live Reg=d Int=i Dst => \
 	gen_increment(Reg, Int, Live, Dst)
 
-gc_bif2 p Live u$bif:erlang:sminus/2 Reg=d Int=i Dst | \
-	negation_is_small(Int) => \
+gen_minus p Live Reg=d Int=i Dst | negation_is_small(Int) => \
 	gen_increment_from_minus(Reg, Int, Live, Dst)
 
 #
 # GCing arithmetic instructions.
 #
 
-gc_bif2 Fail I u$bif:erlang:splus/2 S1=x S2=x Dst=d => i_plus Fail I S1 S2 Dst
-gc_bif2 Fail I u$bif:erlang:splus/2 S1 S2 Dst=d => i_fetch S1 S2 | i_plus Fail I Dst
-gc_bif2 Fail I u$bif:erlang:sminus/2 S1=x S2=x Dst=d => i_minus Fail I S1 S2 Dst
-gc_bif2 Fail I u$bif:erlang:sminus/2 S1 S2 Dst=d => i_fetch S1 S2 | i_minus Fail I Dst
-gc_bif2 Fail I u$bif:erlang:stimes/2 S1 S2 Dst=d => i_fetch S1 S2 | i_times Fail I Dst
-gc_bif2 Fail I u$bif:erlang:div/2 S1 S2 Dst=d => i_fetch S1 S2 | i_m_div Fail I Dst
+gen_plus Fail Live Y=y X=x Dst => i_plus Fail Live X Y Dst
+gen_plus Fail Live S1 S2 Dst => i_plus Fail Live S1 S2 Dst
 
-gc_bif2 Fail I u$bif:erlang:intdiv/2 S1 S2 Dst=d => i_fetch S1 S2 | i_int_div Fail I Dst
-gc_bif2 Fail I u$bif:erlang:rem/2 S1=x S2=x Dst=d => i_rem Fail I S1 S2 Dst
-gc_bif2 Fail I u$bif:erlang:rem/2 S1 S2 Dst=d => i_fetch S1 S2 | i_rem Fail I Dst
+gen_minus Fail Live S1 S2 Dst => i_minus Fail Live S1 S2 Dst
 
-gc_bif2 Fail I u$bif:erlang:bsl/2 S1 S2 Dst=d => i_fetch S1 S2 | i_bsl Fail I Dst
-gc_bif2 Fail I u$bif:erlang:bsr/2 S1 S2 Dst=d => i_fetch S1 S2 | i_bsr Fail I Dst
+gc_bif2 Fail Live u$bif:erlang:stimes/2 S1 S2 Dst => \
+  i_times Fail Live S1 S2 Dst
 
-gc_bif2 Fail I u$bif:erlang:band/2 S1=x S2=c Dst=d => i_band Fail I S1 S2 Dst
-gc_bif2 Fail I u$bif:erlang:band/2 S1 S2 Dst=d => i_fetch S1 S2 | i_band Fail I Dst
-gc_bif2 Fail I u$bif:erlang:bor/2 S1 S2 Dst=d => i_fetch S1 S2 | i_bor Fail I Dst
-gc_bif2 Fail I u$bif:erlang:bxor/2 S1 S2 Dst=d => i_fetch S1 S2 | i_bxor Fail I Dst
+gc_bif2 Fail Live u$bif:erlang:div/2 S1 S2 Dst => \
+  i_m_div Fail Live S1 S2 Dst
+gc_bif2 Fail Live u$bif:erlang:intdiv/2 S1 S2 Dst => \
+  i_int_div Fail Live S1 S2 Dst
 
-gc_bif1 Fail I u$bif:erlang:bnot/1 Src Dst=d => i_int_bnot Fail Src I Dst
+gc_bif2 Fail Live u$bif:erlang:rem/2 S1 S2 Dst => \
+  i_rem Fail Live S1 S2 Dst
+
+gc_bif2 Fail Live u$bif:erlang:bsl/2 S1 S2 Dst => \
+  i_bsl Fail Live S1 S2 Dst
+gc_bif2 Fail Live u$bif:erlang:bsr/2 S1 S2 Dst => \
+  i_bsr Fail Live S1 S2 Dst
+
+gc_bif2 Fail Live u$bif:erlang:band/2 S1 S2 Dst => \
+  i_band Fail Live S1 S2 Dst
+
+gc_bif2 Fail Live u$bif:erlang:bor/2 S1 S2 Dst => \
+  i_bor Fail Live S1 S2 Dst
 
-gc_bif1 Fail I u$bif:erlang:sminus/1 Src Dst=d => i_fetch i Src | i_minus Fail I Dst
-gc_bif1 Fail I u$bif:erlang:splus/1 Src Dst=d => i_fetch i Src | i_plus Fail I Dst
+gc_bif2 Fail Live u$bif:erlang:bxor/2 S1 S2 Dst => \
+  i_bxor Fail Live S1 S2 Dst
+
+gc_bif1 Fail I u$bif:erlang:bnot/1 Src Dst=d => i_int_bnot Fail Src I Dst
 
 i_increment r I I d
 i_increment x I I d
 i_increment y I I d
 
 i_plus j I x x d
-i_plus j I d
+i_plus j I x y d
+i_plus j I s s d
+
 i_minus j I x x d
-i_minus j I d
-i_times j I d
-i_m_div j I d
-i_int_div j I d
+i_minus j I s s d
+
+i_times j I s s d
+
+i_m_div j I s s d
+i_int_div j I s s d
+
 i_rem j I x x d
-i_rem j I d
+i_rem j I s s d
 
-i_bsl j I d
-i_bsr j I d
+i_bsl j I s s d
+i_bsr j I s s d
 
 i_band j I x c d
-i_band j I d
-i_bor j I d
-i_bxor j I d
+i_band j I s s d
+
+i_bor j I s s d
+i_bxor j I s s d
 
 i_int_bnot j s I d
 
@@ -1731,21 +1613,18 @@ gc_bif2 Fail I Bif S1 S2 Dst => \
 gc_bif3 Fail I Bif S1 S2 S3 Dst => \
 	gen_guard_bif3(Fail, I, Bif, S1, S2, S3, Dst)
 
-i_gc_bif1 Fail Bif V=q Live D => move V x | i_gc_bif1 Fail Bif x Live D
-
 i_gc_bif1 j I s I d
 
-ii_gc_bif2/6
-
-ii_gc_bif2 Fail Bif S1 S2 Live D => i_fetch S1 S2 | i_gc_bif2 Fail Bif Live D
-
-i_gc_bif2 j I I d
+i_gc_bif2 j I I s s d
 
 ii_gc_bif3/7
 
-ii_gc_bif3 Fail Bif S1 S2 S3 Live D => move S1 x | i_fetch S2 S3 | i_gc_bif3 Fail Bif x Live D
+# A specific instruction can only have 6 operands, so we must
+# pass one of the arguments in an x register.
+ii_gc_bif3 Fail Bif Live S1 S2 S3 Dst => \
+  move S1 x | i_gc_bif3 Fail Bif Live S2 S3 Dst
 
-i_gc_bif3 j I s I d
+i_gc_bif3 j I I s s d
 
 #
 # The following instruction is specially handled in beam_load.c
diff --git a/erts/emulator/utils/beam_makeops b/erts/emulator/utils/beam_makeops
index e90ed94187..f805e7cc64 100755
--- a/erts/emulator/utils/beam_makeops
+++ b/erts/emulator/utils/beam_makeops
@@ -48,12 +48,17 @@ $pack_shift[4] = ['0', 'BEAM_LOOSE_SHIFT', # Only for 64 bit wordsize
 		  '(3*BEAM_LOOSE_SHIFT)'];
 
 $pack_mask[2]  = ['BEAM_LOOSE_MASK', $WHOLE_WORD];
-$pack_mask[3]  = ['BEAM_TIGHT_MASK', 'BEAM_TIGHT_MASK', 'BEAM_TIGHT_MASK'];
+$pack_mask[3]  = ['BEAM_TIGHT_MASK', 'BEAM_TIGHT_MASK', $WHOLE_WORD];
 $pack_mask[4]  = ['BEAM_LOOSE_MASK', # Only for 64 bit wordsize
 		  'BEAM_LOOSE_MASK',
 		  'BEAM_LOOSE_MASK',
 		  $WHOLE_WORD];
 
+# Mapping from packagable arguments to number of packed arguments per
+# word.  Initialized after the wordsize is known.
+
+my @args_per_word;
+
 # There are two types of instructions: generic and specific.
 # The generic instructions are those generated by the Beam compiler.
 # Corresponding to each generic instruction, there is generally a
@@ -176,11 +181,12 @@ sub define_type_bit {
     }
 
     # Composed types.
-    define_type_bit('d', $type_bit{'x'} | $type_bit{'y'} | $type_bit{'r'});
+    define_type_bit('d', $type_bit{'x'} | $type_bit{'y'});
     define_type_bit('c', $type_bit{'i'} | $type_bit{'a'} |
 		    $type_bit{'n'} | $type_bit{'q'});
     define_type_bit('s', $type_bit{'d'} | $type_bit{'i'} |
-		    $type_bit{'a'} | $type_bit{'n'});
+		    $type_bit{'a'} | $type_bit{'n'} |
+		    $type_bit{'q'});
     define_type_bit('j', $type_bit{'f'} | $type_bit{'p'});
 
     # Aliases (for matching purposes).
@@ -236,6 +242,20 @@ while (@ARGV && $ARGV[0] =~ /^-(.*)/) {
 }
 
 #
+# Initialize number of arguments per packed word.
+#
+
+$args_per_word[2] = 2;
+$args_per_word[3] = 3;
+$args_per_word[4] = 2;
+$args_per_word[5] = 3;
+$args_per_word[6] = 3;
+
+if ($wordsize == 64) {
+    $args_per_word[4] = 4;
+}
+
+#
 # Parse the input files.
 #
 
@@ -527,12 +547,16 @@ sub emulator_output {
 
 	    my(@bits) = (0) x ($max_spec_operands/2);
 	    my($i);
+	    my $involves_r = 0;
 	    for ($i = 0; $i < $max_spec_operands && defined $args[$i]; $i++) {
 		my $t = $args[$i];
-		if (defined $type_bit{$t}) {
-		    my $shift = $max_genop_types * ($i % 2);
-		    $bits[int($i/2)] |= $type_bit{$t} << $shift;
+		my $bits = $type_bit{$t};
+		if ($t eq 'r') {
+		    $bits |= $type_bit{'x'};
+		    $involves_r |= 1 << $i;
 		}
+		my $shift = $max_genop_types * ($i % 2);
+		$bits[int($i/2)] |= $bits << $shift;
 	    }
 
 	    printf "/* %3d */  ", $spec_opnum;
@@ -544,7 +568,7 @@ sub emulator_output {
 		$sep = ",";
 	    }
 	    $init .= "}";
-	    &init_item($print_name, $init, $size, $pack, $sign, 0);
+	    init_item($print_name, $init, $involves_r, $size, $pack, $sign, 0);
 	    $op_to_name[$spec_opnum] = $instr;
 	    $spec_opnum++;
 	}
@@ -601,11 +625,12 @@ sub emulator_output {
     print "#define MAX_GENERIC_OPCODE ", $num_file_opcodes-1, "\n";
     print "#define NUM_GENERIC_OPS ", scalar(@gen_opname), "\n";
     print "#define NUM_SPECIFIC_OPS ", scalar(@op_to_name), "\n";
+    print "#define SCRATCH_X_REG 1023\n";
     print "\n";
     print "#ifdef ARCH_64\n";
     print "#  define BEAM_WIDE_MASK 0xFFFFUL\n";
-    print "#  define BEAM_LOOSE_MASK 0x1FFFUL\n";
-    print "#  define BEAM_TIGHT_MASK 0x1FF8UL\n";
+    print "#  define BEAM_LOOSE_MASK 0xFFFFUL\n";
+    print "#  define BEAM_TIGHT_MASK 0xFFFFUL\n";
     print "#  define BEAM_WIDE_SHIFT 32\n";
     print "#  define BEAM_LOOSE_SHIFT 16\n";
     print "#  define BEAM_TIGHT_SHIFT 16\n";
@@ -730,7 +755,7 @@ sub init_item {
 	    print "${sep}NULL";
 	} elsif (/^\{/) {
 	    print "$sep$_";
-	} elsif (/^-?\d/) {
+	} elsif (/^-?\d+$/) {
 	    print "$sep$_";
 	} else {
 	    print "$sep\"$_\"";
@@ -892,6 +917,7 @@ sub basic_generator {
     my($var_decls) = '';
     my($gen_dest_arg) = 'StoreSimpleDest';
     my($i);
+    my($no_prefetch) = 0;
 
     # The following argument types should be included as macro arguments.
     my(%incl_arg) = ('c' => 1,
@@ -990,6 +1016,7 @@ sub basic_generator {
     #
 
     $flags =~ /-fail_action/ and do {
+	$no_prefetch = 1;
 	if (!defined $fail_type) {
 	    my($i);
 	    for ($i = 0; $i < @f_types; $i++) {
@@ -1036,6 +1063,12 @@ sub basic_generator {
 			 "I += $size + 1;",
 			 "goto $goto;",
 			 "}");
+	} elsif ($no_prefetch) {
+	    $code = join("\n",
+			 "{ $var_decls",
+			 $macro_code,
+			 "Next($size);",
+			 "}", "");
 	} else {
 	    $code = join("\n",
 			 "{ $var_decls",
@@ -1057,6 +1090,7 @@ sub do_pack {
     my($packable_args) = 0;
     my @is_packable;		# Packability (boolean) for each argument.
     my $wide_packing = 0;
+    my(@orig_args) = @args;
 
     #
     # Count the number of packable arguments.  If we encounter any 's' or 'd'
@@ -1077,6 +1111,18 @@ sub do_pack {
 	    }
 	} elsif ($arg =~ /^[sd]/) {
 	    return ('', '', @args);
+	} elsif ($arg =~ /^[scq]/ and $packable_args > 0) {
+	    # When packing, this operand will be picked up from the
+	    # code array, put onto the packing stack, and later put
+	    # back into a different location in the code. The problem
+	    # is that if this operand is a literal, the original
+	    # location in the code would have been remembered in a
+	    # literal patch.  For packing to work, we would have to
+	    # adjust the position in the literal patch. For the
+	    # moment, adding additional instructions to the packing
+	    # engine to handle this does not seem worth it, so we will
+	    # just turn off packing.
+	    return ('', '', @args);
 	} else {
 	    push @is_packable, 0;
 	}
@@ -1086,7 +1132,6 @@ sub do_pack {
     # Get out of here if too few or too many arguments.
     #
     return ('', '', @args) if $packable_args < 2;
-    &error("too many packable arguments") if $packable_args > 4;
 
     my($size) = 0;
     my($pack_prefix) = '';
@@ -1094,14 +1139,8 @@ sub do_pack {
 				# beginning).
     my($up) = '';		# Pack commands (storing back while
 				# moving forward).
-    my $args_per_word;
-    if ($packable_args < 4 or $wordsize == 64) {
-	$args_per_word = $packable_args;
-    } else {
-	# 4 packable argument, 32 bit wordsize. Need 2 words.
-	$args_per_word = 2;
-    }
 
+    my $args_per_word = $args_per_word[$packable_args];
     my @shift;
     my @mask;
     my @instr;
@@ -1295,6 +1334,8 @@ sub tr_parse_op {
 	foreach (split('', $type)) {
 	    &error("bad type in $op")
 		unless defined $type_bit{$_} or $type eq '*';
+	    $_ eq 'r' and
+		error("$op: 'r' is not allowed in transformations")
 	}
     }
 
@@ -1328,7 +1369,10 @@ sub tr_parse_op {
     }
 
     # Get an optional value. (In destination.)
+    $type_val = $type eq 'x' ? 1023 : 0;
     if (/^=(.*)/) {
+	error("value not allowed in source: $op")
+	    if $src;
 	$type_val = $1;
 	$_ = '';
     }
@@ -1347,11 +1391,6 @@ sub tr_parse_op {
 	    if $var && $type;
     }
 
-    # Test that source has no values.
-    if ($src) {
-	error("value not allowed in source: $op")
-	    if $type_val;
-    }
     ($var,$type,$type_val,$cond,$cond_val);
 }