Strengthen alias_ptr_types_compatible_p in LTO mode.

[thirdparty/gcc.git] / gcc / alias.c

/* Alias analysis for GNU C
   Copyright (C) 1997-2019 Free Software Foundation, Inc.
   Contributed by John Carr (jfc@mit.edu).

This file is part of GCC.

GCC is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free
Software Foundation; either version 3, or (at your option) any later
version.

GCC is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
for more details.

You should have received a copy of the GNU General Public License
along with GCC; see the file COPYING3.  If not see
<http://www.gnu.org/licenses/>.  */

#include "config.h"
#include "system.h"
#include "coretypes.h"
#include "backend.h"
#include "target.h"
#include "rtl.h"
#include "tree.h"
#include "gimple.h"
#include "df.h"
#include "memmodel.h"
#include "tm_p.h"
#include "gimple-ssa.h"
#include "emit-rtl.h"
#include "alias.h"
#include "fold-const.h"
#include "varasm.h"
#include "cselib.h"
#include "langhooks.h"
#include "cfganal.h"
#include "rtl-iter.h"
#include "cgraph.h"
#include "ipa-utils.h"

/* The aliasing API provided here solves related but different problems:

   Say there exists (in c)

   struct X {
     struct Y y1;
     struct Z z2;
   } x1, *px1,  *px2;

   struct Y y2, *py;
   struct Z z2, *pz;


   py = &x1.y1;
   px2 = &x1;

   Consider the four questions:

   Can a store to x1 interfere with px2->y1?
   Can a store to x1 interfere with px2->z2?
   Can a store to x1 change the value pointed to by with py?
   Can a store to x1 change the value pointed to by with pz?

   The answer to these questions can be yes, yes, yes, and maybe.

   The first two questions can be answered with a simple examination
   of the type system.  If structure X contains a field of type Y then
   a store through a pointer to an X can overwrite any field that is
   contained (recursively) in an X (unless we know that px1 != px2).

   The last two questions can be solved in the same way as the first
   two questions but this is too conservative.  The observation is
   that in some cases we can know which (if any) fields are addressed
   and if those addresses are used in bad ways.  This analysis may be
   language specific.  In C, arbitrary operations may be applied to
   pointers.  However, there is some indication that this may be too
   conservative for some C++ types.

   The pass ipa-type-escape does this analysis for the types whose
   instances do not escape across the compilation boundary.

   Historically in GCC, these two problems were combined and a single
   data structure that was used to represent the solution to these
   problems.  We now have two similar but different data structures,
   The data structure to solve the last two questions is similar to
   the first, but does not contain the fields whose address are never
   taken.  For types that do escape the compilation unit, the data
   structures will have identical information.
*/

/* The alias sets assigned to MEMs assist the back-end in determining
   which MEMs can alias which other MEMs.  In general, two MEMs in
   different alias sets cannot alias each other, with one important
   exception.  Consider something like:

     struct S { int i; double d; };

   a store to an `S' can alias something of either type `int' or type
   `double'.  (However, a store to an `int' cannot alias a `double'
   and vice versa.)  We indicate this via a tree structure that looks
   like:
           struct S
            /   \
           /     \
         |/_     _\|
         int    double

   (The arrows are directed and point downwards.)
    In this situation we say the alias set for `struct S' is the
   `superset' and that those for `int' and `double' are `subsets'.

   To see whether two alias sets can point to the same memory, we must
   see if either alias set is a subset of the other. We need not trace
   past immediate descendants, however, since we propagate all
   grandchildren up one level.

   Alias set zero is implicitly a superset of all other alias sets.
   However, this is no actual entry for alias set zero.  It is an
   error to attempt to explicitly construct a subset of zero.  */

struct alias_set_hash : int_hash <int, INT_MIN, INT_MIN + 1> {};

struct GTY(()) alias_set_entry {
  /* The alias set number, as stored in MEM_ALIAS_SET.  */
  alias_set_type alias_set;

  /* Nonzero if would have a child of zero: this effectively makes this
     alias set the same as alias set zero.  */
  bool has_zero_child;
  /* Nonzero if alias set corresponds to pointer type itself (i.e. not to
     aggregate contaiing pointer.
     This is used for a special case where we need an universal pointer type
     compatible with all other pointer types.  */
  bool is_pointer;
  /* Nonzero if is_pointer or if one of childs have has_pointer set.  */
  bool has_pointer;

  /* The children of the alias set.  These are not just the immediate
     children, but, in fact, all descendants.  So, if we have:

       struct T { struct S s; float f; }

     continuing our example above, the children here will be all of
     `int', `double', `float', and `struct S'.  */
  hash_map<alias_set_hash, int> *children;
};

static int rtx_equal_for_memref_p (const_rtx, const_rtx);
static void record_set (rtx, const_rtx, void *);
static int base_alias_check (rtx, rtx, rtx, rtx, machine_mode,
                             machine_mode);
static rtx find_base_value (rtx);
static int mems_in_disjoint_alias_sets_p (const_rtx, const_rtx);
static alias_set_entry *get_alias_set_entry (alias_set_type);
static tree decl_for_component_ref (tree);
static int write_dependence_p (const_rtx,
                               const_rtx, machine_mode, rtx,
                               bool, bool, bool);
static int compare_base_symbol_refs (const_rtx, const_rtx);

static void memory_modified_1 (rtx, const_rtx, void *);

/* Query statistics for the different low-level disambiguators.
   A high-level query may trigger multiple of them.  */

static struct {
  unsigned long long num_alias_zero;
  unsigned long long num_same_alias_set;
  unsigned long long num_same_objects;
  unsigned long long num_volatile;
  unsigned long long num_dag;
  unsigned long long num_universal;
  unsigned long long num_disambiguated;
} alias_stats;


/* Set up all info needed to perform alias analysis on memory references.  */

/* Returns the size in bytes of the mode of X.  */
#define SIZE_FOR_MODE(X) (GET_MODE_SIZE (GET_MODE (X)))

/* Cap the number of passes we make over the insns propagating alias
   information through set chains.
   ??? 10 is a completely arbitrary choice.  This should be based on the
   maximum loop depth in the CFG, but we do not have this information
   available (even if current_loops _is_ available).  */
#define MAX_ALIAS_LOOP_PASSES 10

/* reg_base_value[N] gives an address to which register N is related.
   If all sets after the first add or subtract to the current value
   or otherwise modify it so it does not point to a different top level
   object, reg_base_value[N] is equal to the address part of the source
   of the first set.

   A base address can be an ADDRESS, SYMBOL_REF, or LABEL_REF.  ADDRESS
   expressions represent three types of base:

     1. incoming arguments.  There is just one ADDRESS to represent all
        arguments, since we do not know at this level whether accesses
        based on different arguments can alias.  The ADDRESS has id 0.

     2. stack_pointer_rtx, frame_pointer_rtx, hard_frame_pointer_rtx
        (if distinct from frame_pointer_rtx) and arg_pointer_rtx.
        Each of these rtxes has a separate ADDRESS associated with it,
        each with a negative id.

        GCC is (and is required to be) precise in which register it
        chooses to access a particular region of stack.  We can therefore
        assume that accesses based on one of these rtxes do not alias
        accesses based on another of these rtxes.

     3. bases that are derived from malloc()ed memory (REG_NOALIAS).
        Each such piece of memory has a separate ADDRESS associated
        with it, each with an id greater than 0.

   Accesses based on one ADDRESS do not alias accesses based on other
   ADDRESSes.  Accesses based on ADDRESSes in groups (2) and (3) do not
   alias globals either; the ADDRESSes have Pmode to indicate this.
   The ADDRESS in group (1) _may_ alias globals; it has VOIDmode to
   indicate this.  */

static GTY(()) vec<rtx, va_gc> *reg_base_value;
static rtx *new_reg_base_value;

/* The single VOIDmode ADDRESS that represents all argument bases.
   It has id 0.  */
static GTY(()) rtx arg_base_value;

/* Used to allocate unique ids to each REG_NOALIAS ADDRESS.  */
static int unique_id;

/* We preserve the copy of old array around to avoid amount of garbage
   produced.  About 8% of garbage produced were attributed to this
   array.  */
static GTY((deletable)) vec<rtx, va_gc> *old_reg_base_value;

/* Values of XINT (address, 0) of Pmode ADDRESS rtxes for special
   registers.  */
#define UNIQUE_BASE_VALUE_SP    -1
#define UNIQUE_BASE_VALUE_ARGP  -2
#define UNIQUE_BASE_VALUE_FP    -3
#define UNIQUE_BASE_VALUE_HFP   -4

#define static_reg_base_value \
  (this_target_rtl->x_static_reg_base_value)

#define REG_BASE_VALUE(X)                                       \
  (REGNO (X) < vec_safe_length (reg_base_value)                 \
   ? (*reg_base_value)[REGNO (X)] : 0)

/* Vector indexed by N giving the initial (unchanging) value known for
   pseudo-register N.  This vector is initialized in init_alias_analysis,
   and does not change until end_alias_analysis is called.  */
static GTY(()) vec<rtx, va_gc> *reg_known_value;

/* Vector recording for each reg_known_value whether it is due to a
   REG_EQUIV note.  Future passes (viz., reload) may replace the
   pseudo with the equivalent expression and so we account for the
   dependences that would be introduced if that happens.

   The REG_EQUIV notes created in assign_parms may mention the arg
   pointer, and there are explicit insns in the RTL that modify the
   arg pointer.  Thus we must ensure that such insns don't get
   scheduled across each other because that would invalidate the
   REG_EQUIV notes.  One could argue that the REG_EQUIV notes are
   wrong, but solving the problem in the scheduler will likely give
   better code, so we do it here.  */
static sbitmap reg_known_equiv_p;

/* True when scanning insns from the start of the rtl to the
   NOTE_INSN_FUNCTION_BEG note.  */
static bool copying_arguments;


/* The splay-tree used to store the various alias set entries.  */
static GTY (()) vec<alias_set_entry *, va_gc> *alias_sets;
\f
/* Build a decomposed reference object for querying the alias-oracle
   from the MEM rtx and store it in *REF.
   Returns false if MEM is not suitable for the alias-oracle.  */

static bool
ao_ref_from_mem (ao_ref *ref, const_rtx mem)
{
  tree expr = MEM_EXPR (mem);
  tree base;

  if (!expr)
    return false;

  ao_ref_init (ref, expr);

  /* Get the base of the reference and see if we have to reject or
     adjust it.  */
  base = ao_ref_base (ref);
  if (base == NULL_TREE)
    return false;

  /* The tree oracle doesn't like bases that are neither decls
     nor indirect references of SSA names.  */
  if (!(DECL_P (base)
        || (TREE_CODE (base) == MEM_REF
            && TREE_CODE (TREE_OPERAND (base, 0)) == SSA_NAME)
        || (TREE_CODE (base) == TARGET_MEM_REF
            && TREE_CODE (TMR_BASE (base)) == SSA_NAME)))
    return false;

  ref->ref_alias_set = MEM_ALIAS_SET (mem);

  /* If MEM_OFFSET or MEM_SIZE are unknown what we got from MEM_EXPR
     is conservative, so trust it.  */
  if (!MEM_OFFSET_KNOWN_P (mem)
      || !MEM_SIZE_KNOWN_P (mem))
    return true;

  /* If MEM_OFFSET/MEM_SIZE get us outside of ref->offset/ref->max_size
     drop ref->ref.  */
  if (maybe_lt (MEM_OFFSET (mem), 0)
      || (ref->max_size_known_p ()
          && maybe_gt ((MEM_OFFSET (mem) + MEM_SIZE (mem)) * BITS_PER_UNIT,
                       ref->max_size)))
    ref->ref = NULL_TREE;

  /* Refine size and offset we got from analyzing MEM_EXPR by using
     MEM_SIZE and MEM_OFFSET.  */

  ref->offset += MEM_OFFSET (mem) * BITS_PER_UNIT;
  ref->size = MEM_SIZE (mem) * BITS_PER_UNIT;

  /* The MEM may extend into adjacent fields, so adjust max_size if
     necessary.  */
  if (ref->max_size_known_p ())
    ref->max_size = upper_bound (ref->max_size, ref->size);

  /* If MEM_OFFSET and MEM_SIZE might get us outside of the base object of
     the MEM_EXPR punt.  This happens for STRICT_ALIGNMENT targets a lot.  */
  if (MEM_EXPR (mem) != get_spill_slot_decl (false)
      && (maybe_lt (ref->offset, 0)
          || (DECL_P (ref->base)
              && (DECL_SIZE (ref->base) == NULL_TREE
                  || !poly_int_tree_p (DECL_SIZE (ref->base))
                  || maybe_lt (wi::to_poly_offset (DECL_SIZE (ref->base)),
                               ref->offset + ref->size)))))
    return false;

  return true;
}

/* Query the alias-oracle on whether the two memory rtx X and MEM may
   alias.  If TBAA_P is set also apply TBAA.  Returns true if the
   two rtxen may alias, false otherwise.  */

static bool
rtx_refs_may_alias_p (const_rtx x, const_rtx mem, bool tbaa_p)
{
  ao_ref ref1, ref2;

  if (!ao_ref_from_mem (&ref1, x)
      || !ao_ref_from_mem (&ref2, mem))
    return true;

  return refs_may_alias_p_1 (&ref1, &ref2,
                             tbaa_p
                             && MEM_ALIAS_SET (x) != 0
                             && MEM_ALIAS_SET (mem) != 0);
}

/* Returns a pointer to the alias set entry for ALIAS_SET, if there is
   such an entry, or NULL otherwise.  */

static inline alias_set_entry *
get_alias_set_entry (alias_set_type alias_set)
{
  return (*alias_sets)[alias_set];
}

/* Returns nonzero if the alias sets for MEM1 and MEM2 are such that
   the two MEMs cannot alias each other.  */

static inline int
mems_in_disjoint_alias_sets_p (const_rtx mem1, const_rtx mem2)
{
  return (flag_strict_aliasing
          && ! alias_sets_conflict_p (MEM_ALIAS_SET (mem1),
                                      MEM_ALIAS_SET (mem2)));
}

/* Return true if the first alias set is a subset of the second.  */

bool
alias_set_subset_of (alias_set_type set1, alias_set_type set2)
{
  alias_set_entry *ase2;

  /* Disable TBAA oracle with !flag_strict_aliasing.  */
  if (!flag_strict_aliasing)
    return true;

  /* Everything is a subset of the "aliases everything" set.  */
  if (set2 == 0)
    return true;

  /* Check if set1 is a subset of set2.  */
  ase2 = get_alias_set_entry (set2);
  if (ase2 != 0
      && (ase2->has_zero_child
          || (ase2->children && ase2->children->get (set1))))
    return true;

  /* As a special case we consider alias set of "void *" to be both subset
     and superset of every alias set of a pointer.  This extra symmetry does
     not matter for alias_sets_conflict_p but it makes aliasing_component_refs_p
     to return true on the following testcase:

     void *ptr;
     char **ptr2=(char **)&ptr;
     *ptr2 = ...

     Additionally if a set contains universal pointer, we consider every pointer
     to be a subset of it, but we do not represent this explicitely - doing so
     would require us to update transitive closure each time we introduce new
     pointer type.  This makes aliasing_component_refs_p to return true
     on the following testcase:

     struct a {void *ptr;}
     char **ptr = (char **)&a.ptr;
     ptr = ...

     This makes void * truly universal pointer type.  See pointer handling in
     get_alias_set for more details.  */
  if (ase2 && ase2->has_pointer)
    {
      alias_set_entry *ase1 = get_alias_set_entry (set1);

      if (ase1 && ase1->is_pointer)
        {
          alias_set_type voidptr_set = TYPE_ALIAS_SET (ptr_type_node);
          /* If one is ptr_type_node and other is pointer, then we consider
             them subset of each other.  */
          if (set1 == voidptr_set || set2 == voidptr_set)
            return true;
          /* If SET2 contains universal pointer's alias set, then we consdier
             every (non-universal) pointer.  */
          if (ase2->children && set1 != voidptr_set
              && ase2->children->get (voidptr_set))
            return true;
        }
    }
  return false;
}

/* Return 1 if the two specified alias sets may conflict.  */

int
alias_sets_conflict_p (alias_set_type set1, alias_set_type set2)
{
  alias_set_entry *ase1;
  alias_set_entry *ase2;

  /* The easy case.  */
  if (alias_sets_must_conflict_p (set1, set2))
    return 1;

  /* See if the first alias set is a subset of the second.  */
  ase1 = get_alias_set_entry (set1);
  if (ase1 != 0
      && ase1->children && ase1->children->get (set2))
    {
      ++alias_stats.num_dag;
      return 1;
    }

  /* Now do the same, but with the alias sets reversed.  */
  ase2 = get_alias_set_entry (set2);
  if (ase2 != 0
      && ase2->children && ase2->children->get (set1))
    {
      ++alias_stats.num_dag;
      return 1;
    }

  /* We want void * to be compatible with any other pointer without
     really dropping it to alias set 0. Doing so would make it
     compatible with all non-pointer types too.

     This is not strictly necessary by the C/C++ language
     standards, but avoids common type punning mistakes.  In
     addition to that, we need the existence of such universal
     pointer to implement Fortran's C_PTR type (which is defined as
     type compatible with all C pointers).  */
  if (ase1 && ase2 && ase1->has_pointer && ase2->has_pointer)
    {
      alias_set_type voidptr_set = TYPE_ALIAS_SET (ptr_type_node);

      /* If one of the sets corresponds to universal pointer,
         we consider it to conflict with anything that is
         or contains pointer.  */
      if (set1 == voidptr_set || set2 == voidptr_set)
        {
          ++alias_stats.num_universal;
          return true;
        }
     /* If one of sets is (non-universal) pointer and the other
        contains universal pointer, we also get conflict.  */
     if (ase1->is_pointer && set2 != voidptr_set
         && ase2->children && ase2->children->get (voidptr_set))
        {
          ++alias_stats.num_universal;
          return true;
        }
     if (ase2->is_pointer && set1 != voidptr_set
         && ase1->children && ase1->children->get (voidptr_set))
        {
          ++alias_stats.num_universal;
          return true;
        }
    }

  ++alias_stats.num_disambiguated;

  /* The two alias sets are distinct and neither one is the
     child of the other.  Therefore, they cannot conflict.  */
  return 0;
}

/* Return 1 if the two specified alias sets will always conflict.  */

int
alias_sets_must_conflict_p (alias_set_type set1, alias_set_type set2)
{
  /* Disable TBAA oracle with !flag_strict_aliasing.  */
  if (!flag_strict_aliasing)
    return 1;
  if (set1 == 0 || set2 == 0)
    {
      ++alias_stats.num_alias_zero;
      return 1;
    }
  if (set1 == set2)
    {
      ++alias_stats.num_same_alias_set;
      return 1;
    }

  return 0;
}

/* Return 1 if any MEM object of type T1 will always conflict (using the
   dependency routines in this file) with any MEM object of type T2.
   This is used when allocating temporary storage.  If T1 and/or T2 are
   NULL_TREE, it means we know nothing about the storage.  */

int
objects_must_conflict_p (tree t1, tree t2)
{
  alias_set_type set1, set2;

  /* If neither has a type specified, we don't know if they'll conflict
     because we may be using them to store objects of various types, for
     example the argument and local variables areas of inlined functions.  */
  if (t1 == 0 && t2 == 0)
    return 0;

  /* If they are the same type, they must conflict.  */
  if (t1 == t2)
    {
      ++alias_stats.num_same_objects;
      return 1;
    }
  /* Likewise if both are volatile.  */
  if (t1 != 0 && TYPE_VOLATILE (t1) && t2 != 0 && TYPE_VOLATILE (t2))
    {
      ++alias_stats.num_volatile;
      return 1;
    }

  set1 = t1 ? get_alias_set (t1) : 0;
  set2 = t2 ? get_alias_set (t2) : 0;

  /* We can't use alias_sets_conflict_p because we must make sure
     that every subtype of t1 will conflict with every subtype of
     t2 for which a pair of subobjects of these respective subtypes
     overlaps on the stack.  */
  return alias_sets_must_conflict_p (set1, set2);
}
\f
/* Return the outermost parent of component present in the chain of
   component references handled by get_inner_reference in T with the
   following property:
     - the component is non-addressable
   or NULL_TREE if no such parent exists.  In the former cases, the alias
   set of this parent is the alias set that must be used for T itself.  */

tree
component_uses_parent_alias_set_from (const_tree t)
{
  const_tree found = NULL_TREE;

  while (handled_component_p (t))
    {
      switch (TREE_CODE (t))
        {
        case COMPONENT_REF:
          if (DECL_NONADDRESSABLE_P (TREE_OPERAND (t, 1)))
            found = t;
          /* Permit type-punning when accessing a union, provided the access
             is directly through the union.  For example, this code does not
             permit taking the address of a union member and then storing
             through it.  Even the type-punning allowed here is a GCC
             extension, albeit a common and useful one; the C standard says
             that such accesses have implementation-defined behavior.  */
          else if (TREE_CODE (TREE_TYPE (TREE_OPERAND (t, 0))) == UNION_TYPE)
            found = t;
          break;

        case ARRAY_REF:
        case ARRAY_RANGE_REF:
          if (TYPE_NONALIASED_COMPONENT (TREE_TYPE (TREE_OPERAND (t, 0))))
            found = t;
          break;

        case REALPART_EXPR:
        case IMAGPART_EXPR:
          break;

        case BIT_FIELD_REF:
        case VIEW_CONVERT_EXPR:
          /* Bitfields and casts are never addressable.  */
          found = t;
          break;

        default:
          gcc_unreachable ();
        }

      t = TREE_OPERAND (t, 0);
    }
 
  if (found)
    return TREE_OPERAND (found, 0);

  return NULL_TREE;
}


/* Return whether the pointer-type T effective for aliasing may
   access everything and thus the reference has to be assigned
   alias-set zero.  */

static bool
ref_all_alias_ptr_type_p (const_tree t)
{
  return (TREE_CODE (TREE_TYPE (t)) == VOID_TYPE
          || TYPE_REF_CAN_ALIAS_ALL (t));
}

/* Return the alias set for the memory pointed to by T, which may be
   either a type or an expression.  Return -1 if there is nothing
   special about dereferencing T.  */

static alias_set_type
get_deref_alias_set_1 (tree t)
{
  /* All we care about is the type.  */
  if (! TYPE_P (t))
    t = TREE_TYPE (t);

  /* If we have an INDIRECT_REF via a void pointer, we don't
     know anything about what that might alias.  Likewise if the
     pointer is marked that way.  */
  if (ref_all_alias_ptr_type_p (t))
    return 0;

  return -1;
}

/* Return the alias set for the memory pointed to by T, which may be
   either a type or an expression.  */

alias_set_type
get_deref_alias_set (tree t)
{
  /* If we're not doing any alias analysis, just assume everything
     aliases everything else.  */
  if (!flag_strict_aliasing)
    return 0;

  alias_set_type set = get_deref_alias_set_1 (t);

  /* Fall back to the alias-set of the pointed-to type.  */
  if (set == -1)
    {
      if (! TYPE_P (t))
        t = TREE_TYPE (t);
      set = get_alias_set (TREE_TYPE (t));
    }

  return set;
}

/* Return the pointer-type relevant for TBAA purposes from the
   memory reference tree *T or NULL_TREE in which case *T is
   adjusted to point to the outermost component reference that
   can be used for assigning an alias set.  */
 
static tree
reference_alias_ptr_type_1 (tree *t)
{
  tree inner;

  /* Get the base object of the reference.  */
  inner = *t;
  while (handled_component_p (inner))
    {
      /* If there is a VIEW_CONVERT_EXPR in the chain we cannot use
         the type of any component references that wrap it to
         determine the alias-set.  */
      if (TREE_CODE (inner) == VIEW_CONVERT_EXPR)
        *t = TREE_OPERAND (inner, 0);
      inner = TREE_OPERAND (inner, 0);
    }

  /* Handle pointer dereferences here, they can override the
     alias-set.  */
  if (INDIRECT_REF_P (inner)
      && ref_all_alias_ptr_type_p (TREE_TYPE (TREE_OPERAND (inner, 0))))
    return TREE_TYPE (TREE_OPERAND (inner, 0));
  else if (TREE_CODE (inner) == TARGET_MEM_REF)
    return TREE_TYPE (TMR_OFFSET (inner));
  else if (TREE_CODE (inner) == MEM_REF
           && ref_all_alias_ptr_type_p (TREE_TYPE (TREE_OPERAND (inner, 1))))
    return TREE_TYPE (TREE_OPERAND (inner, 1));

  /* If the innermost reference is a MEM_REF that has a
     conversion embedded treat it like a VIEW_CONVERT_EXPR above,
     using the memory access type for determining the alias-set.  */
  if (TREE_CODE (inner) == MEM_REF
      && (TYPE_MAIN_VARIANT (TREE_TYPE (inner))
          != TYPE_MAIN_VARIANT
               (TREE_TYPE (TREE_TYPE (TREE_OPERAND (inner, 1))))))
    return TREE_TYPE (TREE_OPERAND (inner, 1));

  /* Otherwise, pick up the outermost object that we could have
     a pointer to.  */
  tree tem = component_uses_parent_alias_set_from (*t);
  if (tem)
    *t = tem;

  return NULL_TREE;
}

/* Return the pointer-type relevant for TBAA purposes from the
   gimple memory reference tree T.  This is the type to be used for
   the offset operand of MEM_REF or TARGET_MEM_REF replacements of T
   and guarantees that get_alias_set will return the same alias
   set for T and the replacement.  */

tree
reference_alias_ptr_type (tree t)
{
  /* If the frontend assigns this alias-set zero, preserve that.  */
  if (lang_hooks.get_alias_set (t) == 0)
    return ptr_type_node;

  tree ptype = reference_alias_ptr_type_1 (&t);
  /* If there is a given pointer type for aliasing purposes, return it.  */
  if (ptype != NULL_TREE)
    return ptype;

  /* Otherwise build one from the outermost component reference we
     may use.  */
  if (TREE_CODE (t) == MEM_REF
      || TREE_CODE (t) == TARGET_MEM_REF)
    return TREE_TYPE (TREE_OPERAND (t, 1));
  else
    return build_pointer_type (TYPE_MAIN_VARIANT (TREE_TYPE (t)));
}

/* Return whether the pointer-types T1 and T2 used to determine
   two alias sets of two references will yield the same answer
   from get_deref_alias_set.  */

bool
alias_ptr_types_compatible_p (tree t1, tree t2)
{
  if (TYPE_MAIN_VARIANT (t1) == TYPE_MAIN_VARIANT (t2))
    return true;

  if (ref_all_alias_ptr_type_p (t1)
      || ref_all_alias_ptr_type_p (t2))
    return false;

    /* This function originally abstracts from simply comparing
       get_deref_alias_set so that we are sure this still computes
       the same result after LTO type merging is applied.
       When in LTO type merging is done we can actually do this compare.
    */
  if (in_lto_p)
    return get_deref_alias_set (t1) == get_deref_alias_set (t2);
  else
    return (TYPE_MAIN_VARIANT (TREE_TYPE (t1))
            == TYPE_MAIN_VARIANT (TREE_TYPE (t2)));
}

/* Create emptry alias set entry.  */

alias_set_entry *
init_alias_set_entry (alias_set_type set)
{
  alias_set_entry *ase = ggc_alloc<alias_set_entry> ();
  ase->alias_set = set;
  ase->children = NULL;
  ase->has_zero_child = false;
  ase->is_pointer = false;
  ase->has_pointer = false;
  gcc_checking_assert (!get_alias_set_entry (set));
  (*alias_sets)[set] = ase;
  return ase;
}

/* Return the alias set for T, which may be either a type or an
   expression.  Call language-specific routine for help, if needed.  */

alias_set_type
get_alias_set (tree t)
{
  alias_set_type set;

  /* We cannot give up with -fno-strict-aliasing because we need to build
     proper type representation for possible functions which are build with
     -fstrict-aliasing.  */

  /* return 0 if this or its type is an error.  */
  if (t == error_mark_node
      || (! TYPE_P (t)
          && (TREE_TYPE (t) == 0 || TREE_TYPE (t) == error_mark_node)))
    return 0;

  /* We can be passed either an expression or a type.  This and the
     language-specific routine may make mutually-recursive calls to each other
     to figure out what to do.  At each juncture, we see if this is a tree
     that the language may need to handle specially.  First handle things that
     aren't types.  */
  if (! TYPE_P (t))
    {
      /* Give the language a chance to do something with this tree
         before we look at it.  */
      STRIP_NOPS (t);
      set = lang_hooks.get_alias_set (t);
      if (set != -1)
        return set;

      /* Get the alias pointer-type to use or the outermost object
         that we could have a pointer to.  */
      tree ptype = reference_alias_ptr_type_1 (&t);
      if (ptype != NULL)
        return get_deref_alias_set (ptype);

      /* If we've already determined the alias set for a decl, just return
         it.  This is necessary for C++ anonymous unions, whose component
         variables don't look like union members (boo!).  */
      if (VAR_P (t)
          && DECL_RTL_SET_P (t) && MEM_P (DECL_RTL (t)))
        return MEM_ALIAS_SET (DECL_RTL (t));

      /* Now all we care about is the type.  */
      t = TREE_TYPE (t);
    }

  /* Variant qualifiers don't affect the alias set, so get the main
     variant.  */
  t = TYPE_MAIN_VARIANT (t);

  if (AGGREGATE_TYPE_P (t)
      && TYPE_TYPELESS_STORAGE (t))
    return 0;

  /* Always use the canonical type as well.  If this is a type that
     requires structural comparisons to identify compatible types
     use alias set zero.  */
  if (TYPE_STRUCTURAL_EQUALITY_P (t))
    {
      /* Allow the language to specify another alias set for this
         type.  */
      set = lang_hooks.get_alias_set (t);
      if (set != -1)
        return set;
      /* Handle structure type equality for pointer types, arrays and vectors.
         This is easy to do, because the code bellow ignore canonical types on
         these anyway.  This is important for LTO, where TYPE_CANONICAL for
         pointers cannot be meaningfuly computed by the frotnend.  */
      if (canonical_type_used_p (t))
        {
          /* In LTO we set canonical types for all types where it makes
             sense to do so.  Double check we did not miss some type.  */
          gcc_checking_assert (!in_lto_p || !type_with_alias_set_p (t));
          return 0;
        }
    }
  else
    {
      t = TYPE_CANONICAL (t);
      gcc_checking_assert (!TYPE_STRUCTURAL_EQUALITY_P (t));
    }

  /* If this is a type with a known alias set, return it.  */
  gcc_checking_assert (t == TYPE_MAIN_VARIANT (t));
  if (TYPE_ALIAS_SET_KNOWN_P (t))
    return TYPE_ALIAS_SET (t);

  /* We don't want to set TYPE_ALIAS_SET for incomplete types.  */
  if (!COMPLETE_TYPE_P (t))
    {
      /* For arrays with unknown size the conservative answer is the
         alias set of the element type.  */
      if (TREE_CODE (t) == ARRAY_TYPE)
        return get_alias_set (TREE_TYPE (t));

      /* But return zero as a conservative answer for incomplete types.  */
      return 0;
    }

  /* See if the language has special handling for this type.  */
  set = lang_hooks.get_alias_set (t);
  if (set != -1)
    return set;

  /* There are no objects of FUNCTION_TYPE, so there's no point in
     using up an alias set for them.  (There are, of course, pointers
     and references to functions, but that's different.)  */
  else if (TREE_CODE (t) == FUNCTION_TYPE || TREE_CODE (t) == METHOD_TYPE)
    set = 0;

  /* Unless the language specifies otherwise, let vector types alias
     their components.  This avoids some nasty type punning issues in
     normal usage.  And indeed lets vectors be treated more like an
     array slice.  */
  else if (TREE_CODE (t) == VECTOR_TYPE)
    set = get_alias_set (TREE_TYPE (t));

  /* Unless the language specifies otherwise, treat array types the
     same as their components.  This avoids the asymmetry we get
     through recording the components.  Consider accessing a
     character(kind=1) through a reference to a character(kind=1)[1:1].
     Or consider if we want to assign integer(kind=4)[0:D.1387] and
     integer(kind=4)[4] the same alias set or not.
     Just be pragmatic here and make sure the array and its element
     type get the same alias set assigned.  */
  else if (TREE_CODE (t) == ARRAY_TYPE
           && (!TYPE_NONALIASED_COMPONENT (t)
               || TYPE_STRUCTURAL_EQUALITY_P (t)))
    set = get_alias_set (TREE_TYPE (t));

  /* From the former common C and C++ langhook implementation:

     Unfortunately, there is no canonical form of a pointer type.
     In particular, if we have `typedef int I', then `int *', and
     `I *' are different types.  So, we have to pick a canonical
     representative.  We do this below.

     Technically, this approach is actually more conservative that
     it needs to be.  In particular, `const int *' and `int *'
     should be in different alias sets, according to the C and C++
     standard, since their types are not the same, and so,
     technically, an `int **' and `const int **' cannot point at
     the same thing.

     But, the standard is wrong.  In particular, this code is
     legal C++:

     int *ip;
     int **ipp = &ip;
     const int* const* cipp = ipp;
     And, it doesn't make sense for that to be legal unless you
     can dereference IPP and CIPP.  So, we ignore cv-qualifiers on
     the pointed-to types.  This issue has been reported to the
     C++ committee.

     For this reason go to canonical type of the unqalified pointer type.
     Until GCC 6 this code set all pointers sets to have alias set of
     ptr_type_node but that is a bad idea, because it prevents disabiguations
     in between pointers.  For Firefox this accounts about 20% of all
     disambiguations in the program.  */
  else if (POINTER_TYPE_P (t) && t != ptr_type_node)
    {
      tree p;
      auto_vec <bool, 8> reference;

      /* Unnest all pointers and references.
         We also want to make pointer to array/vector equivalent to pointer to
         its element (see the reasoning above). Skip all those types, too.  */
      for (p = t; POINTER_TYPE_P (p)
           || (TREE_CODE (p) == ARRAY_TYPE
               && (!TYPE_NONALIASED_COMPONENT (p)
                   || !COMPLETE_TYPE_P (p)
                   || TYPE_STRUCTURAL_EQUALITY_P (p)))
           || TREE_CODE (p) == VECTOR_TYPE;
           p = TREE_TYPE (p))
        {
          /* Ada supports recusive pointers.  Instead of doing recrusion check
             just give up once the preallocated space of 8 elements is up.
             In this case just punt to void * alias set.  */
          if (reference.length () == 8)
            {
              p = ptr_type_node;
              break;
            }
          if (TREE_CODE (p) == REFERENCE_TYPE)
            /* In LTO we want languages that use references to be compatible
               with languages that use pointers.  */
            reference.safe_push (true && !in_lto_p);
          if (TREE_CODE (p) == POINTER_TYPE)
            reference.safe_push (false);
        }
      p = TYPE_MAIN_VARIANT (p);

      /* In LTO for C++ programs we can turn in complete types to complete
         using ODR name lookup.  */
      if (in_lto_p && TYPE_STRUCTURAL_EQUALITY_P (p) && odr_type_p (p))
        {
          p = prevailing_odr_type (p);
          gcc_checking_assert (TYPE_MAIN_VARIANT (p) == p);
        }

      /* Make void * compatible with char * and also void **.
         Programs are commonly violating TBAA by this.

         We also make void * to conflict with every pointer
         (see record_component_aliases) and thus it is safe it to use it for
         pointers to types with TYPE_STRUCTURAL_EQUALITY_P.  */
      if (TREE_CODE (p) == VOID_TYPE || TYPE_STRUCTURAL_EQUALITY_P (p))
        set = get_alias_set (ptr_type_node);
      else
        {
          /* Rebuild pointer type starting from canonical types using
             unqualified pointers and references only.  This way all such
             pointers will have the same alias set and will conflict with
             each other.

             Most of time we already have pointers or references of a given type.
             If not we build new one just to be sure that if someone later
             (probably only middle-end can, as we should assign all alias
             classes only after finishing translation unit) builds the pointer
             type, the canonical type will match.  */
          p = TYPE_CANONICAL (p);
          while (!reference.is_empty ())
            {
              if (reference.pop ())
                p = build_reference_type (p);
              else
                p = build_pointer_type (p);
              gcc_checking_assert (p == TYPE_MAIN_VARIANT (p));
              /* build_pointer_type should always return the canonical type.
                 For LTO TYPE_CANOINCAL may be NULL, because we do not compute
                 them.  Be sure that frontends do not glob canonical types of
                 pointers in unexpected way and that p == TYPE_CANONICAL (p)
                 in all other cases.  */
              gcc_checking_assert (!TYPE_CANONICAL (p)
                                   || p == TYPE_CANONICAL (p));
            }

          /* Assign the alias set to both p and t.
             We cannot call get_alias_set (p) here as that would trigger
             infinite recursion when p == t.  In other cases it would just
             trigger unnecesary legwork of rebuilding the pointer again.  */
          gcc_checking_assert (p == TYPE_MAIN_VARIANT (p));
          if (TYPE_ALIAS_SET_KNOWN_P (p))
            set = TYPE_ALIAS_SET (p);
          else
            {
              set = new_alias_set ();
              TYPE_ALIAS_SET (p) = set;
            }
        }
    }
  /* Alias set of ptr_type_node is special and serve as universal pointer which
     is TBAA compatible with every other pointer type.  Be sure we have the
     alias set built even for LTO which otherwise keeps all TYPE_CANONICAL
     of pointer types NULL.  */
  else if (t == ptr_type_node)
    set = new_alias_set ();

  /* Otherwise make a new alias set for this type.  */
  else
    {
      /* Each canonical type gets its own alias set, so canonical types
         shouldn't form a tree.  It doesn't really matter for types
         we handle specially above, so only check it where it possibly
         would result in a bogus alias set.  */
      gcc_checking_assert (TYPE_CANONICAL (t) == t);

      set = new_alias_set ();
    }

  TYPE_ALIAS_SET (t) = set;

  /* If this is an aggregate type or a complex type, we must record any
     component aliasing information.  */
  if (AGGREGATE_TYPE_P (t) || TREE_CODE (t) == COMPLEX_TYPE)
    record_component_aliases (t);

  /* We treat pointer types specially in alias_set_subset_of.  */
  if (POINTER_TYPE_P (t) && set)
    {
      alias_set_entry *ase = get_alias_set_entry (set);
      if (!ase)
        ase = init_alias_set_entry (set);
      ase->is_pointer = true;
      ase->has_pointer = true;
    }

  return set;
}

/* Return a brand-new alias set.  */

alias_set_type
new_alias_set (void)
{
  if (alias_sets == 0)
    vec_safe_push (alias_sets, (alias_set_entry *) NULL);
  vec_safe_push (alias_sets, (alias_set_entry *) NULL);
  return alias_sets->length () - 1;
}

/* Indicate that things in SUBSET can alias things in SUPERSET, but that
   not everything that aliases SUPERSET also aliases SUBSET.  For example,
   in C, a store to an `int' can alias a load of a structure containing an
   `int', and vice versa.  But it can't alias a load of a 'double' member
   of the same structure.  Here, the structure would be the SUPERSET and
   `int' the SUBSET.  This relationship is also described in the comment at
   the beginning of this file.

   This function should be called only once per SUPERSET/SUBSET pair.

   It is illegal for SUPERSET to be zero; everything is implicitly a
   subset of alias set zero.  */

void
record_alias_subset (alias_set_type superset, alias_set_type subset)
{
  alias_set_entry *superset_entry;
  alias_set_entry *subset_entry;

  /* It is possible in complex type situations for both sets to be the same,
     in which case we can ignore this operation.  */
  if (superset == subset)
    return;

  gcc_assert (superset);

  superset_entry = get_alias_set_entry (superset);
  if (superset_entry == 0)
    {
      /* Create an entry for the SUPERSET, so that we have a place to
         attach the SUBSET.  */
      superset_entry = init_alias_set_entry (superset);
    }

  if (subset == 0)
    superset_entry->has_zero_child = 1;
  else
    {
      subset_entry = get_alias_set_entry (subset);
      if (!superset_entry->children)
        superset_entry->children
          = hash_map<alias_set_hash, int>::create_ggc (64);
      /* If there is an entry for the subset, enter all of its children
         (if they are not already present) as children of the SUPERSET.  */
      if (subset_entry)
        {
          if (subset_entry->has_zero_child)
            superset_entry->has_zero_child = true;
          if (subset_entry->has_pointer)
            superset_entry->has_pointer = true;

          if (subset_entry->children)
            {
              hash_map<alias_set_hash, int>::iterator iter
                = subset_entry->children->begin ();
              for (; iter != subset_entry->children->end (); ++iter)
                superset_entry->children->put ((*iter).first, (*iter).second);
            }
        }

      /* Enter the SUBSET itself as a child of the SUPERSET.  */
      superset_entry->children->put (subset, 0);
    }
}

/* Record that component types of TYPE, if any, are part of that type for
   aliasing purposes.  For record types, we only record component types
   for fields that are not marked non-addressable.  For array types, we
   only record the component type if it is not marked non-aliased.  */

void
record_component_aliases (tree type)
{
  alias_set_type superset = get_alias_set (type);
  tree field;

  if (superset == 0)
    return;

  switch (TREE_CODE (type))
    {
    case RECORD_TYPE:
    case UNION_TYPE:
    case QUAL_UNION_TYPE:
      {
        /* LTO non-ODR type merging does not make any difference between 
           component pointer types.  We may have

           struct foo {int *a;};

           as TYPE_CANONICAL of 

           struct bar {float *a;};

           Because accesses to int * and float * do not alias, we would get
           false negative when accessing the same memory location by
           float ** and bar *. We thus record the canonical type as:

           struct {void *a;};

           void * is special cased and works as a universal pointer type.
           Accesses to it conflicts with accesses to any other pointer
           type.  */
        bool void_pointers = in_lto_p
                             && (!odr_type_p (type)
                                 || !odr_based_tbaa_p (type));
        for (field = TYPE_FIELDS (type); field != 0; field = DECL_CHAIN (field))
          if (TREE_CODE (field) == FIELD_DECL && !DECL_NONADDRESSABLE_P (field))
            {
              tree t = TREE_TYPE (field);
              if (void_pointers)
                {
                  /* VECTOR_TYPE and ARRAY_TYPE share the alias set with their
                     element type and that type has to be normalized to void *,
                     too, in the case it is a pointer. */
                  while (!canonical_type_used_p (t) && !POINTER_TYPE_P (t))
                    {
                      gcc_checking_assert (TYPE_STRUCTURAL_EQUALITY_P (t));
                      t = TREE_TYPE (t);
                    }
                  if (POINTER_TYPE_P (t))
                    t = ptr_type_node;
                  else if (flag_checking)
                    gcc_checking_assert (get_alias_set (t)
                                         == get_alias_set (TREE_TYPE (field)));
                }

              record_alias_subset (superset, get_alias_set (t));
            }
      }
      break;

    case COMPLEX_TYPE:
      record_alias_subset (superset, get_alias_set (TREE_TYPE (type)));
      break;

    /* VECTOR_TYPE and ARRAY_TYPE share the alias set with their
       element type.  */

    default:
      break;
    }
}

/* Allocate an alias set for use in storing and reading from the varargs
   spill area.  */

static GTY(()) alias_set_type varargs_set = -1;

alias_set_type
get_varargs_alias_set (void)
{
#if 1
  /* We now lower VA_ARG_EXPR, and there's currently no way to attach the
     varargs alias set to an INDIRECT_REF (FIXME!), so we can't
     consistently use the varargs alias set for loads from the varargs
     area.  So don't use it anywhere.  */
  return 0;
#else
  if (varargs_set == -1)
    varargs_set = new_alias_set ();

  return varargs_set;
#endif
}

/* Likewise, but used for the fixed portions of the frame, e.g., register
   save areas.  */

static GTY(()) alias_set_type frame_set = -1;

alias_set_type
get_frame_alias_set (void)
{
  if (frame_set == -1)
    frame_set = new_alias_set ();

  return frame_set;
}

/* Create a new, unique base with id ID.  */

static rtx
unique_base_value (HOST_WIDE_INT id)
{
  return gen_rtx_ADDRESS (Pmode, id);
}

/* Return true if accesses based on any other base value cannot alias
   those based on X.  */

static bool
unique_base_value_p (rtx x)
{
  return GET_CODE (x) == ADDRESS && GET_MODE (x) == Pmode;
}

/* Return true if X is known to be a base value.  */

static bool
known_base_value_p (rtx x)
{
  switch (GET_CODE (x))
    {
    case LABEL_REF:
    case SYMBOL_REF:
      return true;

    case ADDRESS:
      /* Arguments may or may not be bases; we don't know for sure.  */
      return GET_MODE (x) != VOIDmode;

    default:
      return false;
    }
}

/* Inside SRC, the source of a SET, find a base address.  */

static rtx
find_base_value (rtx src)
{
  unsigned int regno;
  scalar_int_mode int_mode;

#if defined (FIND_BASE_TERM)
  /* Try machine-dependent ways to find the base term.  */
  src = FIND_BASE_TERM (src);
#endif

  switch (GET_CODE (src))
    {
    case SYMBOL_REF:
    case LABEL_REF:
      return src;

    case REG:
      regno = REGNO (src);
      /* At the start of a function, argument registers have known base
         values which may be lost later.  Returning an ADDRESS
         expression here allows optimization based on argument values
         even when the argument registers are used for other purposes.  */
      if (regno < FIRST_PSEUDO_REGISTER && copying_arguments)
        return new_reg_base_value[regno];

      /* If a pseudo has a known base value, return it.  Do not do this
         for non-fixed hard regs since it can result in a circular
         dependency chain for registers which have values at function entry.

         The test above is not sufficient because the scheduler may move
         a copy out of an arg reg past the NOTE_INSN_FUNCTION_BEGIN.  */
      if ((regno >= FIRST_PSEUDO_REGISTER || fixed_regs[regno])
          && regno < vec_safe_length (reg_base_value))
        {
          /* If we're inside init_alias_analysis, use new_reg_base_value
             to reduce the number of relaxation iterations.  */
          if (new_reg_base_value && new_reg_base_value[regno]
              && DF_REG_DEF_COUNT (regno) == 1)
            return new_reg_base_value[regno];

          if ((*reg_base_value)[regno])
            return (*reg_base_value)[regno];
        }

      return 0;

    case MEM:
      /* Check for an argument passed in memory.  Only record in the
         copying-arguments block; it is too hard to track changes
         otherwise.  */
      if (copying_arguments
          && (XEXP (src, 0) == arg_pointer_rtx
              || (GET_CODE (XEXP (src, 0)) == PLUS
                  && XEXP (XEXP (src, 0), 0) == arg_pointer_rtx)))
        return arg_base_value;
      return 0;

    case CONST:
      src = XEXP (src, 0);
      if (GET_CODE (src) != PLUS && GET_CODE (src) != MINUS)
        break;

      /* fall through */

    case PLUS:
    case MINUS:
      {
        rtx temp, src_0 = XEXP (src, 0), src_1 = XEXP (src, 1);

        /* If either operand is a REG that is a known pointer, then it
           is the base.  */
        if (REG_P (src_0) && REG_POINTER (src_0))
          return find_base_value (src_0);
        if (REG_P (src_1) && REG_POINTER (src_1))
          return find_base_value (src_1);

        /* If either operand is a REG, then see if we already have
           a known value for it.  */
        if (REG_P (src_0))
          {
            temp = find_base_value (src_0);
            if (temp != 0)
              src_0 = temp;
          }

        if (REG_P (src_1))
          {
            temp = find_base_value (src_1);
            if (temp!= 0)
              src_1 = temp;
          }

        /* If either base is named object or a special address
           (like an argument or stack reference), then use it for the
           base term.  */
        if (src_0 != 0 && known_base_value_p (src_0))
          return src_0;

        if (src_1 != 0 && known_base_value_p (src_1))
          return src_1;

        /* Guess which operand is the base address:
           If either operand is a symbol, then it is the base.  If
           either operand is a CONST_INT, then the other is the base.  */
        if (CONST_INT_P (src_1) || CONSTANT_P (src_0))
          return find_base_value (src_0);
        else if (CONST_INT_P (src_0) || CONSTANT_P (src_1))
          return find_base_value (src_1);

        return 0;
      }

    case LO_SUM:
      /* The standard form is (lo_sum reg sym) so look only at the
         second operand.  */
      return find_base_value (XEXP (src, 1));

    case AND:
      /* If the second operand is constant set the base
         address to the first operand.  */
      if (CONST_INT_P (XEXP (src, 1)) && INTVAL (XEXP (src, 1)) != 0)
        return find_base_value (XEXP (src, 0));
      return 0;

    case TRUNCATE:
      /* As we do not know which address space the pointer is referring to, we can
         handle this only if the target does not support different pointer or
         address modes depending on the address space.  */
      if (!target_default_pointer_address_modes_p ())
        break;
      if (!is_a <scalar_int_mode> (GET_MODE (src), &int_mode)
          || GET_MODE_PRECISION (int_mode) < GET_MODE_PRECISION (Pmode))
        break;
      /* Fall through.  */
    case HIGH:
    case PRE_INC:
    case PRE_DEC:
    case POST_INC:
    case POST_DEC:
    case PRE_MODIFY:
    case POST_MODIFY:
      return find_base_value (XEXP (src, 0));

    case ZERO_EXTEND:
    case SIGN_EXTEND:   /* used for NT/Alpha pointers */
      /* As we do not know which address space the pointer is referring to, we can
         handle this only if the target does not support different pointer or
         address modes depending on the address space.  */
      if (!target_default_pointer_address_modes_p ())
        break;

      {
        rtx temp = find_base_value (XEXP (src, 0));

        if (temp != 0 && CONSTANT_P (temp))
          temp = convert_memory_address (Pmode, temp);

        return temp;
      }

    default:
      break;
    }

  return 0;
}

/* Called from init_alias_analysis indirectly through note_stores,
   or directly if DEST is a register with a REG_NOALIAS note attached.
   SET is null in the latter case.  */

/* While scanning insns to find base values, reg_seen[N] is nonzero if
   register N has been set in this function.  */
static sbitmap reg_seen;

static void
record_set (rtx dest, const_rtx set, void *data ATTRIBUTE_UNUSED)
{
  unsigned regno;
  rtx src;
  int n;

  if (!REG_P (dest))
    return;

  regno = REGNO (dest);

  gcc_checking_assert (regno < reg_base_value->length ());

  n = REG_NREGS (dest);
  if (n != 1)
    {
      while (--n >= 0)
        {
          bitmap_set_bit (reg_seen, regno + n);
          new_reg_base_value[regno + n] = 0;
        }
      return;
    }

  if (set)
    {
      /* A CLOBBER wipes out any old value but does not prevent a previously
         unset register from acquiring a base address (i.e. reg_seen is not
         set).  */
      if (GET_CODE (set) == CLOBBER)
        {
          new_reg_base_value[regno] = 0;
          return;
        }
      /* A CLOBBER_HIGH only wipes out the old value if the mode of the old
         value is greater than that of the clobber.  */
      else if (GET_CODE (set) == CLOBBER_HIGH)
        {
          if (new_reg_base_value[regno] != 0
              && reg_is_clobbered_by_clobber_high (
                   regno, GET_MODE (new_reg_base_value[regno]), XEXP (set, 0)))
            new_reg_base_value[regno] = 0;
          return;
        }

      src = SET_SRC (set);
    }
  else
    {
      /* There's a REG_NOALIAS note against DEST.  */
      if (bitmap_bit_p (reg_seen, regno))
        {
          new_reg_base_value[regno] = 0;
          return;
        }
      bitmap_set_bit (reg_seen, regno);
      new_reg_base_value[regno] = unique_base_value (unique_id++);
      return;
    }

  /* If this is not the first set of REGNO, see whether the new value
     is related to the old one.  There are two cases of interest:

        (1) The register might be assigned an entirely new value
            that has the same base term as the original set.

        (2) The set might be a simple self-modification that
            cannot change REGNO's base value.

     If neither case holds, reject the original base value as invalid.
     Note that the following situation is not detected:

         extern int x, y;  int *p = &x; p += (&y-&x);

     ANSI C does not allow computing the difference of addresses
     of distinct top level objects.  */
  if (new_reg_base_value[regno] != 0
      && find_base_value (src) != new_reg_base_value[regno])
    switch (GET_CODE (src))
      {
      case LO_SUM:
      case MINUS:
        if (XEXP (src, 0) != dest && XEXP (src, 1) != dest)
          new_reg_base_value[regno] = 0;
        break;
      case PLUS:
        /* If the value we add in the PLUS is also a valid base value,
           this might be the actual base value, and the original value
           an index.  */
        {
          rtx other = NULL_RTX;

          if (XEXP (src, 0) == dest)
            other = XEXP (src, 1);
          else if (XEXP (src, 1) == dest)
            other = XEXP (src, 0);

          if (! other || find_base_value (other))
            new_reg_base_value[regno] = 0;
          break;
        }
      case AND:
        if (XEXP (src, 0) != dest || !CONST_INT_P (XEXP (src, 1)))
          new_reg_base_value[regno] = 0;
        break;
      default:
        new_reg_base_value[regno] = 0;
        break;
      }
  /* If this is the first set of a register, record the value.  */
  else if ((regno >= FIRST_PSEUDO_REGISTER || ! fixed_regs[regno])
           && ! bitmap_bit_p (reg_seen, regno) && new_reg_base_value[regno] == 0)
    new_reg_base_value[regno] = find_base_value (src);

  bitmap_set_bit (reg_seen, regno);
}

/* Return REG_BASE_VALUE for REGNO.  Selective scheduler uses this to avoid
   using hard registers with non-null REG_BASE_VALUE for renaming.  */
rtx
get_reg_base_value (unsigned int regno)
{
  return (*reg_base_value)[regno];
}

/* If a value is known for REGNO, return it.  */

rtx
get_reg_known_value (unsigned int regno)
{
  if (regno >= FIRST_PSEUDO_REGISTER)
    {
      regno -= FIRST_PSEUDO_REGISTER;
      if (regno < vec_safe_length (reg_known_value))
        return (*reg_known_value)[regno];
    }
  return NULL;
}

/* Set it.  */

static void
set_reg_known_value (unsigned int regno, rtx val)
{
  if (regno >= FIRST_PSEUDO_REGISTER)
    {
      regno -= FIRST_PSEUDO_REGISTER;
      if (regno < vec_safe_length (reg_known_value))
        (*reg_known_value)[regno] = val;
    }
}

/* Similarly for reg_known_equiv_p.  */

bool
get_reg_known_equiv_p (unsigned int regno)
{
  if (regno >= FIRST_PSEUDO_REGISTER)
    {
      regno -= FIRST_PSEUDO_REGISTER;
      if (regno < vec_safe_length (reg_known_value))
        return bitmap_bit_p (reg_known_equiv_p, regno);
    }
  return false;
}

static void
set_reg_known_equiv_p (unsigned int regno, bool val)
{
  if (regno >= FIRST_PSEUDO_REGISTER)
    {
      regno -= FIRST_PSEUDO_REGISTER;
      if (regno < vec_safe_length (reg_known_value))
        {
          if (val)
            bitmap_set_bit (reg_known_equiv_p, regno);
          else
            bitmap_clear_bit (reg_known_equiv_p, regno);
        }
    }
}


/* Returns a canonical version of X, from the point of view alias
   analysis.  (For example, if X is a MEM whose address is a register,
   and the register has a known value (say a SYMBOL_REF), then a MEM
   whose address is the SYMBOL_REF is returned.)  */

rtx
canon_rtx (rtx x)
{
  /* Recursively look for equivalences.  */
  if (REG_P (x) && REGNO (x) >= FIRST_PSEUDO_REGISTER)
    {
      rtx t = get_reg_known_value (REGNO (x));
      if (t == x)
        return x;
      if (t)
        return canon_rtx (t);
    }

  if (GET_CODE (x) == PLUS)
    {
      rtx x0 = canon_rtx (XEXP (x, 0));
      rtx x1 = canon_rtx (XEXP (x, 1));

      if (x0 != XEXP (x, 0) || x1 != XEXP (x, 1))
        return simplify_gen_binary (PLUS, GET_MODE (x), x0, x1);
    }

  /* This gives us much better alias analysis when called from
     the loop optimizer.   Note we want to leave the original
     MEM alone, but need to return the canonicalized MEM with
     all the flags with their original values.  */
  else if (MEM_P (x))
    x = replace_equiv_address_nv (x, canon_rtx (XEXP (x, 0)));

  return x;
}

/* Return 1 if X and Y are identical-looking rtx's.
   Expect that X and Y has been already canonicalized.

   We use the data in reg_known_value above to see if two registers with
   different numbers are, in fact, equivalent.  */

static int
rtx_equal_for_memref_p (const_rtx x, const_rtx y)
{
  int i;
  int j;
  enum rtx_code code;
  const char *fmt;

  if (x == 0 && y == 0)
    return 1;
  if (x == 0 || y == 0)
    return 0;

  if (x == y)
    return 1;

  code = GET_CODE (x);
  /* Rtx's of different codes cannot be equal.  */
  if (code != GET_CODE (y))
    return 0;

  /* (MULT:SI x y) and (MULT:HI x y) are NOT equivalent.
     (REG:SI x) and (REG:HI x) are NOT equivalent.  */

  if (GET_MODE (x) != GET_MODE (y))
    return 0;

  /* Some RTL can be compared without a recursive examination.  */
  switch (code)
    {
    case REG:
      return REGNO (x) == REGNO (y);

    case LABEL_REF:
      return label_ref_label (x) == label_ref_label (y);

    case SYMBOL_REF:
      return compare_base_symbol_refs (x, y) == 1;

    case ENTRY_VALUE:
      /* This is magic, don't go through canonicalization et al.  */
      return rtx_equal_p (ENTRY_VALUE_EXP (x), ENTRY_VALUE_EXP (y));

    case VALUE:
    CASE_CONST_UNIQUE:
      /* Pointer equality guarantees equality for these nodes.  */
      return 0;

    default:
      break;
    }

  /* canon_rtx knows how to handle plus.  No need to canonicalize.  */
  if (code == PLUS)
    return ((rtx_equal_for_memref_p (XEXP (x, 0), XEXP (y, 0))
             && rtx_equal_for_memref_p (XEXP (x, 1), XEXP (y, 1)))
            || (rtx_equal_for_memref_p (XEXP (x, 0), XEXP (y, 1))
                && rtx_equal_for_memref_p (XEXP (x, 1), XEXP (y, 0))));
  /* For commutative operations, the RTX match if the operand match in any
     order.  Also handle the simple binary and unary cases without a loop.  */
  if (COMMUTATIVE_P (x))
    {
      rtx xop0 = canon_rtx (XEXP (x, 0));
      rtx yop0 = canon_rtx (XEXP (y, 0));
      rtx yop1 = canon_rtx (XEXP (y, 1));

      return ((rtx_equal_for_memref_p (xop0, yop0)
               && rtx_equal_for_memref_p (canon_rtx (XEXP (x, 1)), yop1))
              || (rtx_equal_for_memref_p (xop0, yop1)
                  && rtx_equal_for_memref_p (canon_rtx (XEXP (x, 1)), yop0)));
    }
  else if (NON_COMMUTATIVE_P (x))
    {
      return (rtx_equal_for_memref_p (canon_rtx (XEXP (x, 0)),
                                      canon_rtx (XEXP (y, 0)))
              && rtx_equal_for_memref_p (canon_rtx (XEXP (x, 1)),
                                         canon_rtx (XEXP (y, 1))));
    }
  else if (UNARY_P (x))
    return rtx_equal_for_memref_p (canon_rtx (XEXP (x, 0)),
                                   canon_rtx (XEXP (y, 0)));

  /* Compare the elements.  If any pair of corresponding elements
     fail to match, return 0 for the whole things.

     Limit cases to types which actually appear in addresses.  */

  fmt = GET_RTX_FORMAT (code);
  for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
    {
      switch (fmt[i])
        {
        case 'i':
          if (XINT (x, i) != XINT (y, i))
            return 0;
          break;

        case 'p':
          if (maybe_ne (SUBREG_BYTE (x), SUBREG_BYTE (y)))
            return 0;
          break;

        case 'E':
          /* Two vectors must have the same length.  */
          if (XVECLEN (x, i) != XVECLEN (y, i))
            return 0;

          /* And the corresponding elements must match.  */
          for (j = 0; j < XVECLEN (x, i); j++)
            if (rtx_equal_for_memref_p (canon_rtx (XVECEXP (x, i, j)),
                                        canon_rtx (XVECEXP (y, i, j))) == 0)
              return 0;
          break;

        case 'e':
          if (rtx_equal_for_memref_p (canon_rtx (XEXP (x, i)),
                                      canon_rtx (XEXP (y, i))) == 0)
            return 0;
          break;

          /* This can happen for asm operands.  */
        case 's':
          if (strcmp (XSTR (x, i), XSTR (y, i)))
            return 0;
          break;

        /* This can happen for an asm which clobbers memory.  */
        case '0':
          break;

          /* It is believed that rtx's at this level will never
             contain anything but integers and other rtx's,
             except for within LABEL_REFs and SYMBOL_REFs.  */
        default:
          gcc_unreachable ();
        }
    }
  return 1;
}

static rtx
find_base_term (rtx x, vec<std::pair<cselib_val *,
                                     struct elt_loc_list *> > &visited_vals)
{
  cselib_val *val;
  struct elt_loc_list *l, *f;
  rtx ret;
  scalar_int_mode int_mode;

#if defined (FIND_BASE_TERM)
  /* Try machine-dependent ways to find the base term.  */
  x = FIND_BASE_TERM (x);
#endif

  switch (GET_CODE (x))
    {
    case REG:
      return REG_BASE_VALUE (x);

    case TRUNCATE:
      /* As we do not know which address space the pointer is referring to, we can
         handle this only if the target does not support different pointer or
         address modes depending on the address space.  */
      if (!target_default_pointer_address_modes_p ())
        return 0;
      if (!is_a <scalar_int_mode> (GET_MODE (x), &int_mode)
          || GET_MODE_PRECISION (int_mode) < GET_MODE_PRECISION (Pmode))
        return 0;
      /* Fall through.  */
    case HIGH:
    case PRE_INC:
    case PRE_DEC:
    case POST_INC:
    case POST_DEC:
    case PRE_MODIFY:
    case POST_MODIFY:
      return find_base_term (XEXP (x, 0), visited_vals);

    case ZERO_EXTEND:
    case SIGN_EXTEND:   /* Used for Alpha/NT pointers */
      /* As we do not know which address space the pointer is referring to, we can
         handle this only if the target does not support different pointer or
         address modes depending on the address space.  */
      if (!target_default_pointer_address_modes_p ())
        return 0;

      {
        rtx temp = find_base_term (XEXP (x, 0), visited_vals);

        if (temp != 0 && CONSTANT_P (temp))
          temp = convert_memory_address (Pmode, temp);

        return temp;
      }

    case VALUE:
      val = CSELIB_VAL_PTR (x);
      ret = NULL_RTX;

      if (!val)
        return ret;

      if (cselib_sp_based_value_p (val))
        return static_reg_base_value[STACK_POINTER_REGNUM];

      f = val->locs;
      /* Reset val->locs to avoid infinite recursion.  */
      if (f)
        visited_vals.safe_push (std::make_pair (val, f));
      val->locs = NULL;

      for (l = f; l; l = l->next)
        if (GET_CODE (l->loc) == VALUE
            && CSELIB_VAL_PTR (l->loc)->locs
            && !CSELIB_VAL_PTR (l->loc)->locs->next
            && CSELIB_VAL_PTR (l->loc)->locs->loc == x)
          continue;
        else if ((ret = find_base_term (l->loc, visited_vals)) != 0)
          break;

      return ret;

    case LO_SUM:
      /* The standard form is (lo_sum reg sym) so look only at the
         second operand.  */
      return find_base_term (XEXP (x, 1), visited_vals);

    case CONST:
      x = XEXP (x, 0);
      if (GET_CODE (x) != PLUS && GET_CODE (x) != MINUS)
        return 0;
      /* Fall through.  */
    case PLUS:
    case MINUS:
      {
        rtx tmp1 = XEXP (x, 0);
        rtx tmp2 = XEXP (x, 1);

        /* This is a little bit tricky since we have to determine which of
           the two operands represents the real base address.  Otherwise this
           routine may return the index register instead of the base register.

           That may cause us to believe no aliasing was possible, when in
           fact aliasing is possible.

           We use a few simple tests to guess the base register.  Additional
           tests can certainly be added.  For example, if one of the operands
           is a shift or multiply, then it must be the index register and the
           other operand is the base register.  */

        if (tmp1 == pic_offset_table_rtx && CONSTANT_P (tmp2))
          return find_base_term (tmp2, visited_vals);

        /* If either operand is known to be a pointer, then prefer it
           to determine the base term.  */
        if (REG_P (tmp1) && REG_POINTER (tmp1))
          ;
        else if (REG_P (tmp2) && REG_POINTER (tmp2))
          std::swap (tmp1, tmp2);
        /* If second argument is constant which has base term, prefer it
           over variable tmp1.  See PR64025.  */
        else if (CONSTANT_P (tmp2) && !CONST_INT_P (tmp2))
          std::swap (tmp1, tmp2);

        /* Go ahead and find the base term for both operands.  If either base
           term is from a pointer or is a named object or a special address
           (like an argument or stack reference), then use it for the
           base term.  */
        rtx base = find_base_term (tmp1, visited_vals);
        if (base != NULL_RTX
            && ((REG_P (tmp1) && REG_POINTER (tmp1))
                 || known_base_value_p (base)))
          return base;
        base = find_base_term (tmp2, visited_vals);
        if (base != NULL_RTX
            && ((REG_P (tmp2) && REG_POINTER (tmp2))
                 || known_base_value_p (base)))
          return base;

        /* We could not determine which of the two operands was the
           base register and which was the index.  So we can determine
           nothing from the base alias check.  */
        return 0;
      }

    case AND:
      if (CONST_INT_P (XEXP (x, 1)) && INTVAL (XEXP (x, 1)) != 0)
        return find_base_term (XEXP (x, 0), visited_vals);
      return 0;

    case SYMBOL_REF:
    case LABEL_REF:
      return x;

    default:
      return 0;
    }
}

/* Wrapper around the worker above which removes locs from visited VALUEs
   to avoid visiting them multiple times.  We unwind that changes here.  */

static rtx
find_base_term (rtx x)
{
  auto_vec<std::pair<cselib_val *, struct elt_loc_list *>, 32> visited_vals;
  rtx res = find_base_term (x, visited_vals);
  for (unsigned i = 0; i < visited_vals.length (); ++i)
    visited_vals[i].first->locs = visited_vals[i].second;
  return res;
}

/* Return true if accesses to address X may alias accesses based
   on the stack pointer.  */

bool
may_be_sp_based_p (rtx x)
{
  rtx base = find_base_term (x);
  return !base || base == static_reg_base_value[STACK_POINTER_REGNUM];
}

/* BASE1 and BASE2 are decls.  Return 1 if they refer to same object, 0
   if they refer to different objects and -1 if we cannot decide.  */

int
compare_base_decls (tree base1, tree base2)
{
  int ret;
  gcc_checking_assert (DECL_P (base1) && DECL_P (base2));
  if (base1 == base2)
    return 1;

  /* If we have two register decls with register specification we
     cannot decide unless their assembler names are the same.  */
  if (DECL_REGISTER (base1)
      && DECL_REGISTER (base2)
      && HAS_DECL_ASSEMBLER_NAME_P (base1)
      && HAS_DECL_ASSEMBLER_NAME_P (base2)
      && DECL_ASSEMBLER_NAME_SET_P (base1)
      && DECL_ASSEMBLER_NAME_SET_P (base2))
    {
      if (DECL_ASSEMBLER_NAME_RAW (base1) == DECL_ASSEMBLER_NAME_RAW (base2))
        return 1;
      return -1;
    }

  /* Declarations of non-automatic variables may have aliases.  All other
     decls are unique.  */
  if (!decl_in_symtab_p (base1)
      || !decl_in_symtab_p (base2))
    return 0;

  /* Don't cause symbols to be inserted by the act of checking.  */
  symtab_node *node1 = symtab_node::get (base1);
  if (!node1)
    return 0;
  symtab_node *node2 = symtab_node::get (base2);
  if (!node2)
    return 0;
  
  ret = node1->equal_address_to (node2, true);
  return ret;
}

/* Same as compare_base_decls but for SYMBOL_REF.  */

static int
compare_base_symbol_refs (const_rtx x_base, const_rtx y_base)
{
  tree x_decl = SYMBOL_REF_DECL (x_base);
  tree y_decl = SYMBOL_REF_DECL (y_base);
  bool binds_def = true;

  if (XSTR (x_base, 0) == XSTR (y_base, 0))
    return 1;
  if (x_decl && y_decl)
    return compare_base_decls (x_decl, y_decl);
  if (x_decl || y_decl)
    {
      if (!x_decl)
        {
          std::swap (x_decl, y_decl);
          std::swap (x_base, y_base);
        }
      /* We handle specially only section anchors and assume that other
         labels may overlap with user variables in an arbitrary way.  */
      if (!SYMBOL_REF_HAS_BLOCK_INFO_P (y_base))
        return -1;
      /* Anchors contains static VAR_DECLs and CONST_DECLs.  We are safe
         to ignore CONST_DECLs because they are readonly.  */
      if (!VAR_P (x_decl)
          || (!TREE_STATIC (x_decl) && !TREE_PUBLIC (x_decl)))
        return 0;

      symtab_node *x_node = symtab_node::get_create (x_decl)
                            ->ultimate_alias_target ();
      /* External variable cannot be in section anchor.  */
      if (!x_node->definition)
        return 0;
      x_base = XEXP (DECL_RTL (x_node->decl), 0);
      /* If not in anchor, we can disambiguate.  */
      if (!SYMBOL_REF_HAS_BLOCK_INFO_P (x_base))
        return 0;

      /* We have an alias of anchored variable.  If it can be interposed;
         we must assume it may or may not alias its anchor.  */
      binds_def = decl_binds_to_current_def_p (x_decl);
    }
  /* If we have variable in section anchor, we can compare by offset.  */
  if (SYMBOL_REF_HAS_BLOCK_INFO_P (x_base)
      && SYMBOL_REF_HAS_BLOCK_INFO_P (y_base))
    {
      if (SYMBOL_REF_BLOCK (x_base) != SYMBOL_REF_BLOCK (y_base))
        return 0;
      if (SYMBOL_REF_BLOCK_OFFSET (x_base) == SYMBOL_REF_BLOCK_OFFSET (y_base))
        return binds_def ? 1 : -1;
      if (SYMBOL_REF_ANCHOR_P (x_base) != SYMBOL_REF_ANCHOR_P (y_base))
        return -1;
      return 0;
    }
  /* In general we assume that memory locations pointed to by different labels
     may overlap in undefined ways.  */
  return -1;
}

/* Return 0 if the addresses X and Y are known to point to different
   objects, 1 if they might be pointers to the same object.  */

static int
base_alias_check (rtx x, rtx x_base, rtx y, rtx y_base,
                  machine_mode x_mode, machine_mode y_mode)
{
  /* If the address itself has no known base see if a known equivalent
     value has one.  If either address still has no known base, nothing
     is known about aliasing.  */
  if (x_base == 0)
    {
      rtx x_c;

      if (! flag_expensive_optimizations || (x_c = canon_rtx (x)) == x)
        return 1;

      x_base = find_base_term (x_c);
      if (x_base == 0)
        return 1;
    }

  if (y_base == 0)
    {
      rtx y_c;
      if (! flag_expensive_optimizations || (y_c = canon_rtx (y)) == y)
        return 1;

      y_base = find_base_term (y_c);
      if (y_base == 0)
        return 1;
    }

  /* If the base addresses are equal nothing is known about aliasing.  */
  if (rtx_equal_p (x_base, y_base))
    return 1;

  /* The base addresses are different expressions.  If they are not accessed
     via AND, there is no conflict.  We can bring knowledge of object
     alignment into play here.  For example, on alpha, "char a, b;" can
     alias one another, though "char a; long b;" cannot.  AND addresses may
     implicitly alias surrounding objects; i.e. unaligned access in DImode
     via AND address can alias all surrounding object types except those
     with aligment 8 or higher.  */
  if (GET_CODE (x) == AND && GET_CODE (y) == AND)
    return 1;
  if (GET_CODE (x) == AND
      && (!CONST_INT_P (XEXP (x, 1))
          || (int) GET_MODE_UNIT_SIZE (y_mode) < -INTVAL (XEXP (x, 1))))
    return 1;
  if (GET_CODE (y) == AND
      && (!CONST_INT_P (XEXP (y, 1))
          || (int) GET_MODE_UNIT_SIZE (x_mode) < -INTVAL (XEXP (y, 1))))
    return 1;

  /* Differing symbols not accessed via AND never alias.  */
  if (GET_CODE (x_base) == SYMBOL_REF && GET_CODE (y_base) == SYMBOL_REF)
    return compare_base_symbol_refs (x_base, y_base) != 0;

  if (GET_CODE (x_base) != ADDRESS && GET_CODE (y_base) != ADDRESS)
    return 0;

  if (unique_base_value_p (x_base) || unique_base_value_p (y_base))
    return 0;

  return 1;
}

/* Return TRUE if EXPR refers to a VALUE whose uid is greater than
   (or equal to) that of V.  */

static bool
refs_newer_value_p (const_rtx expr, rtx v)
{
  int minuid = CSELIB_VAL_PTR (v)->uid;
  subrtx_iterator::array_type array;
  FOR_EACH_SUBRTX (iter, array, expr, NONCONST)
    if (GET_CODE (*iter) == VALUE && CSELIB_VAL_PTR (*iter)->uid >= minuid)
      return true;
  return false;
}

/* Convert the address X into something we can use.  This is done by returning
   it unchanged unless it is a VALUE or VALUE +/- constant; for VALUE
   we call cselib to get a more useful rtx.  */

rtx
get_addr (rtx x)
{
  cselib_val *v;
  struct elt_loc_list *l;

  if (GET_CODE (x) != VALUE)
    {
      if ((GET_CODE (x) == PLUS || GET_CODE (x) == MINUS)
          && GET_CODE (XEXP (x, 0)) == VALUE
          && CONST_SCALAR_INT_P (XEXP (x, 1)))
        {
          rtx op0 = get_addr (XEXP (x, 0));
          if (op0 != XEXP (x, 0))
            {
              poly_int64 c;
              if (GET_CODE (x) == PLUS
                  && poly_int_rtx_p (XEXP (x, 1), &c))
                return plus_constant (GET_MODE (x), op0, c);
              return simplify_gen_binary (GET_CODE (x), GET_MODE (x),
                                          op0, XEXP (x, 1));
            }
        }
      return x;
    }
  v = CSELIB_VAL_PTR (x);
  if (v)
    {
      bool have_equivs = cselib_have_permanent_equivalences ();
      if (have_equivs)
        v = canonical_cselib_val (v);
      for (l = v->locs; l; l = l->next)
        if (CONSTANT_P (l->loc))
          return l->loc;
      for (l = v->locs; l; l = l->next)
        if (!REG_P (l->loc) && !MEM_P (l->loc)
            /* Avoid infinite recursion when potentially dealing with
               var-tracking artificial equivalences, by skipping the
               equivalences themselves, and not choosing expressions
               that refer to newer VALUEs.  */
            && (!have_equivs
                || (GET_CODE (l->loc) != VALUE
                    && !refs_newer_value_p (l->loc, x))))
          return l->loc;
      if (have_equivs)
        {
          for (l = v->locs; l; l = l->next)
            if (REG_P (l->loc)
                || (GET_CODE (l->loc) != VALUE
                    && !refs_newer_value_p (l->loc, x)))
              return l->loc;
          /* Return the canonical value.  */
          return v->val_rtx;
        }
      if (v->locs)
        return v->locs->loc;
    }
  return x;
}

/*  Return the address of the (N_REFS + 1)th memory reference to ADDR
    where SIZE is the size in bytes of the memory reference.  If ADDR
    is not modified by the memory reference then ADDR is returned.  */

static rtx
addr_side_effect_eval (rtx addr, poly_int64 size, int n_refs)
{
  poly_int64 offset = 0;

  switch (GET_CODE (addr))
    {
    case PRE_INC:
      offset = (n_refs + 1) * size;
      break;
    case PRE_DEC:
      offset = -(n_refs + 1) * size;
      break;
    case POST_INC:
      offset = n_refs * size;
      break;
    case POST_DEC:
      offset = -n_refs * size;
      break;

    default:
      return addr;
    }

  addr = plus_constant (GET_MODE (addr), XEXP (addr, 0), offset);
  addr = canon_rtx (addr);

  return addr;
}

/* Return TRUE if an object X sized at XSIZE bytes and another object
   Y sized at YSIZE bytes, starting C bytes after X, may overlap.  If
   any of the sizes is zero, assume an overlap, otherwise use the
   absolute value of the sizes as the actual sizes.  */

static inline bool
offset_overlap_p (poly_int64 c, poly_int64 xsize, poly_int64 ysize)
{
  if (known_eq (xsize, 0) || known_eq (ysize, 0))
    return true;

  if (maybe_ge (c, 0))
    return maybe_gt (maybe_lt (xsize, 0) ? -xsize : xsize, c);
  else
    return maybe_gt (maybe_lt (ysize, 0) ? -ysize : ysize, -c);
}

/* Return one if X and Y (memory addresses) reference the
   same location in memory or if the references overlap.
   Return zero if they do not overlap, else return
   minus one in which case they still might reference the same location.

   C is an offset accumulator.  When
   C is nonzero, we are testing aliases between X and Y + C.
   XSIZE is the size in bytes of the X reference,
   similarly YSIZE is the size in bytes for Y.
   Expect that canon_rtx has been already called for X and Y.

   If XSIZE or YSIZE is zero, we do not know the amount of memory being
   referenced (the reference was BLKmode), so make the most pessimistic
   assumptions.

   If XSIZE or YSIZE is negative, we may access memory outside the object
   being referenced as a side effect.  This can happen when using AND to
   align memory references, as is done on the Alpha.

   Nice to notice that varying addresses cannot conflict with fp if no
   local variables had their addresses taken, but that's too hard now.

   ???  Contrary to the tree alias oracle this does not return
   one for X + non-constant and Y + non-constant when X and Y are equal.
   If that is fixed the TBAA hack for union type-punning can be removed.  */

static int
memrefs_conflict_p (poly_int64 xsize, rtx x, poly_int64 ysize, rtx y,
                    poly_int64 c)
{
  if (GET_CODE (x) == VALUE)
    {
      if (REG_P (y))
        {
          struct elt_loc_list *l = NULL;
          if (CSELIB_VAL_PTR (x))
            for (l = canonical_cselib_val (CSELIB_VAL_PTR (x))->locs;
                 l; l = l->next)
              if (REG_P (l->loc) && rtx_equal_for_memref_p (l->loc, y))
                break;
          if (l)
            x = y;
          else
            x = get_addr (x);
        }
      /* Don't call get_addr if y is the same VALUE.  */
      else if (x != y)
        x = get_addr (x);
    }
  if (GET_CODE (y) == VALUE)
    {
      if (REG_P (x))
        {
          struct elt_loc_list *l = NULL;
          if (CSELIB_VAL_PTR (y))
            for (l = canonical_cselib_val (CSELIB_VAL_PTR (y))->locs;
                 l; l = l->next)
              if (REG_P (l->loc) && rtx_equal_for_memref_p (l->loc, x))
                break;
          if (l)
            y = x;
          else
            y = get_addr (y);
        }
      /* Don't call get_addr if x is the same VALUE.  */
      else if (y != x)
        y = get_addr (y);
    }
  if (GET_CODE (x) == HIGH)
    x = XEXP (x, 0);
  else if (GET_CODE (x) == LO_SUM)
    x = XEXP (x, 1);
  else
    x = addr_side_effect_eval (x, maybe_lt (xsize, 0) ? -xsize : xsize, 0);
  if (GET_CODE (y) == HIGH)
    y = XEXP (y, 0);
  else if (GET_CODE (y) == LO_SUM)
    y = XEXP (y, 1);
  else
    y = addr_side_effect_eval (y, maybe_lt (ysize, 0) ? -ysize : ysize, 0);

  if (GET_CODE (x) == SYMBOL_REF && GET_CODE (y) == SYMBOL_REF)
    {
      int cmp = compare_base_symbol_refs (x,y);

      /* If both decls are the same, decide by offsets.  */
      if (cmp == 1)
        return offset_overlap_p (c, xsize, ysize);
      /* Assume a potential overlap for symbolic addresses that went
         through alignment adjustments (i.e., that have negative
         sizes), because we can't know how far they are from each
         other.  */
      if (maybe_lt (xsize, 0) || maybe_lt (ysize, 0))
        return -1;
      /* If decls are different or we know by offsets that there is no overlap,
         we win.  */
      if (!cmp || !offset_overlap_p (c, xsize, ysize))
        return 0;
      /* Decls may or may not be different and offsets overlap....*/
      return -1;
    }
  else if (rtx_equal_for_memref_p (x, y))
    {
      return offset_overlap_p (c, xsize, ysize);
    }

  /* This code used to check for conflicts involving stack references and
     globals but the base address alias code now handles these cases.  */

  if (GET_CODE (x) == PLUS)
    {
      /* The fact that X is canonicalized means that this
         PLUS rtx is canonicalized.  */
      rtx x0 = XEXP (x, 0);
      rtx x1 = XEXP (x, 1);

      /* However, VALUEs might end up in different positions even in
         canonical PLUSes.  Comparing their addresses is enough.  */
      if (x0 == y)
        return memrefs_conflict_p (xsize, x1, ysize, const0_rtx, c);
      else if (x1 == y)
        return memrefs_conflict_p (xsize, x0, ysize, const0_rtx, c);

      poly_int64 cx1, cy1;
      if (GET_CODE (y) == PLUS)
        {
          /* The fact that Y is canonicalized means that this
             PLUS rtx is canonicalized.  */
          rtx y0 = XEXP (y, 0);
          rtx y1 = XEXP (y, 1);

          if (x0 == y1)
            return memrefs_conflict_p (xsize, x1, ysize, y0, c);
          if (x1 == y0)
            return memrefs_conflict_p (xsize, x0, ysize, y1, c);

          if (rtx_equal_for_memref_p (x1, y1))
            return memrefs_conflict_p (xsize, x0, ysize, y0, c);
          if (rtx_equal_for_memref_p (x0, y0))
            return memrefs_conflict_p (xsize, x1, ysize, y1, c);
          if (poly_int_rtx_p (x1, &cx1))
            {
              if (poly_int_rtx_p (y1, &cy1))
                return memrefs_conflict_p (xsize, x0, ysize, y0,
                                           c - cx1 + cy1);
              else
                return memrefs_conflict_p (xsize, x0, ysize, y, c - cx1);
            }
          else if (poly_int_rtx_p (y1, &cy1))
            return memrefs_conflict_p (xsize, x, ysize, y0, c + cy1);

          return -1;
        }
      else if (poly_int_rtx_p (x1, &cx1))
        return memrefs_conflict_p (xsize, x0, ysize, y, c - cx1);
    }
  else if (GET_CODE (y) == PLUS)
    {
      /* The fact that Y is canonicalized means that this
         PLUS rtx is canonicalized.  */
      rtx y0 = XEXP (y, 0);
      rtx y1 = XEXP (y, 1);

      if (x == y0)
        return memrefs_conflict_p (xsize, const0_rtx, ysize, y1, c);
      if (x == y1)
        return memrefs_conflict_p (xsize, const0_rtx, ysize, y0, c);

      poly_int64 cy1;
      if (poly_int_rtx_p (y1, &cy1))
        return memrefs_conflict_p (xsize, x, ysize, y0, c + cy1);
      else
        return -1;
    }

  if (GET_CODE (x) == GET_CODE (y))
    switch (GET_CODE (x))
      {
      case MULT:
        {
          /* Handle cases where we expect the second operands to be the
             same, and check only whether the first operand would conflict
             or not.  */
          rtx x0, y0;
          rtx x1 = canon_rtx (XEXP (x, 1));
          rtx y1 = canon_rtx (XEXP (y, 1));
          if (! rtx_equal_for_memref_p (x1, y1))
            return -1;
          x0 = canon_rtx (XEXP (x, 0));
          y0 = canon_rtx (XEXP (y, 0));
          if (rtx_equal_for_memref_p (x0, y0))
            return offset_overlap_p (c, xsize, ysize);

          /* Can't properly adjust our sizes.  */
          poly_int64 c1;
          if (!poly_int_rtx_p (x1, &c1)
              || !can_div_trunc_p (xsize, c1, &xsize)
              || !can_div_trunc_p (ysize, c1, &ysize)
              || !can_div_trunc_p (c, c1, &c))
            return -1;
          return memrefs_conflict_p (xsize, x0, ysize, y0, c);
        }

      default:
        break;
      }

  /* Deal with alignment ANDs by adjusting offset and size so as to
     cover the maximum range, without taking any previously known
     alignment into account.  Make a size negative after such an
     adjustments, so that, if we end up with e.g. two SYMBOL_REFs, we
     assume a potential overlap, because they may end up in contiguous
     memory locations and the stricter-alignment access may span over
     part of both.  */
  if (GET_CODE (x) == AND && CONST_INT_P (XEXP (x, 1)))
    {
      HOST_WIDE_INT sc = INTVAL (XEXP (x, 1));
      unsigned HOST_WIDE_INT uc = sc;
      if (sc < 0 && pow2_or_zerop (-uc))
        {
          if (maybe_gt (xsize, 0))
            xsize = -xsize;
          if (maybe_ne (xsize, 0))
            xsize += sc + 1;
          c -= sc + 1;
          return memrefs_conflict_p (xsize, canon_rtx (XEXP (x, 0)),
                                     ysize, y, c);
        }
    }
  if (GET_CODE (y) == AND && CONST_INT_P (XEXP (y, 1)))
    {
      HOST_WIDE_INT sc = INTVAL (XEXP (y, 1));
      unsigned HOST_WIDE_INT uc = sc;
      if (sc < 0 && pow2_or_zerop (-uc))
        {
          if (maybe_gt (ysize, 0))
            ysize = -ysize;
          if (maybe_ne (ysize, 0))
            ysize += sc + 1;
          c += sc + 1;
          return memrefs_conflict_p (xsize, x,
                                     ysize, canon_rtx (XEXP (y, 0)), c);
        }
    }

  if (CONSTANT_P (x))
    {
      poly_int64 cx, cy;
      if (poly_int_rtx_p (x, &cx) && poly_int_rtx_p (y, &cy))
        {
          c += cy - cx;
          return offset_overlap_p (c, xsize, ysize);
        }

      if (GET_CODE (x) == CONST)
        {
          if (GET_CODE (y) == CONST)
            return memrefs_conflict_p (xsize, canon_rtx (XEXP (x, 0)),
                                       ysize, canon_rtx (XEXP (y, 0)), c);
          else
            return memrefs_conflict_p (xsize, canon_rtx (XEXP (x, 0)),
                                       ysize, y, c);
        }
      if (GET_CODE (y) == CONST)
        return memrefs_conflict_p (xsize, x, ysize,
                                   canon_rtx (XEXP (y, 0)), c);

      /* Assume a potential overlap for symbolic addresses that went
         through alignment adjustments (i.e., that have negative
         sizes), because we can't know how far they are from each
         other.  */
      if (CONSTANT_P (y))
        return (maybe_lt (xsize, 0)
                || maybe_lt (ysize, 0)
                || offset_overlap_p (c, xsize, ysize));

      return -1;
    }

  return -1;
}

/* Functions to compute memory dependencies.

   Since we process the insns in execution order, we can build tables
   to keep track of what registers are fixed (and not aliased), what registers
   are varying in known ways, and what registers are varying in unknown
   ways.

   If both memory references are volatile, then there must always be a
   dependence between the two references, since their order cannot be
   changed.  A volatile and non-volatile reference can be interchanged
   though.

   We also must allow AND addresses, because they may generate accesses
   outside the object being referenced.  This is used to generate aligned
   addresses from unaligned addresses, for instance, the alpha
   storeqi_unaligned pattern.  */

/* Read dependence: X is read after read in MEM takes place.  There can
   only be a dependence here if both reads are volatile, or if either is
   an explicit barrier.  */

int
read_dependence (const_rtx mem, const_rtx x)
{
  if (MEM_VOLATILE_P (x) && MEM_VOLATILE_P (mem))
    return true;
  if (MEM_ALIAS_SET (x) == ALIAS_SET_MEMORY_BARRIER
      || MEM_ALIAS_SET (mem) == ALIAS_SET_MEMORY_BARRIER)
    return true;
  return false;
}

/* Look at the bottom of the COMPONENT_REF list for a DECL, and return it.  */

static tree
decl_for_component_ref (tree x)
{
  do
    {
      x = TREE_OPERAND (x, 0);
    }
  while (x && TREE_CODE (x) == COMPONENT_REF);

  return x && DECL_P (x) ? x : NULL_TREE;
}

/* Walk up the COMPONENT_REF list in X and adjust *OFFSET to compensate
   for the offset of the field reference.  *KNOWN_P says whether the
   offset is known.  */

static void
adjust_offset_for_component_ref (tree x, bool *known_p,
                                 poly_int64 *offset)
{
  if (!*known_p)
    return;
  do
    {
      tree xoffset = component_ref_field_offset (x);
      tree field = TREE_OPERAND (x, 1);
      if (!poly_int_tree_p (xoffset))
        {
          *known_p = false;
          return;
        }

      poly_offset_int woffset
        = (wi::to_poly_offset (xoffset)
           + (wi::to_offset (DECL_FIELD_BIT_OFFSET (field))
              >> LOG2_BITS_PER_UNIT)
           + *offset);
      if (!woffset.to_shwi (offset))
        {
          *known_p = false;
          return;
        }

      x = TREE_OPERAND (x, 0);
    }
  while (x && TREE_CODE (x) == COMPONENT_REF);
}

/* Return nonzero if we can determine the exprs corresponding to memrefs
   X and Y and they do not overlap. 
   If LOOP_VARIANT is set, skip offset-based disambiguation */

int
nonoverlapping_memrefs_p (const_rtx x, const_rtx y, bool loop_invariant)
{
  tree exprx = MEM_EXPR (x), expry = MEM_EXPR (y);
  rtx rtlx, rtly;
  rtx basex, basey;
  bool moffsetx_known_p, moffsety_known_p;
  poly_int64 moffsetx = 0, moffsety = 0;
  poly_int64 offsetx = 0, offsety = 0, sizex, sizey;

  /* Unless both have exprs, we can't tell anything.  */
  if (exprx == 0 || expry == 0)
    return 0;

  /* For spill-slot accesses make sure we have valid offsets.  */
  if ((exprx == get_spill_slot_decl (false)
       && ! MEM_OFFSET_KNOWN_P (x))
      || (expry == get_spill_slot_decl (false)
          && ! MEM_OFFSET_KNOWN_P (y)))
    return 0;

  /* If the field reference test failed, look at the DECLs involved.  */
  moffsetx_known_p = MEM_OFFSET_KNOWN_P (x);
  if (moffsetx_known_p)
    moffsetx = MEM_OFFSET (x);
  if (TREE_CODE (exprx) == COMPONENT_REF)
    {
      tree t = decl_for_component_ref (exprx);
      if (! t)
        return 0;
      adjust_offset_for_component_ref (exprx, &moffsetx_known_p, &moffsetx);
      exprx = t;
    }

  moffsety_known_p = MEM_OFFSET_KNOWN_P (y);
  if (moffsety_known_p)
    moffsety = MEM_OFFSET (y);
  if (TREE_CODE (expry) == COMPONENT_REF)
    {
      tree t = decl_for_component_ref (expry);
      if (! t)
        return 0;
      adjust_offset_for_component_ref (expry, &moffsety_known_p, &moffsety);
      expry = t;
    }

  if (! DECL_P (exprx) || ! DECL_P (expry))
    return 0;

  /* If we refer to different gimple registers, or one gimple register
     and one non-gimple-register, we know they can't overlap.  First,
     gimple registers don't have their addresses taken.  Now, there
     could be more than one stack slot for (different versions of) the
     same gimple register, but we can presumably tell they don't
     overlap based on offsets from stack base addresses elsewhere.
     It's important that we don't proceed to DECL_RTL, because gimple
     registers may not pass DECL_RTL_SET_P, and make_decl_rtl won't be
     able to do anything about them since no SSA information will have
     remained to guide it.  */
  if (is_gimple_reg (exprx) || is_gimple_reg (expry))
    return exprx != expry
      || (moffsetx_known_p && moffsety_known_p
          && MEM_SIZE_KNOWN_P (x) && MEM_SIZE_KNOWN_P (y)
          && !offset_overlap_p (moffsety - moffsetx,
                                MEM_SIZE (x), MEM_SIZE (y)));

  /* With invalid code we can end up storing into the constant pool.
     Bail out to avoid ICEing when creating RTL for this.
     See gfortran.dg/lto/20091028-2_0.f90.  */
  if (TREE_CODE (exprx) == CONST_DECL
      || TREE_CODE (expry) == CONST_DECL)
    return 1;

  /* If one decl is known to be a function or label in a function and
     the other is some kind of data, they can't overlap.  */
  if ((TREE_CODE (exprx) == FUNCTION_DECL
       || TREE_CODE (exprx) == LABEL_DECL)
      != (TREE_CODE (expry) == FUNCTION_DECL
          || TREE_CODE (expry) == LABEL_DECL))
    return 1;

  /* If either of the decls doesn't have DECL_RTL set (e.g. marked as
     living in multiple places), we can't tell anything.  Exception
     are FUNCTION_DECLs for which we can create DECL_RTL on demand.  */
  if ((!DECL_RTL_SET_P (exprx) && TREE_CODE (exprx) != FUNCTION_DECL)
      || (!DECL_RTL_SET_P (expry) && TREE_CODE (expry) != FUNCTION_DECL))
    return 0;

  rtlx = DECL_RTL (exprx);
  rtly = DECL_RTL (expry);

  /* If either RTL is not a MEM, it must be a REG or CONCAT, meaning they
     can't overlap unless they are the same because we never reuse that part
     of the stack frame used for locals for spilled pseudos.  */
  if ((!MEM_P (rtlx) || !MEM_P (rtly))
      && ! rtx_equal_p (rtlx, rtly))
    return 1;

  /* If we have MEMs referring to different address spaces (which can
     potentially overlap), we cannot easily tell from the addresses
     whether the references overlap.  */
  if (MEM_P (rtlx) && MEM_P (rtly)
      && MEM_ADDR_SPACE (rtlx) != MEM_ADDR_SPACE (rtly))
    return 0;

  /* Get the base and offsets of both decls.  If either is a register, we
     know both are and are the same, so use that as the base.  The only
     we can avoid overlap is if we can deduce that they are nonoverlapping
     pieces of that decl, which is very rare.  */
  basex = MEM_P (rtlx) ? XEXP (rtlx, 0) : rtlx;
  basex = strip_offset_and_add (basex, &offsetx);

  basey = MEM_P (rtly) ? XEXP (rtly, 0) : rtly;
  basey = strip_offset_and_add (basey, &offsety);

  /* If the bases are different, we know they do not overlap if both
     are constants or if one is a constant and the other a pointer into the
     stack frame.  Otherwise a different base means we can't tell if they
     overlap or not.  */
  if (compare_base_decls (exprx, expry) == 0)
    return ((CONSTANT_P (basex) && CONSTANT_P (basey))
            || (CONSTANT_P (basex) && REG_P (basey)
                && REGNO_PTR_FRAME_P (REGNO (basey)))
            || (CONSTANT_P (basey) && REG_P (basex)
                && REGNO_PTR_FRAME_P (REGNO (basex))));

  /* Offset based disambiguation not appropriate for loop invariant */
  if (loop_invariant)
    return 0;

  /* Offset based disambiguation is OK even if we do not know that the
     declarations are necessarily different
    (i.e. compare_base_decls (exprx, expry) == -1)  */

  sizex = (!MEM_P (rtlx) ? poly_int64 (GET_MODE_SIZE (GET_MODE (rtlx)))
           : MEM_SIZE_KNOWN_P (rtlx) ? MEM_SIZE (rtlx)
           : -1);
  sizey = (!MEM_P (rtly) ? poly_int64 (GET_MODE_SIZE (GET_MODE (rtly)))
           : MEM_SIZE_KNOWN_P (rtly) ? MEM_SIZE (rtly)
           : -1);

  /* If we have an offset for either memref, it can update the values computed
     above.  */
  if (moffsetx_known_p)
    offsetx += moffsetx, sizex -= moffsetx;
  if (moffsety_known_p)
    offsety += moffsety, sizey -= moffsety;

  /* If a memref has both a size and an offset, we can use the smaller size.
     We can't do this if the offset isn't known because we must view this
     memref as being anywhere inside the DECL's MEM.  */
  if (MEM_SIZE_KNOWN_P (x) && moffsetx_known_p)
    sizex = MEM_SIZE (x);
  if (MEM_SIZE_KNOWN_P (y) && moffsety_known_p)
    sizey = MEM_SIZE (y);

  return !ranges_maybe_overlap_p (offsetx, sizex, offsety, sizey);
}

/* Helper for true_dependence and canon_true_dependence.
   Checks for true dependence: X is read after store in MEM takes place.

   If MEM_CANONICALIZED is FALSE, then X_ADDR and MEM_ADDR should be
   NULL_RTX, and the canonical addresses of MEM and X are both computed
   here.  If MEM_CANONICALIZED, then MEM must be already canonicalized.

   If X_ADDR is non-NULL, it is used in preference of XEXP (x, 0).

   Returns 1 if there is a true dependence, 0 otherwise.  */

static int
true_dependence_1 (const_rtx mem, machine_mode mem_mode, rtx mem_addr,
                   const_rtx x, rtx x_addr, bool mem_canonicalized)
{
  rtx true_mem_addr;
  rtx base;
  int ret;

  gcc_checking_assert (mem_canonicalized ? (mem_addr != NULL_RTX)
                       : (mem_addr == NULL_RTX && x_addr == NULL_RTX));

  if (MEM_VOLATILE_P (x) && MEM_VOLATILE_P (mem))
    return 1;

  /* (mem:BLK (scratch)) is a special mechanism to conflict with everything.
     This is used in epilogue deallocation functions, and in cselib.  */
  if (GET_MODE (x) == BLKmode && GET_CODE (XEXP (x, 0)) == SCRATCH)
    return 1;
  if (GET_MODE (mem) == BLKmode && GET_CODE (XEXP (mem, 0)) == SCRATCH)
    return 1;
  if (MEM_ALIAS_SET (x) == ALIAS_SET_MEMORY_BARRIER
      || MEM_ALIAS_SET (mem) == ALIAS_SET_MEMORY_BARRIER)
    return 1;

  if (! x_addr)
    x_addr = XEXP (x, 0);
  x_addr = get_addr (x_addr);

  if (! mem_addr)
    {
      mem_addr = XEXP (mem, 0);
      if (mem_mode == VOIDmode)
        mem_mode = GET_MODE (mem);
    }
  true_mem_addr = get_addr (mem_addr);

  /* Read-only memory is by definition never modified, and therefore can't
     conflict with anything.  However, don't assume anything when AND
     addresses are involved and leave to the code below to determine
     dependence.  We don't expect to find read-only set on MEM, but
     stupid user tricks can produce them, so don't die.  */
  if (MEM_READONLY_P (x)
      && GET_CODE (x_addr) != AND
      && GET_CODE (true_mem_addr) != AND)
    return 0;

  /* If we have MEMs referring to different address spaces (which can
     potentially overlap), we cannot easily tell from the addresses
     whether the references overlap.  */
  if (MEM_ADDR_SPACE (mem) != MEM_ADDR_SPACE (x))
    return 1;

  base = find_base_term (x_addr);
  if (base && (GET_CODE (base) == LABEL_REF
               || (GET_CODE (base) == SYMBOL_REF
                   && CONSTANT_POOL_ADDRESS_P (base))))
    return 0;

  rtx mem_base = find_base_term (true_mem_addr);
  if (! base_alias_check (x_addr, base, true_mem_addr, mem_base,
                          GET_MODE (x), mem_mode))
    return 0;

  x_addr = canon_rtx (x_addr);
  if (!mem_canonicalized)
    mem_addr = canon_rtx (true_mem_addr);

  if ((ret = memrefs_conflict_p (GET_MODE_SIZE (mem_mode), mem_addr,
                                 SIZE_FOR_MODE (x), x_addr, 0)) != -1)
    return ret;

  if (mems_in_disjoint_alias_sets_p (x, mem))
    return 0;

  if (nonoverlapping_memrefs_p (mem, x, false))
    return 0;

  return rtx_refs_may_alias_p (x, mem, true);
}

/* True dependence: X is read after store in MEM takes place.  */

int
true_dependence (const_rtx mem, machine_mode mem_mode, const_rtx x)
{
  return true_dependence_1 (mem, mem_mode, NULL_RTX,
                            x, NULL_RTX, /*mem_canonicalized=*/false);
}

/* Canonical true dependence: X is read after store in MEM takes place.
   Variant of true_dependence which assumes MEM has already been
   canonicalized (hence we no longer do that here).
   The mem_addr argument has been added, since true_dependence_1 computed
   this value prior to canonicalizing.  */

int
canon_true_dependence (const_rtx mem, machine_mode mem_mode, rtx mem_addr,
                       const_rtx x, rtx x_addr)
{
  return true_dependence_1 (mem, mem_mode, mem_addr,
                            x, x_addr, /*mem_canonicalized=*/true);
}

/* Returns nonzero if a write to X might alias a previous read from
   (or, if WRITEP is true, a write to) MEM.
   If X_CANONCALIZED is true, then X_ADDR is the canonicalized address of X,
   and X_MODE the mode for that access.
   If MEM_CANONICALIZED is true, MEM is canonicalized.  */

static int
write_dependence_p (const_rtx mem,
                    const_rtx x, machine_mode x_mode, rtx x_addr,
                    bool mem_canonicalized, bool x_canonicalized, bool writep)
{
  rtx mem_addr;
  rtx true_mem_addr, true_x_addr;
  rtx base;
  int ret;

  gcc_checking_assert (x_canonicalized
                       ? (x_addr != NULL_RTX
                          && (x_mode != VOIDmode || GET_MODE (x) == VOIDmode))
                       : (x_addr == NULL_RTX && x_mode == VOIDmode));

  if (MEM_VOLATILE_P (x) && MEM_VOLATILE_P (mem))
    return 1;

  /* (mem:BLK (scratch)) is a special mechanism to conflict with everything.
     This is used in epilogue deallocation functions.  */
  if (GET_MODE (x) == BLKmode && GET_CODE (XEXP (x, 0)) == SCRATCH)
    return 1;
  if (GET_MODE (mem) == BLKmode && GET_CODE (XEXP (mem, 0)) == SCRATCH)
    return 1;
  if (MEM_ALIAS_SET (x) == ALIAS_SET_MEMORY_BARRIER
      || MEM_ALIAS_SET (mem) == ALIAS_SET_MEMORY_BARRIER)
    return 1;

  if (!x_addr)
    x_addr = XEXP (x, 0);
  true_x_addr = get_addr (x_addr);

  mem_addr = XEXP (mem, 0);
  true_mem_addr = get_addr (mem_addr);

  /* A read from read-only memory can't conflict with read-write memory.
     Don't assume anything when AND addresses are involved and leave to
     the code below to determine dependence.  */
  if (!writep
      && MEM_READONLY_P (mem)
      && GET_CODE (true_x_addr) != AND
      && GET_CODE (true_mem_addr) != AND)
    return 0;

  /* If we have MEMs referring to different address spaces (which can
     potentially overlap), we cannot easily tell from the addresses
     whether the references overlap.  */
  if (MEM_ADDR_SPACE (mem) != MEM_ADDR_SPACE (x))
    return 1;

  base = find_base_term (true_mem_addr);
  if (! writep
      && base
      && (GET_CODE (base) == LABEL_REF
          || (GET_CODE (base) == SYMBOL_REF
              && CONSTANT_POOL_ADDRESS_P (base))))
    return 0;

  rtx x_base = find_base_term (true_x_addr);
  if (! base_alias_check (true_x_addr, x_base, true_mem_addr, base,
                          GET_MODE (x), GET_MODE (mem)))
    return 0;

  if (!x_canonicalized)
    {
      x_addr = canon_rtx (true_x_addr);
      x_mode = GET_MODE (x);
    }
  if (!mem_canonicalized)
    mem_addr = canon_rtx (true_mem_addr);

  if ((ret = memrefs_conflict_p (SIZE_FOR_MODE (mem), mem_addr,
                                 GET_MODE_SIZE (x_mode), x_addr, 0)) != -1)
    return ret;

  if (nonoverlapping_memrefs_p (x, mem, false))
    return 0;

  return rtx_refs_may_alias_p (x, mem, false);
}

/* Anti dependence: X is written after read in MEM takes place.  */

int
anti_dependence (const_rtx mem, const_rtx x)
{
  return write_dependence_p (mem, x, VOIDmode, NULL_RTX,
                             /*mem_canonicalized=*/false,
                             /*x_canonicalized*/false, /*writep=*/false);
}

/* Likewise, but we already have a canonicalized MEM, and X_ADDR for X.
   Also, consider X in X_MODE (which might be from an enclosing
   STRICT_LOW_PART / ZERO_EXTRACT).
   If MEM_CANONICALIZED is true, MEM is canonicalized.  */

int
canon_anti_dependence (const_rtx mem, bool mem_canonicalized,
                       const_rtx x, machine_mode x_mode, rtx x_addr)
{
  return write_dependence_p (mem, x, x_mode, x_addr,
                             mem_canonicalized, /*x_canonicalized=*/true,
                             /*writep=*/false);
}

/* Output dependence: X is written after store in MEM takes place.  */

int
output_dependence (const_rtx mem, const_rtx x)
{
  return write_dependence_p (mem, x, VOIDmode, NULL_RTX,
                             /*mem_canonicalized=*/false,
                             /*x_canonicalized*/false, /*writep=*/true);
}

/* Likewise, but we already have a canonicalized MEM, and X_ADDR for X.
   Also, consider X in X_MODE (which might be from an enclosing
   STRICT_LOW_PART / ZERO_EXTRACT).
   If MEM_CANONICALIZED is true, MEM is canonicalized.  */

int
canon_output_dependence (const_rtx mem, bool mem_canonicalized,
                         const_rtx x, machine_mode x_mode, rtx x_addr)
{
  return write_dependence_p (mem, x, x_mode, x_addr,
                             mem_canonicalized, /*x_canonicalized=*/true,
                             /*writep=*/true);
}
\f


/* Check whether X may be aliased with MEM.  Don't do offset-based
  memory disambiguation & TBAA.  */
int
may_alias_p (const_rtx mem, const_rtx x)
{
  rtx x_addr, mem_addr;

  if (MEM_VOLATILE_P (x) && MEM_VOLATILE_P (mem))
    return 1;

  /* (mem:BLK (scratch)) is a special mechanism to conflict with everything.
     This is used in epilogue deallocation functions.  */
  if (GET_MODE (x) == BLKmode && GET_CODE (XEXP (x, 0)) == SCRATCH)
    return 1;
  if (GET_MODE (mem) == BLKmode && GET_CODE (XEXP (mem, 0)) == SCRATCH)
    return 1;
  if (MEM_ALIAS_SET (x) == ALIAS_SET_MEMORY_BARRIER
      || MEM_ALIAS_SET (mem) == ALIAS_SET_MEMORY_BARRIER)
    return 1;

  x_addr = XEXP (x, 0);
  x_addr = get_addr (x_addr);

  mem_addr = XEXP (mem, 0);
  mem_addr = get_addr (mem_addr);

  /* Read-only memory is by definition never modified, and therefore can't
     conflict with anything.  However, don't assume anything when AND
     addresses are involved and leave to the code below to determine
     dependence.  We don't expect to find read-only set on MEM, but
     stupid user tricks can produce them, so don't die.  */
  if (MEM_READONLY_P (x)
      && GET_CODE (x_addr) != AND
      && GET_CODE (mem_addr) != AND)
    return 0;

  /* If we have MEMs referring to different address spaces (which can
     potentially overlap), we cannot easily tell from the addresses
     whether the references overlap.  */
  if (MEM_ADDR_SPACE (mem) != MEM_ADDR_SPACE (x))
    return 1;

  rtx x_base = find_base_term (x_addr);
  rtx mem_base = find_base_term (mem_addr);
  if (! base_alias_check (x_addr, x_base, mem_addr, mem_base,
                          GET_MODE (x), GET_MODE (mem_addr)))
    return 0;

  if (nonoverlapping_memrefs_p (mem, x, true))
    return 0;

  /* TBAA not valid for loop_invarint */
  return rtx_refs_may_alias_p (x, mem, false);
}

void
init_alias_target (void)
{
  int i;

  if (!arg_base_value)
    arg_base_value = gen_rtx_ADDRESS (VOIDmode, 0);

  memset (static_reg_base_value, 0, sizeof static_reg_base_value);

  for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
    /* Check whether this register can hold an incoming pointer
       argument.  FUNCTION_ARG_REGNO_P tests outgoing register
       numbers, so translate if necessary due to register windows.  */
    if (FUNCTION_ARG_REGNO_P (OUTGOING_REGNO (i))
        && targetm.hard_regno_mode_ok (i, Pmode))
      static_reg_base_value[i] = arg_base_value;

  /* RTL code is required to be consistent about whether it uses the
     stack pointer, the frame pointer or the argument pointer to
     access a given area of the frame.  We can therefore use the
     base address to distinguish between the different areas.  */
  static_reg_base_value[STACK_POINTER_REGNUM]
    = unique_base_value (UNIQUE_BASE_VALUE_SP);
  static_reg_base_value[ARG_POINTER_REGNUM]
    = unique_base_value (UNIQUE_BASE_VALUE_ARGP);
  static_reg_base_value[FRAME_POINTER_REGNUM]
    = unique_base_value (UNIQUE_BASE_VALUE_FP);

  /* The above rules extend post-reload, with eliminations applying
     consistently to each of the three pointers.  Cope with cases in
     which the frame pointer is eliminated to the hard frame pointer
     rather than the stack pointer.  */
  if (!HARD_FRAME_POINTER_IS_FRAME_POINTER)
    static_reg_base_value[HARD_FRAME_POINTER_REGNUM]
      = unique_base_value (UNIQUE_BASE_VALUE_HFP);
}

/* Set MEMORY_MODIFIED when X modifies DATA (that is assumed
   to be memory reference.  */
static bool memory_modified;
static void
memory_modified_1 (rtx x, const_rtx pat ATTRIBUTE_UNUSED, void *data)
{
  if (MEM_P (x))
    {
      if (anti_dependence (x, (const_rtx)data) || output_dependence (x, (const_rtx)data))
        memory_modified = true;
    }
}


/* Return true when INSN possibly modify memory contents of MEM
   (i.e. address can be modified).  */
bool
memory_modified_in_insn_p (const_rtx mem, const_rtx insn)
{
  if (!INSN_P (insn))
    return false;
  /* Conservatively assume all non-readonly MEMs might be modified in
     calls.  */
  if (CALL_P (insn))
    return true;
  memory_modified = false;
  note_stores (PATTERN (insn), memory_modified_1, CONST_CAST_RTX(mem));
  return memory_modified;
}

/* Initialize the aliasing machinery.  Initialize the REG_KNOWN_VALUE
   array.  */

void
init_alias_analysis (void)
{
  unsigned int maxreg = max_reg_num ();
  int changed, pass;
  int i;
  unsigned int ui;
  rtx_insn *insn;
  rtx val;
  int rpo_cnt;
  int *rpo;

  timevar_push (TV_ALIAS_ANALYSIS);

  vec_safe_grow_cleared (reg_known_value, maxreg - FIRST_PSEUDO_REGISTER);
  reg_known_equiv_p = sbitmap_alloc (maxreg - FIRST_PSEUDO_REGISTER);
  bitmap_clear (reg_known_equiv_p);

  /* If we have memory allocated from the previous run, use it.  */
  if (old_reg_base_value)
    reg_base_value = old_reg_base_value;

  if (reg_base_value)
    reg_base_value->truncate (0);

  vec_safe_grow_cleared (reg_base_value, maxreg);

  new_reg_base_value = XNEWVEC (rtx, maxreg);
  reg_seen = sbitmap_alloc (maxreg);

  /* The basic idea is that each pass through this loop will use the
     "constant" information from the previous pass to propagate alias
     information through another level of assignments.

     The propagation is done on the CFG in reverse post-order, to propagate
     things forward as far as possible in each iteration.

     This could get expensive if the assignment chains are long.  Maybe
     we should throttle the number of iterations, possibly based on
     the optimization level or flag_expensive_optimizations.

     We could propagate more information in the first pass by making use
     of DF_REG_DEF_COUNT to determine immediately that the alias information
     for a pseudo is "constant".

     A program with an uninitialized variable can cause an infinite loop
     here.  Instead of doing a full dataflow analysis to detect such problems
     we just cap the number of iterations for the loop.

     The state of the arrays for the set chain in question does not matter
     since the program has undefined behavior.  */

  rpo = XNEWVEC (int, n_basic_blocks_for_fn (cfun));
  rpo_cnt = pre_and_rev_post_order_compute (NULL, rpo, false);

  /* The prologue/epilogue insns are not threaded onto the
     insn chain until after reload has completed.  Thus,
     there is no sense wasting time checking if INSN is in
     the prologue/epilogue until after reload has completed.  */
  bool could_be_prologue_epilogue = ((targetm.have_prologue ()
                                      || targetm.have_epilogue ())
                                     && reload_completed);

  pass = 0;
  do
    {
      /* Assume nothing will change this iteration of the loop.  */
      changed = 0;

      /* We want to assign the same IDs each iteration of this loop, so
         start counting from one each iteration of the loop.  */
      unique_id = 1;

      /* We're at the start of the function each iteration through the
         loop, so we're copying arguments.  */
      copying_arguments = true;

      /* Wipe the potential alias information clean for this pass.  */
      memset (new_reg_base_value, 0, maxreg * sizeof (rtx));

      /* Wipe the reg_seen array clean.  */
      bitmap_clear (reg_seen);

      /* Initialize the alias information for this pass.  */
      for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
        if (static_reg_base_value[i]
            /* Don't treat the hard frame pointer as special if we
               eliminated the frame pointer to the stack pointer instead.  */
            && !(i == HARD_FRAME_POINTER_REGNUM
                 && reload_completed
                 && !frame_pointer_needed
                 && targetm.can_eliminate (FRAME_POINTER_REGNUM,
                                           STACK_POINTER_REGNUM)))
          {
            new_reg_base_value[i] = static_reg_base_value[i];
            bitmap_set_bit (reg_seen, i);
          }

      /* Walk the insns adding values to the new_reg_base_value array.  */
      for (i = 0; i < rpo_cnt; i++)
        {
          basic_block bb = BASIC_BLOCK_FOR_FN (cfun, rpo[i]);
          FOR_BB_INSNS (bb, insn)
            {
              if (NONDEBUG_INSN_P (insn))
                {
                  rtx note, set;

                  if (could_be_prologue_epilogue
                      && prologue_epilogue_contains (insn))
                    continue;

                  /* If this insn has a noalias note, process it,  Otherwise,
                     scan for sets.  A simple set will have no side effects
                     which could change the base value of any other register.  */

                  if (GET_CODE (PATTERN (insn)) == SET
                      && REG_NOTES (insn) != 0
                      && find_reg_note (insn, REG_NOALIAS, NULL_RTX))
                    record_set (SET_DEST (PATTERN (insn)), NULL_RTX, NULL);
                  else
                    note_stores (PATTERN (insn), record_set, NULL);

                  set = single_set (insn);

                  if (set != 0
                      && REG_P (SET_DEST (set))
                      && REGNO (SET_DEST (set)) >= FIRST_PSEUDO_REGISTER)
                    {
                      unsigned int regno = REGNO (SET_DEST (set));
                      rtx src = SET_SRC (set);
                      rtx t;

                      note = find_reg_equal_equiv_note (insn);
                      if (note && REG_NOTE_KIND (note) == REG_EQUAL
                          && DF_REG_DEF_COUNT (regno) != 1)
                        note = NULL_RTX;

                      poly_int64 offset;
                      if (note != NULL_RTX
                          && GET_CODE (XEXP (note, 0)) != EXPR_LIST
                          && ! rtx_varies_p (XEXP (note, 0), 1)
                          && ! reg_overlap_mentioned_p (SET_DEST (set),
                                                        XEXP (note, 0)))
                        {
                          set_reg_known_value (regno, XEXP (note, 0));
                          set_reg_known_equiv_p (regno,
                                                 REG_NOTE_KIND (note) == REG_EQUIV);
                        }
                      else if (DF_REG_DEF_COUNT (regno) == 1
                               && GET_CODE (src) == PLUS
                               && REG_P (XEXP (src, 0))
                               && (t = get_reg_known_value (REGNO (XEXP (src, 0))))
                               && poly_int_rtx_p (XEXP (src, 1), &offset))
                        {
                          t = plus_constant (GET_MODE (src), t, offset);
                          set_reg_known_value (regno, t);
                          set_reg_known_equiv_p (regno, false);
                        }
                      else if (DF_REG_DEF_COUNT (regno) == 1
                               && ! rtx_varies_p (src, 1))
                        {
                          set_reg_known_value (regno, src);
                          set_reg_known_equiv_p (regno, false);
                        }
                    }
                }
              else if (NOTE_P (insn)
                       && NOTE_KIND (insn) == NOTE_INSN_FUNCTION_BEG)
                copying_arguments = false;
            }
        }

      /* Now propagate values from new_reg_base_value to reg_base_value.  */
      gcc_assert (maxreg == (unsigned int) max_reg_num ());

      for (ui = 0; ui < maxreg; ui++)
        {
          if (new_reg_base_value[ui]
              && new_reg_base_value[ui] != (*reg_base_value)[ui]
              && ! rtx_equal_p (new_reg_base_value[ui], (*reg_base_value)[ui]))
            {
              (*reg_base_value)[ui] = new_reg_base_value[ui];
              changed = 1;
            }
        }
    }
  while (changed && ++pass < MAX_ALIAS_LOOP_PASSES);
  XDELETEVEC (rpo);

  /* Fill in the remaining entries.  */
  FOR_EACH_VEC_ELT (*reg_known_value, i, val)
    {
      int regno = i + FIRST_PSEUDO_REGISTER;
      if (! val)
        set_reg_known_value (regno, regno_reg_rtx[regno]);
    }

  /* Clean up.  */
  free (new_reg_base_value);
  new_reg_base_value = 0;
  sbitmap_free (reg_seen);
  reg_seen = 0;
  timevar_pop (TV_ALIAS_ANALYSIS);
}

/* Equate REG_BASE_VALUE (reg1) to REG_BASE_VALUE (reg2).
   Special API for var-tracking pass purposes.  */

void
vt_equate_reg_base_value (const_rtx reg1, const_rtx reg2)
{
  (*reg_base_value)[REGNO (reg1)] = REG_BASE_VALUE (reg2);
}

void
end_alias_analysis (void)
{
  old_reg_base_value = reg_base_value;
  vec_free (reg_known_value);
  sbitmap_free (reg_known_equiv_p);
}

void
dump_alias_stats_in_alias_c (FILE *s)
{
  fprintf (s, "  TBAA oracle: %llu disambiguations %llu queries\n"
              "               %llu are in alias set 0\n"
              "               %llu queries asked about the same object\n"
              "               %llu queries asked about the same alias set\n"
              "               %llu access volatile\n"
              "               %llu are dependent in the DAG\n"
              "               %llu are aritificially in conflict with void *\n",
           alias_stats.num_disambiguated,
           alias_stats.num_alias_zero + alias_stats.num_same_alias_set
           + alias_stats.num_same_objects + alias_stats.num_volatile
           + alias_stats.num_dag + alias_stats.num_disambiguated
           + alias_stats.num_universal,
           alias_stats.num_alias_zero, alias_stats.num_same_alias_set,
           alias_stats.num_same_objects, alias_stats.num_volatile,
           alias_stats.num_dag, alias_stats.num_universal);
}
#include "gt-alias.h"