/erts/emulator/beam/beam_emu.c

Large files files are truncated, but you can click here to view the full file

/*
 * %CopyrightBegin%
 *
 * Copyright Ericsson AB 1996-2010. All Rights Reserved.
 *
 * The contents of this file are subject to the Erlang Public License,
 * Version 1.1, (the "License"); you may not use this file except in
 * compliance with the License. You should have received a copy of the
 * Erlang Public License along with this software. If not, it can be
 * retrieved online at http://www.erlang.org/.
 *
 * Software distributed under the License is distributed on an "AS IS"
 * basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
 * the License for the specific language governing rights and limitations
 * under the License.
 *
 * %CopyrightEnd%
 */

#ifdef HAVE_CONFIG_H
#  include "config.h"
#endif

#include <stddef.h> /* offsetof() */
#include "sys.h"
#include "erl_vm.h"
#include "global.h"
#include "erl_process.h"
#include "erl_nmgc.h"
#include "error.h"
#include "bif.h"
#include "big.h"
#include "beam_load.h"
#include "erl_binary.h"
#include "erl_bits.h"
#include "dist.h"
#include "beam_bp.h"
#include "beam_catches.h"
#ifdef HIPE
#include "hipe_mode_switch.h"
#include "hipe_bif1.h"
#endif

/* #define HARDDEBUG 1 */

#if defined(NO_JUMP_TABLE)
#  define OpCase(OpCode)    case op_##OpCode: lb_##OpCode
#  define CountCase(OpCode) case op_count_##OpCode
#  define OpCode(OpCode)    ((Uint*)op_##OpCode)
#  define Goto(Rel) {Go = (int)(Rel); goto emulator_loop;}
#  define LabelAddr(Addr) &&##Addr
#else
#  define OpCase(OpCode)    lb_##OpCode
#  define CountCase(OpCode) lb_count_##OpCode
#  define Goto(Rel) goto *(Rel)
#  define LabelAddr(Label) &&Label
#  define OpCode(OpCode)  (&&lb_##OpCode)
#endif

#ifdef ERTS_ENABLE_LOCK_CHECK
#  ifdef ERTS_SMP
#    define PROCESS_MAIN_CHK_LOCKS(P)					\
do {									\
    if ((P)) {								\
	erts_pix_lock_t *pix_lock__ = ERTS_PIX2PIXLOCK(internal_pid_index((P)->id));\
	erts_proc_lc_chk_only_proc_main((P));				\
	erts_pix_lock(pix_lock__);					\
	ASSERT(0 < (P)->lock.refc && (P)->lock.refc < erts_no_schedulers*5);\
	erts_pix_unlock(pix_lock__);					\
    }									\
    else								\
	erts_lc_check_exact(NULL, 0);					\
    ERTS_SMP_LC_ASSERT(!ERTS_LC_IS_BLOCKING);				\
} while (0)
#    define ERTS_SMP_REQ_PROC_MAIN_LOCK(P) \
        if ((P)) erts_proc_lc_require_lock((P), ERTS_PROC_LOCK_MAIN)
#    define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P) \
        if ((P)) erts_proc_lc_unrequire_lock((P), ERTS_PROC_LOCK_MAIN)
#  else
#    define ERTS_SMP_REQ_PROC_MAIN_LOCK(P)
#    define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P)
#    define PROCESS_MAIN_CHK_LOCKS(P) erts_lc_check_exact(NULL, 0)
#  endif
#else
#  define PROCESS_MAIN_CHK_LOCKS(P)
#  define ERTS_SMP_REQ_PROC_MAIN_LOCK(P)
#  define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P)
#endif

/*
 * Define macros for deep checking of terms.
 */

#if defined(HARDDEBUG)

#  define CHECK_TERM(T) size_object(T)

#  define CHECK_ARGS(PC)                 \
do {                                     \
  int i_;                                \
  int Arity_ = PC[-1];                   \
  if (Arity_ > 0) {                      \
	CHECK_TERM(r(0));                \
  }                                      \
  for (i_ = 1; i_ < Arity_; i_++) {      \
	CHECK_TERM(x(i_));               \
  }                                      \
} while (0)
    
#else
#  define CHECK_TERM(T) ASSERT(!is_CP(T))
#  define CHECK_ARGS(T)
#endif

#ifndef MAX
#define MAX(x, y) (((x) > (y)) ? (x) : (y))
#endif

#define GET_BIF_ADDRESS(p) ((BifFunction) (((Export *) p)->code[4]))
#define TermWords(t) (((t) / (sizeof(BeamInstr)/sizeof(Eterm))) + !!((t) % (sizeof(BeamInstr)/sizeof(Eterm))))


/*
 * We reuse some of fields in the save area in the process structure.
 * This is safe to do, since this space is only activly used when
 * the process is switched out.
 */
#define REDS_IN(p)  ((p)->def_arg_reg[5])

/*
 * Add a byte offset to a pointer to Eterm.  This is useful when the
 * the loader has precalculated a byte offset.
 */
#define ADD_BYTE_OFFSET(ptr, offset) \
   ((Eterm *) (((unsigned char *)ptr) + (offset)))

/* We don't check the range if an ordinary switch is used */
#ifdef NO_JUMP_TABLE
#define VALID_INSTR(IP) (0 <= (int)(IP) && ((int)(IP) < (NUMBER_OF_OPCODES*2+10)))
#else
#define VALID_INSTR(IP) \
   ((SWord)LabelAddr(emulator_loop) <= (SWord)(IP) && \
    (SWord)(IP) < (SWord)LabelAddr(end_emulator_loop))
#endif /* NO_JUMP_TABLE */

#define SET_CP(p, ip)           \
   ASSERT(VALID_INSTR(*(ip)));  \
   (p)->cp = (ip)

#define SET_I(ip) \
   ASSERT(VALID_INSTR(* (Eterm *)(ip))); \
   I = (ip)

#define FetchArgs(S1, S2) tmp_arg1 = (S1); tmp_arg2 = (S2)

/*
 * 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;  \
    }                                               \
  } while (0)

#define StoreSimpleDest(Src, Dest) Dest = (Src)

/*
 * Store a result into a register and execute the next instruction.
 * Dst points to the word with a destination descriptor, which MUST
 * 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);  \
    }                                                        \
  } while (0)

#define ClauseFail() goto lb_jump_f

#define SAVE_CP(X)				\
   do {						\
      *(X) = make_cp(c_p->cp);			\
      c_p->cp = 0;				\
   } while(0)

#define RESTORE_CP(X)		SET_CP(c_p, (BeamInstr *) cp_val(*(X)))

#define ISCATCHEND(instr) ((Eterm *) *(instr) == OpCode(catch_end_y))

/*
 * Special Beam instructions.
 */

BeamInstr beam_apply[2];
BeamInstr beam_exit[1];
BeamInstr beam_continue_exit[1];

BeamInstr* em_call_error_handler;
BeamInstr* em_apply_bif;
BeamInstr* em_call_traced_function;


/* NOTE These should be the only variables containing trace instructions.
**      Sometimes tests are form the instruction value, and sometimes
**      for the refering variable (one of these), and rouge references
**      will most likely cause chaos.
*/
BeamInstr beam_return_to_trace[1];   /* OpCode(i_return_to_trace) */
BeamInstr beam_return_trace[1];      /* OpCode(i_return_trace) */
BeamInstr beam_exception_trace[1];   /* UGLY also OpCode(i_return_trace) */
BeamInstr beam_return_time_trace[1]; /* OpCode(i_return_time_trace) */

/*
 * All Beam instructions in numerical order.
 */

#ifndef NO_JUMP_TABLE
void** beam_ops;
#endif

#ifndef ERTS_SMP /* Not supported with smp emulator */
extern int count_instructions;
#endif

#if defined(HYBRID)
#define SWAPIN             \
    g_htop = global_htop;  \
    g_hend = global_hend;  \
    HTOP = HEAP_TOP(c_p);  \
    E = c_p->stop

#define SWAPOUT            \
    global_htop = g_htop;  \
    global_hend = g_hend;  \
    HEAP_TOP(c_p) = HTOP;  \
    c_p->stop = E

#else
#define SWAPIN             \
    HTOP = HEAP_TOP(c_p);  \
    E = c_p->stop

#define SWAPOUT            \
    HEAP_TOP(c_p) = HTOP;  \
    c_p->stop = E

/*
 * Use LIGHT_SWAPOUT when the called function
 * will call HeapOnlyAlloc() (and never HAlloc()).
 */
#ifdef DEBUG
#  /* The stack pointer is used in an assertion. */
#  define LIGHT_SWAPOUT SWAPOUT
#else
#  define LIGHT_SWAPOUT HEAP_TOP(c_p) = HTOP
#endif

/*
 * Use LIGHT_SWAPIN when we know that c_p->stop cannot
 * have been updated (i.e. if there cannot have been
 * a garbage-collection).
 */

#define LIGHT_SWAPIN HTOP = HEAP_TOP(c_p)

#endif

#ifdef FORCE_HEAP_FRAGS
#  define HEAP_SPACE_VERIFIED(Words) do { \
      c_p->space_verified = (Words);	  \
      c_p->space_verified_from = HTOP;	  \
    }while(0)
#else
#  define HEAP_SPACE_VERIFIED(Words) ((void)0)
#endif

#define PRE_BIF_SWAPOUT(P)						\
     HEAP_TOP((P)) = HTOP;  						\
     (P)->stop = E;  							\
     PROCESS_MAIN_CHK_LOCKS((P));					\
     ERTS_SMP_UNREQ_PROC_MAIN_LOCK((P))

#if defined(HYBRID)
#  define POST_BIF_GC_SWAPIN_0(_p, _res)				\
     if (((_p)->mbuf) || (MSO(_p).overhead >= BIN_VHEAP_SZ(_p)) ) {	\
       _res = erts_gc_after_bif_call((_p), (_res), NULL, 0);		\
     }									\
     SWAPIN

#  define POST_BIF_GC_SWAPIN(_p, _res, _regs, _arity)			\
     if (((_p)->mbuf) || (MSO(_p).overhead >= BIN_VHEAP_SZ(_p)) ) {	\
       _regs[0] = r(0);							\
       _res = erts_gc_after_bif_call((_p), (_res), _regs, (_arity));	\
       r(0) = _regs[0];							\
     }									\
     SWAPIN
#else
#  define POST_BIF_GC_SWAPIN_0(_p, _res)				\
     ERTS_SMP_REQ_PROC_MAIN_LOCK((_p));					\
     PROCESS_MAIN_CHK_LOCKS((_p));					\
     if (((_p)->mbuf) || (MSO(_p).overhead >= BIN_VHEAP_SZ(_p)) ) {	\
       _res = erts_gc_after_bif_call((_p), (_res), NULL, 0);		\
       E = (_p)->stop;							\
     }									\
     HTOP = HEAP_TOP((_p))

#  define POST_BIF_GC_SWAPIN(_p, _res, _regs, _arity)			\
     ERTS_SMP_REQ_PROC_MAIN_LOCK((_p));					\
     PROCESS_MAIN_CHK_LOCKS((_p));					\
     if (((_p)->mbuf) || (MSO(_p).overhead >= BIN_VHEAP_SZ(_p)) ) {	\
       _regs[0] = r(0);							\
       _res = erts_gc_after_bif_call((_p), (_res), _regs, (_arity));	\
       r(0) = _regs[0];							\
       E = (_p)->stop;							\
     }									\
     HTOP = HEAP_TOP((_p))
#endif

#define db(N) (N)
#define tb(N) (N)
#define xb(N) (*(Eterm *) (((unsigned char *)reg) + (N)))
#define yb(N) (*(Eterm *) (((unsigned char *)E) + (N)))
#define fb(N) (*(double *) (((unsigned char *)&(freg[0].fd)) + (N)))
#define x(N) reg[N]
#define y(N) E[N]
#define r(N) x##N

/*
 * Makes sure that there are StackNeed + HeapNeed + 1 words available
 * on the combined heap/stack segment, then allocates StackNeed + 1
 * words on the stack and saves CP.
 *
 * M is number of live registers to preserve during garbage collection
 */

#define AH(StackNeed, HeapNeed, M) \
  do { \
     int needed; \
     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)); \
           PROCESS_MAIN_CHK_LOCKS(c_p); \
           r(0) = reg[0]; \
           SWAPIN; \
     } \
     E -= needed; \
     SAVE_CP(E); \
  } while (0)

#define Allocate(Ns, Live) AH(Ns, 0, Live)

#define AllocateZero(Ns, Live)             \
 do { Eterm* ptr;                          \
      int i = (Ns);                        \
      AH(i, 0, Live);                      \
      for (ptr = E + i; ptr > E; ptr--) {  \
	 make_blank(*ptr);                 \
     }                                     \
  } while (0)

#define AllocateHeap(Ns, Nh, Live) AH(Ns, Nh, Live)

#define AllocateHeapZero(Ns, Nh, Live)     \
 do { Eterm* ptr;                          \
      int i = (Ns);                        \
      AH(i, Nh, Live);                     \
      for (ptr = E + i; ptr > E; ptr--) {  \
	 make_blank(*ptr);                 \
     }                                     \
  } while (0)

#define AllocateInit(Ns, Live, Y) \
  do { AH(Ns, 0, Live); make_blank(Y); } while (0)

/*
 * Like the AH macro, but allocates no additional heap space.
 */

#define A(StackNeed, M) AH(StackNeed, 0, M)

#define D(N)             \
     RESTORE_CP(E);      \
     E += (N) + 1;



#define TestBinVHeap(VNh, Nh, Live)                             		\
  do {                                                          		\
    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));  		\
       PROCESS_MAIN_CHK_LOCKS(c_p);                             		\
       r(0) = reg[0];                                           		\
       SWAPIN;                                                  		\
    }                                                           		\
    HEAP_SPACE_VERIFIED(need);                                                  \
  } while (0)



/*
 * Check if Nh words of heap are available; if not, do a garbage collection.
 * Live is number of active argument registers to be preserved.
 */

#define TestHeap(Nh, Live)                                      \
  do {                                                          \
    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));  \
       PROCESS_MAIN_CHK_LOCKS(c_p);                             \
       r(0) = reg[0];                                           \
       SWAPIN;                                                  \
    }                                                           \
    HEAP_SPACE_VERIFIED(need);                             \
  } while (0)

/*
 * Check if Nh words of heap are available; if not, do a garbage collection.
 * Live is number of active argument registers to be preserved.
 * Takes special care to preserve Extra if a garbage collection occurs.
 */

#define TestHeapPreserve(Nh, Live, Extra)				\
  do {									\
    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);	\
       PROCESS_MAIN_CHK_LOCKS(c_p);					\
       if (Live > 0) {							\
	   r(0) = reg[0];						\
       }								\
       Extra = reg[Live];						\
       SWAPIN;								\
    }									\
    HEAP_SPACE_VERIFIED(need);                                      \
  } while (0)

#ifdef HYBRID
#ifdef INCREMENTAL
#define TestGlobalHeap(Nh, Live, hp)                                    \
  do {                                                                  \
    unsigned need = (Nh);                                               \
    ASSERT(global_heap <= g_htop && g_htop <= global_hend);             \
    SWAPOUT;                                                            \
    reg[0] = r(0);                                                      \
    FCALLS -= need;                                                     \
    (hp) = IncAlloc(c_p,need,reg,(Live));                               \
    r(0) = reg[0];                                                      \
    SWAPIN;                                                             \
  } while (0)
#else
#define TestGlobalHeap(Nh, Live, hp)                                    \
  do {                                                                  \
    unsigned need = (Nh);                                               \
    ASSERT(global_heap <= g_htop && g_htop <= global_hend);             \
    if (g_hend - g_htop < need) {                                       \
       SWAPOUT;                                                         \
       reg[0] = r(0);                                                   \
       FCALLS -= erts_global_garbage_collect(c_p, need, reg, (Live));   \
       r(0) = reg[0];                                                   \
       SWAPIN;                                                          \
    }                                                                   \
    (hp) = global_htop;                                                 \
  } while (0)
#endif
#endif /* HYBRID */

#define Init(N) make_blank(yb(N))

#define Init2(Y1, Y2) do { make_blank(Y1); make_blank(Y2); } while (0)
#define Init3(Y1, Y2, Y3) \
   do { make_blank(Y1); make_blank(Y2); make_blank(Y3); } while (0)

#define MakeFun(FunP, NumFree)					\
  do {								\
     SWAPOUT;							\
     reg[0] = r(0);						\
     r(0) = new_fun(c_p, reg, (ErlFunEntry *) FunP, NumFree);	\
     SWAPIN;							\
  } while (0)


/*
 * Check that we haven't used the reductions and jump to function pointed to by
 * the I register.  If we are out of reductions, do a context switch.
 */

#define DispatchMacro()				\
  do {						\
     BeamInstr* dis_next;				\
     dis_next = (BeamInstr *) *I;			\
     CHECK_ARGS(I);				\
     if (FCALLS > 0 || FCALLS > neg_o_reds) {	\
        FCALLS--;				\
        Goto(dis_next);				\
     } else {					\
	goto context_switch;			\
     }						\
 } while (0)

#define DispatchMacroFun()			\
  do {						\
     BeamInstr* dis_next;				\
     dis_next = (BeamInstr *) *I;			\
     CHECK_ARGS(I);				\
     if (FCALLS > 0 || FCALLS > neg_o_reds) {	\
        FCALLS--;				\
        Goto(dis_next);				\
     } else {					\
	goto context_switch_fun;		\
     }						\
 } while (0)

#define DispatchMacrox()					\
  do {								\
     if (FCALLS > 0) {						\
        Eterm* dis_next;					\
        SET_I(((Export *) Arg(0))->address);			\
        dis_next = (Eterm *) *I;				\
        FCALLS--;						\
        CHECK_ARGS(I);						\
        Goto(dis_next);						\
     } else if (ERTS_PROC_GET_SAVED_CALLS_BUF(c_p)		\
		&& FCALLS > neg_o_reds) {			\
        goto save_calls1;					\
     } else {							\
        SET_I(((Export *) Arg(0))->address);			\
        CHECK_ARGS(I);						\
	goto context_switch;					\
     }								\
 } while (0)

#ifdef DEBUG
/*
 * To simplify breakpoint setting, put the code in one place only and jump to it.
 */
#  define Dispatch() goto do_dispatch
#  define Dispatchx() goto do_dispatchx
#  define Dispatchfun() goto do_dispatchfun
#else
/*
 * Inline for speed.
 */
#  define Dispatch() DispatchMacro()
#  define Dispatchx() DispatchMacrox()
#  define Dispatchfun() DispatchMacroFun()
#endif

#define Self(R) R = c_p->id
#define Node(R) R = erts_this_node->sysname

#define Arg(N)       I[(N)+1]
#define Next(N)                \
    I += (N) + 1;              \
    ASSERT(VALID_INSTR(*I));   \
    Goto(*I)

#define PreFetch(N, Dst) do { Dst = (BeamInstr *) *(I + N + 1); } while (0)
#define NextPF(N, Dst)         \
    I += N + 1;                \
    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); \
   } while (0)

#define GetArg1(N, Dst) GetR((N), Dst)

#define GetArg2(N, Dst1, Dst2)     \
   do {                            \
     GetR(N, Dst1);                \
     GetR((N)+1, Dst2);            \
   } while (0)

#define PutList(H, T, Dst, Store)  \
  do {                             \
   HTOP[0] = (H); HTOP[1] = (T);   \
   Store(make_list(HTOP), Dst);    \
   HTOP += 2;                      \
  } while (0)

#define Move(Src, Dst, Store)      \
   do {                            \
       Eterm term = (Src);         \
       Store(term, Dst);           \
   } while (0)

#define Move2(src1, dst1, src2, dst2) dst1 = (src1); dst2 = (src2)

#define MoveGenDest(src, dstp) \
   if ((dstp) == NULL) { r(0) = (src); } else { *(dstp) = src; }

#define MoveReturn(Src, Dest)       \
    (Dest) = (Src);                 \
    I = c_p->cp;                    \
    ASSERT(VALID_INSTR(*c_p->cp));  \
    c_p->cp = 0;                    \
    CHECK_TERM(r(0));               \
    Goto(*I)

#define DeallocateReturn(Deallocate)       \
  do {                                     \
    int words_to_pop = (Deallocate);       \
    SET_I((BeamInstr *) cp_val(*E));                     \
    E = ADD_BYTE_OFFSET(E, words_to_pop);  \
    CHECK_TERM(r(0));                      \
    Goto(*I);                              \
  } while (0)

#define MoveDeallocateReturn(Src, Dest, Deallocate)  \
    (Dest) = (Src);                                  \
    DeallocateReturn(Deallocate)

#define MoveCall(Src, Dest, CallDest, Size)	\
    (Dest) = (Src);				\
    SET_CP(c_p, I+Size+1);			\
    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);				\
    Dispatch();

#define MoveCallOnly(Src, Dest, CallDest)	\
    (Dest) = (Src);				\
    SET_I((BeamInstr *) CallDest);			\
    Dispatch();

#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) COMPRESS_POINTER(((unsigned char *) tuple_val(Src)) + 	\
				(Element));				\
    (Dest) = (*(Eterm *) EXPAND_POINTER(tmp_arg1));			\
  } while (0)

#define ExtractNextElement(Dest)					  \
    tmp_arg1 += sizeof(Eterm);						  \
    (Dest) = (* (Eterm *) (((unsigned char *) EXPAND_POINTER(tmp_arg1))))

#define ExtractNextElement2(Dest)				\
  do {								\
    Eterm* ene_dstp = &(Dest);					\
    ene_dstp[0] = ((Eterm *) EXPAND_POINTER(tmp_arg1))[1];	\
    ene_dstp[1] = ((Eterm *) EXPAND_POINTER(tmp_arg1))[2];	\
    tmp_arg1 += sizeof(Eterm) + sizeof(Eterm);			\
  } while (0)

#define ExtractNextElement3(Dest)		\
  do {						\
    Eterm* ene_dstp = &(Dest);			\
    ene_dstp[0] = ((Eterm *) EXPAND_POINTER(tmp_arg1))[1];	\
    ene_dstp[1] = ((Eterm *) EXPAND_POINTER(tmp_arg1))[2];	\
    ene_dstp[2] = ((Eterm *) EXPAND_POINTER(tmp_arg1))[3];	\
    tmp_arg1 += 3*sizeof(Eterm);		\
  } while (0)

#define ExtractNextElement4(Dest)		\
  do {						\
    Eterm* ene_dstp = &(Dest);			\
    ene_dstp[0] = ((Eterm *) EXPAND_POINTER(tmp_arg1))[1];	\
    ene_dstp[1] = ((Eterm *) EXPAND_POINTER(tmp_arg1))[2];	\
    ene_dstp[2] = ((Eterm *) EXPAND_POINTER(tmp_arg1))[3];	\
    ene_dstp[3] = ((Eterm *) EXPAND_POINTER(tmp_arg1))[4];	\
    tmp_arg1 += 4*sizeof(Eterm);		\
  } while (0)

#define ExtractElement(Element, Dest)		\
  do {						\
     tmp_arg1 += (Element);			\
     (Dest) = (* (Eterm *) EXPAND_POINTER(tmp_arg1));		\
  } while (0)

#define PutTuple(Arity, Src, Dest)		\
     ASSERT(is_arity_value(Arity));		\
     Dest = make_tuple(HTOP);			\
     HTOP[0] = (Arity);				\
     HTOP[1] = (Src);				\
     HTOP += 2

#define Put(Word) *HTOP++ = (Word)

#define EqualImmed(X, Y, Action) if (X != Y) { Action; }

#define IsFloat(Src, Fail) if (is_not_float(Src)) { Fail; }

#define IsInteger(Src, Fail) if (is_not_integer(Src)) { Fail; }

#define IsNumber(X, Fail) if (is_not_integer(X) && is_not_float(X)) { Fail; }

#define IsAtom(Src, Fail) if (is_not_atom(Src)) { Fail; }

#define IsIntegerAllocate(Src, Need, Alive, Fail)  \
    if (is_not_integer(Src)) { Fail; }             \
    A(Need, Alive)

#define IsNil(Src, Fail) if (is_not_nil(Src)) { Fail; }

#define IsList(Src, Fail) if (is_not_list(Src) && is_not_nil(Src)) { Fail; }

#define IsNonemptyList(Src, Fail) if (is_not_list(Src)) { Fail; }

#define IsNonemptyListAllocate(Src, Need, Alive, Fail)  \
    if (is_not_list(Src)) { Fail; }                     \
    A(Need, Alive)

#define IsNonemptyListTestHeap(Src, Need, Alive, Fail)  \
    if (is_not_list(Src)) { Fail; }                     \
    TestHeap(Need, Alive)

#define IsTuple(X, Action) if (is_not_tuple(X)) Action

#define IsArity(Pointer, Arity, Fail)					  \
    if (*(Eterm *)							  \
	EXPAND_POINTER(tmp_arg1 = (Eterm) 				  \
		       COMPRESS_POINTER(tuple_val(Pointer))) != (Arity))  \
    { 									  \
        Fail; 								  \
    }

#define IsFunction(X, Action)			\
  do {						\
     if ( !(is_any_fun(X)) ) {			\
          Action;				\
     }						\
  } while (0)

#define IsFunction2(F, A, Action)		\
  do {						\
     if (is_function_2(c_p, F, A) != am_true ) {\
          Action;				\
     }						\
  } while (0)

#define IsTupleOfArity(Src, Arity, Fail)				      \
  do {									      \
    if (is_not_tuple(Src) || 						      \
	*(Eterm *)							      \
	EXPAND_POINTER(tmp_arg1 = 					      \
		       (Eterm) COMPRESS_POINTER(tuple_val(Src))) != Arity) { \
        Fail;								      \
    }									      \
  } while (0)

#define IsBoolean(X, Fail) if ((X) != am_true && (X) != am_false) { Fail; }

#define IsBinary(Src, Fail) \
 if (is_not_binary(Src) || binary_bitsize(Src) != 0) { Fail; }

#define IsBitstring(Src, Fail) \
  if (is_not_binary(Src)) { Fail; }

#if defined(ARCH_64) && !HALFWORD_HEAP
#define BsSafeMul(A, B, Fail, Target)		\
   do { Uint64 _res = (A) * (B);		\
      if (_res / B != A) { Fail; }		\
      Target = _res;				\
   } while (0)
#else
#define BsSafeMul(A, B, Fail, Target)			\
   do { Uint64 _res = (Uint64)(A) * (Uint64)(B);	\
      if ((_res >> (8*sizeof(Uint))) != 0) { Fail; }	\
      Target = _res;					\
   } while (0)
#endif

#define BsGetFieldSize(Bits, Unit, Fail, Target)	\
   do {							\
      Sint _signed_size; Uint _uint_size;		\
      if (is_small(Bits)) {				\
        _signed_size = signed_val(Bits);		\
         if (_signed_size < 0) { Fail; }		\
         _uint_size = (Uint) _signed_size;		\
      } else {						\
         if (!term_to_Uint(Bits, &temp_bits)) { Fail; }	\
         _uint_size = temp_bits;			\
      }							\
      BsSafeMul(_uint_size, Unit, Fail, Target);	\
   } while (0)

#define BsGetUncheckedFieldSize(Bits, Unit, Fail, Target)	\
   do {								\
      Sint _signed_size; Uint _uint_size;			\
      if (is_small(Bits)) {					\
        _signed_size = signed_val(Bits);			\
         if (_signed_size < 0) { Fail; }			\
         _uint_size = (Uint) _signed_size;			\
      } else {							\
         if (!term_to_Uint(Bits, &temp_bits)) { Fail; }		\
         _uint_size = (Uint) temp_bits;				\
      }								\
      Target = _uint_size * Unit;				\
   } while (0)

#define BsGetFloat2(Ms, Live, Sz, Flags, Dst, Store, Fail)		\
 do {									\
   ErlBinMatchBuffer *_mb;						\
   Eterm _result; Sint _size;						\
   if (!is_small(Sz) || (_size = unsigned_val(Sz)) > 64) { Fail; }	\
   _size *= ((Flags) >> 3);						\
   TestHeap(FLOAT_SIZE_OBJECT, Live);					\
   _mb = ms_matchbuffer(Ms);						\
   LIGHT_SWAPOUT;							\
   _result = erts_bs_get_float_2(c_p, _size, (Flags), _mb);		\
   LIGHT_SWAPIN;							\
   HEAP_SPACE_VERIFIED(0);                                          \
   if (is_non_value(_result)) { Fail; }					\
   else { Store(_result, Dst); }					\
 } while (0)

#define BsGetBinaryImm_2(Ms, Live, Sz, Flags, Dst, Store, Fail)	\
  do {								\
    ErlBinMatchBuffer *_mb;					\
    Eterm _result;						\
    TestHeap(heap_bin_size(ERL_ONHEAP_BIN_LIMIT), Live);	\
    _mb = ms_matchbuffer(Ms);					\
    LIGHT_SWAPOUT;						\
    _result = erts_bs_get_binary_2(c_p, (Sz), (Flags), _mb);	\
    LIGHT_SWAPIN;						\
    HEAP_SPACE_VERIFIED(0);                                 \
    if (is_non_value(_result)) { Fail; }			\
    else { Store(_result, Dst); }				\
  } while (0)

#define BsGetBinary_2(Ms, Live, Sz, Flags, Dst, Store, Fail)	\
  do {								\
    ErlBinMatchBuffer *_mb;					\
    Eterm _result; Uint _size;					\
    BsGetFieldSize(Sz, ((Flags) >> 3), Fail, _size);		\
    TestHeap(ERL_SUB_BIN_SIZE, Live);				\
    _mb = ms_matchbuffer(Ms);					\
    LIGHT_SWAPOUT;						\
    _result = erts_bs_get_binary_2(c_p, _size, (Flags), _mb);	\
    LIGHT_SWAPIN;						\
    HEAP_SPACE_VERIFIED(0);                                 \
    if (is_non_value(_result)) { Fail; }			\
    else { Store(_result, Dst); }				\
  } while (0)

#define BsGetBinaryAll_2(Ms, Live, Unit, Dst, Store, Fail)	\
  do {								\
    ErlBinMatchBuffer *_mb;					\
    Eterm _result;						\
    TestHeap(ERL_SUB_BIN_SIZE, Live);				\
    _mb = ms_matchbuffer(Ms);					\
    if (((_mb->size - _mb->offset) % Unit) == 0) {		\
      LIGHT_SWAPOUT;						\
      _result = erts_bs_get_binary_all_2(c_p, _mb);		\
      LIGHT_SWAPIN;						\
      HEAP_SPACE_VERIFIED(0);                               \
      ASSERT(is_value(_result));				\
      Store(_result, Dst);					\
    } else {                                                    \
	HEAP_SPACE_VERIFIED(0);                             \
	Fail; }						        \
 } while (0)

#define BsSkipBits2(Ms, Bits, Unit, Fail)			\
 do {								\
   ErlBinMatchBuffer *_mb;					\
   size_t new_offset;						\
   Uint _size;							\
    _mb = ms_matchbuffer(Ms);					\
   BsGetFieldSize(Bits, Unit, Fail, _size);			\
   new_offset = _mb->offset + _size;				\
   if (new_offset <= _mb->size) { _mb->offset = new_offset; }	\
   else { Fail; }						\
 } while (0)

#define BsSkipBitsAll2(Ms, Unit, Fail)		\
 do {						\
    ErlBinMatchBuffer *_mb;			\
   _mb = ms_matchbuffer(Ms);			\
   if (((_mb->size - _mb->offset) % Unit) == 0) {_mb->offset = _mb->size; } \
   else { Fail; }					\
 } while (0)

#define BsSkipBitsImm2(Ms, Bits, Fail)				\
 do {								\
   ErlBinMatchBuffer *_mb;					\
   size_t new_offset;						\
   _mb = ms_matchbuffer(Ms);					\
   new_offset = _mb->offset + (Bits);				\
   if (new_offset <= _mb->size) { _mb->offset = new_offset; }	\
   else { Fail; }						\
 } while (0)

#define NewBsPutIntegerImm(Sz, Flags, Src)					\
 do {									\
    if (!erts_new_bs_put_integer(ERL_BITS_ARGS_3((Src), (Sz), (Flags)))) { goto badarg; }	\
 } while (0)

#define NewBsPutInteger(Sz, Flags, Src)						\
 do {										\
    Sint _size;									\
    BsGetUncheckedFieldSize(Sz, ((Flags) >> 3), goto badarg, _size);		\
    if (!erts_new_bs_put_integer(ERL_BITS_ARGS_3((Src), _size, (Flags))))	\
       { goto badarg; }								\
 } while (0)

#define NewBsPutFloatImm(Sz, Flags, Src)					\
 do {									\
    if (!erts_new_bs_put_float(c_p, (Src), (Sz), (Flags))) { goto badarg; }	\
 } while (0)

#define NewBsPutFloat(Sz, Flags, Src)						\
 do {										\
    Sint _size;									\
    BsGetUncheckedFieldSize(Sz, ((Flags) >> 3), goto badarg, _size);		\
    if (!erts_new_bs_put_float(c_p, (Src), _size, (Flags))) { goto badarg; }	\
 } while (0)

#define NewBsPutBinary(Sz, Flags, Src)							\
 do {											\
    Sint _size;										\
    BsGetUncheckedFieldSize(Sz, ((Flags) >> 3), goto badarg, _size);			\
    if (!erts_new_bs_put_binary(ERL_BITS_ARGS_2((Src), _size))) { goto badarg; }	\
 } while (0)

#define NewBsPutBinaryImm(Sz, Src)				        \
 do {							        \
    if (!erts_new_bs_put_binary(ERL_BITS_ARGS_2((Src), (Sz)))) { goto badarg; }	\
 } while (0)

#define NewBsPutBinaryAll(Src, Unit)							\
 do {											\
    if (!erts_new_bs_put_binary_all(ERL_BITS_ARGS_2((Src), (Unit)))) { goto badarg; }	\
 } while (0)


#define IsPort(Src, Fail) if (is_not_port(Src)) { Fail; }
#define IsPid(Src, Fail) if (is_not_pid(Src)) { Fail; }
#define IsRef(Src, Fail) if (is_not_ref(Src)) { Fail; }

static BifFunction translate_gc_bif(void* gcf);
static BeamInstr* handle_error(Process* c_p, BeamInstr* pc, Eterm* reg, BifFunction bf);
static BeamInstr* next_catch(Process* c_p, Eterm *reg);
static void terminate_proc(Process* c_p, Eterm Value);
static Eterm add_stacktrace(Process* c_p, Eterm Value, Eterm exc);
static void save_stacktrace(Process* c_p, BeamInstr* pc, Eterm* reg,
			     BifFunction bf, Eterm args);
static struct StackTrace * get_trace_from_exc(Eterm exc);
static Eterm make_arglist(Process* c_p, Eterm* reg, int a);
static Eterm call_error_handler(Process* p, BeamInstr* ip, Eterm* reg);
static Eterm call_breakpoint_handler(Process* p, BeamInstr* fi, Eterm* reg);
static BeamInstr* fixed_apply(Process* p, Eterm* reg, Uint arity);
static BeamInstr* apply(Process* p, Eterm module, Eterm function,
		     Eterm args, Eterm* reg);
static int hibernate(Process* c_p, Eterm module, Eterm function,
		     Eterm args, Eterm* reg);
static BeamInstr* call_fun(Process* p, int arity, Eterm* reg, Eterm args);
static BeamInstr* apply_fun(Process* p, Eterm fun, Eterm args, Eterm* reg);
static Eterm new_fun(Process* p, Eterm* reg, ErlFunEntry* fe, int num_free);

#if defined(VXWORKS)
static int init_done;
#endif

void
init_emulator(void)
{
#if defined(VXWORKS)
    init_done = 0;
#endif
    process_main();
}

/*
 * On certain platforms, make sure that the main variables really are placed
 * in registers.
 */

#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

/*
 * process_main() is called twice:
 * The first call performs some initialisation, including exporting
 * the instructions' C labels to the loader.
 * The second call starts execution of BEAM code. This call never returns.
 */
void process_main(void)
{
#if !defined(VXWORKS)
    static int init_done = 0;
#endif
    Process* c_p = NULL;
    int reds_used;
#ifdef DEBUG
    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.
     */
    register Eterm* reg REG_xregs = NULL;

    /*
     * Top of heap (next free location); grows upwards.
     */
    register Eterm* HTOP REG_htop = NULL;


#ifdef HYBRID
     Eterm *g_htop;
     Eterm *g_hend;
#endif

    /* Stack pointer.  Grows downwards; points
     * to last item pushed (normally a saved
     * continuation pointer).
     */
    register Eterm* E REG_stop = NULL;

    /*
     * Pointer to next threaded instruction.
     */
    register BeamInstr *I REG_I = NULL;

    /* Number of reductions left.  This function
     * returns to the scheduler when FCALLS reaches zero.
     */
    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;
#if HEAP_ON_C_STACK
    Eterm tmp_big[2];           /* Temporary buffer for small bignums if HEAP_ON_C_STACK. */
#else
    Eterm *tmp_big;		/* Temporary buffer for small bignums if !HEAP_ON_C_STACK. */
#endif

#ifndef ERTS_SMP
#if !HALFWORD_HEAP
    static Eterm save_reg[ERTS_X_REGS_ALLOCATED];
    /* X registers -- not used directly, but
     * through 'reg', because using it directly
     * needs two instructions on a SPARC,
     * while using it through reg needs only
     * one.
     */
#endif
    /*
     * Floating point registers.
     */
    static FloatDef freg[MAX_REG];
#else
    /* X regisers and floating point registers are located in
     * scheduler specific data.
     */
    register FloatDef *freg;
#endif

    /*
     * For keeping the negative old value of 'reds' when call saving is active.
     */
    int neg_o_reds = 0;

    Eterm (*arith_func)(Process* p, Eterm* reg, Uint live);

#ifndef NO_JUMP_TABLE
    static void* opcodes[] = { DEFINE_OPCODES };
#ifdef ERTS_OPCODE_COUNTER_SUPPORT
    static void* counting_opcodes[] = { DEFINE_COUNTING_OPCODES };
#endif
#else
    int Go;
#endif

    Uint temp_bits; /* Temporary used by BsSkipBits2 & BsGetInteger2 */

    ERL_BITS_DECLARE_STATEP; /* Has to be last declaration */


    /*
     * Note: In this function, we attempt to place rarely executed code towards
     * the end of the function, in the hope that the cache hit rate will be better.
     * The initialization code is only run once, so it is at the very end.
     *
     * Note: c_p->arity must be set to reflect the number of useful terms in
     * c_p->arg_reg before calling the scheduler.
     */
    if (!init_done) {
	init_done = 1;
	goto init_emulator;
    }
#ifndef ERTS_SMP
#if !HALFWORD_HEAP
    reg = save_reg;	/* XXX: probably wastes a register on x86 */
#else
    /* Registers need to be heap allocated (correct memory range) for tracing to work */
    reg = erts_alloc(ERTS_ALC_T_BEAM_REGISTER, ERTS_X_REGS_ALLOCATED * sizeof(Eterm));
#endif
#endif
    c_p = NULL;
    reds_used = 0;
    goto do_schedule1;

 do_schedule:
    reds_used = REDS_IN(c_p) - FCALLS;
 do_schedule1:
    PROCESS_MAIN_CHK_LOCKS(c_p);
    ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
    c_p = schedule(c_p, reds_used);
#ifdef DEBUG
    pid = c_p->id;
#endif
    ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
    PROCESS_MAIN_CHK_LOCKS(c_p);
#ifdef ERTS_SMP
    reg = c_p->scheduler_data->save_reg;
    freg = c_p->scheduler_data->freg;
#endif
#if !HEAP_ON_C_STACK
    tmp_big = ERTS_PROC_GET_SCHDATA(c_p)->beam_emu_tmp_heap;
#endif
    ERL_BITS_RELOAD_STATEP(c_p);
    {
	int reds;
	Eterm* argp;
	BeamInstr *next;
	int i;

	argp = c_p->arg_reg;
	for (i = c_p->arity - 1; i > 0; i--) {
	    reg[i] = argp[i];
	    CHECK_TERM(reg[i]);
	}

	/*
	 * We put the original reduction count in the process structure, to reduce
	 * the code size (referencing a field in a struct through a pointer stored
	 * in a register gives smaller code than referencing a global variable).
	 */

	SET_I(c_p->i);

	reds = c_p->fcalls;
	if (ERTS_PROC_GET_SAVED_CALLS_BUF(c_p)
	    && (c_p->trace_flags & F_SENSITIVE) == 0) {
	    neg_o_reds = -reds;
	    FCALLS = REDS_IN(c_p) = 0;
	} else {
	    neg_o_reds = 0;
	    FCALLS = REDS_IN(c_p) = reds;
	}

	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));
	Goto(next);
    }

#if defined(DEBUG) || defined(NO_JUMP_TABLE)
 emulator_loop:
#endif

#ifdef NO_JUMP_TABLE
    switch (Go) {
#endif
#include "beam_hot.h"

#define STORE_ARITH_RESULT(res) StoreBifResult(2, (res));
#define ARITH_FUNC(name) erts_gc_##name

 OpCase(i_plus_jId):
 {
     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_jId):
 {
     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_minus);
     goto do_big_arith2;
 }

 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);

 OpCase(i_move_call_only_fcr): {
     r(0) = Arg(1);
 }
 /* FALL THROUGH */
 OpCase(i_call_only_f): {
     SET_I((BeamInstr *) Arg(0));
     Dispatch();
 }

 OpCase(i_move_call_last_fPcr): {
     r(0) = Arg(2);
 }
 /* FALL THROUGH */
 OpCase(i_call_last_fP): {
     RESTORE_CP(E);
     E = ADD_BYTE_OFFSET(E, Arg(1));
     SET_I((BeamInstr *) Arg(0));
     Dispatch();
 }

 OpCase(i_move_call_crf): {
     r(0) = Arg(0);
     I++;
 }
 /* FALL THROUGH */
 OpCase(i_call_f): {
     SET_CP(c_p, I+2);
     SET_I((BeamInstr *) Arg(0));
     Dispatch();
 }

 OpCase(i_move_call_ext_last_ePcr): {
     r(0) = Arg(2);
 }
 /* FALL THROUGH */
 OpCase(i_call_ext_last_eP):
    RESTORE_CP(E);
    E = ADD_BYTE_OFFSET(E, Arg(1));

    /*
     * Note: The pointer to the export entry is never NULL; if the module
     * is not loaded, it points to code which will invoke the error handler
     * (see lb_call_error_handler below).
     */
    Dispatchx();

 OpCase(i_move_call_ext_cre): {
     r(0) = Arg(0);
     I++;
 }
 /* FALL THROUGH */
 OpCase(i_call_ext_e):
    SET_CP(c_p, I+2);
    Dispatchx();

 OpCase(i_move_call_ext_only_ecr): {
     r(0) = Arg(1);
 }
 /* FALL THROUGH */
 OpCase(i_call_ext_only_e):
    Dispatchx();

 OpCase(init_y): {
     BeamInstr *next;

     PreFetch(1, next);
     make_blank(yb(Arg(0)));
     NextPF(1, next);
 }

 OpCase(i_trim_I): {
     BeamInstr *next;
     Uint words;
     Uint cp;

     words = Arg(0);
     cp = E[0];
     PreFetch(1, next);
     E += words;
     E[0] = cp;
     NextPF(1, next);
 }

 OpCase(return): {
    SET_I(c_p->cp);
    /*
     * We must clear the CP to make sure that a stale value do not
     * create a false module dependcy preventing code upgrading.
     * It also means that we can use the CP in stack backtraces.
     */
    c_p->cp = 0;
    CHECK_TERM(r(0));
    HEAP_SPACE_VERIFIED(0);
    Goto(*I);
 }

 OpCase(test_heap_1_put_list_Iy): {
     BeamInstr *next;

     PreFetch(2, next);
     TestHeap(Arg(0), 1);
     PutList(yb(Arg(1)), r(0), r(0), StoreSimpleDest);
     CHECK_TERM(r(0));
     NextPF(2, next);
 }

    /*
     * Send is almost a standard call-BIF with two arguments, except for:
     *    1) It cannot be traced.
     *	  2) There is no pointer to the send_2 function stored in
     *       the instruction.
     */

 OpCase(send): {
     BeamInstr *next;
     Eterm result;

     PRE_BIF_SWAPOUT(c_p);
     c_p->fcalls = FCALLS - 1;
     result = send_2(c_p, r(0), x(1));
     PreFetch(0, next);
     POST_BIF_GC_SWAPIN(c_p, result, reg, 2);
     FCALLS = c_p->fcalls;
     if (is_value(result)) {
	 r(0) = result;
	 CHECK_TERM(r(0));
	 NextPF(0, next);
     } else if (c_p->freason == TRAP) {
	 SET_CP(c_p, I+1);
	 SET_I(*((BeamInstr **) (BeamInstr) ((c_p)->def_arg_reg + 3)));
	 SWAPIN;
	 r(0) = c_p->def_arg_reg[0];
	 x(1) = c_p->def_arg_reg[1];
	 Dispatch();
     }
     goto find_func_info;
 }

 OpCase(i_element_jssd): {
     Eterm index;
     Eterm tuple;

     /*
      * Inlined version of element/2 for speed.
      */
     GetArg2(1, index, tuple);
     if (is_small(index) && is_tuple(tuple)) {
	 Eterm* tp = tuple_val(tuple);

	 if ((signed_val(index) >= 1) &&
	     (signed_val(index) <= arityval(*tp))) {
	     Eterm result = tp[signed_val(index)];
	     StoreBifResult(3, result);
	 }
     }
 }
 /* Fall through */

 OpCase(badarg_j):
 badarg:
    c_p->freason = BADARG;
    goto lb_Cl_error;

 OpCase(i_fast_element_jIsd): {
     Eterm tuple;

     /*
      * Inlined version of element/2 for even more speed.
      * The first argument is an untagged integer >= 1.
      * The second argument is guaranteed to be a register operand.
      */
     GetArg1(2, tuple);
     if (is_tuple(tuple)) {
	 Eterm* tp = tuple_val(tuple);
	 tmp_arg2 = Arg(1);
	 if (tmp_arg2 <= arityval(*tp)) {
	     Eterm result = tp[tmp_arg2];
	     StoreBifResult(3, result);
	 }
     }
     goto badarg;
 }

 OpCase(catch_yf):
     c_p->catches++;
     yb(Arg(0)) = Arg(1);
     Next(2);

 OpCase(catch_end_y): {
     c_p->catches--;
     make_blank(yb(Arg(0)));
     if (is_non_value(r(0))) {
	 if (x(1) == am_throw) {
	     r(0) = x(2);
	 } else {
	     if (x(1) == am_error) {
	         SWAPOUT;
		 x(2) = add_stacktrace(c_p, x(2), x(3));
		 SWAPIN;
	     }
	     /* only x(2) is included in the rootset here */
	     if (E - HTOP < 3 || c_p->mbuf) {	/* Force GC in case add_stacktrace()
						 * created heap fragments */
		 SWAPOUT;
		 PROCESS_MAIN_CHK_LOCKS(c_p);
		 FCALLS -= erts_garbage_collect(c_p, 3, reg+2, 1);
		 PROCESS_MAIN_CHK_LOCKS(c_p);
		 SWAPIN;
	     }
	     r(0) = TUPLE2(HTOP, am_EXIT, x(2));
	     HTOP += 3;
	 }
     }
     CHECK_TERM(r(0));
     Next(1);
 }

 OpCase(try_end_y): {
     c_p->catches--;
     make_blank(yb(Arg(0)));
     if (is_non_value(r(0))) {
	 r(0) = x(1);
	 x(1) = x(2);
	 x(2) = x(3);
     }
     Next(1);
 }

 /*
  * Skeleton for receive statement:
  *
  *             recv_mark L1                     Optional
  *             call make_ref/monitor            Optional
  *             ...
  *             recv_set L1                      Optional
  *      L1:          <-------------------+
  *                   <-----------+       |
  *     	     	       	  |   	  |
  *             loop_rec L2 ------+---+   |
  *             ...               |   |   |
  *             remove_message 	  |   |	  |
  *             jump L3           |   |   |
  *		...	          |   |   |
  *		loop_rec_end L1 --+   |   |
  *      L2:          <---------------+   |
  *	   	wait L1  -----------------+      or wait_timeout
  *		timeout
  *
  *	 L3:    Code after receive...
  *
  *
  */

 OpCase(recv_mark_f): {
     /*
      * Save the current position in message buffer and the
      * the label for the loop_rec/2 instruction for the
      * the receive statement.
      */
     c_p->msg.mark = (BeamInstr *) Arg(0);
     c_p->msg.saved_last = c_p->msg.last;
     Next(1);
 }

 OpCase(i_recv_set): {
     /*
      * If the mark is valid (points to the loop_rec/2
      * instruction that follows), we know that the saved
      * position points to the first message that could
      * possibly be matched out.
      *
      * If the mark is invalid, we do nothing, meaning that
      * we will look through all messages in the message queue.
      */
     if (c_p->msg.mark == (BeamInstr *) (I+1)) {
	 c_p->msg.save = c_p->msg.saved_last;
     }
     I++;
     /* Fall through to the loop_rec/2 instruction */
 }

    /*
     * 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):
 {
     BeamInstr *next;
     ErlMessage* msgp;

 loop_rec__:

     PROCESS_MAIN_CHK_LOCKS(c_p);

     msgp = PEEK_MESSAGE(c_p);

     if (!msgp) {
#ifdef ERTS_SMP
	 erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
	 /* Make sure messages wont pass exit signals... */
	 if (ERTS_PROC_PENDING_EXIT(c_p)) {
	     erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
	     SWAPOUT;
	     goto do_schedule; /* Will be rescheduled for exit */
	 }
	 ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
	 msgp = PEEK_MESSAGE(c_p);
	 if (msgp)
	     erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
	 else {
#endif
	     SET_I((BeamInstr *) Arg(0));
	     Goto(*I);		/* Jump to a wait or wait_timeout instruction */
#ifdef ERTS_SMP
	 }
#endif
     }
     ErtsMoveMsgAttachmentIntoProc(msgp, c_p, E, HTOP, FCALLS,
				   {
				       SWAPOUT;
				       reg[0] = r(0);
				       PROCESS_MAIN_CHK_LOCKS(c_p);
				   },
				   {
				       PROCESS_MAIN_CHK_LOCKS(c_p);
				       r(0) = reg[0];
				       SWAPIN;
				   });
     if (is_non_value(ERL_MESSAGE_TERM(msgp))) {
	 /*
	  * A corrupt distribution message that we weren't able to decode;
	  * remove it...
	  */
	 ASSERT(!msgp->data.attached);
	 UNLINK_MESSAGE(c_p, msgp);
	 free_message(msgp);
	 goto loop_rec__;
     }
     PreFetch(1, next);
     r(0) = ERL_MESSAGE_TERM(msgp);
     NextPF(1, next);
 }

 /*
  * Remove a (matched) message from the message queue.
  */
 OpCase(remove_message): {
     BeamInstr *next;
     ErlMessage* msgp;

     PROCESS_MAIN_CHK_LOCKS(c_p);

     PreFetch(0, next);
     msgp = PEEK_MESSAGE(c_p);

     if (ERTS_PROC_GET_SAVED_CALLS_BUF(c_p)) {
	 save_calls(c_p, &exp_receive);
     }
     if (ERL_MESSAGE_TOKEN(msgp) == NIL) {
	 SEQ_TRACE_TOKEN(c_p) = NIL;
     } else if (ERL_MESSAGE_TOKEN(msgp) != am_undefined) {
	 Eterm msg;
	 SEQ_TRACE_TOKEN(c_p) = ERL_MESSAGE_TOKEN(msgp);
	 ASSERT(is_tuple(SEQ_TRACE_TOKEN(c_p)));
	 ASSERT(SEQ_TRACE_TOKEN_ARITY(c_p) == 5);
	 ASSERT(is_small(SEQ_TRACE_TOKEN_SERIAL(c_p)));
	 ASSERT(is_small(SEQ_TRACE_TOKEN_LASTCNT(c_p)));
	 ASSERT(is_small(SEQ_TRACE_TOKEN_FLAGS(c_p)));
	 ASSERT(is_pid(SEQ_TRACE_TOKEN_SENDER(c_p)));
	 c_p->seq_trace_lastcnt = unsigned_val(SEQ_TRACE_TOKEN_SERIAL(c_p));
	 if (c_p->seq_trace_clock < unsigned_val(SEQ_TRACE_TOKEN_SERIAL(c_p))) {
	     c_p->seq_trace_clock = unsigned_val(SEQ_TRACE_TOKEN_SERIAL(c_p));
	 }
	 msg = ERL_MESSAGE_TERM(msgp);
	 seq_trace_output(SEQ_TRACE_TOKEN(c_p), msg, SEQ_TRACE_RECEIVE, 
			  c_p->id, c_p);
     }
     UNLINK_MESSAGE(c_p, msgp);
     JOIN_MESSAGE(c_p);
     CANCEL_TIMER(c_p);
     free_message(msgp);

     PROCESS_MAIN_CHK_LOCKS(c_p);

     NextPF(0, next);
 }

    /*
     * Advance the save pointer to the next message (the current
     * message didn't match), then jump to the loop_rec instruction.
     */
 OpCase(loop_rec_end_f): {
     SET_I((BeamInstr *) Arg(0));
     SAVE_MESSAGE(c_p);
     goto loop_rec__;
 }
    /*
     * Prepare to wait for a message or a timeout, whichever occurs first.
     *
     * Note: In order to keep the compatibility between 32 and 64 bits
     * emulators, only timeout values that can be represented in 32 bits
     * (unsigned) or less…